From 5231eb9773c61a2a17590eabcadf4aecf44100bd Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Mon, 3 Aug 2015 13:02:41 -0400 Subject: xprtrdma: Make xprt_setup_rdma() agnostic to family of server address In particular, recognize when an IPv6 connection is bound. Signed-off-by: Chuck Lever Tested-by: Devesh Sharma Reviewed-by: Christoph Hellwig Signed-off-by: Anna Schumaker --- net/sunrpc/xprtrdma/transport.c | 28 +++++++++++----------------- 1 file changed, 11 insertions(+), 17 deletions(-) diff --git a/net/sunrpc/xprtrdma/transport.c b/net/sunrpc/xprtrdma/transport.c index 680f888a9ddd..d737300ebc42 100644 --- a/net/sunrpc/xprtrdma/transport.c +++ b/net/sunrpc/xprtrdma/transport.c @@ -175,10 +175,8 @@ xprt_rdma_format_addresses6(struct rpc_xprt *xprt, struct sockaddr *sap) } static void -xprt_rdma_format_addresses(struct rpc_xprt *xprt) +xprt_rdma_format_addresses(struct rpc_xprt *xprt, struct sockaddr *sap) { - struct sockaddr *sap = (struct sockaddr *) - &rpcx_to_rdmad(xprt).addr; char buf[128]; switch (sap->sa_family) { @@ -302,7 +300,7 @@ xprt_setup_rdma(struct xprt_create *args) struct rpc_xprt *xprt; struct rpcrdma_xprt *new_xprt; struct rpcrdma_ep *new_ep; - struct sockaddr_in *sin; + struct sockaddr *sap; int rc; if (args->addrlen > sizeof(xprt->addr)) { @@ -333,26 +331,20 @@ xprt_setup_rdma(struct xprt_create *args) * Set up RDMA-specific connect data. */ - /* Put server RDMA address in local cdata */ - memcpy(&cdata.addr, args->dstaddr, args->addrlen); + sap = (struct sockaddr *)&cdata.addr; + memcpy(sap, args->dstaddr, args->addrlen); /* Ensure xprt->addr holds valid server TCP (not RDMA) * address, for any side protocols which peek at it */ xprt->prot = IPPROTO_TCP; xprt->addrlen = args->addrlen; - memcpy(&xprt->addr, &cdata.addr, xprt->addrlen); + memcpy(&xprt->addr, sap, xprt->addrlen); - sin = (struct sockaddr_in *)&cdata.addr; - if (ntohs(sin->sin_port) != 0) + if (rpc_get_port(sap)) xprt_set_bound(xprt); - dprintk("RPC: %s: %pI4:%u\n", - __func__, &sin->sin_addr.s_addr, ntohs(sin->sin_port)); - - /* Set max requests */ cdata.max_requests = xprt->max_reqs; - /* Set some length limits */ cdata.rsize = RPCRDMA_MAX_SEGS * PAGE_SIZE; /* RDMA write max */ cdata.wsize = RPCRDMA_MAX_SEGS * PAGE_SIZE; /* RDMA read max */ @@ -375,8 +367,7 @@ xprt_setup_rdma(struct xprt_create *args) new_xprt = rpcx_to_rdmax(xprt); - rc = rpcrdma_ia_open(new_xprt, (struct sockaddr *) &cdata.addr, - xprt_rdma_memreg_strategy); + rc = rpcrdma_ia_open(new_xprt, sap, xprt_rdma_memreg_strategy); if (rc) goto out1; @@ -409,7 +400,7 @@ xprt_setup_rdma(struct xprt_create *args) INIT_DELAYED_WORK(&new_xprt->rx_connect_worker, xprt_rdma_connect_worker); - xprt_rdma_format_addresses(xprt); + xprt_rdma_format_addresses(xprt, sap); xprt->max_payload = new_xprt->rx_ia.ri_ops->ro_maxpages(new_xprt); if (xprt->max_payload == 0) goto out4; @@ -420,6 +411,9 @@ xprt_setup_rdma(struct xprt_create *args) if (!try_module_get(THIS_MODULE)) goto out4; + dprintk("RPC: %s: %s:%s\n", __func__, + xprt->address_strings[RPC_DISPLAY_ADDR], + xprt->address_strings[RPC_DISPLAY_PORT]); return xprt; out4: -- cgit v1.2.3 From 864be126fe5f325ec4b3e28173ca5cad2b8ee28c Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Mon, 3 Aug 2015 13:02:50 -0400 Subject: xprtrdma: Raise maximum payload size to one megabyte The point of larger rsize and wsize is to reduce the per-byte cost of memory registration and deregistration. Modern HCAs can typically handle a megabyte or more with a single registration operation. Signed-off-by: Chuck Lever Reviewed-by: Devesh Sharma Reviewed-By: Sagi Grimberg Tested-by: Devesh Sharma Signed-off-by: Anna Schumaker --- net/sunrpc/xprtrdma/xprt_rdma.h | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/net/sunrpc/xprtrdma/xprt_rdma.h b/net/sunrpc/xprtrdma/xprt_rdma.h index f49dd8b38122..abee4728f885 100644 --- a/net/sunrpc/xprtrdma/xprt_rdma.h +++ b/net/sunrpc/xprtrdma/xprt_rdma.h @@ -165,8 +165,7 @@ rdmab_to_msg(struct rpcrdma_regbuf *rb) * struct rpcrdma_buffer. N is the max number of outstanding requests. */ -/* temporary static scatter/gather max */ -#define RPCRDMA_MAX_DATA_SEGS (64) /* max scatter/gather */ +#define RPCRDMA_MAX_DATA_SEGS ((1 * 1024 * 1024) / PAGE_SIZE) #define RPCRDMA_MAX_SEGS (RPCRDMA_MAX_DATA_SEGS + 2) /* head+tail = 2 */ struct rpcrdma_buffer; -- cgit v1.2.3 From 061dff29f8f62c21c9222897e4d121b4a5fa50da Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Mon, 3 Aug 2015 13:02:59 -0400 Subject: xprtrdma: Increase default credit limit In preparation for similar increases on NFS/RDMA servers, bump the advertised credit limit for RPC/RDMA to 128. This allocates some extra resources, but the client will continue to allow only the number of RPCs in flight that the server requests via its advertised credit limit. Signed-off-by: Chuck Lever Reviewed-By: Sagi Grimberg Tested-by: Devesh Sharma Signed-off-by: Anna Schumaker --- include/linux/sunrpc/xprtrdma.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/include/linux/sunrpc/xprtrdma.h b/include/linux/sunrpc/xprtrdma.h index b17613052cc3..b7b279b54504 100644 --- a/include/linux/sunrpc/xprtrdma.h +++ b/include/linux/sunrpc/xprtrdma.h @@ -49,7 +49,7 @@ * a single chunk type per message is supported currently. */ #define RPCRDMA_MIN_SLOT_TABLE (2U) -#define RPCRDMA_DEF_SLOT_TABLE (32U) +#define RPCRDMA_DEF_SLOT_TABLE (128U) #define RPCRDMA_MAX_SLOT_TABLE (256U) #define RPCRDMA_DEF_INLINE (1024) /* default inline max */ -- cgit v1.2.3 From d23109302390d61d83675a86453674446eccb776 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Mon, 3 Aug 2015 13:03:09 -0400 Subject: xprtrdma: Don't fall back to PHYSICAL memory registration PHYSICAL memory registration uses a single rkey for all of the client's memory, thus is insecure. It is still useful in some cases for testing. Retain the ability to select PHYSICAL memory registration capability via /proc/sys/sunrpc/rdma_memreg_strategy, but don't fall back to it if the HCA does not support FRWR or FMR. This means amso1100 no longer works out of the box with NFS/RDMA. When using amso1100 HCAs, set the memreg_strategy sysctl to 6 before performing NFS/RDMA mounts. Signed-off-by: Chuck Lever Tested-by: Devesh Sharma Reviewed-by: Christoph Hellwig Signed-off-by: Anna Schumaker --- net/sunrpc/xprtrdma/verbs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c index 891c4ede2c20..1065808d25b4 100644 --- a/net/sunrpc/xprtrdma/verbs.c +++ b/net/sunrpc/xprtrdma/verbs.c @@ -539,7 +539,7 @@ rpcrdma_ia_open(struct rpcrdma_xprt *xprt, struct sockaddr *addr, int memreg) if (!ia->ri_device->alloc_fmr) { dprintk("RPC: %s: MTHCAFMR registration " "not supported by HCA\n", __func__); - memreg = RPCRDMA_ALLPHYSICAL; + goto out3; } } -- cgit v1.2.3 From e531dcabec8dc2ee141aab01ddf20ca87c52d916 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Mon, 3 Aug 2015 13:03:20 -0400 Subject: xprtrdma: Remove last ib_reg_phys_mr() call site All HCA providers have an ib_get_dma_mr() verb. Thus rpcrdma_ia_open() will either grab the device's local_dma_key if one is available, or it will call ib_get_dma_mr(). If ib_get_dma_mr() fails, rpcrdma_ia_open() fails and no transport is created. Therefore execution never reaches the ib_reg_phys_mr() call site in rpcrdma_register_internal(), so it can be removed. The remaining logic in rpcrdma_{de}register_internal() is folded into rpcrdma_{alloc,free}_regbuf(). This is clean up only. No behavior change is expected. Signed-off-by: Chuck Lever Reviewed-by: Devesh Sharma Reviewed-By: Sagi Grimberg Tested-by: Devesh Sharma Signed-off-by: Anna Schumaker --- net/sunrpc/xprtrdma/verbs.c | 102 +++++++++------------------------------- net/sunrpc/xprtrdma/xprt_rdma.h | 1 - 2 files changed, 21 insertions(+), 82 deletions(-) diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c index 1065808d25b4..da184f98fdf9 100644 --- a/net/sunrpc/xprtrdma/verbs.c +++ b/net/sunrpc/xprtrdma/verbs.c @@ -1229,75 +1229,6 @@ rpcrdma_mapping_error(struct rpcrdma_mr_seg *seg) (unsigned long long)seg->mr_dma, seg->mr_dmalen); } -static int -rpcrdma_register_internal(struct rpcrdma_ia *ia, void *va, int len, - struct ib_mr **mrp, struct ib_sge *iov) -{ - struct ib_phys_buf ipb; - struct ib_mr *mr; - int rc; - - /* - * All memory passed here was kmalloc'ed, therefore phys-contiguous. - */ - iov->addr = ib_dma_map_single(ia->ri_device, - va, len, DMA_BIDIRECTIONAL); - if (ib_dma_mapping_error(ia->ri_device, iov->addr)) - return -ENOMEM; - - iov->length = len; - - if (ia->ri_have_dma_lkey) { - *mrp = NULL; - iov->lkey = ia->ri_dma_lkey; - return 0; - } else if (ia->ri_bind_mem != NULL) { - *mrp = NULL; - iov->lkey = ia->ri_bind_mem->lkey; - return 0; - } - - ipb.addr = iov->addr; - ipb.size = iov->length; - mr = ib_reg_phys_mr(ia->ri_pd, &ipb, 1, - IB_ACCESS_LOCAL_WRITE, &iov->addr); - - dprintk("RPC: %s: phys convert: 0x%llx " - "registered 0x%llx length %d\n", - __func__, (unsigned long long)ipb.addr, - (unsigned long long)iov->addr, len); - - if (IS_ERR(mr)) { - *mrp = NULL; - rc = PTR_ERR(mr); - dprintk("RPC: %s: failed with %i\n", __func__, rc); - } else { - *mrp = mr; - iov->lkey = mr->lkey; - rc = 0; - } - - return rc; -} - -static int -rpcrdma_deregister_internal(struct rpcrdma_ia *ia, - struct ib_mr *mr, struct ib_sge *iov) -{ - int rc; - - ib_dma_unmap_single(ia->ri_device, - iov->addr, iov->length, DMA_BIDIRECTIONAL); - - if (NULL == mr) - return 0; - - rc = ib_dereg_mr(mr); - if (rc) - dprintk("RPC: %s: ib_dereg_mr failed %i\n", __func__, rc); - return rc; -} - /** * rpcrdma_alloc_regbuf - kmalloc and register memory for SEND/RECV buffers * @ia: controlling rpcrdma_ia @@ -1317,26 +1248,30 @@ struct rpcrdma_regbuf * rpcrdma_alloc_regbuf(struct rpcrdma_ia *ia, size_t size, gfp_t flags) { struct rpcrdma_regbuf *rb; - int rc; + struct ib_sge *iov; - rc = -ENOMEM; rb = kmalloc(sizeof(*rb) + size, flags); if (rb == NULL) goto out; - rb->rg_size = size; - rb->rg_owner = NULL; - rc = rpcrdma_register_internal(ia, rb->rg_base, size, - &rb->rg_mr, &rb->rg_iov); - if (rc) + iov = &rb->rg_iov; + iov->addr = ib_dma_map_single(ia->ri_device, + (void *)rb->rg_base, size, + DMA_BIDIRECTIONAL); + if (ib_dma_mapping_error(ia->ri_device, iov->addr)) goto out_free; + iov->length = size; + iov->lkey = ia->ri_have_dma_lkey ? + ia->ri_dma_lkey : ia->ri_bind_mem->lkey; + rb->rg_size = size; + rb->rg_owner = NULL; return rb; out_free: kfree(rb); out: - return ERR_PTR(rc); + return ERR_PTR(-ENOMEM); } /** @@ -1347,10 +1282,15 @@ out: void rpcrdma_free_regbuf(struct rpcrdma_ia *ia, struct rpcrdma_regbuf *rb) { - if (rb) { - rpcrdma_deregister_internal(ia, rb->rg_mr, &rb->rg_iov); - kfree(rb); - } + struct ib_sge *iov; + + if (!rb) + return; + + iov = &rb->rg_iov; + ib_dma_unmap_single(ia->ri_device, + iov->addr, iov->length, DMA_BIDIRECTIONAL); + kfree(rb); } /* diff --git a/net/sunrpc/xprtrdma/xprt_rdma.h b/net/sunrpc/xprtrdma/xprt_rdma.h index abee4728f885..ce4e79e4cf1d 100644 --- a/net/sunrpc/xprtrdma/xprt_rdma.h +++ b/net/sunrpc/xprtrdma/xprt_rdma.h @@ -119,7 +119,6 @@ struct rpcrdma_ep { struct rpcrdma_regbuf { size_t rg_size; struct rpcrdma_req *rg_owner; - struct ib_mr *rg_mr; struct ib_sge rg_iov; __be32 rg_base[0] __attribute__ ((aligned(256))); }; -- cgit v1.2.3 From d1ed857e5707e073973cfb1b8df801053a356518 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Mon, 3 Aug 2015 13:03:30 -0400 Subject: xprtrdma: Clean up rpcrdma_ia_open() Untangle the end of rpcrdma_ia_open() by moving DMA MR set-up, which is different for each registration method, to the .ro_open functions. This is refactoring only. No behavior change is expected. Signed-off-by: Chuck Lever Tested-by: Devesh Sharma Signed-off-by: Anna Schumaker --- net/sunrpc/xprtrdma/fmr_ops.c | 19 ++++++++++++ net/sunrpc/xprtrdma/frwr_ops.c | 5 ++++ net/sunrpc/xprtrdma/physical_ops.c | 25 +++++++++++++++- net/sunrpc/xprtrdma/verbs.c | 60 ++++++++++++-------------------------- net/sunrpc/xprtrdma/xprt_rdma.h | 3 +- 5 files changed, 67 insertions(+), 45 deletions(-) diff --git a/net/sunrpc/xprtrdma/fmr_ops.c b/net/sunrpc/xprtrdma/fmr_ops.c index f1e8dafbd507..cb25c89da623 100644 --- a/net/sunrpc/xprtrdma/fmr_ops.c +++ b/net/sunrpc/xprtrdma/fmr_ops.c @@ -39,6 +39,25 @@ static int fmr_op_open(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep, struct rpcrdma_create_data_internal *cdata) { + struct ib_device_attr *devattr = &ia->ri_devattr; + struct ib_mr *mr; + + /* Obtain an lkey to use for the regbufs, which are + * protected from remote access. + */ + if (devattr->device_cap_flags & IB_DEVICE_LOCAL_DMA_LKEY) { + ia->ri_dma_lkey = ia->ri_device->local_dma_lkey; + } else { + mr = ib_get_dma_mr(ia->ri_pd, IB_ACCESS_LOCAL_WRITE); + if (IS_ERR(mr)) { + pr_err("%s: ib_get_dma_mr for failed with %lX\n", + __func__, PTR_ERR(mr)); + return -ENOMEM; + } + ia->ri_dma_lkey = ia->ri_dma_mr->lkey; + ia->ri_dma_mr = mr; + } + return 0; } diff --git a/net/sunrpc/xprtrdma/frwr_ops.c b/net/sunrpc/xprtrdma/frwr_ops.c index 04ea914201b2..63f282e770b8 100644 --- a/net/sunrpc/xprtrdma/frwr_ops.c +++ b/net/sunrpc/xprtrdma/frwr_ops.c @@ -189,6 +189,11 @@ frwr_op_open(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep, struct ib_device_attr *devattr = &ia->ri_devattr; int depth, delta; + /* Obtain an lkey to use for the regbufs, which are + * protected from remote access. + */ + ia->ri_dma_lkey = ia->ri_device->local_dma_lkey; + ia->ri_max_frmr_depth = min_t(unsigned int, RPCRDMA_MAX_DATA_SEGS, devattr->max_fast_reg_page_list_len); diff --git a/net/sunrpc/xprtrdma/physical_ops.c b/net/sunrpc/xprtrdma/physical_ops.c index 41985d07fdb7..72cf8b15bbb4 100644 --- a/net/sunrpc/xprtrdma/physical_ops.c +++ b/net/sunrpc/xprtrdma/physical_ops.c @@ -23,6 +23,29 @@ static int physical_op_open(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep, struct rpcrdma_create_data_internal *cdata) { + struct ib_device_attr *devattr = &ia->ri_devattr; + struct ib_mr *mr; + + /* Obtain an rkey to use for RPC data payloads. + */ + mr = ib_get_dma_mr(ia->ri_pd, + IB_ACCESS_LOCAL_WRITE | + IB_ACCESS_REMOTE_WRITE | + IB_ACCESS_REMOTE_READ); + if (IS_ERR(mr)) { + pr_err("%s: ib_get_dma_mr for failed with %lX\n", + __func__, PTR_ERR(mr)); + return -ENOMEM; + } + ia->ri_dma_mr = mr; + + /* Obtain an lkey to use for regbufs. + */ + if (devattr->device_cap_flags & IB_DEVICE_LOCAL_DMA_LKEY) + ia->ri_dma_lkey = ia->ri_device->local_dma_lkey; + else + ia->ri_dma_lkey = ia->ri_dma_mr->lkey; + return 0; } @@ -51,7 +74,7 @@ physical_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg, struct rpcrdma_ia *ia = &r_xprt->rx_ia; rpcrdma_map_one(ia->ri_device, seg, rpcrdma_data_dir(writing)); - seg->mr_rkey = ia->ri_bind_mem->rkey; + seg->mr_rkey = ia->ri_dma_mr->rkey; seg->mr_base = seg->mr_dma; seg->mr_nsegs = 1; return 1; diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c index da184f98fdf9..8516d9894599 100644 --- a/net/sunrpc/xprtrdma/verbs.c +++ b/net/sunrpc/xprtrdma/verbs.c @@ -493,9 +493,11 @@ rpcrdma_clean_cq(struct ib_cq *cq) int rpcrdma_ia_open(struct rpcrdma_xprt *xprt, struct sockaddr *addr, int memreg) { - int rc, mem_priv; struct rpcrdma_ia *ia = &xprt->rx_ia; struct ib_device_attr *devattr = &ia->ri_devattr; + int rc; + + ia->ri_dma_mr = NULL; ia->ri_id = rpcrdma_create_id(xprt, ia, addr); if (IS_ERR(ia->ri_id)) { @@ -519,11 +521,6 @@ rpcrdma_ia_open(struct rpcrdma_xprt *xprt, struct sockaddr *addr, int memreg) goto out3; } - if (devattr->device_cap_flags & IB_DEVICE_LOCAL_DMA_LKEY) { - ia->ri_have_dma_lkey = 1; - ia->ri_dma_lkey = ia->ri_device->local_dma_lkey; - } - if (memreg == RPCRDMA_FRMR) { /* Requires both frmr reg and local dma lkey */ if (((devattr->device_cap_flags & @@ -543,38 +540,15 @@ rpcrdma_ia_open(struct rpcrdma_xprt *xprt, struct sockaddr *addr, int memreg) } } - /* - * Optionally obtain an underlying physical identity mapping in - * order to do a memory window-based bind. This base registration - * is protected from remote access - that is enabled only by binding - * for the specific bytes targeted during each RPC operation, and - * revoked after the corresponding completion similar to a storage - * adapter. - */ switch (memreg) { case RPCRDMA_FRMR: ia->ri_ops = &rpcrdma_frwr_memreg_ops; break; case RPCRDMA_ALLPHYSICAL: ia->ri_ops = &rpcrdma_physical_memreg_ops; - mem_priv = IB_ACCESS_LOCAL_WRITE | - IB_ACCESS_REMOTE_WRITE | - IB_ACCESS_REMOTE_READ; - goto register_setup; + break; case RPCRDMA_MTHCAFMR: ia->ri_ops = &rpcrdma_fmr_memreg_ops; - if (ia->ri_have_dma_lkey) - break; - mem_priv = IB_ACCESS_LOCAL_WRITE; - register_setup: - ia->ri_bind_mem = ib_get_dma_mr(ia->ri_pd, mem_priv); - if (IS_ERR(ia->ri_bind_mem)) { - printk(KERN_ALERT "%s: ib_get_dma_mr for " - "phys register failed with %lX\n", - __func__, PTR_ERR(ia->ri_bind_mem)); - rc = -ENOMEM; - goto out3; - } break; default: printk(KERN_ERR "RPC: Unsupported memory " @@ -606,15 +580,7 @@ out1: void rpcrdma_ia_close(struct rpcrdma_ia *ia) { - int rc; - dprintk("RPC: %s: entering\n", __func__); - if (ia->ri_bind_mem != NULL) { - rc = ib_dereg_mr(ia->ri_bind_mem); - dprintk("RPC: %s: ib_dereg_mr returned %i\n", - __func__, rc); - } - if (ia->ri_id != NULL && !IS_ERR(ia->ri_id)) { if (ia->ri_id->qp) rdma_destroy_qp(ia->ri_id); @@ -661,8 +627,10 @@ rpcrdma_ep_create(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia, if (cdata->padding) { ep->rep_padbuf = rpcrdma_alloc_regbuf(ia, cdata->padding, GFP_KERNEL); - if (IS_ERR(ep->rep_padbuf)) - return PTR_ERR(ep->rep_padbuf); + if (IS_ERR(ep->rep_padbuf)) { + rc = PTR_ERR(ep->rep_padbuf); + goto out0; + } } else ep->rep_padbuf = NULL; @@ -749,6 +717,9 @@ out2: __func__, err); out1: rpcrdma_free_regbuf(ia, ep->rep_padbuf); +out0: + if (ia->ri_dma_mr) + ib_dereg_mr(ia->ri_dma_mr); return rc; } @@ -788,6 +759,12 @@ rpcrdma_ep_destroy(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia) if (rc) dprintk("RPC: %s: ib_destroy_cq returned %i\n", __func__, rc); + + if (ia->ri_dma_mr) { + rc = ib_dereg_mr(ia->ri_dma_mr); + dprintk("RPC: %s: ib_dereg_mr returned %i\n", + __func__, rc); + } } /* @@ -1262,8 +1239,7 @@ rpcrdma_alloc_regbuf(struct rpcrdma_ia *ia, size_t size, gfp_t flags) goto out_free; iov->length = size; - iov->lkey = ia->ri_have_dma_lkey ? - ia->ri_dma_lkey : ia->ri_bind_mem->lkey; + iov->lkey = ia->ri_dma_lkey; rb->rg_size = size; rb->rg_owner = NULL; return rb; diff --git a/net/sunrpc/xprtrdma/xprt_rdma.h b/net/sunrpc/xprtrdma/xprt_rdma.h index ce4e79e4cf1d..82190118b8d9 100644 --- a/net/sunrpc/xprtrdma/xprt_rdma.h +++ b/net/sunrpc/xprtrdma/xprt_rdma.h @@ -65,9 +65,8 @@ struct rpcrdma_ia { struct ib_device *ri_device; struct rdma_cm_id *ri_id; struct ib_pd *ri_pd; - struct ib_mr *ri_bind_mem; + struct ib_mr *ri_dma_mr; u32 ri_dma_lkey; - int ri_have_dma_lkey; struct completion ri_done; int ri_async_rc; unsigned int ri_max_frmr_depth; -- cgit v1.2.3 From b3221d6a53c44cd572a3a400abdd1e2a24bea587 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Mon, 3 Aug 2015 13:03:39 -0400 Subject: xprtrdma: Remove logic that constructs RDMA_MSGP type calls RDMA_MSGP type calls insert a zero pad in the middle of the RPC message to align the RPC request's data payload to the server's alignment preferences. A server can then "page flip" the payload into place to avoid a data copy in certain circumstances. However: 1. The client has to have a priori knowledge of the server's preferred alignment 2. Requests eligible for RDMA_MSGP are requests that are small enough to have been sent inline, and convey a data payload at the _end_ of the RPC message Today 1. is done with a sysctl, and is a global setting that is copied during mount. Linux does not support CCP to query the server's preferences (RFC 5666, Section 6). A small-ish NFSv3 WRITE might use RDMA_MSGP, but no NFSv4 compound fits bullet 2. Thus the Linux client currently leaves RDMA_MSGP disabled. The Linux server handles RDMA_MSGP, but does not use any special page flipping, so it confers no benefit. Clean up the marshaling code by removing the logic that constructs RDMA_MSGP type calls. This also reduces the maximum send iovec size from four to just two elements. /proc/sys/sunrpc/rdma_inline_write_padding is a kernel API, and thus is left in place. Signed-off-by: Chuck Lever Tested-by: Devesh Sharma Signed-off-by: Anna Schumaker --- net/sunrpc/xprtrdma/rpc_rdma.c | 92 +++++++++++------------------------------ net/sunrpc/xprtrdma/verbs.c | 47 ++++++++------------- net/sunrpc/xprtrdma/xprt_rdma.h | 19 +++++---- 3 files changed, 51 insertions(+), 107 deletions(-) diff --git a/net/sunrpc/xprtrdma/rpc_rdma.c b/net/sunrpc/xprtrdma/rpc_rdma.c index 84ea37daef36..8e9c56429ada 100644 --- a/net/sunrpc/xprtrdma/rpc_rdma.c +++ b/net/sunrpc/xprtrdma/rpc_rdma.c @@ -297,8 +297,7 @@ out: * pre-registered memory buffer for this request. For small amounts * of data, this is efficient. The cutoff value is tunable. */ -static int -rpcrdma_inline_pullup(struct rpc_rqst *rqst, int pad) +static void rpcrdma_inline_pullup(struct rpc_rqst *rqst) { int i, npages, curlen; int copy_len; @@ -310,16 +309,9 @@ rpcrdma_inline_pullup(struct rpc_rqst *rqst, int pad) destp = rqst->rq_svec[0].iov_base; curlen = rqst->rq_svec[0].iov_len; destp += curlen; - /* - * Do optional padding where it makes sense. Alignment of write - * payload can help the server, if our setting is accurate. - */ - pad -= (curlen + 36/*sizeof(struct rpcrdma_msg_padded)*/); - if (pad < 0 || rqst->rq_slen - curlen < RPCRDMA_INLINE_PAD_THRESH) - pad = 0; /* don't pad this request */ - dprintk("RPC: %s: pad %d destp 0x%p len %d hdrlen %d\n", - __func__, pad, destp, rqst->rq_slen, curlen); + dprintk("RPC: %s: destp 0x%p len %d hdrlen %d\n", + __func__, destp, rqst->rq_slen, curlen); copy_len = rqst->rq_snd_buf.page_len; @@ -355,7 +347,6 @@ rpcrdma_inline_pullup(struct rpc_rqst *rqst, int pad) page_base = 0; } /* header now contains entire send message */ - return pad; } /* @@ -380,7 +371,7 @@ rpcrdma_marshal_req(struct rpc_rqst *rqst) struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt); struct rpcrdma_req *req = rpcr_to_rdmar(rqst); char *base; - size_t rpclen, padlen; + size_t rpclen; ssize_t hdrlen; enum rpcrdma_chunktype rtype, wtype; struct rpcrdma_msg *headerp; @@ -458,7 +449,6 @@ rpcrdma_marshal_req(struct rpc_rqst *rqst) } hdrlen = RPCRDMA_HDRLEN_MIN; - padlen = 0; /* * Pull up any extra send data into the preregistered buffer. @@ -467,43 +457,24 @@ rpcrdma_marshal_req(struct rpc_rqst *rqst) */ if (rtype == rpcrdma_noch) { - padlen = rpcrdma_inline_pullup(rqst, - RPCRDMA_INLINE_PAD_VALUE(rqst)); - - if (padlen) { - headerp->rm_type = rdma_msgp; - headerp->rm_body.rm_padded.rm_align = - cpu_to_be32(RPCRDMA_INLINE_PAD_VALUE(rqst)); - headerp->rm_body.rm_padded.rm_thresh = - cpu_to_be32(RPCRDMA_INLINE_PAD_THRESH); - headerp->rm_body.rm_padded.rm_pempty[0] = xdr_zero; - headerp->rm_body.rm_padded.rm_pempty[1] = xdr_zero; - headerp->rm_body.rm_padded.rm_pempty[2] = xdr_zero; - hdrlen += 2 * sizeof(u32); /* extra words in padhdr */ - if (wtype != rpcrdma_noch) { - dprintk("RPC: %s: invalid chunk list\n", - __func__); - return -EIO; - } - } else { - headerp->rm_body.rm_nochunks.rm_empty[0] = xdr_zero; - headerp->rm_body.rm_nochunks.rm_empty[1] = xdr_zero; - headerp->rm_body.rm_nochunks.rm_empty[2] = xdr_zero; - /* new length after pullup */ - rpclen = rqst->rq_svec[0].iov_len; - /* - * Currently we try to not actually use read inline. - * Reply chunks have the desirable property that - * they land, packed, directly in the target buffers - * without headers, so they require no fixup. The - * additional RDMA Write op sends the same amount - * of data, streams on-the-wire and adds no overhead - * on receive. Therefore, we request a reply chunk - * for non-writes wherever feasible and efficient. - */ - if (wtype == rpcrdma_noch) - wtype = rpcrdma_replych; - } + rpcrdma_inline_pullup(rqst); + + headerp->rm_body.rm_nochunks.rm_empty[0] = xdr_zero; + headerp->rm_body.rm_nochunks.rm_empty[1] = xdr_zero; + headerp->rm_body.rm_nochunks.rm_empty[2] = xdr_zero; + /* new length after pullup */ + rpclen = rqst->rq_svec[0].iov_len; + /* Currently we try to not actually use read inline. + * Reply chunks have the desirable property that + * they land, packed, directly in the target buffers + * without headers, so they require no fixup. The + * additional RDMA Write op sends the same amount + * of data, streams on-the-wire and adds no overhead + * on receive. Therefore, we request a reply chunk + * for non-writes wherever feasible and efficient. + */ + if (wtype == rpcrdma_noch) + wtype = rpcrdma_replych; } if (rtype != rpcrdma_noch) { @@ -518,9 +489,9 @@ rpcrdma_marshal_req(struct rpc_rqst *rqst) if (hdrlen < 0) return hdrlen; - dprintk("RPC: %s: %s: hdrlen %zd rpclen %zd padlen %zd" + dprintk("RPC: %s: %s: hdrlen %zd rpclen %zd" " headerp 0x%p base 0x%p lkey 0x%x\n", - __func__, transfertypes[wtype], hdrlen, rpclen, padlen, + __func__, transfertypes[wtype], hdrlen, rpclen, headerp, base, rdmab_lkey(req->rl_rdmabuf)); /* @@ -539,21 +510,6 @@ rpcrdma_marshal_req(struct rpc_rqst *rqst) req->rl_send_iov[1].lkey = rdmab_lkey(req->rl_sendbuf); req->rl_niovs = 2; - - if (padlen) { - struct rpcrdma_ep *ep = &r_xprt->rx_ep; - - req->rl_send_iov[2].addr = rdmab_addr(ep->rep_padbuf); - req->rl_send_iov[2].length = padlen; - req->rl_send_iov[2].lkey = rdmab_lkey(ep->rep_padbuf); - - req->rl_send_iov[3].addr = req->rl_send_iov[1].addr + rpclen; - req->rl_send_iov[3].length = rqst->rq_slen - rpclen; - req->rl_send_iov[3].lkey = rdmab_lkey(req->rl_sendbuf); - - req->rl_niovs = 4; - } - return 0; } diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c index 8516d9894599..b4d4f6300fbc 100644 --- a/net/sunrpc/xprtrdma/verbs.c +++ b/net/sunrpc/xprtrdma/verbs.c @@ -605,6 +605,12 @@ rpcrdma_ep_create(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia, struct ib_cq_init_attr cq_attr = {}; int rc, err; + if (devattr->max_sge < RPCRDMA_MAX_IOVS) { + dprintk("RPC: %s: insufficient sge's available\n", + __func__); + return -ENOMEM; + } + /* check provider's send/recv wr limits */ if (cdata->max_requests > devattr->max_qp_wr) cdata->max_requests = devattr->max_qp_wr; @@ -617,23 +623,13 @@ rpcrdma_ep_create(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia, if (rc) return rc; ep->rep_attr.cap.max_recv_wr = cdata->max_requests; - ep->rep_attr.cap.max_send_sge = (cdata->padding ? 4 : 2); + ep->rep_attr.cap.max_send_sge = RPCRDMA_MAX_IOVS; ep->rep_attr.cap.max_recv_sge = 1; ep->rep_attr.cap.max_inline_data = 0; ep->rep_attr.sq_sig_type = IB_SIGNAL_REQ_WR; ep->rep_attr.qp_type = IB_QPT_RC; ep->rep_attr.port_num = ~0; - if (cdata->padding) { - ep->rep_padbuf = rpcrdma_alloc_regbuf(ia, cdata->padding, - GFP_KERNEL); - if (IS_ERR(ep->rep_padbuf)) { - rc = PTR_ERR(ep->rep_padbuf); - goto out0; - } - } else - ep->rep_padbuf = NULL; - dprintk("RPC: %s: requested max: dtos: send %d recv %d; " "iovs: send %d recv %d\n", __func__, @@ -716,8 +712,6 @@ out2: dprintk("RPC: %s: ib_destroy_cq returned %i\n", __func__, err); out1: - rpcrdma_free_regbuf(ia, ep->rep_padbuf); -out0: if (ia->ri_dma_mr) ib_dereg_mr(ia->ri_dma_mr); return rc; @@ -746,8 +740,6 @@ rpcrdma_ep_destroy(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia) ia->ri_id->qp = NULL; } - rpcrdma_free_regbuf(ia, ep->rep_padbuf); - rpcrdma_clean_cq(ep->rep_attr.recv_cq); rc = ib_destroy_cq(ep->rep_attr.recv_cq); if (rc) @@ -1279,9 +1271,11 @@ rpcrdma_ep_post(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep, struct rpcrdma_req *req) { + struct ib_device *device = ia->ri_device; struct ib_send_wr send_wr, *send_wr_fail; struct rpcrdma_rep *rep = req->rl_reply; - int rc; + struct ib_sge *iov = req->rl_send_iov; + int i, rc; if (rep) { rc = rpcrdma_ep_post_recv(ia, ep, rep); @@ -1292,22 +1286,15 @@ rpcrdma_ep_post(struct rpcrdma_ia *ia, send_wr.next = NULL; send_wr.wr_id = RPCRDMA_IGNORE_COMPLETION; - send_wr.sg_list = req->rl_send_iov; + send_wr.sg_list = iov; send_wr.num_sge = req->rl_niovs; send_wr.opcode = IB_WR_SEND; - if (send_wr.num_sge == 4) /* no need to sync any pad (constant) */ - ib_dma_sync_single_for_device(ia->ri_device, - req->rl_send_iov[3].addr, - req->rl_send_iov[3].length, - DMA_TO_DEVICE); - ib_dma_sync_single_for_device(ia->ri_device, - req->rl_send_iov[1].addr, - req->rl_send_iov[1].length, - DMA_TO_DEVICE); - ib_dma_sync_single_for_device(ia->ri_device, - req->rl_send_iov[0].addr, - req->rl_send_iov[0].length, - DMA_TO_DEVICE); + + for (i = 0; i < send_wr.num_sge; i++) + ib_dma_sync_single_for_device(device, iov[i].addr, + iov[i].length, DMA_TO_DEVICE); + dprintk("RPC: %s: posting %d s/g entries\n", + __func__, send_wr.num_sge); if (DECR_CQCOUNT(ep) > 0) send_wr.send_flags = 0; diff --git a/net/sunrpc/xprtrdma/xprt_rdma.h b/net/sunrpc/xprtrdma/xprt_rdma.h index 82190118b8d9..8422c09043b0 100644 --- a/net/sunrpc/xprtrdma/xprt_rdma.h +++ b/net/sunrpc/xprtrdma/xprt_rdma.h @@ -88,7 +88,6 @@ struct rpcrdma_ep { int rep_connected; struct ib_qp_init_attr rep_attr; wait_queue_head_t rep_connect_wait; - struct rpcrdma_regbuf *rep_padbuf; struct rdma_conn_param rep_remote_cma; struct sockaddr_storage rep_remote_addr; struct delayed_work rep_connect_worker; @@ -255,16 +254,18 @@ struct rpcrdma_mr_seg { /* chunk descriptors */ char *mr_offset; /* kva if no page, else offset */ }; +#define RPCRDMA_MAX_IOVS (2) + struct rpcrdma_req { - unsigned int rl_niovs; /* 0, 2 or 4 */ - unsigned int rl_nchunks; /* non-zero if chunks */ - unsigned int rl_connect_cookie; /* retry detection */ - struct rpcrdma_buffer *rl_buffer; /* home base for this structure */ + unsigned int rl_niovs; + unsigned int rl_nchunks; + unsigned int rl_connect_cookie; + struct rpcrdma_buffer *rl_buffer; struct rpcrdma_rep *rl_reply;/* holder for reply buffer */ - struct ib_sge rl_send_iov[4]; /* for active requests */ - struct rpcrdma_regbuf *rl_rdmabuf; - struct rpcrdma_regbuf *rl_sendbuf; - struct rpcrdma_mr_seg rl_segments[RPCRDMA_MAX_SEGS]; + struct ib_sge rl_send_iov[RPCRDMA_MAX_IOVS]; + struct rpcrdma_regbuf *rl_rdmabuf; + struct rpcrdma_regbuf *rl_sendbuf; + struct rpcrdma_mr_seg rl_segments[RPCRDMA_MAX_SEGS]; }; static inline struct rpcrdma_req * -- cgit v1.2.3 From 5457ced0b504b41afe9439a6533066dea2fc0e1a Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Mon, 3 Aug 2015 13:03:49 -0400 Subject: xprtrdma: Account for RPC/RDMA header size when deciding to inline MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When the size of the RPC message is near the inline threshold (1KB), the client would allow messages to be sent that were a few bytes too large. When marshaling RPC/RDMA requests, ensure the combined size of RPC/RDMA header and RPC header do not exceed the inline threshold. Endpoints typically reject RPC/RDMA messages that exceed the size of their receive buffers. The two server implementations I test with (Linux and Solaris) use receive buffers that are larger than the client’s inline threshold. Thus so far this has been benign, observed only by code inspection. Signed-off-by: Chuck Lever Reviewed-by: Devesh Sharma Tested-by: Devesh Sharma Signed-off-by: Anna Schumaker --- net/sunrpc/xprtrdma/rpc_rdma.c | 29 +++++++++++++++++++++++++++-- 1 file changed, 27 insertions(+), 2 deletions(-) diff --git a/net/sunrpc/xprtrdma/rpc_rdma.c b/net/sunrpc/xprtrdma/rpc_rdma.c index 8e9c56429ada..950b654bad80 100644 --- a/net/sunrpc/xprtrdma/rpc_rdma.c +++ b/net/sunrpc/xprtrdma/rpc_rdma.c @@ -71,6 +71,31 @@ static const char transfertypes[][12] = { }; #endif +/* The client can send a request inline as long as the RPCRDMA header + * plus the RPC call fit under the transport's inline limit. If the + * combined call message size exceeds that limit, the client must use + * the read chunk list for this operation. + */ +static bool rpcrdma_args_inline(struct rpc_rqst *rqst) +{ + unsigned int callsize = RPCRDMA_HDRLEN_MIN + rqst->rq_snd_buf.len; + + return callsize <= RPCRDMA_INLINE_WRITE_THRESHOLD(rqst); +} + +/* The client can't know how large the actual reply will be. Thus it + * plans for the largest possible reply for that particular ULP + * operation. If the maximum combined reply message size exceeds that + * limit, the client must provide a write list or a reply chunk for + * this request. + */ +static bool rpcrdma_results_inline(struct rpc_rqst *rqst) +{ + unsigned int repsize = RPCRDMA_HDRLEN_MIN + rqst->rq_rcv_buf.buflen; + + return repsize <= RPCRDMA_INLINE_READ_THRESHOLD(rqst); +} + /* * Chunk assembly from upper layer xdr_buf. * @@ -409,7 +434,7 @@ rpcrdma_marshal_req(struct rpc_rqst *rqst) * a READ, then use write chunks to separate the file data * into pages; otherwise use reply chunks. */ - if (rqst->rq_rcv_buf.buflen <= RPCRDMA_INLINE_READ_THRESHOLD(rqst)) + if (rpcrdma_results_inline(rqst)) wtype = rpcrdma_noch; else if (rqst->rq_rcv_buf.page_len == 0) wtype = rpcrdma_replych; @@ -432,7 +457,7 @@ rpcrdma_marshal_req(struct rpc_rqst *rqst) * implies the op is a write. * TBD check NFSv4 setacl */ - if (rqst->rq_snd_buf.len <= RPCRDMA_INLINE_WRITE_THRESHOLD(rqst)) + if (rpcrdma_args_inline(rqst)) rtype = rpcrdma_noch; else if (rqst->rq_snd_buf.page_len == 0) rtype = rpcrdma_areadch; -- cgit v1.2.3 From 02eb57d8f44caa582e297f51f3555d47767c5fe9 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Mon, 3 Aug 2015 13:03:58 -0400 Subject: xprtrdma: Always provide a write list when sending NFS READ The client has been setting up a reply chunk for NFS READs that are smaller than the inline threshold. This is not efficient: both the server and client CPUs have to copy the reply's data payload into and out of the memory region that is then transferred via RDMA. Using the write list, the data payload is moved by the device and no extra data copying is necessary. Signed-off-by: Chuck Lever Reviewed-by: Devesh Sharma Reviewed-By: Sagi Grimberg Tested-by: Devesh Sharma Signed-off-by: Anna Schumaker --- net/sunrpc/xprtrdma/rpc_rdma.c | 21 ++++----------------- 1 file changed, 4 insertions(+), 17 deletions(-) diff --git a/net/sunrpc/xprtrdma/rpc_rdma.c b/net/sunrpc/xprtrdma/rpc_rdma.c index 950b654bad80..e7cf976aff47 100644 --- a/net/sunrpc/xprtrdma/rpc_rdma.c +++ b/net/sunrpc/xprtrdma/rpc_rdma.c @@ -418,28 +418,15 @@ rpcrdma_marshal_req(struct rpc_rqst *rqst) /* * Chunks needed for results? * + * o Read ops return data as write chunk(s), header as inline. * o If the expected result is under the inline threshold, all ops * return as inline (but see later). * o Large non-read ops return as a single reply chunk. - * o Large read ops return data as write chunk(s), header as inline. - * - * Note: the NFS code sending down multiple result segments implies - * the op is one of read, readdir[plus], readlink or NFSv4 getacl. - */ - - /* - * This code can handle read chunks, write chunks OR reply - * chunks -- only one type. If the request is too big to fit - * inline, then we will choose read chunks. If the request is - * a READ, then use write chunks to separate the file data - * into pages; otherwise use reply chunks. */ - if (rpcrdma_results_inline(rqst)) - wtype = rpcrdma_noch; - else if (rqst->rq_rcv_buf.page_len == 0) - wtype = rpcrdma_replych; - else if (rqst->rq_rcv_buf.flags & XDRBUF_READ) + if (rqst->rq_rcv_buf.flags & XDRBUF_READ) wtype = rpcrdma_writech; + else if (rpcrdma_results_inline(rqst)) + wtype = rpcrdma_noch; else wtype = rpcrdma_replych; -- cgit v1.2.3 From 33943b2974734ca5e5bef583d09ddd1eded6a77b Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Mon, 3 Aug 2015 13:04:08 -0400 Subject: xprtrdma: Don't provide a reply chunk when expecting a short reply Currently Linux always offers a reply chunk, even when the reply can be sent inline (ie. is smaller than 1KB). On the client, registering a memory region can be expensive. A server may choose not to use the reply chunk, wasting the cost of the registration. This is a change only for RPC replies smaller than 1KB which the server constructs in the RPC reply send buffer. Because the elements of the reply must be XDR encoded, a copy-free data transfer has no benefit in this case. Signed-off-by: Chuck Lever Reviewed-by: Sagi Grimberg Tested-by: Devesh Sharma Signed-off-by: Anna Schumaker --- net/sunrpc/xprtrdma/rpc_rdma.c | 13 +------------ 1 file changed, 1 insertion(+), 12 deletions(-) diff --git a/net/sunrpc/xprtrdma/rpc_rdma.c b/net/sunrpc/xprtrdma/rpc_rdma.c index e7cf976aff47..62150ae2dc09 100644 --- a/net/sunrpc/xprtrdma/rpc_rdma.c +++ b/net/sunrpc/xprtrdma/rpc_rdma.c @@ -420,7 +420,7 @@ rpcrdma_marshal_req(struct rpc_rqst *rqst) * * o Read ops return data as write chunk(s), header as inline. * o If the expected result is under the inline threshold, all ops - * return as inline (but see later). + * return as inline. * o Large non-read ops return as a single reply chunk. */ if (rqst->rq_rcv_buf.flags & XDRBUF_READ) @@ -476,17 +476,6 @@ rpcrdma_marshal_req(struct rpc_rqst *rqst) headerp->rm_body.rm_nochunks.rm_empty[2] = xdr_zero; /* new length after pullup */ rpclen = rqst->rq_svec[0].iov_len; - /* Currently we try to not actually use read inline. - * Reply chunks have the desirable property that - * they land, packed, directly in the target buffers - * without headers, so they require no fixup. The - * additional RDMA Write op sends the same amount - * of data, streams on-the-wire and adds no overhead - * on receive. Therefore, we request a reply chunk - * for non-writes wherever feasible and efficient. - */ - if (wtype == rpcrdma_noch) - wtype = rpcrdma_replych; } if (rtype != rpcrdma_noch) { -- cgit v1.2.3 From 677eb17e94edfbbea3b7e628d8aa046930f102c3 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Mon, 3 Aug 2015 13:04:17 -0400 Subject: xprtrdma: Fix XDR tail buffer marshalling Currently xprtrdma appends an extra chunk element to the RPC/RDMA read chunk list of each NFSv4 WRITE compound. The extra element contains the final GETATTR operation in the compound. The result is an extra RDMA READ operation to transfer a very short piece of each NFS WRITE compound (typically 16 bytes). This is inefficient. It is also incorrect. The client is sending the trailing GETATTR at the same Position as the preceding WRITE data payload. Whether or not RFC 5667 allows the GETATTR to appear in a read chunk, RFC 5666 requires that these two separate RPC arguments appear at two distinct Positions. It can also be argued that the GETATTR operation is not bulk data, and therefore RFC 5667 forbids its appearance in a read chunk at all. Although RFC 5667 is not precise about when using a read list with NFSv4 COMPOUND is allowed, the intent is that only data arguments not touched by NFS (ie, read and write payloads) are to be sent using RDMA READ or WRITE. The NFS client constructs GETATTR arguments itself, and therefore is required to send the trailing GETATTR operation as additional inline content, not as a data payload. NB: This change is not backwards compatible. Some older servers do not accept inline content following the read list. The Linux NFS server should handle this content correctly as of commit a97c331f9aa9 ("svcrdma: Handle additional inline content"). Signed-off-by: Chuck Lever Tested-by: Devesh Sharma Signed-off-by: Anna Schumaker --- net/sunrpc/xprtrdma/rpc_rdma.c | 44 ++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 42 insertions(+), 2 deletions(-) diff --git a/net/sunrpc/xprtrdma/rpc_rdma.c b/net/sunrpc/xprtrdma/rpc_rdma.c index 62150ae2dc09..1dd48f269986 100644 --- a/net/sunrpc/xprtrdma/rpc_rdma.c +++ b/net/sunrpc/xprtrdma/rpc_rdma.c @@ -96,6 +96,42 @@ static bool rpcrdma_results_inline(struct rpc_rqst *rqst) return repsize <= RPCRDMA_INLINE_READ_THRESHOLD(rqst); } +static int +rpcrdma_tail_pullup(struct xdr_buf *buf) +{ + size_t tlen = buf->tail[0].iov_len; + size_t skip = tlen & 3; + + /* Do not include the tail if it is only an XDR pad */ + if (tlen < 4) + return 0; + + /* xdr_write_pages() adds a pad at the beginning of the tail + * if the content in "buf->pages" is unaligned. Force the + * tail's actual content to land at the next XDR position + * after the head instead. + */ + if (skip) { + unsigned char *src, *dst; + unsigned int count; + + src = buf->tail[0].iov_base; + dst = buf->head[0].iov_base; + dst += buf->head[0].iov_len; + + src += skip; + tlen -= skip; + + dprintk("RPC: %s: skip=%zu, memmove(%p, %p, %zu)\n", + __func__, skip, dst, src, tlen); + + for (count = tlen; count; count--) + *dst++ = *src++; + } + + return tlen; +} + /* * Chunk assembly from upper layer xdr_buf. * @@ -147,6 +183,10 @@ rpcrdma_convert_iovs(struct xdr_buf *xdrbuf, unsigned int pos, if (len && n == nsegs) return -EIO; + /* When encoding the read list, the tail is always sent inline */ + if (type == rpcrdma_readch) + return n; + if (xdrbuf->tail[0].iov_len) { /* the rpcrdma protocol allows us to omit any trailing * xdr pad bytes, saving the server an RDMA operation. */ @@ -476,8 +516,8 @@ rpcrdma_marshal_req(struct rpc_rqst *rqst) headerp->rm_body.rm_nochunks.rm_empty[2] = xdr_zero; /* new length after pullup */ rpclen = rqst->rq_svec[0].iov_len; - } - + } else if (rtype == rpcrdma_readch) + rpclen += rpcrdma_tail_pullup(&rqst->rq_snd_buf); if (rtype != rpcrdma_noch) { hdrlen = rpcrdma_create_chunks(rqst, &rqst->rq_snd_buf, headerp, rtype); -- cgit v1.2.3 From 2fcc213a18644610c79edbb5e847d73c6c5d5ded Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Mon, 3 Aug 2015 13:04:26 -0400 Subject: xprtrdma: Fix large NFS SYMLINK calls Repair how rpcrdma_marshal_req() chooses which RDMA message type to use for large non-WRITE operations so that it picks RDMA_NOMSG in the correct situations, and sets up the marshaling logic to SEND only the RPC/RDMA header. Large NFSv2 SYMLINK requests now use RDMA_NOMSG calls. The Linux NFS server XDR decoder for NFSv2 SYMLINK does not handle having the pathname argument arrive in a separate buffer. The decoder could be fixed, but this is simpler and RDMA_NOMSG can be used in a variety of other situations. Ensure that the Linux client continues to use "RDMA_MSG + read list" when sending large NFSv3 SYMLINK requests, which is more efficient than using RDMA_NOMSG. Large NFSv4 CREATE(NF4LNK) requests are changed to use "RDMA_MSG + read list" just like NFSv3 (see Section 5 of RFC 5667). Before, these did not work at all. Signed-off-by: Chuck Lever Tested-by: Devesh Sharma Signed-off-by: Anna Schumaker --- fs/nfs/nfs3xdr.c | 1 + fs/nfs/nfs4xdr.c | 4 +++- net/sunrpc/xprtrdma/rpc_rdma.c | 25 ++++++++++++++++--------- 3 files changed, 20 insertions(+), 10 deletions(-) diff --git a/fs/nfs/nfs3xdr.c b/fs/nfs/nfs3xdr.c index 9b04c2e6fffc..267126d32ec0 100644 --- a/fs/nfs/nfs3xdr.c +++ b/fs/nfs/nfs3xdr.c @@ -1103,6 +1103,7 @@ static void nfs3_xdr_enc_symlink3args(struct rpc_rqst *req, { encode_diropargs3(xdr, args->fromfh, args->fromname, args->fromlen); encode_symlinkdata3(xdr, args); + xdr->buf->flags |= XDRBUF_WRITE; } /* diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 558cd65dbdb7..c42459e45f62 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -1154,7 +1154,9 @@ static void encode_create(struct xdr_stream *xdr, const struct nfs4_create_arg * case NF4LNK: p = reserve_space(xdr, 4); *p = cpu_to_be32(create->u.symlink.len); - xdr_write_pages(xdr, create->u.symlink.pages, 0, create->u.symlink.len); + xdr_write_pages(xdr, create->u.symlink.pages, 0, + create->u.symlink.len); + xdr->buf->flags |= XDRBUF_WRITE; break; case NF4BLK: case NF4CHR: diff --git a/net/sunrpc/xprtrdma/rpc_rdma.c b/net/sunrpc/xprtrdma/rpc_rdma.c index 1dd48f269986..272158623b00 100644 --- a/net/sunrpc/xprtrdma/rpc_rdma.c +++ b/net/sunrpc/xprtrdma/rpc_rdma.c @@ -475,21 +475,24 @@ rpcrdma_marshal_req(struct rpc_rqst *rqst) * * o If the total request is under the inline threshold, all ops * are sent as inline. - * o Large non-write ops are sent with the entire message as a - * single read chunk (protocol 0-position special case). * o Large write ops transmit data as read chunk(s), header as * inline. + * o Large non-write ops are sent with the entire message as a + * single read chunk (protocol 0-position special case). * - * Note: the NFS code sending down multiple argument segments - * implies the op is a write. - * TBD check NFSv4 setacl + * This assumes that the upper layer does not present a request + * that both has a data payload, and whose non-data arguments + * by themselves are larger than the inline threshold. */ - if (rpcrdma_args_inline(rqst)) + if (rpcrdma_args_inline(rqst)) { rtype = rpcrdma_noch; - else if (rqst->rq_snd_buf.page_len == 0) - rtype = rpcrdma_areadch; - else + } else if (rqst->rq_snd_buf.flags & XDRBUF_WRITE) { rtype = rpcrdma_readch; + } else { + headerp->rm_type = htonl(RDMA_NOMSG); + rtype = rpcrdma_areadch; + rpclen = 0; + } /* The following simplification is not true forever */ if (rtype != rpcrdma_noch && wtype == rpcrdma_replych) @@ -546,6 +549,10 @@ rpcrdma_marshal_req(struct rpc_rqst *rqst) req->rl_send_iov[0].length = hdrlen; req->rl_send_iov[0].lkey = rdmab_lkey(req->rl_rdmabuf); + req->rl_niovs = 1; + if (rtype == rpcrdma_areadch) + return 0; + req->rl_send_iov[1].addr = rdmab_addr(req->rl_sendbuf); req->rl_send_iov[1].length = rpclen; req->rl_send_iov[1].lkey = rdmab_lkey(req->rl_sendbuf); -- cgit v1.2.3 From 763f7e4e4b9033ac6a0f13aa37ba43636fc3c0af Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Mon, 3 Aug 2015 13:04:36 -0400 Subject: xprtrdma: Clean up xprt_rdma_print_stats() checkpatch.pl complained about the seq_printf() format string split across lines and the use of %Lu. Signed-off-by: Chuck Lever Tested-by: Devesh Sharma Signed-off-by: Anna Schumaker --- net/sunrpc/xprtrdma/transport.c | 48 ++++++++++++++++++++--------------------- 1 file changed, 23 insertions(+), 25 deletions(-) diff --git a/net/sunrpc/xprtrdma/transport.c b/net/sunrpc/xprtrdma/transport.c index d737300ebc42..0985b2b5de9d 100644 --- a/net/sunrpc/xprtrdma/transport.c +++ b/net/sunrpc/xprtrdma/transport.c @@ -647,31 +647,29 @@ static void xprt_rdma_print_stats(struct rpc_xprt *xprt, struct seq_file *seq) if (xprt_connected(xprt)) idle_time = (long)(jiffies - xprt->last_used) / HZ; - seq_printf(seq, - "\txprt:\trdma %u %lu %lu %lu %ld %lu %lu %lu %Lu %Lu " - "%lu %lu %lu %Lu %Lu %Lu %Lu %lu %lu %lu\n", - - 0, /* need a local port? */ - xprt->stat.bind_count, - xprt->stat.connect_count, - xprt->stat.connect_time, - idle_time, - xprt->stat.sends, - xprt->stat.recvs, - xprt->stat.bad_xids, - xprt->stat.req_u, - xprt->stat.bklog_u, - - r_xprt->rx_stats.read_chunk_count, - r_xprt->rx_stats.write_chunk_count, - r_xprt->rx_stats.reply_chunk_count, - r_xprt->rx_stats.total_rdma_request, - r_xprt->rx_stats.total_rdma_reply, - r_xprt->rx_stats.pullup_copy_count, - r_xprt->rx_stats.fixup_copy_count, - r_xprt->rx_stats.hardway_register_count, - r_xprt->rx_stats.failed_marshal_count, - r_xprt->rx_stats.bad_reply_count); + seq_puts(seq, "\txprt:\trdma "); + seq_printf(seq, "%u %lu %lu %lu %ld %lu %lu %lu %llu %llu ", + 0, /* need a local port? */ + xprt->stat.bind_count, + xprt->stat.connect_count, + xprt->stat.connect_time, + idle_time, + xprt->stat.sends, + xprt->stat.recvs, + xprt->stat.bad_xids, + xprt->stat.req_u, + xprt->stat.bklog_u); + seq_printf(seq, "%lu %lu %lu %llu %llu %llu %llu %lu %lu %lu\n", + r_xprt->rx_stats.read_chunk_count, + r_xprt->rx_stats.write_chunk_count, + r_xprt->rx_stats.reply_chunk_count, + r_xprt->rx_stats.total_rdma_request, + r_xprt->rx_stats.total_rdma_reply, + r_xprt->rx_stats.pullup_copy_count, + r_xprt->rx_stats.fixup_copy_count, + r_xprt->rx_stats.hardway_register_count, + r_xprt->rx_stats.failed_marshal_count, + r_xprt->rx_stats.bad_reply_count); } static int -- cgit v1.2.3 From 860477d1ff176549f2bf438b61e5c1ec6b1d43e5 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Mon, 3 Aug 2015 13:04:45 -0400 Subject: xprtrdma: Count RDMA_NOMSG type calls RDMA_NOMSG type calls are less efficient than RDMA_MSG. Count NOMSG calls so administrators can tell if they happen to be used more than expected. Signed-off-by: Chuck Lever Tested-by: Devesh Sharma Signed-off-by: Anna Schumaker --- net/sunrpc/xprtrdma/rpc_rdma.c | 1 + net/sunrpc/xprtrdma/transport.c | 5 +++-- net/sunrpc/xprtrdma/xprt_rdma.h | 1 + 3 files changed, 5 insertions(+), 2 deletions(-) diff --git a/net/sunrpc/xprtrdma/rpc_rdma.c b/net/sunrpc/xprtrdma/rpc_rdma.c index 272158623b00..bc8bd6577467 100644 --- a/net/sunrpc/xprtrdma/rpc_rdma.c +++ b/net/sunrpc/xprtrdma/rpc_rdma.c @@ -489,6 +489,7 @@ rpcrdma_marshal_req(struct rpc_rqst *rqst) } else if (rqst->rq_snd_buf.flags & XDRBUF_WRITE) { rtype = rpcrdma_readch; } else { + r_xprt->rx_stats.nomsg_call_count++; headerp->rm_type = htonl(RDMA_NOMSG); rtype = rpcrdma_areadch; rpclen = 0; diff --git a/net/sunrpc/xprtrdma/transport.c b/net/sunrpc/xprtrdma/transport.c index 0985b2b5de9d..64443eb754ad 100644 --- a/net/sunrpc/xprtrdma/transport.c +++ b/net/sunrpc/xprtrdma/transport.c @@ -659,7 +659,7 @@ static void xprt_rdma_print_stats(struct rpc_xprt *xprt, struct seq_file *seq) xprt->stat.bad_xids, xprt->stat.req_u, xprt->stat.bklog_u); - seq_printf(seq, "%lu %lu %lu %llu %llu %llu %llu %lu %lu %lu\n", + seq_printf(seq, "%lu %lu %lu %llu %llu %llu %llu %lu %lu %lu %lu\n", r_xprt->rx_stats.read_chunk_count, r_xprt->rx_stats.write_chunk_count, r_xprt->rx_stats.reply_chunk_count, @@ -669,7 +669,8 @@ static void xprt_rdma_print_stats(struct rpc_xprt *xprt, struct seq_file *seq) r_xprt->rx_stats.fixup_copy_count, r_xprt->rx_stats.hardway_register_count, r_xprt->rx_stats.failed_marshal_count, - r_xprt->rx_stats.bad_reply_count); + r_xprt->rx_stats.bad_reply_count, + r_xprt->rx_stats.nomsg_call_count); } static int diff --git a/net/sunrpc/xprtrdma/xprt_rdma.h b/net/sunrpc/xprtrdma/xprt_rdma.h index 8422c09043b0..d252457ff21a 100644 --- a/net/sunrpc/xprtrdma/xprt_rdma.h +++ b/net/sunrpc/xprtrdma/xprt_rdma.h @@ -340,6 +340,7 @@ struct rpcrdma_stats { unsigned long hardway_register_count; unsigned long failed_marshal_count; unsigned long bad_reply_count; + unsigned long nomsg_call_count; }; /* -- cgit v1.2.3 From 1241d7bf2ac8838d0d2d0b54a6173ac3eb3747a4 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Mon, 3 Aug 2015 13:04:54 -0400 Subject: core: Remove the ib_reg_phys_mr() and ib_rereg_phys_mr() verbs The verbs are obsolete. The ib_rereg_phys_mr() verb is not used by kernel ULPs, and the last ib_reg_phys_mr() call site in the kernel tree has now been removed. Two staging tree call sites remain in the Lustre client. The Lustre team has been notified of the deprecation of reg_phys_mr. Signed-off-by: Chuck Lever Acked-by: Doug Ledford Signed-off-by: Anna Schumaker --- drivers/infiniband/core/verbs.c | 67 ----------------------------------------- include/rdma/ib_verbs.h | 46 ---------------------------- 2 files changed, 113 deletions(-) diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c index bac3fb406a74..30eb2457000c 100644 --- a/drivers/infiniband/core/verbs.c +++ b/drivers/infiniband/core/verbs.c @@ -1144,73 +1144,6 @@ struct ib_mr *ib_get_dma_mr(struct ib_pd *pd, int mr_access_flags) } EXPORT_SYMBOL(ib_get_dma_mr); -struct ib_mr *ib_reg_phys_mr(struct ib_pd *pd, - struct ib_phys_buf *phys_buf_array, - int num_phys_buf, - int mr_access_flags, - u64 *iova_start) -{ - struct ib_mr *mr; - int err; - - err = ib_check_mr_access(mr_access_flags); - if (err) - return ERR_PTR(err); - - if (!pd->device->reg_phys_mr) - return ERR_PTR(-ENOSYS); - - mr = pd->device->reg_phys_mr(pd, phys_buf_array, num_phys_buf, - mr_access_flags, iova_start); - - if (!IS_ERR(mr)) { - mr->device = pd->device; - mr->pd = pd; - mr->uobject = NULL; - atomic_inc(&pd->usecnt); - atomic_set(&mr->usecnt, 0); - } - - return mr; -} -EXPORT_SYMBOL(ib_reg_phys_mr); - -int ib_rereg_phys_mr(struct ib_mr *mr, - int mr_rereg_mask, - struct ib_pd *pd, - struct ib_phys_buf *phys_buf_array, - int num_phys_buf, - int mr_access_flags, - u64 *iova_start) -{ - struct ib_pd *old_pd; - int ret; - - ret = ib_check_mr_access(mr_access_flags); - if (ret) - return ret; - - if (!mr->device->rereg_phys_mr) - return -ENOSYS; - - if (atomic_read(&mr->usecnt)) - return -EBUSY; - - old_pd = mr->pd; - - ret = mr->device->rereg_phys_mr(mr, mr_rereg_mask, pd, - phys_buf_array, num_phys_buf, - mr_access_flags, iova_start); - - if (!ret && (mr_rereg_mask & IB_MR_REREG_PD)) { - atomic_dec(&old_pd->usecnt); - atomic_inc(&pd->usecnt); - } - - return ret; -} -EXPORT_SYMBOL(ib_rereg_phys_mr); - int ib_query_mr(struct ib_mr *mr, struct ib_mr_attr *mr_attr) { return mr->device->query_mr ? diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h index b0f898e3b2e7..43c1cf01c84b 100644 --- a/include/rdma/ib_verbs.h +++ b/include/rdma/ib_verbs.h @@ -2759,52 +2759,6 @@ static inline void ib_dma_free_coherent(struct ib_device *dev, dma_free_coherent(dev->dma_device, size, cpu_addr, dma_handle); } -/** - * ib_reg_phys_mr - Prepares a virtually addressed memory region for use - * by an HCA. - * @pd: The protection domain associated assigned to the registered region. - * @phys_buf_array: Specifies a list of physical buffers to use in the - * memory region. - * @num_phys_buf: Specifies the size of the phys_buf_array. - * @mr_access_flags: Specifies the memory access rights. - * @iova_start: The offset of the region's starting I/O virtual address. - */ -struct ib_mr *ib_reg_phys_mr(struct ib_pd *pd, - struct ib_phys_buf *phys_buf_array, - int num_phys_buf, - int mr_access_flags, - u64 *iova_start); - -/** - * ib_rereg_phys_mr - Modifies the attributes of an existing memory region. - * Conceptually, this call performs the functions deregister memory region - * followed by register physical memory region. Where possible, - * resources are reused instead of deallocated and reallocated. - * @mr: The memory region to modify. - * @mr_rereg_mask: A bit-mask used to indicate which of the following - * properties of the memory region are being modified. - * @pd: If %IB_MR_REREG_PD is set in mr_rereg_mask, this field specifies - * the new protection domain to associated with the memory region, - * otherwise, this parameter is ignored. - * @phys_buf_array: If %IB_MR_REREG_TRANS is set in mr_rereg_mask, this - * field specifies a list of physical buffers to use in the new - * translation, otherwise, this parameter is ignored. - * @num_phys_buf: If %IB_MR_REREG_TRANS is set in mr_rereg_mask, this - * field specifies the size of the phys_buf_array, otherwise, this - * parameter is ignored. - * @mr_access_flags: If %IB_MR_REREG_ACCESS is set in mr_rereg_mask, this - * field specifies the new memory access rights, otherwise, this - * parameter is ignored. - * @iova_start: The offset of the region's starting I/O virtual address. - */ -int ib_rereg_phys_mr(struct ib_mr *mr, - int mr_rereg_mask, - struct ib_pd *pd, - struct ib_phys_buf *phys_buf_array, - int num_phys_buf, - int mr_access_flags, - u64 *iova_start); - /** * ib_query_mr - Retrieves information about a specific memory region. * @mr: The memory region to retrieve information about. -- cgit v1.2.3 From d0f36c46deea97bd16b9277be2f1acac74d76037 Mon Sep 17 00:00:00 2001 From: Devesh Sharma Date: Mon, 3 Aug 2015 13:05:04 -0400 Subject: xprtrdma: take HCA driver refcount at client This is a rework of the following patch sent almost a year back: http://www.mail-archive.com/linux-rdma%40vger.kernel.org/msg20730.html In presence of active mount if someone tries to rmmod vendor-driver, the command remains stuck forever waiting for destruction of all rdma-cm-id. in worst case client can crash during shutdown with active mounts. The existing code assumes that ia->ri_id->device cannot change during the lifetime of a transport. xprtrdma do not have support for DEVICE_REMOVAL event either. Lifting that assumption and adding support for DEVICE_REMOVAL event is a long chain of work, and is in plan. The community decided that preventing the hang right now is more important than waiting for architectural changes. Thus, this patch introduces a temporary workaround to acquire HCA driver module reference count during the mount of a nfs-rdma mount point. Signed-off-by: Devesh Sharma Signed-off-by: Chuck Lever Reviewed-by: Sagi Grimberg Signed-off-by: Anna Schumaker --- net/sunrpc/xprtrdma/verbs.c | 37 +++++++++++++++++++++++++++++-------- 1 file changed, 29 insertions(+), 8 deletions(-) diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c index b4d4f6300fbc..f73d7a71035c 100644 --- a/net/sunrpc/xprtrdma/verbs.c +++ b/net/sunrpc/xprtrdma/verbs.c @@ -52,6 +52,7 @@ #include #include #include +#include /* try_module_get()/module_put() */ #include "xprt_rdma.h" @@ -414,6 +415,14 @@ connected: return 0; } +static void rpcrdma_destroy_id(struct rdma_cm_id *id) +{ + if (id) { + module_put(id->device->owner); + rdma_destroy_id(id); + } +} + static struct rdma_cm_id * rpcrdma_create_id(struct rpcrdma_xprt *xprt, struct rpcrdma_ia *ia, struct sockaddr *addr) @@ -440,6 +449,17 @@ rpcrdma_create_id(struct rpcrdma_xprt *xprt, } wait_for_completion_interruptible_timeout(&ia->ri_done, msecs_to_jiffies(RDMA_RESOLVE_TIMEOUT) + 1); + + /* FIXME: + * Until xprtrdma supports DEVICE_REMOVAL, the provider must + * be pinned while there are active NFS/RDMA mounts to prevent + * hangs and crashes at umount time. + */ + if (!ia->ri_async_rc && !try_module_get(id->device->owner)) { + dprintk("RPC: %s: Failed to get device module\n", + __func__); + ia->ri_async_rc = -ENODEV; + } rc = ia->ri_async_rc; if (rc) goto out; @@ -449,16 +469,17 @@ rpcrdma_create_id(struct rpcrdma_xprt *xprt, if (rc) { dprintk("RPC: %s: rdma_resolve_route() failed %i\n", __func__, rc); - goto out; + goto put; } wait_for_completion_interruptible_timeout(&ia->ri_done, msecs_to_jiffies(RDMA_RESOLVE_TIMEOUT) + 1); rc = ia->ri_async_rc; if (rc) - goto out; + goto put; return id; - +put: + module_put(id->device->owner); out: rdma_destroy_id(id); return ERR_PTR(rc); @@ -566,7 +587,7 @@ out3: ib_dealloc_pd(ia->ri_pd); ia->ri_pd = NULL; out2: - rdma_destroy_id(ia->ri_id); + rpcrdma_destroy_id(ia->ri_id); ia->ri_id = NULL; out1: return rc; @@ -584,7 +605,7 @@ rpcrdma_ia_close(struct rpcrdma_ia *ia) if (ia->ri_id != NULL && !IS_ERR(ia->ri_id)) { if (ia->ri_id->qp) rdma_destroy_qp(ia->ri_id); - rdma_destroy_id(ia->ri_id); + rpcrdma_destroy_id(ia->ri_id); ia->ri_id = NULL; } @@ -794,7 +815,7 @@ retry: if (ia->ri_device != id->device) { printk("RPC: %s: can't reconnect on " "different device!\n", __func__); - rdma_destroy_id(id); + rpcrdma_destroy_id(id); rc = -ENETUNREACH; goto out; } @@ -803,7 +824,7 @@ retry: if (rc) { dprintk("RPC: %s: rdma_create_qp failed %i\n", __func__, rc); - rdma_destroy_id(id); + rpcrdma_destroy_id(id); rc = -ENETUNREACH; goto out; } @@ -814,7 +835,7 @@ retry: write_unlock(&ia->ri_qplock); rdma_destroy_qp(old); - rdma_destroy_id(old); + rpcrdma_destroy_id(old); } else { dprintk("RPC: %s: connecting...\n", __func__); rc = rdma_create_qp(ia->ri_id, ia->ri_pd, &ep->rep_attr); -- cgit v1.2.3 From 86d80f973434de24d8a807a92cd59d5ced7bd519 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 31 Jul 2015 16:24:30 -0400 Subject: NFSv4.1/pnfs: Fix atomicity of commit list updates pnfs_layout_mark_request_commit() needs to ensure that it adds the request to the commit list atomically with all the other updates in order to prevent corruption to buckets[ds_commit_idx].wlseg due to races with pnfs_generic_clear_request_commit(). Fixes: 338d00cfef07d ("pnfs: Refactor the *_layout_mark_request_commit...") Cc: stable@vger.kernel.org # v4.0+ Signed-off-by: Trond Myklebust --- fs/nfs/internal.h | 15 ++++++++++----- fs/nfs/pnfs_nfs.c | 5 +++-- fs/nfs/write.c | 29 ++++++++++++++++++++++++----- 3 files changed, 37 insertions(+), 12 deletions(-) diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h index 9b372b845f6a..1dad18105ed0 100644 --- a/fs/nfs/internal.h +++ b/fs/nfs/internal.h @@ -490,6 +490,9 @@ void nfs_retry_commit(struct list_head *page_list, void nfs_commitdata_release(struct nfs_commit_data *data); void nfs_request_add_commit_list(struct nfs_page *req, struct list_head *dst, struct nfs_commit_info *cinfo); +void nfs_request_add_commit_list_locked(struct nfs_page *req, + struct list_head *dst, + struct nfs_commit_info *cinfo); void nfs_request_remove_commit_list(struct nfs_page *req, struct nfs_commit_info *cinfo); void nfs_init_cinfo(struct nfs_commit_info *cinfo, @@ -623,13 +626,15 @@ void nfs_super_set_maxbytes(struct super_block *sb, __u64 maxfilesize) * Record the page as unstable and mark its inode as dirty. */ static inline -void nfs_mark_page_unstable(struct page *page) +void nfs_mark_page_unstable(struct page *page, struct nfs_commit_info *cinfo) { - struct inode *inode = page_file_mapping(page)->host; + if (!cinfo->dreq) { + struct inode *inode = page_file_mapping(page)->host; - inc_zone_page_state(page, NR_UNSTABLE_NFS); - inc_wb_stat(&inode_to_bdi(inode)->wb, WB_RECLAIMABLE); - __mark_inode_dirty(inode, I_DIRTY_DATASYNC); + inc_zone_page_state(page, NR_UNSTABLE_NFS); + inc_wb_stat(&inode_to_bdi(inode)->wb, WB_RECLAIMABLE); + __mark_inode_dirty(inode, I_DIRTY_DATASYNC); + } } /* diff --git a/fs/nfs/pnfs_nfs.c b/fs/nfs/pnfs_nfs.c index f37e25b6311c..7a282876662f 100644 --- a/fs/nfs/pnfs_nfs.c +++ b/fs/nfs/pnfs_nfs.c @@ -863,9 +863,10 @@ pnfs_layout_mark_request_commit(struct nfs_page *req, } set_bit(PG_COMMIT_TO_DS, &req->wb_flags); cinfo->ds->nwritten++; - spin_unlock(cinfo->lock); - nfs_request_add_commit_list(req, list, cinfo); + nfs_request_add_commit_list_locked(req, list, cinfo); + spin_unlock(cinfo->lock); + nfs_mark_page_unstable(req->wb_page, cinfo); } EXPORT_SYMBOL_GPL(pnfs_layout_mark_request_commit); diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 75a35a1afa79..fdee9270ca15 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -767,6 +767,28 @@ nfs_page_search_commits_for_head_request_locked(struct nfs_inode *nfsi, return NULL; } +/** + * nfs_request_add_commit_list_locked - add request to a commit list + * @req: pointer to a struct nfs_page + * @dst: commit list head + * @cinfo: holds list lock and accounting info + * + * This sets the PG_CLEAN bit, updates the cinfo count of + * number of outstanding requests requiring a commit as well as + * the MM page stats. + * + * The caller must hold the cinfo->lock, and the nfs_page lock. + */ +void +nfs_request_add_commit_list_locked(struct nfs_page *req, struct list_head *dst, + struct nfs_commit_info *cinfo) +{ + set_bit(PG_CLEAN, &req->wb_flags); + nfs_list_add_request(req, dst); + cinfo->mds->ncommit++; +} +EXPORT_SYMBOL_GPL(nfs_request_add_commit_list_locked); + /** * nfs_request_add_commit_list - add request to a commit list * @req: pointer to a struct nfs_page @@ -784,13 +806,10 @@ void nfs_request_add_commit_list(struct nfs_page *req, struct list_head *dst, struct nfs_commit_info *cinfo) { - set_bit(PG_CLEAN, &(req)->wb_flags); spin_lock(cinfo->lock); - nfs_list_add_request(req, dst); - cinfo->mds->ncommit++; + nfs_request_add_commit_list_locked(req, dst, cinfo); spin_unlock(cinfo->lock); - if (!cinfo->dreq) - nfs_mark_page_unstable(req->wb_page); + nfs_mark_page_unstable(req->wb_page, cinfo); } EXPORT_SYMBOL_GPL(nfs_request_add_commit_list); -- cgit v1.2.3 From d099d7b8316f3ebd63472d207c4801a464330016 Mon Sep 17 00:00:00 2001 From: Peng Tao Date: Mon, 10 Aug 2015 16:47:32 +0800 Subject: pnfs/flexfiles: LAYOUTSTATS ii_count should be ops instead of bytes Turned out I misinterpreted the spec... Cc: Tom Haynes Reported-by: Jean Spector Signed-off-by: Peng Tao Signed-off-by: Trond Myklebust --- fs/nfs/flexfilelayout/flexfilelayout.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index b3289d701eea..ba74b15bf0ba 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -1863,10 +1863,9 @@ ff_layout_mirror_prepare_stats(struct nfs42_layoutstat_args *args, memcpy(&devinfo->dev_id, &dev->deviceid, NFS4_DEVICEID4_SIZE); devinfo->offset = pls->pls_range.offset; devinfo->length = pls->pls_range.length; - /* well, we don't really know if IO is continuous or not! */ - devinfo->read_count = mirror->read_stat.io_stat.bytes_completed; + devinfo->read_count = mirror->read_stat.io_stat.ops_completed; devinfo->read_bytes = mirror->read_stat.io_stat.bytes_completed; - devinfo->write_count = mirror->write_stat.io_stat.bytes_completed; + devinfo->write_count = mirror->write_stat.io_stat.ops_completed; devinfo->write_bytes = mirror->write_stat.io_stat.bytes_completed; devinfo->layout_type = LAYOUT_FLEX_FILES; devinfo->layoutstats_encode = ff_layout_encode_layoutstats; -- cgit v1.2.3 From c8ad8894e92b853df5a766061ee9cde7e10e682f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 5 Aug 2015 17:31:58 -0400 Subject: NFSv4.2/pnfs: Use GFP_NOIO for layoutstat reporting in the writeback path Prevent a potential deadlock. Signed-off-by: Trond Myklebust --- fs/nfs/flexfilelayout/flexfilelayout.c | 6 ++++-- fs/nfs/pnfs.c | 4 ++-- fs/nfs/pnfs.h | 4 ++-- 3 files changed, 8 insertions(+), 6 deletions(-) diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index ba74b15bf0ba..2a93bec7e6dd 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -510,7 +510,8 @@ nfs4_ff_layout_stat_io_start_read(struct nfs4_ff_layout_mirror *mirror, spin_unlock(&mirror->lock); if (report) - pnfs_report_layoutstat(mirror->lseg->pls_layout->plh_inode); + pnfs_report_layoutstat(mirror->lseg->pls_layout->plh_inode, + GFP_KERNEL); } static void @@ -538,7 +539,8 @@ nfs4_ff_layout_stat_io_start_write(struct nfs4_ff_layout_mirror *mirror, spin_unlock(&mirror->lock); if (report) - pnfs_report_layoutstat(mirror->lseg->pls_layout->plh_inode); + pnfs_report_layoutstat(mirror->lseg->pls_layout->plh_inode, + GFP_NOIO); } static void diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 70bf706b1090..a6ec420983d1 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -2267,7 +2267,7 @@ struct nfs4_threshold *pnfs_mdsthreshold_alloc(void) #if IS_ENABLED(CONFIG_NFS_V4_2) int -pnfs_report_layoutstat(struct inode *inode) +pnfs_report_layoutstat(struct inode *inode, gfp_t gfp_flags) { struct pnfs_layoutdriver_type *ld = NFS_SERVER(inode)->pnfs_curr_ld; struct nfs_server *server = NFS_SERVER(inode); @@ -2294,7 +2294,7 @@ pnfs_report_layoutstat(struct inode *inode) pnfs_get_layout_hdr(hdr); spin_unlock(&inode->i_lock); - data = kzalloc(sizeof(*data), GFP_KERNEL); + data = kzalloc(sizeof(*data), gfp_flags); if (!data) { status = -ENOMEM; goto out_put; diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h index 3e6ab7bfbabd..738672a0f8da 100644 --- a/fs/nfs/pnfs.h +++ b/fs/nfs/pnfs.h @@ -691,10 +691,10 @@ static inline void nfs4_pnfs_v3_ds_connect_unload(void) #endif /* CONFIG_NFS_V4_1 */ #if IS_ENABLED(CONFIG_NFS_V4_2) -int pnfs_report_layoutstat(struct inode *inode); +int pnfs_report_layoutstat(struct inode *inode, gfp_t gfp_flags); #else static inline int -pnfs_report_layoutstat(struct inode *inode) +pnfs_report_layoutstat(struct inode *inode, gfp_t gfp_flags) { return 0; } -- cgit v1.2.3 From 24a9a9610ce3ba36fd87c1d2f2c9106de6b7e832 Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Mon, 3 Aug 2015 07:44:53 -0400 Subject: sunrpc: increase UNX_MAXNODENAME from 32 to __NEW_UTS_LEN bytes The current limit of 32 bytes artificially limits the name string that we end up stuffing into NFSv4.x client ID blobs. If you have multiple hosts with long hostnames that only differ near the end, then this can cause NFSv4 client ID collisions. Linux nodenames are actually limited to __NEW_UTS_LEN bytes (64), so use that as the limit instead. Also, use XDR_QUADLEN to specify the slack length, just for clarity and in case someone in the future changes this to something not evenly divisible by 4. Reported-by: Michael Skralivetsky Signed-off-by: Jeff Layton Signed-off-by: Trond Myklebust --- include/linux/sunrpc/auth.h | 8 ++++++-- net/sunrpc/auth_unix.c | 2 +- 2 files changed, 7 insertions(+), 3 deletions(-) diff --git a/include/linux/sunrpc/auth.h b/include/linux/sunrpc/auth.h index a7cbb570cc5c..1ecf13e148b8 100644 --- a/include/linux/sunrpc/auth.h +++ b/include/linux/sunrpc/auth.h @@ -18,9 +18,13 @@ #include #include #include +#include -/* size of the nodename buffer */ -#define UNX_MAXNODENAME 32 +/* + * Size of the nodename buffer. RFC1831 specifies a hard limit of 255 bytes, + * but Linux hostnames are actually limited to __NEW_UTS_LEN bytes. + */ +#define UNX_MAXNODENAME __NEW_UTS_LEN struct rpcsec_gss_info; diff --git a/net/sunrpc/auth_unix.c b/net/sunrpc/auth_unix.c index 4feda2d0a833..548240dd15fc 100644 --- a/net/sunrpc/auth_unix.c +++ b/net/sunrpc/auth_unix.c @@ -23,7 +23,7 @@ struct unx_cred { }; #define uc_uid uc_base.cr_uid -#define UNX_WRITESLACK (21 + (UNX_MAXNODENAME >> 2)) +#define UNX_WRITESLACK (21 + XDR_QUADLEN(UNX_MAXNODENAME)) #if IS_ENABLED(CONFIG_SUNRPC_DEBUG) # define RPCDBG_FACILITY RPCDBG_AUTH -- cgit v1.2.3 From 5ef8d792fabedeb932375b23735bc7a1a3e8684d Mon Sep 17 00:00:00 2001 From: Kinglong Mee Date: Thu, 30 Jul 2015 21:41:08 +0800 Subject: NFS: Error out when register_shrinker fail in register_nfs_fs Commit 1d3d4437ea "vmscan: per-node deferred work" have made register_shrinker can return an intergater error. If register_shrinker() fail, the later unregister_shrinker() will cause a NULL pointer access. v2, same as v1. Signed-off-by: Kinglong Mee Signed-off-by: Trond Myklebust --- fs/nfs/super.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) diff --git a/fs/nfs/super.c b/fs/nfs/super.c index aa62004f1706..383a027de452 100644 --- a/fs/nfs/super.c +++ b/fs/nfs/super.c @@ -381,9 +381,12 @@ int __init register_nfs_fs(void) ret = nfs_register_sysctl(); if (ret < 0) goto error_2; - register_shrinker(&acl_shrinker); + ret = register_shrinker(&acl_shrinker); + if (ret < 0) + goto error_3; return 0; - +error_3: + nfs_unregister_sysctl(); error_2: unregister_nfs4_fs(); error_1: -- cgit v1.2.3 From efcbc04e16dfa95fef76309f89710dd1d99a5453 Mon Sep 17 00:00:00 2001 From: NeilBrown Date: Thu, 30 Jul 2015 13:00:56 +1000 Subject: NFSv4: don't set SETATTR for O_RDONLY|O_EXCL It is unusual to combine the open flags O_RDONLY and O_EXCL, but it appears that libre-office does just that. [pid 3250] stat("/home/USER/.config", {st_mode=S_IFDIR|0700, st_size=8192, ...}) = 0 [pid 3250] open("/home/USER/.config/libreoffice/4-suse/user/extensions/buildid", O_RDONLY|O_EXCL NFSv4 takes O_EXCL as a sign that a setattr command should be sent, probably to reset the timestamps. When it was an O_RDONLY open, the SETATTR command does not identify any actual attributes to change. If no delegation was provided to the open, the SETATTR uses the all-zeros stateid and the request is accepted (at least by the Linux NFS server - no harm, no foul). If a read-delegation was provided, this is used in the SETATTR request, and a Netapp filer will justifiably claim NFS4ERR_BAD_STATEID, which the Linux client takes as a sign to retry - indefinitely. So only treat O_EXCL specially if O_CREAT was also given. Signed-off-by: NeilBrown Cc: stable@vger.kernel.org Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 3acb1eb72930..15ee8bd99b61 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -2425,7 +2425,7 @@ static int _nfs4_do_open(struct inode *dir, goto err_free_label; state = ctx->state; - if ((opendata->o_arg.open_flags & O_EXCL) && + if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) && (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) { nfs4_exclusive_attrset(opendata, sattr); -- cgit v1.2.3 From 0847ef88c3c9318d85e92fc42369df0e0190e1ab Mon Sep 17 00:00:00 2001 From: Kinglong Mee Date: Thu, 30 Jul 2015 21:40:06 +0800 Subject: NFS: Remove duplicate svc_xprt_put from nfs41_callback_up The xprt created by svc_create_xprt have be added to serv->sv_permsocks. So putting the xprt directly is useless. Otherwise, there is a more svc_xprt_put after the xprt be freed. v2, same as v1. Signed-off-by: Kinglong Mee Signed-off-by: Trond Myklebust --- fs/nfs/callback.c | 4 ---- 1 file changed, 4 deletions(-) diff --git a/fs/nfs/callback.c b/fs/nfs/callback.c index 682529c00996..6d27d7215a21 100644 --- a/fs/nfs/callback.c +++ b/fs/nfs/callback.c @@ -162,10 +162,6 @@ nfs41_callback_up(struct svc_serv *serv) spin_lock_init(&serv->sv_cb_lock); init_waitqueue_head(&serv->sv_cb_waitq); rqstp = svc_prepare_thread(serv, &serv->sv_pools[0], NUMA_NO_NODE); - if (IS_ERR(rqstp)) { - svc_xprt_put(serv->sv_bc_xprt); - serv->sv_bc_xprt = NULL; - } dprintk("--> %s return %d\n", __func__, PTR_ERR_OR_ZERO(rqstp)); return rqstp; } -- cgit v1.2.3 From 27571297a7e9a2a845c232813a7ba7e1227f5ec6 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 3 Aug 2015 17:38:33 -0400 Subject: pNFS: Tighten up locking around DS commit buckets I'm not aware of any bugreports around this issue, but the locking around the pnfs_commit_bucket is inconsistent at best. This patch tightens it up by ensuring that the 'bucket->committing' list is always changed atomically w.r.t. the 'bucket->clseg' layout segment tracking. Signed-off-by: Trond Myklebust --- fs/nfs/pnfs_nfs.c | 50 ++++++++++++++++++++++++++++++++------------------ 1 file changed, 32 insertions(+), 18 deletions(-) diff --git a/fs/nfs/pnfs_nfs.c b/fs/nfs/pnfs_nfs.c index 7a282876662f..cd3c0403101b 100644 --- a/fs/nfs/pnfs_nfs.c +++ b/fs/nfs/pnfs_nfs.c @@ -124,11 +124,12 @@ pnfs_generic_scan_ds_commit_list(struct pnfs_commit_bucket *bucket, if (ret) { cinfo->ds->nwritten -= ret; cinfo->ds->ncommitting += ret; - bucket->clseg = bucket->wlseg; - if (list_empty(src)) + if (bucket->clseg == NULL) + bucket->clseg = pnfs_get_lseg(bucket->wlseg); + if (list_empty(src)) { + pnfs_put_lseg_locked(bucket->wlseg); bucket->wlseg = NULL; - else - pnfs_get_lseg(bucket->clseg); + } } return ret; } @@ -182,19 +183,23 @@ static void pnfs_generic_retry_commit(struct nfs_commit_info *cinfo, int idx) struct pnfs_ds_commit_info *fl_cinfo = cinfo->ds; struct pnfs_commit_bucket *bucket; struct pnfs_layout_segment *freeme; + LIST_HEAD(pages); int i; + spin_lock(cinfo->lock); for (i = idx; i < fl_cinfo->nbuckets; i++) { bucket = &fl_cinfo->buckets[i]; if (list_empty(&bucket->committing)) continue; - nfs_retry_commit(&bucket->committing, bucket->clseg, cinfo, i); - spin_lock(cinfo->lock); freeme = bucket->clseg; bucket->clseg = NULL; + list_splice_init(&bucket->committing, &pages); spin_unlock(cinfo->lock); + nfs_retry_commit(&pages, freeme, cinfo, i); pnfs_put_lseg(freeme); + spin_lock(cinfo->lock); } + spin_unlock(cinfo->lock); } static unsigned int @@ -216,10 +221,6 @@ pnfs_generic_alloc_ds_commits(struct nfs_commit_info *cinfo, if (!data) break; data->ds_commit_index = i; - spin_lock(cinfo->lock); - data->lseg = bucket->clseg; - bucket->clseg = NULL; - spin_unlock(cinfo->lock); list_add(&data->pages, list); nreq++; } @@ -229,6 +230,22 @@ pnfs_generic_alloc_ds_commits(struct nfs_commit_info *cinfo, return nreq; } +static inline +void pnfs_fetch_commit_bucket_list(struct list_head *pages, + struct nfs_commit_data *data, + struct nfs_commit_info *cinfo) +{ + struct pnfs_commit_bucket *bucket; + + bucket = &cinfo->ds->buckets[data->ds_commit_index]; + spin_lock(cinfo->lock); + list_splice_init(pages, &bucket->committing); + data->lseg = bucket->clseg; + bucket->clseg = NULL; + spin_unlock(cinfo->lock); + +} + /* This follows nfs_commit_list pretty closely */ int pnfs_generic_commit_pagelist(struct inode *inode, struct list_head *mds_pages, @@ -243,7 +260,7 @@ pnfs_generic_commit_pagelist(struct inode *inode, struct list_head *mds_pages, if (!list_empty(mds_pages)) { data = nfs_commitdata_alloc(); if (data != NULL) { - data->lseg = NULL; + data->ds_commit_index = -1; list_add(&data->pages, &list); nreq++; } else { @@ -265,19 +282,16 @@ pnfs_generic_commit_pagelist(struct inode *inode, struct list_head *mds_pages, list_for_each_entry_safe(data, tmp, &list, pages) { list_del_init(&data->pages); - if (!data->lseg) { + if (data->ds_commit_index < 0) { nfs_init_commit(data, mds_pages, NULL, cinfo); nfs_initiate_commit(NFS_CLIENT(inode), data, NFS_PROTO(data->inode), data->mds_ops, how, 0); } else { - struct pnfs_commit_bucket *buckets; + LIST_HEAD(pages); - buckets = cinfo->ds->buckets; - nfs_init_commit(data, - &buckets[data->ds_commit_index].committing, - data->lseg, - cinfo); + pnfs_fetch_commit_bucket_list(&pages, data, cinfo); + nfs_init_commit(data, &pages, data->lseg, cinfo); initiate_commit(data, how); } } -- cgit v1.2.3 From a4497a58e4043a925b7b308bd2c32f0744eca440 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 4 Aug 2015 15:41:50 -0400 Subject: NFSv4.1/pnfs: Remove redundant checks in pnfs_layoutgets_blocked() If there are no valid layout segments, then we should already have checked in pnfs_update_layout() whether or not this is the first layoutget. Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 10 ++++------ 1 file changed, 4 insertions(+), 6 deletions(-) diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index a6ec420983d1..8855b322d127 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -829,12 +829,10 @@ pnfs_layout_returning(const struct pnfs_layout_hdr *lo, /* lget is set to 1 if called from inside send_layoutget call chain */ static bool pnfs_layoutgets_blocked(const struct pnfs_layout_hdr *lo, - struct pnfs_layout_range *range, int lget) + struct pnfs_layout_range *range) { return lo->plh_block_lgets || test_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags) || - (list_empty(&lo->plh_segs) && - (atomic_read(&lo->plh_outstanding) > lget)) || pnfs_layout_returning(lo, range); } @@ -847,7 +845,7 @@ pnfs_choose_layoutget_stateid(nfs4_stateid *dst, struct pnfs_layout_hdr *lo, dprintk("--> %s\n", __func__); spin_lock(&lo->plh_inode->i_lock); - if (pnfs_layoutgets_blocked(lo, range, 1)) { + if (pnfs_layoutgets_blocked(lo, range)) { status = -EAGAIN; } else if (!nfs4_valid_open_stateid(open_state)) { status = -EBADF; @@ -1547,7 +1545,7 @@ lookup_again: goto out_put_layout_hdr; } - if (pnfs_layoutgets_blocked(lo, &arg, 0)) + if (pnfs_layoutgets_blocked(lo, &arg)) goto out_unlock; atomic_inc(&lo->plh_outstanding); spin_unlock(&ino->i_lock); @@ -1624,7 +1622,7 @@ pnfs_layout_process(struct nfs4_layoutget *lgp) goto out_forget_reply; } - if (pnfs_layoutgets_blocked(lo, &lgp->args.range, 1)) { + if (pnfs_layoutgets_blocked(lo, &lgp->args.range)) { dprintk("%s forget reply due to state\n", __func__); goto out_forget_reply; } -- cgit v1.2.3 From 8f70f53a87007bdbb34c79d11178a153914f5db1 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 4 Aug 2015 16:09:44 -0400 Subject: NFSv4.1/pnfs: Fix serialisation of layout return and layoutget We should always test for outstanding layout returns, whether or not pnfs_should_retry_layoutget() is true. Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 8855b322d127..e3a47ee15474 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1436,6 +1436,8 @@ static int pnfs_layoutget_retry_bit_wait(struct wait_bit_key *key) static bool pnfs_prepare_to_retry_layoutget(struct pnfs_layout_hdr *lo) { + if (!pnfs_should_retry_layoutget(lo)) + return false; /* * send layoutcommit as it can hold up layoutreturn due to lseg * reference @@ -1531,8 +1533,7 @@ lookup_again: * Because we free lsegs before sending LAYOUTRETURN, we need to wait * for LAYOUTRETURN even if first is true. */ - if (!lseg && pnfs_should_retry_layoutget(lo) && - test_bit(NFS_LAYOUT_RETURN, &lo->plh_flags)) { + if (test_bit(NFS_LAYOUT_RETURN, &lo->plh_flags)) { spin_unlock(&ino->i_lock); dprintk("%s wait for layoutreturn\n", __func__); if (pnfs_prepare_to_retry_layoutget(lo)) { -- cgit v1.2.3 From 5c4a79fb2b1cd80cb58986f6acf402721901c545 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 4 Aug 2015 15:57:13 -0400 Subject: NFSv4.1/pnfs: Don't prevent layoutgets when doing return-on-close If there is an outstanding return-on-close, then we just want new layoutget requests to wait rather than fail. Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index e3a47ee15474..8c5f9f59efbb 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1105,7 +1105,9 @@ bool pnfs_roc(struct inode *ino) } if (!found) goto out_noroc; - lo->plh_block_lgets++; + if (test_and_set_bit(NFS_LAYOUT_RETURN, &lo->plh_flags)) + goto out_noroc; + lo->plh_return_iomode = IOMODE_ANY; pnfs_get_layout_hdr(lo); /* matched in pnfs_roc_release */ spin_unlock(&ino->i_lock); pnfs_free_lseg_list(&tmp_list); @@ -1133,7 +1135,7 @@ void pnfs_roc_release(struct inode *ino) spin_lock(&ino->i_lock); lo = NFS_I(ino)->layout; - lo->plh_block_lgets--; + pnfs_clear_layoutreturn_waitbit(lo); if (atomic_dec_and_test(&lo->plh_refcount)) { pnfs_detach_layout_hdr(lo); spin_unlock(&ino->i_lock); -- cgit v1.2.3 From 2d8ae84fbc32a14bba176cf9c20d5eb2a3d42791 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 4 Aug 2015 16:15:48 -0400 Subject: NFSv4.1/pnfs: Remove redundant lo->plh_block_lgets in layoutreturn The NFS_LAYOUT_RETURN bit already suffices to ensure that layoutget is blocked. Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 2 -- 1 file changed, 2 deletions(-) diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 8c5f9f59efbb..ada07376a6c7 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -368,7 +368,6 @@ pnfs_prepare_layoutreturn(struct pnfs_layout_hdr *lo) if (test_and_set_bit(NFS_LAYOUT_RETURN, &lo->plh_flags)) return false; lo->plh_return_iomode = 0; - lo->plh_block_lgets++; pnfs_get_layout_hdr(lo); clear_bit(NFS_LAYOUT_RETURN_BEFORE_CLOSE, &lo->plh_flags); return true; @@ -954,7 +953,6 @@ pnfs_send_layoutreturn(struct pnfs_layout_hdr *lo, nfs4_stateid stateid, if (unlikely(lrp == NULL)) { status = -ENOMEM; spin_lock(&ino->i_lock); - lo->plh_block_lgets--; pnfs_clear_layoutreturn_waitbit(lo); rpc_wake_up(&NFS_SERVER(ino)->roc_rpcwaitq); spin_unlock(&ino->i_lock); -- cgit v1.2.3 From e1c06f80dcca54cd323d1b98eb385a2c05c6e06b Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 4 Aug 2015 16:40:08 -0400 Subject: NFSv4.1/pnfs: Remove redundant check in pnfs_layoutgets_blocked() layoutget now should already be serialised w.r.t. layout returns Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 26 +++++--------------------- 1 file changed, 5 insertions(+), 21 deletions(-) diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index ada07376a6c7..037126a8c216 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -816,23 +816,12 @@ pnfs_layout_stateid_blocked(const struct pnfs_layout_hdr *lo, return !pnfs_seqid_is_newer(seqid, lo->plh_barrier); } -static bool -pnfs_layout_returning(const struct pnfs_layout_hdr *lo, - struct pnfs_layout_range *range) -{ - return test_bit(NFS_LAYOUT_RETURN, &lo->plh_flags) && - (lo->plh_return_iomode == IOMODE_ANY || - lo->plh_return_iomode == range->iomode); -} - /* lget is set to 1 if called from inside send_layoutget call chain */ static bool -pnfs_layoutgets_blocked(const struct pnfs_layout_hdr *lo, - struct pnfs_layout_range *range) +pnfs_layoutgets_blocked(const struct pnfs_layout_hdr *lo) { return lo->plh_block_lgets || - test_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags) || - pnfs_layout_returning(lo, range); + test_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags); } int @@ -844,7 +833,7 @@ pnfs_choose_layoutget_stateid(nfs4_stateid *dst, struct pnfs_layout_hdr *lo, dprintk("--> %s\n", __func__); spin_lock(&lo->plh_inode->i_lock); - if (pnfs_layoutgets_blocked(lo, range)) { + if (pnfs_layoutgets_blocked(lo)) { status = -EAGAIN; } else if (!nfs4_valid_open_stateid(open_state)) { status = -EBADF; @@ -1546,7 +1535,7 @@ lookup_again: goto out_put_layout_hdr; } - if (pnfs_layoutgets_blocked(lo, &arg)) + if (pnfs_layoutgets_blocked(lo)) goto out_unlock; atomic_inc(&lo->plh_outstanding); spin_unlock(&ino->i_lock); @@ -1618,12 +1607,7 @@ pnfs_layout_process(struct nfs4_layoutget *lgp) lseg->pls_range = res->range; spin_lock(&ino->i_lock); - if (test_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags)) { - dprintk("%s forget reply due to recall\n", __func__); - goto out_forget_reply; - } - - if (pnfs_layoutgets_blocked(lo, &lgp->args.range)) { + if (pnfs_layoutgets_blocked(lo)) { dprintk("%s forget reply due to state\n", __func__); goto out_forget_reply; } -- cgit v1.2.3 From 58830550f009c5f60f702c9d3021f8c3be0012b0 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 4 Aug 2015 17:18:10 -0400 Subject: NFSv4.1/pnfs: Remove redundant wakeup in pnfs_send_layoutreturn() pnfs_clear_layoutreturn_waitbit() should already be calling rpc_wake_up(&NFS_SERVER(ino)->roc_rpcwaitq) for us. Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 1 - 1 file changed, 1 deletion(-) diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 037126a8c216..6151f39c8291 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -943,7 +943,6 @@ pnfs_send_layoutreturn(struct pnfs_layout_hdr *lo, nfs4_stateid stateid, status = -ENOMEM; spin_lock(&ino->i_lock); pnfs_clear_layoutreturn_waitbit(lo); - rpc_wake_up(&NFS_SERVER(ino)->roc_rpcwaitq); spin_unlock(&ino->i_lock); pnfs_put_layout_hdr(lo); goto out; -- cgit v1.2.3 From e9ae58aeee8842a50f7e199d602a5ccb2e41a95f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 17 Aug 2015 12:57:07 -0500 Subject: NFS: nfs_set_pgio_error sometimes misses errors We should ensure that we always set the pgio_header's error field if a READ or WRITE RPC call returns an error. The current code depends on 'hdr->good_bytes' always being initialised to a large value, which is not always done correctly by callers. When this happens, applications may end up missing important errors. Cc: stable@vger.kernel.org Signed-off-by: Trond Myklebust --- fs/nfs/pagelist.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c index 4984bbe55ff1..7c5718ba625e 100644 --- a/fs/nfs/pagelist.c +++ b/fs/nfs/pagelist.c @@ -77,8 +77,8 @@ EXPORT_SYMBOL_GPL(nfs_pgheader_init); void nfs_set_pgio_error(struct nfs_pgio_header *hdr, int error, loff_t pos) { spin_lock(&hdr->lock); - if (pos < hdr->io_start + hdr->good_bytes) { - set_bit(NFS_IOHDR_ERROR, &hdr->flags); + if (!test_and_set_bit(NFS_IOHDR_ERROR, &hdr->flags) + || pos < hdr->io_start + hdr->good_bytes) { clear_bit(NFS_IOHDR_EOF, &hdr->flags); hdr->good_bytes = pos - hdr->io_start; hdr->error = error; -- cgit v1.2.3 From 6f536936b79bd4b5cea8fb0e5b8b0bce8cd1ea4a Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 13 Aug 2015 10:59:07 -0400 Subject: NFSv4.1/pNFS: Fix borken function _same_data_server_addrs_locked() - Switch back to using list_for_each_entry(). Fixes an incorrect test for list NULL termination. - Do not assume that lists are sorted. - Finally, consider an existing entry to match if it consists of a subset of the addresses in the new entry. Cc: stable@vger.kernel.org # 4.0+ Signed-off-by: Trond Myklebust --- fs/nfs/pnfs_nfs.c | 33 +++++++++++++++++++-------------- 1 file changed, 19 insertions(+), 14 deletions(-) diff --git a/fs/nfs/pnfs_nfs.c b/fs/nfs/pnfs_nfs.c index 7a282876662f..e5c679f04099 100644 --- a/fs/nfs/pnfs_nfs.c +++ b/fs/nfs/pnfs_nfs.c @@ -359,26 +359,31 @@ same_sockaddr(struct sockaddr *addr1, struct sockaddr *addr2) return false; } +/* + * Checks if 'dsaddrs1' contains a subset of 'dsaddrs2'. If it does, + * declare a match. + */ static bool _same_data_server_addrs_locked(const struct list_head *dsaddrs1, const struct list_head *dsaddrs2) { struct nfs4_pnfs_ds_addr *da1, *da2; - - /* step through both lists, comparing as we go */ - for (da1 = list_first_entry(dsaddrs1, typeof(*da1), da_node), - da2 = list_first_entry(dsaddrs2, typeof(*da2), da_node); - da1 != NULL && da2 != NULL; - da1 = list_entry(da1->da_node.next, typeof(*da1), da_node), - da2 = list_entry(da2->da_node.next, typeof(*da2), da_node)) { - if (!same_sockaddr((struct sockaddr *)&da1->da_addr, - (struct sockaddr *)&da2->da_addr)) - return false; + struct sockaddr *sa1, *sa2; + bool match = false; + + list_for_each_entry(da1, dsaddrs1, da_node) { + sa1 = (struct sockaddr *)&da1->da_addr; + match = false; + list_for_each_entry(da2, dsaddrs2, da_node) { + sa2 = (struct sockaddr *)&da2->da_addr; + match = same_sockaddr(sa1, sa2); + if (match) + break; + } + if (!match) + break; } - if (da1 == NULL && da2 == NULL) - return true; - - return false; + return match; } /* -- cgit v1.2.3 From 99b1a4c32ad22024ac6198a4337aaec5ea23168f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 13 Aug 2015 15:33:51 -0400 Subject: SUNRPC: Fix a thinko in xs_connect() It is rather pointless to test the value of transport->inet after calling xs_reset_transport(), since it will always be zero, and so we will never see any exponential back off behaviour. Also don't force early connections for SOFTCONN tasks. If the server disconnects us, we should respect the exponential backoff. Cc: stable@vger.kernel.org # 4.0+ Signed-off-by: Trond Myklebust --- net/sunrpc/xprtsock.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c index 0030376327b7..86ed778bb18b 100644 --- a/net/sunrpc/xprtsock.c +++ b/net/sunrpc/xprtsock.c @@ -2279,13 +2279,14 @@ static void xs_connect(struct rpc_xprt *xprt, struct rpc_task *task) WARN_ON_ONCE(!xprt_lock_connect(xprt, task, transport)); - /* Start by resetting any existing state */ - xs_reset_transport(transport); - - if (transport->sock != NULL && !RPC_IS_SOFTCONN(task)) { + if (transport->sock != NULL) { dprintk("RPC: xs_connect delayed xprt %p for %lu " "seconds\n", xprt, xprt->reestablish_timeout / HZ); + + /* Start by resetting any existing state */ + xs_reset_transport(transport); + queue_delayed_work(rpciod_workqueue, &transport->connect_worker, xprt->reestablish_timeout); -- cgit v1.2.3 From 18e3b739fdc826481c6a1335ce0c5b19b3d415da Mon Sep 17 00:00:00 2001 From: Kinglong Mee Date: Sat, 15 Aug 2015 21:52:10 +0800 Subject: NFS: Fix a NULL pointer dereference of migration recovery ops for v4.2 client ---Steps to Reproduce-- # cat /etc/exports /nfs/referal *(rw,insecure,no_subtree_check,no_root_squash,crossmnt) /nfs/old *(ro,insecure,subtree_check,root_squash,crossmnt) # mount -t nfs nfs-server:/nfs/ /mnt/ # ll /mnt/*/ # cat /etc/exports /nfs/referal *(rw,insecure,no_subtree_check,no_root_squash,crossmnt,refer=/nfs/old/@nfs-server) /nfs/old *(ro,insecure,subtree_check,root_squash,crossmnt) # service nfs restart # ll /mnt/*/ --->>>>> oops here [ 5123.102925] BUG: unable to handle kernel NULL pointer dereference at (null) [ 5123.103363] IP: [] nfs4_proc_get_locations+0x9b/0x120 [nfsv4] [ 5123.103752] PGD 587b9067 PUD 3cbf5067 PMD 0 [ 5123.104131] Oops: 0000 [#1] [ 5123.104529] Modules linked in: nfsv4(OE) nfs(OE) fscache(E) nfsd(OE) xfs libcrc32c iscsi_tcp libiscsi_tcp libiscsi scsi_transport_iscsi coretemp crct10dif_pclmul crc32_pclmul crc32c_intel ghash_clmulni_intel ppdev vmw_balloon parport_pc parport i2c_piix4 shpchp auth_rpcgss nfs_acl vmw_vmci lockd grace sunrpc vmwgfx drm_kms_helper ttm drm mptspi serio_raw scsi_transport_spi e1000 mptscsih mptbase ata_generic pata_acpi [last unloaded: nfsd] [ 5123.105887] CPU: 0 PID: 15853 Comm: ::1-manager Tainted: G OE 4.2.0-rc6+ #214 [ 5123.106358] Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 05/20/2014 [ 5123.106860] task: ffff88007620f300 ti: ffff88005877c000 task.ti: ffff88005877c000 [ 5123.107363] RIP: 0010:[] [] nfs4_proc_get_locations+0x9b/0x120 [nfsv4] [ 5123.107909] RSP: 0018:ffff88005877fdb8 EFLAGS: 00010246 [ 5123.108435] RAX: ffff880053f3bc00 RBX: ffff88006ce6c908 RCX: ffff880053a0d240 [ 5123.108968] RDX: ffffea0000e6d940 RSI: ffff8800399a0000 RDI: ffff88006ce6c908 [ 5123.109503] RBP: ffff88005877fe28 R08: ffffffff81c708a0 R09: 0000000000000000 [ 5123.110045] R10: 00000000000001a2 R11: ffff88003ba7f5c8 R12: ffff880054c55800 [ 5123.110618] R13: 0000000000000000 R14: ffff880053a0d240 R15: ffff880053a0d240 [ 5123.111169] FS: 0000000000000000(0000) GS:ffffffff81c27000(0000) knlGS:0000000000000000 [ 5123.111726] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 5123.112286] CR2: 0000000000000000 CR3: 0000000054cac000 CR4: 00000000001406f0 [ 5123.112888] Stack: [ 5123.113458] ffffea0000e6d940 ffff8800399a0000 00000000000167d0 0000000000000000 [ 5123.114049] 0000000000000000 0000000000000000 0000000000000000 00000000a7ec82c6 [ 5123.114662] ffff88005877fe18 ffffea0000e6d940 ffff8800399a0000 ffff880054c55800 [ 5123.115264] Call Trace: [ 5123.115868] [] nfs4_try_migration+0xbb/0x220 [nfsv4] [ 5123.116487] [] nfs4_run_state_manager+0x4ab/0x7b0 [nfsv4] [ 5123.117104] [] ? nfs4_do_reclaim+0x510/0x510 [nfsv4] [ 5123.117813] [] kthread+0xd7/0xf0 [ 5123.118456] [] ? kthread_worker_fn+0x160/0x160 [ 5123.119108] [] ret_from_fork+0x3f/0x70 [ 5123.119723] [] ? kthread_worker_fn+0x160/0x160 [ 5123.120329] Code: 4c 8b 6a 58 74 17 eb 52 48 8d 55 a8 89 c6 4c 89 e7 e8 4a b5 ff ff 8b 45 b0 85 c0 74 1c 4c 89 f9 48 8b 55 90 48 8b 75 98 48 89 df <41> ff 55 00 3d e8 d8 ff ff 41 89 c6 74 cf 48 8b 4d c8 65 48 33 [ 5123.121643] RIP [] nfs4_proc_get_locations+0x9b/0x120 [nfsv4] [ 5123.122308] RSP [ 5123.122942] CR2: 0000000000000000 Fixes: ec011fe847 ("NFS: Introduce a vector of migration recovery ops") Cc: stable@vger.kernel.org # v3.13+ Signed-off-by: Kinglong Mee Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 1 + 1 file changed, 1 insertion(+) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 15ee8bd99b61..43bace809be7 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -8661,6 +8661,7 @@ static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = { .reboot_recovery_ops = &nfs41_reboot_recovery_ops, .nograce_recovery_ops = &nfs41_nograce_recovery_ops, .state_renewal_ops = &nfs41_state_renewal_ops, + .mig_recovery_ops = &nfs41_mig_recovery_ops, }; #endif -- cgit v1.2.3 From 29662fa646b41492a9c298a83399126f94847e93 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Aug 2015 18:40:57 +0200 Subject: pnfs/blocklayout: calculate layoutupdate size correctly We need to include the first u32 for the number of entries. Add a helper for the calculation instead of opencoding it so that it's in one place. Signed-off-by: Christoph Hellwig Signed-off-by: Trond Myklebust --- fs/nfs/blocklayout/extent_tree.c | 12 +++++++++--- 1 file changed, 9 insertions(+), 3 deletions(-) diff --git a/fs/nfs/blocklayout/extent_tree.c b/fs/nfs/blocklayout/extent_tree.c index 31d0b5e53dfd..7536036fb526 100644 --- a/fs/nfs/blocklayout/extent_tree.c +++ b/fs/nfs/blocklayout/extent_tree.c @@ -462,6 +462,12 @@ out: return err; } +static size_t ext_tree_layoutupdate_size(size_t count) +{ + return sizeof(__be32) /* number of entries */ + + BL_EXTENT_SIZE * count; +} + static void ext_tree_free_commitdata(struct nfs4_layoutcommit_args *arg, size_t buffer_size) { @@ -489,7 +495,7 @@ static int ext_tree_encode_commit(struct pnfs_block_layout *bl, __be32 *p, continue; (*count)++; - if (*count * BL_EXTENT_SIZE > buffer_size) { + if (ext_tree_layoutupdate_size(*count) > buffer_size) { /* keep counting.. */ ret = -ENOSPC; continue; @@ -530,7 +536,7 @@ retry: if (unlikely(ret)) { ext_tree_free_commitdata(arg, buffer_size); - buffer_size = sizeof(__be32) + BL_EXTENT_SIZE * count; + buffer_size = ext_tree_layoutupdate_size(count); count = 0; arg->layoutupdate_pages = @@ -549,7 +555,7 @@ retry: } *start_p = cpu_to_be32(count); - arg->layoutupdate_len = sizeof(__be32) + BL_EXTENT_SIZE * count; + arg->layoutupdate_len = ext_tree_layoutupdate_size(count); if (unlikely(arg->layoutupdate_pages != &arg->layoutupdate_page)) { __be32 *p = start_p; -- cgit v1.2.3 From 68596bd188e5e621c28a2f6fc0a3dd80a606d16b Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Aug 2015 18:40:58 +0200 Subject: pnfs/blocklayout: set up layoutupdate_pages properly We need to replace the __be32 with a void pointer to do proper arithmentics on the virtual addresses so that we can get the right page pointers. Reported-by: Dan Carpenter Signed-off-by: Christoph Hellwig Signed-off-by: Trond Myklebust --- fs/nfs/blocklayout/extent_tree.c | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) diff --git a/fs/nfs/blocklayout/extent_tree.c b/fs/nfs/blocklayout/extent_tree.c index 7536036fb526..a11b759d294a 100644 --- a/fs/nfs/blocklayout/extent_tree.c +++ b/fs/nfs/blocklayout/extent_tree.c @@ -558,14 +558,11 @@ retry: arg->layoutupdate_len = ext_tree_layoutupdate_size(count); if (unlikely(arg->layoutupdate_pages != &arg->layoutupdate_page)) { - __be32 *p = start_p; + void *p = start_p, *end = p + arg->layoutupdate_len; int i = 0; - for (p = start_p; - p < start_p + arg->layoutupdate_len; - p += PAGE_SIZE) { + for ( ; p < end; p += PAGE_SIZE) arg->layoutupdate_pages[i++] = vmalloc_to_page(p); - } } dprintk("%s found %zu ranges\n", __func__, count); -- cgit v1.2.3 From 2bd3c63a333c364cfde4701b51f82e1fae106a88 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Aug 2015 18:40:59 +0200 Subject: pnfs/blocklayout: reject too long signatures Instead of overwriting kernel memory reject too long signatures. Signed-off-by: Christoph Hellwig Signed-off-by: Trond Myklebust --- fs/nfs/blocklayout/dev.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/fs/nfs/blocklayout/dev.c b/fs/nfs/blocklayout/dev.c index e535599a0719..d76993a42432 100644 --- a/fs/nfs/blocklayout/dev.c +++ b/fs/nfs/blocklayout/dev.c @@ -65,6 +65,11 @@ nfs4_block_decode_volume(struct xdr_stream *xdr, struct pnfs_block_volume *b) return -EIO; p = xdr_decode_hyper(p, &b->simple.sigs[i].offset); b->simple.sigs[i].sig_len = be32_to_cpup(p++); + if (b->simple.sigs[i].sig_len > PNFS_BLOCK_UUID_LEN) { + pr_info("signature too long: %d\n", + b->simple.sigs[i].sig_len); + return -EIO; + } p = xdr_inline_decode(xdr, b->simple.sigs[i].sig_len); if (!p) -- cgit v1.2.3 From 513d6d7a9591d7c5ebb5f05465942185db9299a4 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Aug 2015 18:41:00 +0200 Subject: pnfs/blocklayout: pass proper file mode to blkdev_get/put We generally want to read and write to a block device that's used by the pNFS block layout client (and even if it's read only the server has no way of telling us). Add FMODE_WRITE to the mode argument so that we don't incorrectly tell the block driver that we want a read-only open. Signed-off-by: Christoph Hellwig Signed-off-by: Trond Myklebust --- fs/nfs/blocklayout/dev.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/fs/nfs/blocklayout/dev.c b/fs/nfs/blocklayout/dev.c index d76993a42432..a861bbdfe577 100644 --- a/fs/nfs/blocklayout/dev.c +++ b/fs/nfs/blocklayout/dev.c @@ -22,7 +22,7 @@ bl_free_device(struct pnfs_block_dev *dev) kfree(dev->children); } else { if (dev->bdev) - blkdev_put(dev->bdev, FMODE_READ); + blkdev_put(dev->bdev, FMODE_READ | FMODE_WRITE); } } @@ -200,7 +200,7 @@ bl_parse_simple(struct nfs_server *server, struct pnfs_block_dev *d, if (!dev) return -EIO; - d->bdev = blkdev_get_by_dev(dev, FMODE_READ, NULL); + d->bdev = blkdev_get_by_dev(dev, FMODE_READ | FMODE_WRITE, NULL); if (IS_ERR(d->bdev)) { printk(KERN_WARNING "pNFS: failed to open device %d:%d (%ld)\n", MAJOR(dev), MINOR(dev), PTR_ERR(d->bdev)); -- cgit v1.2.3 From 8bb28975823aee062f82b99ddacc499601c0cfd1 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Aug 2015 18:41:01 +0200 Subject: pnfs: move common blocklayout XDR defintions to nfs4.h Signed-off-by: Christoph Hellwig Signed-off-by: Trond Myklebust --- fs/nfs/blocklayout/blocklayout.h | 19 +------------------ fs/nfs/blocklayout/extent_tree.c | 2 +- fs/nfsd/blocklayoutxdr.c | 2 +- fs/nfsd/blocklayoutxdr.h | 15 --------------- include/linux/nfs4.h | 18 ++++++++++++++++++ 5 files changed, 21 insertions(+), 35 deletions(-) diff --git a/fs/nfs/blocklayout/blocklayout.h b/fs/nfs/blocklayout/blocklayout.h index 92dca9e90d8d..c556640dcf3b 100644 --- a/fs/nfs/blocklayout/blocklayout.h +++ b/fs/nfs/blocklayout/blocklayout.h @@ -46,13 +46,6 @@ struct pnfs_block_dev; -enum pnfs_block_volume_type { - PNFS_BLOCK_VOLUME_SIMPLE = 0, - PNFS_BLOCK_VOLUME_SLICE = 1, - PNFS_BLOCK_VOLUME_CONCAT = 2, - PNFS_BLOCK_VOLUME_STRIPE = 3, -}; - #define PNFS_BLOCK_MAX_UUIDS 4 #define PNFS_BLOCK_MAX_DEVICES 64 @@ -117,13 +110,6 @@ struct pnfs_block_dev { struct pnfs_block_dev_map *map); }; -enum exstate4 { - PNFS_BLOCK_READWRITE_DATA = 0, - PNFS_BLOCK_READ_DATA = 1, - PNFS_BLOCK_INVALID_DATA = 2, /* mapped, but data is invalid */ - PNFS_BLOCK_NONE_DATA = 3 /* unmapped, it's a hole */ -}; - /* sector_t fields are all in 512-byte sectors */ struct pnfs_block_extent { union { @@ -134,15 +120,12 @@ struct pnfs_block_extent { sector_t be_f_offset; /* the starting offset in the file */ sector_t be_length; /* the size of the extent */ sector_t be_v_offset; /* the starting offset in the volume */ - enum exstate4 be_state; /* the state of this extent */ + enum pnfs_block_extent_state be_state; /* the state of this extent */ #define EXTENT_WRITTEN 1 #define EXTENT_COMMITTING 2 unsigned int be_tag; }; -/* on the wire size of the extent */ -#define BL_EXTENT_SIZE (7 * sizeof(__be32) + NFS4_DEVICEID4_SIZE) - struct pnfs_block_layout { struct pnfs_layout_hdr bl_layout; struct rb_root bl_ext_rw; diff --git a/fs/nfs/blocklayout/extent_tree.c b/fs/nfs/blocklayout/extent_tree.c index a11b759d294a..c59a59c37f3d 100644 --- a/fs/nfs/blocklayout/extent_tree.c +++ b/fs/nfs/blocklayout/extent_tree.c @@ -465,7 +465,7 @@ out: static size_t ext_tree_layoutupdate_size(size_t count) { return sizeof(__be32) /* number of entries */ + - BL_EXTENT_SIZE * count; + PNFS_BLOCK_EXTENT_SIZE * count; } static void ext_tree_free_commitdata(struct nfs4_layoutcommit_args *arg, diff --git a/fs/nfsd/blocklayoutxdr.c b/fs/nfsd/blocklayoutxdr.c index 9aa2796da90d..6d834dc9bbc8 100644 --- a/fs/nfsd/blocklayoutxdr.c +++ b/fs/nfsd/blocklayoutxdr.c @@ -101,7 +101,7 @@ nfsd4_block_decode_layoutupdate(__be32 *p, u32 len, struct iomap **iomapp, } nr_iomaps = be32_to_cpup(p++); - expected = sizeof(__be32) + nr_iomaps * NFS4_BLOCK_EXTENT_SIZE; + expected = sizeof(__be32) + nr_iomaps * PNFS_BLOCK_EXTENT_SIZE; if (len != expected) { dprintk("%s: extent array size mismatch: %u/%u\n", __func__, len, expected); diff --git a/fs/nfsd/blocklayoutxdr.h b/fs/nfsd/blocklayoutxdr.h index fdc79037c0e7..6de925fe8499 100644 --- a/fs/nfsd/blocklayoutxdr.h +++ b/fs/nfsd/blocklayoutxdr.h @@ -7,13 +7,6 @@ struct iomap; struct xdr_stream; -enum pnfs_block_extent_state { - PNFS_BLOCK_READWRITE_DATA = 0, - PNFS_BLOCK_READ_DATA = 1, - PNFS_BLOCK_INVALID_DATA = 2, - PNFS_BLOCK_NONE_DATA = 3, -}; - struct pnfs_block_extent { struct nfsd4_deviceid vol_id; u64 foff; @@ -21,14 +14,6 @@ struct pnfs_block_extent { u64 soff; enum pnfs_block_extent_state es; }; -#define NFS4_BLOCK_EXTENT_SIZE 44 - -enum pnfs_block_volume_type { - PNFS_BLOCK_VOLUME_SIMPLE = 0, - PNFS_BLOCK_VOLUME_SLICE = 1, - PNFS_BLOCK_VOLUME_CONCAT = 2, - PNFS_BLOCK_VOLUME_STRIPE = 3, -}; /* * Random upper cap for the uuid length to avoid unbounded allocation. diff --git a/include/linux/nfs4.h b/include/linux/nfs4.h index b8e72aad919c..00121f298269 100644 --- a/include/linux/nfs4.h +++ b/include/linux/nfs4.h @@ -547,6 +547,24 @@ enum pnfs_notify_deviceid_type4 { NOTIFY_DEVICEID4_DELETE = 1 << 2, }; +enum pnfs_block_volume_type { + PNFS_BLOCK_VOLUME_SIMPLE = 0, + PNFS_BLOCK_VOLUME_SLICE = 1, + PNFS_BLOCK_VOLUME_CONCAT = 2, + PNFS_BLOCK_VOLUME_STRIPE = 3, +}; + +enum pnfs_block_extent_state { + PNFS_BLOCK_READWRITE_DATA = 0, + PNFS_BLOCK_READ_DATA = 1, + PNFS_BLOCK_INVALID_DATA = 2, + PNFS_BLOCK_NONE_DATA = 3, +}; + +/* on the wire size of a block layout extent */ +#define PNFS_BLOCK_EXTENT_SIZE \ + (7 * sizeof(__be32) + NFS4_DEVICEID4_SIZE) + #define NFL4_UFLG_MASK 0x0000003F #define NFL4_UFLG_DENSE 0x00000001 #define NFL4_UFLG_COMMIT_THRU_MDS 0x00000002 -- cgit v1.2.3 From ce603281468cf1be0a70a45fdaef761db134b900 Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Fri, 10 Jul 2015 15:59:26 -0400 Subject: nfs: remove some dead code in ff_layout_pg_get_mirror_count_write We already know that pg_lseg is NULL here. Signed-off-by: Jeff Layton Signed-off-by: Trond Myklebust --- fs/nfs/flexfilelayout/flexfilelayout.c | 2 -- 1 file changed, 2 deletions(-) diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index 2a93bec7e6dd..13fe64b4e259 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -730,8 +730,6 @@ ff_layout_pg_get_mirror_count_write(struct nfs_pageio_descriptor *pgio, return FF_LAYOUT_MIRROR_COUNT(pgio->pg_lseg); /* no lseg means that pnfs is not in use, so no mirroring here */ - pnfs_put_lseg(pgio->pg_lseg); - pgio->pg_lseg = NULL; nfs_pageio_reset_write_mds(pgio); return 1; } -- cgit v1.2.3 From 0b936e37df8111ea727263a5a246b58e05b7aa1b Mon Sep 17 00:00:00 2001 From: Anna Schumaker Date: Mon, 13 Jul 2015 14:01:24 -0400 Subject: NFS: Remove unused variable "pages_ptr" This variable is initialized to NULL and is never modified before being passed to nfs_readdir_free_large_page(). But that's okay, because nfs_readdir_free_large_page() only seems to exist as a way of calling nfs_readdir_free_pagearray() without this parameter. Let's simplify by removing pages_ptr and nfs_readdir_free_pagearray(). Signed-off-by: Anna Schumaker Signed-off-by: Trond Myklebust --- fs/nfs/dir.c | 12 ++---------- 1 file changed, 2 insertions(+), 10 deletions(-) diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index 547308a5ec6f..26c5d63cfc3d 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -590,16 +590,9 @@ void nfs_readdir_free_pagearray(struct page **pages, unsigned int npages) put_page(pages[i]); } -static -void nfs_readdir_free_large_page(void *ptr, struct page **pages, - unsigned int npages) -{ - nfs_readdir_free_pagearray(pages, npages); -} - /* * nfs_readdir_large_page will allocate pages that must be freed with a call - * to nfs_readdir_free_large_page + * to nfs_readdir_free_pagearray */ static int nfs_readdir_large_page(struct page **pages, unsigned int npages) @@ -623,7 +616,6 @@ static int nfs_readdir_xdr_to_array(nfs_readdir_descriptor_t *desc, struct page *page, struct inode *inode) { struct page *pages[NFS_MAX_READDIR_PAGES]; - void *pages_ptr = NULL; struct nfs_entry entry; struct file *file = desc->file; struct nfs_cache_array *array; @@ -671,7 +663,7 @@ int nfs_readdir_xdr_to_array(nfs_readdir_descriptor_t *desc, struct page *page, } } while (array->eof_index < 0); - nfs_readdir_free_large_page(pages_ptr, pages, array_size); + nfs_readdir_free_pagearray(pages, array_size); out_release_array: nfs_readdir_release_array(page); out_label_free: -- cgit v1.2.3 From c7e9668e78eab69b7ade9897bea1fbd77dd18775 Mon Sep 17 00:00:00 2001 From: Anna Schumaker Date: Mon, 13 Jul 2015 14:01:25 -0400 Subject: NFS: Rename nfs_readdir_free_pagearray() and nfs_readdir_large_page() nfs_readdir_xdr_to_array() uses both a cache array and an array of pages, so I rename these functions to make it clearer how the code works. nfs_readdir_large_page() becomes nfs_readdir_alloc_pages() because this function has absolutely nothing to do with setting up a large page. nfs_readdir_free_pagearray() becomes nfs_readdir_free_pages() to stay consistent with the new alloc_pages() function. Signed-off-by: Anna Schumaker Signed-off-by: Trond Myklebust --- fs/nfs/dir.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index 26c5d63cfc3d..3d8e4ffa0a33 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -583,7 +583,7 @@ out_nopages: } static -void nfs_readdir_free_pagearray(struct page **pages, unsigned int npages) +void nfs_readdir_free_pages(struct page **pages, unsigned int npages) { unsigned int i; for (i = 0; i < npages; i++) @@ -595,7 +595,7 @@ void nfs_readdir_free_pagearray(struct page **pages, unsigned int npages) * to nfs_readdir_free_pagearray */ static -int nfs_readdir_large_page(struct page **pages, unsigned int npages) +int nfs_readdir_alloc_pages(struct page **pages, unsigned int npages) { unsigned int i; @@ -608,7 +608,7 @@ int nfs_readdir_large_page(struct page **pages, unsigned int npages) return 0; out_freepages: - nfs_readdir_free_pagearray(pages, i); + nfs_readdir_free_pages(pages, i); return -ENOMEM; } @@ -645,7 +645,7 @@ int nfs_readdir_xdr_to_array(nfs_readdir_descriptor_t *desc, struct page *page, memset(array, 0, sizeof(struct nfs_cache_array)); array->eof_index = -1; - status = nfs_readdir_large_page(pages, array_size); + status = nfs_readdir_alloc_pages(pages, array_size); if (status < 0) goto out_release_array; do { @@ -663,7 +663,7 @@ int nfs_readdir_xdr_to_array(nfs_readdir_descriptor_t *desc, struct page *page, } } while (array->eof_index < 0); - nfs_readdir_free_pagearray(pages, array_size); + nfs_readdir_free_pages(pages, array_size); out_release_array: nfs_readdir_release_array(page); out_label_free: -- cgit v1.2.3 From 7b0ce60c0b203a2a037f308e387152acbf99c445 Mon Sep 17 00:00:00 2001 From: Anna Schumaker Date: Mon, 13 Jul 2015 14:01:26 -0400 Subject: SUNRPC: Drop double-underscores from rpc_cmp_addr{4|6}() I'm planning on using these functions inside the client, so remove the underscores to make it feel like I'm using a public interface. Signed-off-by: Anna Schumaker Signed-off-by: Trond Myklebust --- include/linux/sunrpc/addr.h | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/include/linux/sunrpc/addr.h b/include/linux/sunrpc/addr.h index 07d8e53bedfc..772faefc15aa 100644 --- a/include/linux/sunrpc/addr.h +++ b/include/linux/sunrpc/addr.h @@ -46,8 +46,8 @@ static inline void rpc_set_port(struct sockaddr *sap, #define IPV6_SCOPE_DELIMITER '%' #define IPV6_SCOPE_ID_LEN sizeof("%nnnnnnnnnn") -static inline bool __rpc_cmp_addr4(const struct sockaddr *sap1, - const struct sockaddr *sap2) +static inline bool rpc_cmp_addr4(const struct sockaddr *sap1, + const struct sockaddr *sap2) { const struct sockaddr_in *sin1 = (const struct sockaddr_in *)sap1; const struct sockaddr_in *sin2 = (const struct sockaddr_in *)sap2; @@ -67,8 +67,8 @@ static inline bool __rpc_copy_addr4(struct sockaddr *dst, } #if IS_ENABLED(CONFIG_IPV6) -static inline bool __rpc_cmp_addr6(const struct sockaddr *sap1, - const struct sockaddr *sap2) +static inline bool rpc_cmp_addr6(const struct sockaddr *sap1, + const struct sockaddr *sap2) { const struct sockaddr_in6 *sin1 = (const struct sockaddr_in6 *)sap1; const struct sockaddr_in6 *sin2 = (const struct sockaddr_in6 *)sap2; @@ -122,9 +122,9 @@ static inline bool rpc_cmp_addr(const struct sockaddr *sap1, if (sap1->sa_family == sap2->sa_family) { switch (sap1->sa_family) { case AF_INET: - return __rpc_cmp_addr4(sap1, sap2); + return rpc_cmp_addr4(sap1, sap2); case AF_INET6: - return __rpc_cmp_addr6(sap1, sap2); + return rpc_cmp_addr6(sap1, sap2); } } return false; -- cgit v1.2.3 From 58cc8a55aa38655d472a4c381df465c9ab97b4d6 Mon Sep 17 00:00:00 2001 From: Anna Schumaker Date: Mon, 13 Jul 2015 14:01:27 -0400 Subject: SUNRPC: Add an rpc_cmp_addr_port() function This function is to help determine if two sockaddrs are really the same socket. Signed-off-by: Anna Schumaker Signed-off-by: Trond Myklebust --- include/linux/sunrpc/addr.h | 13 +++++++++++++ 1 file changed, 13 insertions(+) diff --git a/include/linux/sunrpc/addr.h b/include/linux/sunrpc/addr.h index 772faefc15aa..9bc3538d09a9 100644 --- a/include/linux/sunrpc/addr.h +++ b/include/linux/sunrpc/addr.h @@ -130,6 +130,19 @@ static inline bool rpc_cmp_addr(const struct sockaddr *sap1, return false; } +/** + * rpc_cmp_addr_port - compare the address and port number of two sockaddrs. + * @sap1: first sockaddr + * @sap2: second sockaddr + */ +static inline bool rpc_cmp_addr_port(const struct sockaddr *sap1, + const struct sockaddr *sap2) +{ + if (!rpc_cmp_addr(sap1, sap2)) + return false; + return rpc_get_port(sap1) == rpc_get_port(sap2); +} + /** * rpc_copy_addr - copy the address portion of one sockaddr to another * @dst: destination sockaddr -- cgit v1.2.3 From d8efa4e62505f5113e363572b5438b7be0d08b12 Mon Sep 17 00:00:00 2001 From: Anna Schumaker Date: Mon, 13 Jul 2015 14:01:28 -0400 Subject: NFS: Use RPC functions for matching sockaddrs They already exist and do the exact same thing. Let's save ourselves several lines of code! Signed-off-by: Anna Schumaker Signed-off-by: Trond Myklebust --- fs/nfs/client.c | 113 +--------------------------------------------------- fs/nfs/internal.h | 4 -- fs/nfs/nfs4client.c | 5 +-- 3 files changed, 3 insertions(+), 119 deletions(-) diff --git a/fs/nfs/client.c b/fs/nfs/client.c index 4a90c9bb3135..57c5a02f6213 100644 --- a/fs/nfs/client.c +++ b/fs/nfs/client.c @@ -20,6 +20,7 @@ #include #include #include +#include #include #include #include @@ -285,116 +286,6 @@ void nfs_put_client(struct nfs_client *clp) } EXPORT_SYMBOL_GPL(nfs_put_client); -#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) -/* - * Test if two ip6 socket addresses refer to the same socket by - * comparing relevant fields. The padding bytes specifically, are not - * compared. sin6_flowinfo is not compared because it only affects QoS - * and sin6_scope_id is only compared if the address is "link local" - * because "link local" addresses need only be unique to a specific - * link. Conversely, ordinary unicast addresses might have different - * sin6_scope_id. - * - * The caller should ensure both socket addresses are AF_INET6. - */ -static int nfs_sockaddr_match_ipaddr6(const struct sockaddr *sa1, - const struct sockaddr *sa2) -{ - const struct sockaddr_in6 *sin1 = (const struct sockaddr_in6 *)sa1; - const struct sockaddr_in6 *sin2 = (const struct sockaddr_in6 *)sa2; - - if (!ipv6_addr_equal(&sin1->sin6_addr, &sin2->sin6_addr)) - return 0; - else if (ipv6_addr_type(&sin1->sin6_addr) & IPV6_ADDR_LINKLOCAL) - return sin1->sin6_scope_id == sin2->sin6_scope_id; - - return 1; -} -#else /* !defined(CONFIG_IPV6) && !defined(CONFIG_IPV6_MODULE) */ -static int nfs_sockaddr_match_ipaddr6(const struct sockaddr *sa1, - const struct sockaddr *sa2) -{ - return 0; -} -#endif - -/* - * Test if two ip4 socket addresses refer to the same socket, by - * comparing relevant fields. The padding bytes specifically, are - * not compared. - * - * The caller should ensure both socket addresses are AF_INET. - */ -static int nfs_sockaddr_match_ipaddr4(const struct sockaddr *sa1, - const struct sockaddr *sa2) -{ - const struct sockaddr_in *sin1 = (const struct sockaddr_in *)sa1; - const struct sockaddr_in *sin2 = (const struct sockaddr_in *)sa2; - - return sin1->sin_addr.s_addr == sin2->sin_addr.s_addr; -} - -static int nfs_sockaddr_cmp_ip6(const struct sockaddr *sa1, - const struct sockaddr *sa2) -{ - const struct sockaddr_in6 *sin1 = (const struct sockaddr_in6 *)sa1; - const struct sockaddr_in6 *sin2 = (const struct sockaddr_in6 *)sa2; - - return nfs_sockaddr_match_ipaddr6(sa1, sa2) && - (sin1->sin6_port == sin2->sin6_port); -} - -static int nfs_sockaddr_cmp_ip4(const struct sockaddr *sa1, - const struct sockaddr *sa2) -{ - const struct sockaddr_in *sin1 = (const struct sockaddr_in *)sa1; - const struct sockaddr_in *sin2 = (const struct sockaddr_in *)sa2; - - return nfs_sockaddr_match_ipaddr4(sa1, sa2) && - (sin1->sin_port == sin2->sin_port); -} - -#if defined(CONFIG_NFS_V4_1) -/* - * Test if two socket addresses represent the same actual socket, - * by comparing (only) relevant fields, excluding the port number. - */ -int nfs_sockaddr_match_ipaddr(const struct sockaddr *sa1, - const struct sockaddr *sa2) -{ - if (sa1->sa_family != sa2->sa_family) - return 0; - - switch (sa1->sa_family) { - case AF_INET: - return nfs_sockaddr_match_ipaddr4(sa1, sa2); - case AF_INET6: - return nfs_sockaddr_match_ipaddr6(sa1, sa2); - } - return 0; -} -EXPORT_SYMBOL_GPL(nfs_sockaddr_match_ipaddr); -#endif /* CONFIG_NFS_V4_1 */ - -/* - * Test if two socket addresses represent the same actual socket, - * by comparing (only) relevant fields, including the port number. - */ -static int nfs_sockaddr_cmp(const struct sockaddr *sa1, - const struct sockaddr *sa2) -{ - if (sa1->sa_family != sa2->sa_family) - return 0; - - switch (sa1->sa_family) { - case AF_INET: - return nfs_sockaddr_cmp_ip4(sa1, sa2); - case AF_INET6: - return nfs_sockaddr_cmp_ip6(sa1, sa2); - } - return 0; -} - /* * Find an nfs_client on the list that matches the initialisation data * that is supplied. @@ -421,7 +312,7 @@ static struct nfs_client *nfs_match_client(const struct nfs_client_initdata *dat if (clp->cl_minorversion != data->minorversion) continue; /* Match the full socket address */ - if (!nfs_sockaddr_cmp(sap, clap)) + if (!rpc_cmp_addr_port(sap, clap)) continue; atomic_inc(&clp->cl_count); diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h index 1dad18105ed0..9ab3b1c21bb4 100644 --- a/fs/nfs/internal.h +++ b/fs/nfs/internal.h @@ -219,10 +219,6 @@ static inline void nfs_fs_proc_exit(void) } #endif -#ifdef CONFIG_NFS_V4_1 -int nfs_sockaddr_match_ipaddr(const struct sockaddr *, const struct sockaddr *); -#endif - /* callback_xdr.c */ extern struct svc_version nfs4_callback_version1; extern struct svc_version nfs4_callback_version4; diff --git a/fs/nfs/nfs4client.c b/fs/nfs/nfs4client.c index 3aa6a9ba5113..223bedda64ae 100644 --- a/fs/nfs/nfs4client.c +++ b/fs/nfs/nfs4client.c @@ -729,10 +729,7 @@ static bool nfs4_cb_match_client(const struct sockaddr *addr, return false; /* Match only the IP address, not the port number */ - if (!nfs_sockaddr_match_ipaddr(addr, clap)) - return false; - - return true; + return rpc_cmp_addr(addr, clap); } /* -- cgit v1.2.3 From fb2a525cf086bb080f84b0b748ee823ef79e539c Mon Sep 17 00:00:00 2001 From: Anna Schumaker Date: Mon, 13 Jul 2015 14:01:29 -0400 Subject: NFS: Combine nfs_idmap_{init|quit}() and nfs_idmap_{init|quit}_keyring() The idmap_init() and idmap_quit() functions only exist to call the _keyring() version. Let's just call the keyring() functions directly. Signed-off-by: Anna Schumaker Signed-off-by: Trond Myklebust --- fs/nfs/nfs4idmap.c | 14 ++------------ 1 file changed, 2 insertions(+), 12 deletions(-) diff --git a/fs/nfs/nfs4idmap.c b/fs/nfs/nfs4idmap.c index 535dfc69c628..2e4902203c35 100644 --- a/fs/nfs/nfs4idmap.c +++ b/fs/nfs/nfs4idmap.c @@ -184,7 +184,7 @@ static struct key_type key_type_id_resolver = { .read = user_read, }; -static int nfs_idmap_init_keyring(void) +int nfs_idmap_init(void) { struct cred *cred; struct key *keyring; @@ -230,7 +230,7 @@ failed_put_cred: return ret; } -static void nfs_idmap_quit_keyring(void) +void nfs_idmap_quit(void) { key_revoke(id_resolver_cache->thread_keyring); unregister_key_type(&key_type_id_resolver); @@ -492,16 +492,6 @@ nfs_idmap_delete(struct nfs_client *clp) kfree(idmap); } -int nfs_idmap_init(void) -{ - return nfs_idmap_init_keyring(); -} - -void nfs_idmap_quit(void) -{ - nfs_idmap_quit_keyring(); -} - static int nfs_idmap_prepare_message(char *desc, struct idmap *idmap, struct idmap_msg *im, struct rpc_pipe_msg *msg) -- cgit v1.2.3 From 3f10a6af4b577da2117907ac8420f27af81d57de Mon Sep 17 00:00:00 2001 From: Anna Schumaker Date: Mon, 13 Jul 2015 14:01:31 -0400 Subject: NFS: Remove nfs41_server_notify_{target|highest}_slotid_update() All these functions do is call nfs41_ping_server() without adding anything. Let's remove them and give nfs41_ping_server() a better name instead. Signed-off-by: Anna Schumaker Signed-off-by: Trond Myklebust --- fs/nfs/callback_proc.c | 2 +- fs/nfs/nfs4_fs.h | 4 +--- fs/nfs/nfs4proc.c | 2 +- fs/nfs/nfs4state.c | 12 +----------- 4 files changed, 4 insertions(+), 16 deletions(-) diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c index 29e3c1b011b7..624bef79ba7c 100644 --- a/fs/nfs/callback_proc.c +++ b/fs/nfs/callback_proc.c @@ -554,7 +554,7 @@ __be32 nfs4_callback_recallslot(struct cb_recallslotargs *args, void *dummy, status = htonl(NFS4_OK); nfs41_set_target_slotid(fc_tbl, args->crsa_target_highest_slotid); - nfs41_server_notify_target_slotid_update(cps->clp); + nfs41_notify_server(cps->clp); out: dprintk("%s: exit with status = %d\n", __func__, ntohl(status)); return status; diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index ea3bee919a76..50cfc4ca7a02 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -405,9 +405,7 @@ int nfs40_discover_server_trunking(struct nfs_client *clp, int nfs41_discover_server_trunking(struct nfs_client *clp, struct nfs_client **, struct rpc_cred *); extern void nfs4_schedule_session_recovery(struct nfs4_session *, int); -extern void nfs41_server_notify_target_slotid_update(struct nfs_client *clp); -extern void nfs41_server_notify_highest_slotid_update(struct nfs_client *clp); - +extern void nfs41_notify_server(struct nfs_client *); #else static inline void nfs4_schedule_session_recovery(struct nfs4_session *session, int err) { diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 43bace809be7..f4e5816a77b0 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -586,7 +586,7 @@ out_unlock: spin_unlock(&tbl->slot_tbl_lock); res->sr_slot = NULL; if (send_new_highest_used_slotid) - nfs41_server_notify_highest_slotid_update(session->clp); + nfs41_notify_server(session->clp); } int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res) diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index f2e2ad894461..da73bc443238 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -2152,23 +2152,13 @@ void nfs4_schedule_session_recovery(struct nfs4_session *session, int err) } EXPORT_SYMBOL_GPL(nfs4_schedule_session_recovery); -static void nfs41_ping_server(struct nfs_client *clp) +void nfs41_notify_server(struct nfs_client *clp) { /* Use CHECK_LEASE to ping the server with a SEQUENCE */ set_bit(NFS4CLNT_CHECK_LEASE, &clp->cl_state); nfs4_schedule_state_manager(clp); } -void nfs41_server_notify_target_slotid_update(struct nfs_client *clp) -{ - nfs41_ping_server(clp); -} - -void nfs41_server_notify_highest_slotid_update(struct nfs_client *clp) -{ - nfs41_ping_server(clp); -} - static void nfs4_reset_all_state(struct nfs_client *clp) { if (test_and_set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) == 0) { -- cgit v1.2.3 From ae09c31f66cc673bb8c64a5dbfdc04ab67f66d7e Mon Sep 17 00:00:00 2001 From: Anna Schumaker Date: Mon, 13 Jul 2015 14:01:32 -0400 Subject: NFS: Rename nfs_commit_unstable_pages() to nfs_write_inode() All nfs_write_inode() does is pass its arguments to nfs_commit_unstable_pages(). Let's cut out the middle man and have nfs_write_pages() do the work directly. Signed-off-by: Anna Schumaker Signed-off-by: Trond Myklebust --- fs/nfs/write.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) diff --git a/fs/nfs/write.c b/fs/nfs/write.c index fdee9270ca15..388f48079c43 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -1812,7 +1812,7 @@ out_mark_dirty: return res; } -static int nfs_commit_unstable_pages(struct inode *inode, struct writeback_control *wbc) +int nfs_write_inode(struct inode *inode, struct writeback_control *wbc) { struct nfs_inode *nfsi = NFS_I(inode); int flags = FLUSH_SYNC; @@ -1847,11 +1847,6 @@ out_mark_dirty: __mark_inode_dirty(inode, I_DIRTY_DATASYNC); return ret; } - -int nfs_write_inode(struct inode *inode, struct writeback_control *wbc) -{ - return nfs_commit_unstable_pages(inode, wbc); -} EXPORT_SYMBOL_GPL(nfs_write_inode); /* -- cgit v1.2.3 From aff8d8dc4c34804c6a1de04f1b3313aa9063bf46 Mon Sep 17 00:00:00 2001 From: Anna Schumaker Date: Mon, 13 Jul 2015 14:01:33 -0400 Subject: NFS: Remove nfs_release() And call nfs_file_clear_open_context() directly. This makes it obvious that nfs_file_release() will always return 0. Signed-off-by: Anna Schumaker Signed-off-by: Trond Myklebust --- fs/nfs/file.c | 3 ++- fs/nfs/inode.c | 8 +------- include/linux/nfs_fs.h | 2 +- 3 files changed, 4 insertions(+), 9 deletions(-) diff --git a/fs/nfs/file.c b/fs/nfs/file.c index cc4fa1ed61fc..7538a8582384 100644 --- a/fs/nfs/file.c +++ b/fs/nfs/file.c @@ -82,7 +82,8 @@ nfs_file_release(struct inode *inode, struct file *filp) dprintk("NFS: release(%pD2)\n", filp); nfs_inc_stats(inode, NFSIOS_VFSRELEASE); - return nfs_release(inode, filp); + nfs_file_clear_open_context(filp); + return 0; } EXPORT_SYMBOL_GPL(nfs_file_release); diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 0adc7d245b3d..382c8a46c078 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -888,7 +888,7 @@ struct nfs_open_context *nfs_find_open_context(struct inode *inode, struct rpc_c return ctx; } -static void nfs_file_clear_open_context(struct file *filp) +void nfs_file_clear_open_context(struct file *filp) { struct nfs_open_context *ctx = nfs_file_open_context(filp); @@ -919,12 +919,6 @@ int nfs_open(struct inode *inode, struct file *filp) return 0; } -int nfs_release(struct inode *inode, struct file *filp) -{ - nfs_file_clear_open_context(filp); - return 0; -} - /* * This function is called whenever some part of NFS notices that * the cached attributes have to be refreshed. diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h index 874b77228fb9..c0e961474a52 100644 --- a/include/linux/nfs_fs.h +++ b/include/linux/nfs_fs.h @@ -353,7 +353,6 @@ extern void nfs_access_add_cache(struct inode *, struct nfs_access_entry *); extern void nfs_access_set_mask(struct nfs_access_entry *, u32); extern int nfs_permission(struct inode *, int); extern int nfs_open(struct inode *, struct file *); -extern int nfs_release(struct inode *, struct file *); extern int nfs_attribute_timeout(struct inode *inode); extern int nfs_attribute_cache_expired(struct inode *inode); extern int nfs_revalidate_inode(struct nfs_server *server, struct inode *inode); @@ -371,6 +370,7 @@ extern struct nfs_open_context *nfs_find_open_context(struct inode *inode, struc extern struct nfs_open_context *alloc_nfs_open_context(struct dentry *dentry, fmode_t f_mode); extern void nfs_inode_attach_open_context(struct nfs_open_context *ctx); extern void nfs_file_set_open_context(struct file *filp, struct nfs_open_context *ctx); +extern void nfs_file_clear_open_context(struct file *flip); extern struct nfs_lock_context *nfs_get_lock_context(struct nfs_open_context *ctx); extern void nfs_put_lock_context(struct nfs_lock_context *l_ctx); extern u64 nfs_compat_user_ino64(u64 fileid); -- cgit v1.2.3 From 7c2dad99d60c86ec686b3bfdcb787c450a7ea89f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 6 Aug 2015 12:06:30 -0400 Subject: NFS: Don't let the ctime override attribute barriers. Chuck reports seeing cases where a GETATTR that happens to race with an asynchronous WRITE is overriding the file size, despite the attribute barrier being set by the writeback code. The culprit turns out to be the check in nfs_ctime_need_update(), which sees that the ctime is newer than the cached ctime, and assumes that it is safe to override the attribute barrier. This patch removes that override, and ensures that attribute barriers are always respected. Reported-by: Chuck Lever Fixes: a08a8cd375db9 ("NFS: Add attribute update barriers to NFS writebacks") Cc: stable@vger.kernel.org # v4.0+ Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 8 -------- 1 file changed, 8 deletions(-) diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 382c8a46c078..2744d48bbbfe 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -1267,13 +1267,6 @@ static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fat return 0; } -static int nfs_ctime_need_update(const struct inode *inode, const struct nfs_fattr *fattr) -{ - if (!(fattr->valid & NFS_ATTR_FATTR_CTIME)) - return 0; - return timespec_compare(&fattr->ctime, &inode->i_ctime) > 0; -} - static atomic_long_t nfs_attr_generation_counter; static unsigned long nfs_read_attr_generation_counter(void) @@ -1422,7 +1415,6 @@ static int nfs_inode_attrs_need_update(const struct inode *inode, const struct n const struct nfs_inode *nfsi = NFS_I(inode); return ((long)fattr->gencount - (long)nfsi->attr_gencount) > 0 || - nfs_ctime_need_update(inode, fattr) || ((long)nfsi->attr_gencount - (long)nfs_read_attr_generation_counter() > 0); } -- cgit v1.2.3 From 9fba8e30f704d49627ab8a4a57213862206b3e5f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 17 Aug 2015 14:46:51 -0500 Subject: SUNRPC: Drop double-underscores from __rpc_cmp_addr6() Reported-by: kbuild test robot Fixes: 7b0ce60c0b20 ("SUNRPC: Drop double-underscores from rpc_cmp_addr{4|6}()") Signed-off-by: Trond Myklebust --- include/linux/sunrpc/addr.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/include/linux/sunrpc/addr.h b/include/linux/sunrpc/addr.h index 9bc3538d09a9..5c9c6cd08d3b 100644 --- a/include/linux/sunrpc/addr.h +++ b/include/linux/sunrpc/addr.h @@ -93,7 +93,7 @@ static inline bool __rpc_copy_addr6(struct sockaddr *dst, return true; } #else /* !(IS_ENABLED(CONFIG_IPV6) */ -static inline bool __rpc_cmp_addr6(const struct sockaddr *sap1, +static inline bool rpc_cmp_addr6(const struct sockaddr *sap1, const struct sockaddr *sap2) { return false; -- cgit v1.2.3 From 7e94d6c4ab69562423fdc18801050a84462c2787 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 17 Aug 2015 16:55:18 -0500 Subject: NFS: Don't fsync twice for O_SYNC/IS_SYNC files generic_file_write_iter() will already do an fsync on our behalf if the file descriptor is O_SYNC or the file is marked as IS_SYNC. Signed-off-by: Trond Myklebust --- fs/nfs/file.c | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-) diff --git a/fs/nfs/file.c b/fs/nfs/file.c index 7538a8582384..526a2681d975 100644 --- a/fs/nfs/file.c +++ b/fs/nfs/file.c @@ -645,12 +645,10 @@ static const struct vm_operations_struct nfs_file_vm_ops = { .page_mkwrite = nfs_vm_page_mkwrite, }; -static int nfs_need_sync_write(struct file *filp, struct inode *inode) +static int nfs_need_check_write(struct file *filp, struct inode *inode) { struct nfs_open_context *ctx; - if (IS_SYNC(inode) || (filp->f_flags & O_DSYNC)) - return 1; ctx = nfs_file_open_context(filp); if (test_bit(NFS_CONTEXT_ERROR_WRITE, &ctx->flags) || nfs_ctx_key_to_expire(ctx)) @@ -700,8 +698,8 @@ ssize_t nfs_file_write(struct kiocb *iocb, struct iov_iter *from) if (result > 0) written = result; - /* Return error values for O_DSYNC and IS_SYNC() */ - if (result >= 0 && nfs_need_sync_write(file, inode)) { + /* Return error values */ + if (result >= 0 && nfs_need_check_write(file, inode)) { int err = vfs_fsync(file, 0); if (err < 0) result = err; -- cgit v1.2.3 From 4ff376feaf57af94e08c8df769e7c48b805ac897 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 18 Aug 2015 23:23:21 -0500 Subject: NFSv4.1/pnfs: Fix a close/delegreturn hang when return-on-close is set The helper pnfs_roc() has already verified that we have no delegations, and no further open files, hence no outstanding I/O and it has marked all the return-on-close lsegs as being invalid. Furthermore, it sets the NFS_LAYOUT_RETURN bit, thus serialising the close/delegreturn with all future layoutget calls on this inode. The checks in pnfs_roc_drain() for valid layout segments are therefore redundant: those cannot exist until another layoutget completes. The other check for whether or not NFS_LAYOUT_RETURN is set, actually causes a hang, since we already know that we hold that flag. To fix, we therefore strip out all the functionality in pnfs_roc_drain() except the retrieval of the barrier state, and then rename the function accordingly. Reported-by: Christoph Hellwig Fixes: 5c4a79fb2b1c ("Don't prevent layoutgets when doing return-on-close") Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 12 ++++-------- fs/nfs/pnfs.c | 24 +----------------------- fs/nfs/pnfs.h | 7 +++---- 3 files changed, 8 insertions(+), 35 deletions(-) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index f4e5816a77b0..bda7837dfe6b 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -2737,11 +2737,8 @@ static void nfs4_close_prepare(struct rpc_task *task, void *data) if (calldata->arg.fmode == 0) { task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE]; - if (calldata->roc && - pnfs_roc_drain(inode, &calldata->roc_barrier, task)) { - nfs_release_seqid(calldata->arg.seqid); - goto out_wait; - } + if (calldata->roc) + pnfs_roc_get_barrier(inode, &calldata->roc_barrier); } calldata->arg.share_access = nfs4_map_atomic_open_share(NFS_SERVER(inode), @@ -5289,9 +5286,8 @@ static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data) d_data = (struct nfs4_delegreturndata *)data; - if (d_data->roc && - pnfs_roc_drain(d_data->inode, &d_data->roc_barrier, task)) - return; + if (d_data->roc) + pnfs_roc_get_barrier(d_data->inode, &d_data->roc_barrier); nfs4_setup_sequence(d_data->res.server, &d_data->args.seq_args, diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 6151f39c8291..6aabbb654021 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1141,25 +1141,14 @@ void pnfs_roc_set_barrier(struct inode *ino, u32 barrier) spin_unlock(&ino->i_lock); } -bool pnfs_roc_drain(struct inode *ino, u32 *barrier, struct rpc_task *task) +void pnfs_roc_get_barrier(struct inode *ino, u32 *barrier) { struct nfs_inode *nfsi = NFS_I(ino); struct pnfs_layout_hdr *lo; - struct pnfs_layout_segment *lseg; nfs4_stateid stateid; u32 current_seqid; - bool layoutreturn = false; spin_lock(&ino->i_lock); - list_for_each_entry(lseg, &nfsi->layout->plh_segs, pls_list) { - if (!test_bit(NFS_LSEG_ROC, &lseg->pls_flags)) - continue; - if (test_bit(NFS_LSEG_VALID, &lseg->pls_flags)) - continue; - rpc_sleep_on(&NFS_SERVER(ino)->roc_rpcwaitq, task, NULL); - spin_unlock(&ino->i_lock); - return true; - } lo = nfsi->layout; current_seqid = be32_to_cpu(lo->plh_stateid.seqid); @@ -1168,18 +1157,7 @@ bool pnfs_roc_drain(struct inode *ino, u32 *barrier, struct rpc_task *task) */ *barrier = current_seqid + atomic_read(&lo->plh_outstanding); stateid = lo->plh_stateid; - if (test_and_clear_bit(NFS_LAYOUT_RETURN_BEFORE_CLOSE, - &lo->plh_flags)) - layoutreturn = pnfs_prepare_layoutreturn(lo); - if (test_bit(NFS_LAYOUT_RETURN, &lo->plh_flags)) - rpc_sleep_on(&NFS_SERVER(ino)->roc_rpcwaitq, task, NULL); - spin_unlock(&ino->i_lock); - if (layoutreturn) { - pnfs_send_layoutreturn(lo, stateid, IOMODE_ANY, false); - return true; - } - return false; } /* diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h index 738672a0f8da..a3d57a8fac76 100644 --- a/fs/nfs/pnfs.h +++ b/fs/nfs/pnfs.h @@ -267,7 +267,7 @@ int pnfs_mark_matching_lsegs_invalid(struct pnfs_layout_hdr *lo, bool pnfs_roc(struct inode *ino); void pnfs_roc_release(struct inode *ino); void pnfs_roc_set_barrier(struct inode *ino, u32 barrier); -bool pnfs_roc_drain(struct inode *ino, u32 *barrier, struct rpc_task *task); +void pnfs_roc_get_barrier(struct inode *ino, u32 *barrier); void pnfs_set_layoutcommit(struct inode *, struct pnfs_layout_segment *, loff_t); void pnfs_cleanup_layoutcommit(struct nfs4_layoutcommit_data *data); int pnfs_layoutcommit_inode(struct inode *inode, bool sync); @@ -605,10 +605,9 @@ pnfs_roc_set_barrier(struct inode *ino, u32 barrier) { } -static inline bool -pnfs_roc_drain(struct inode *ino, u32 *barrier, struct rpc_task *task) +static inline void +pnfs_roc_get_barrier(struct inode *ino, u32 *barrier) { - return false; } static inline void set_pnfs_layoutdriver(struct nfs_server *s, -- cgit v1.2.3 From 3c13cb5b647ebe36fb79128bc8b917d2a3317b65 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 18 Aug 2015 23:45:13 -0500 Subject: NFSv4.1/pnfs: Play safe w.r.t. close() races when return-on-close is set If we have an OPEN_DOWNGRADE and CLOSE race with one another, we want to ensure that the layout is forgotten by the client, so that we start afresh with a new layoutget. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index bda7837dfe6b..9e9f7816cf24 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -2661,7 +2661,7 @@ static void nfs4_close_done(struct rpc_task *task, void *data) switch (task->tk_status) { case 0: res_stateid = &calldata->res.stateid; - if (calldata->arg.fmode == 0 && calldata->roc) + if (calldata->roc) pnfs_roc_set_barrier(state->inode, calldata->roc_barrier); renew_lease(server, calldata->timestamp); @@ -2735,11 +2735,11 @@ static void nfs4_close_prepare(struct rpc_task *task, void *data) goto out_no_action; } - if (calldata->arg.fmode == 0) { + if (calldata->arg.fmode == 0) task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE]; - if (calldata->roc) - pnfs_roc_get_barrier(inode, &calldata->roc_barrier); - } + if (calldata->roc) + pnfs_roc_get_barrier(inode, &calldata->roc_barrier); + calldata->arg.share_access = nfs4_map_atomic_open_share(NFS_SERVER(inode), calldata->arg.fmode, 0); -- cgit v1.2.3 From 36319608e28701c07cad80ae3be8b0fdfb1ab40f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 19 Aug 2015 00:14:20 -0500 Subject: Revert "NFSv4: Remove incorrect check in can_open_delegated()" This reverts commit 4e379d36c050b0117b5d10048be63a44f5036115. This commit opens up a race between the recovery code and the open code. Reported-by: Olga Kornievskaia Cc: stable@vger.kernel # v4.0+ Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 9e9f7816cf24..95c5e8d39bef 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -1156,6 +1156,8 @@ static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode) return 0; if ((delegation->type & fmode) != fmode) return 0; + if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags)) + return 0; if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags)) return 0; nfs_mark_delegation_referenced(delegation); -- cgit v1.2.3 From e755d638e91be254d441602e8d7d9f1d9c944556 Mon Sep 17 00:00:00 2001 From: Peng Tao Date: Wed, 19 Aug 2015 13:49:19 +0800 Subject: NFS41: make sure sending LAYOUTRETURN before close if marked so If layout is marked by NFS_LAYOUT_RETURN_BEFORE_CLOSE, we should always send LAYOUTRETURN before close, and we don't need to do ROC drain if we do send LAYOUTRETURN. Signed-off-by: Peng Tao Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 51 ++++++++++++++++++++++++++++----------------------- 1 file changed, 28 insertions(+), 23 deletions(-) diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 6aabbb654021..e101a491e4e7 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1064,7 +1064,7 @@ bool pnfs_roc(struct inode *ino) struct pnfs_layout_segment *lseg, *tmp; nfs4_stateid stateid; LIST_HEAD(tmp_list); - bool found = false, layoutreturn = false; + bool found = false, layoutreturn = false, roc = false; spin_lock(&ino->i_lock); lo = nfsi->layout; @@ -1072,7 +1072,7 @@ bool pnfs_roc(struct inode *ino) test_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags)) goto out_noroc; - /* Don't return layout if we hold a delegation */ + /* no roc if we hold a delegation */ if (nfs4_check_delegation(ino, FMODE_READ)) goto out_noroc; @@ -1083,36 +1083,41 @@ bool pnfs_roc(struct inode *ino) goto out_noroc; } + stateid = lo->plh_stateid; + /* always send layoutreturn if being marked so */ + if (test_and_clear_bit(NFS_LAYOUT_RETURN_BEFORE_CLOSE, + &lo->plh_flags)) + layoutreturn = pnfs_prepare_layoutreturn(lo); + pnfs_clear_retry_layoutget(lo); list_for_each_entry_safe(lseg, tmp, &lo->plh_segs, pls_list) - if (test_bit(NFS_LSEG_ROC, &lseg->pls_flags)) { + /* If we are sending layoutreturn, invalidate all valid lsegs */ + if (layoutreturn || test_bit(NFS_LSEG_ROC, &lseg->pls_flags)) { mark_lseg_invalid(lseg, &tmp_list); found = true; } - if (!found) - goto out_noroc; - if (test_and_set_bit(NFS_LAYOUT_RETURN, &lo->plh_flags)) - goto out_noroc; - lo->plh_return_iomode = IOMODE_ANY; - pnfs_get_layout_hdr(lo); /* matched in pnfs_roc_release */ - spin_unlock(&ino->i_lock); - pnfs_free_lseg_list(&tmp_list); - pnfs_layoutcommit_inode(ino, true); - return true; + /* pnfs_prepare_layoutreturn() grabs lo ref and it will be put + * in pnfs_roc_release(). We don't really send a layoutreturn but + * still want others to view us like we are sending one! + * + * If pnfs_prepare_layoutreturn() fails, it means someone else is doing + * LAYOUTRETURN, so we proceed like there are no layouts to return. + * + * ROC in three conditions: + * 1. there are ROC lsegs + * 2. we don't send layoutreturn + * 3. no others are sending layoutreturn + */ + if (found && !layoutreturn && pnfs_prepare_layoutreturn(lo)) + roc = true; out_noroc: - if (lo) { - stateid = lo->plh_stateid; - if (test_and_clear_bit(NFS_LAYOUT_RETURN_BEFORE_CLOSE, - &lo->plh_flags)) - layoutreturn = pnfs_prepare_layoutreturn(lo); - } spin_unlock(&ino->i_lock); - if (layoutreturn) { - pnfs_layoutcommit_inode(ino, true); + pnfs_free_lseg_list(&tmp_list); + pnfs_layoutcommit_inode(ino, true); + if (layoutreturn) pnfs_send_layoutreturn(lo, stateid, IOMODE_ANY, true); - } - return false; + return roc; } void pnfs_roc_release(struct inode *ino) -- cgit v1.2.3 From 69f230d907e8c1ca3f9bd528993eeb98f712b0dd Mon Sep 17 00:00:00 2001 From: Peng Tao Date: Thu, 20 Aug 2015 01:52:59 +0800 Subject: NFS41/flexfiles: update inode after write finishes Otherwise we break fstest case tests/read_write/mctime.t Does files layout need the same fix as well? Cc: stable@vger.kernel.org # v4.0+ Cc: Anna Schumaker Signed-off-by: Peng Tao Signed-off-by: Trond Myklebust --- fs/nfs/flexfilelayout/flexfilelayout.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index 13fe64b4e259..4bd3cff94190 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -1199,6 +1199,9 @@ static int ff_layout_write_done_cb(struct rpc_task *task, hdr->res.verf->committed == NFS_DATA_SYNC) ff_layout_set_layoutcommit(hdr); + if (task->tk_status >= 0) + nfs_writeback_update_inode(hdr); + return 0; } -- cgit v1.2.3 From c2126157ea3c4f72b315749e0c07a1b162a2fe2b Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 19 Aug 2015 21:46:15 -0500 Subject: SUNRPC: Allow sockets to do GFP_NOIO allocations Follow up to commit c4a7ca774949 ("SUNRPC: Allow waiting on memory allocation"). Allows the RPC socket code to do non-IO blocking. Signed-off-by: Trond Myklebust --- net/sunrpc/xprtsock.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c index 86ed778bb18b..723df7cb096d 100644 --- a/net/sunrpc/xprtsock.c +++ b/net/sunrpc/xprtsock.c @@ -1866,7 +1866,7 @@ static int xs_local_finish_connecting(struct rpc_xprt *xprt, sk->sk_data_ready = xs_local_data_ready; sk->sk_write_space = xs_udp_write_space; sk->sk_error_report = xs_error_report; - sk->sk_allocation = GFP_ATOMIC; + sk->sk_allocation = GFP_NOIO; xprt_clear_connected(xprt); @@ -2051,7 +2051,7 @@ static void xs_udp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock) sk->sk_user_data = xprt; sk->sk_data_ready = xs_udp_data_ready; sk->sk_write_space = xs_udp_write_space; - sk->sk_allocation = GFP_ATOMIC; + sk->sk_allocation = GFP_NOIO; xprt_set_connected(xprt); @@ -2153,7 +2153,7 @@ static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock) sk->sk_state_change = xs_tcp_state_change; sk->sk_write_space = xs_tcp_write_space; sk->sk_error_report = xs_error_report; - sk->sk_allocation = GFP_ATOMIC; + sk->sk_allocation = GFP_NOIO; /* socket options */ sock_reset_flag(sk, SOCK_LINGER); -- cgit v1.2.3 From c740624989eb87fa7cbd1b5338cef01dd49f1f29 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 19 Aug 2015 23:00:50 -0500 Subject: pNFS: Fix an unused variable warning in pnfs_roc_get_barrier Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 2 -- 1 file changed, 2 deletions(-) diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index e101a491e4e7..8a3f30b695e2 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1150,7 +1150,6 @@ void pnfs_roc_get_barrier(struct inode *ino, u32 *barrier) { struct nfs_inode *nfsi = NFS_I(ino); struct pnfs_layout_hdr *lo; - nfs4_stateid stateid; u32 current_seqid; spin_lock(&ino->i_lock); @@ -1161,7 +1160,6 @@ void pnfs_roc_get_barrier(struct inode *ino, u32 *barrier) * a barrier, we choose the worst-case barrier. */ *barrier = current_seqid + atomic_read(&lo->plh_outstanding); - stateid = lo->plh_stateid; spin_unlock(&ino->i_lock); } -- cgit v1.2.3 From 2a606188c55990fa65cba3fd9b64f2b7542b7692 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 19 Aug 2015 22:30:00 -0500 Subject: NFSv4: Enable delegated opens even when reboot recovery is pending Unlike the previous attempt, this takes into account the fact that we may be calling it from the recovery thread itself. Detect this by looking at what kind of open we're doing, and checking the state of the NFS_DELEGATION_NEED_RECLAIM if it turns out we're doing a reboot reclaim-type open. Cc: Olga Kornievskaia Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 27 +++++++++++++++++++-------- include/linux/nfs_xdr.h | 2 +- 2 files changed, 20 insertions(+), 9 deletions(-) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 95c5e8d39bef..6e988fd92f69 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -1150,16 +1150,25 @@ out: return ret; } -static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode) +static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode, + enum open_claim_type4 claim) { if (delegation == NULL) return 0; if ((delegation->type & fmode) != fmode) return 0; - if (test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags)) - return 0; if (test_bit(NFS_DELEGATION_RETURNING, &delegation->flags)) return 0; + switch (claim) { + case NFS4_OPEN_CLAIM_NULL: + case NFS4_OPEN_CLAIM_FH: + break; + case NFS4_OPEN_CLAIM_PREVIOUS: + if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags)) + break; + default: + return 0; + } nfs_mark_delegation_referenced(delegation); return 1; } @@ -1378,6 +1387,7 @@ static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata) struct nfs_delegation *delegation; int open_mode = opendata->o_arg.open_flags; fmode_t fmode = opendata->o_arg.fmode; + enum open_claim_type4 claim = opendata->o_arg.claim; nfs4_stateid stateid; int ret = -EAGAIN; @@ -1391,7 +1401,7 @@ static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata) spin_unlock(&state->owner->so_lock); rcu_read_lock(); delegation = rcu_dereference(nfsi->delegation); - if (!can_open_delegated(delegation, fmode)) { + if (!can_open_delegated(delegation, fmode, claim)) { rcu_read_unlock(); break; } @@ -1854,6 +1864,7 @@ static void nfs4_open_prepare(struct rpc_task *task, void *calldata) struct nfs4_opendata *data = calldata; struct nfs4_state_owner *sp = data->owner; struct nfs_client *clp = sp->so_server->nfs_client; + enum open_claim_type4 claim = data->o_arg.claim; if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0) goto out_wait; @@ -1868,15 +1879,15 @@ static void nfs4_open_prepare(struct rpc_task *task, void *calldata) goto out_no_action; rcu_read_lock(); delegation = rcu_dereference(NFS_I(data->state->inode)->delegation); - if (data->o_arg.claim != NFS4_OPEN_CLAIM_DELEGATE_CUR && - data->o_arg.claim != NFS4_OPEN_CLAIM_DELEG_CUR_FH && - can_open_delegated(delegation, data->o_arg.fmode)) + if (can_open_delegated(delegation, data->o_arg.fmode, claim)) goto unlock_no_action; rcu_read_unlock(); } /* Update client id. */ data->o_arg.clientid = clp->cl_clientid; - switch (data->o_arg.claim) { + switch (claim) { + default: + break; case NFS4_OPEN_CLAIM_PREVIOUS: case NFS4_OPEN_CLAIM_DELEG_CUR_FH: case NFS4_OPEN_CLAIM_DELEG_PREV_FH: diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index 7bbe50504211..b9b530409ff7 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -389,7 +389,7 @@ struct nfs_openargs { const struct nfs_server *server; /* Needed for ID mapping */ const u32 * bitmask; const u32 * open_bitmap; - __u32 claim; + enum open_claim_type4 claim; enum createmode4 createmode; const struct nfs4_label *label; }; -- cgit v1.2.3 From 046be74da8f257c4f1925ed4b5d4ee4c822ef9c6 Mon Sep 17 00:00:00 2001 From: Peng Tao Date: Fri, 21 Aug 2015 10:32:50 +0800 Subject: NFS41: fix list splice type We want to move commiting pages to pages list instead. Otherwise it causes pnfs small writes crash like: [34560.037692] BUG: unable to handle kernel NULL pointer dereference at 0000000000000068 [34560.038557] IP: [] nfs_init_commit+0x26/0x130 [nfs] [34560.039400] PGD 69f5a067 PUD 69f59067 PMD 0 [34560.040207] Oops: 0000 [#1] SMP [34560.041014] Modules linked in: nfsv3(OE) nfs_layout_flexfiles(OE) nfsv4(OE) nfs(OE) fscache(E) rpcsec_gss_krb5(E) xt_addrtype(E) xt_conntrack(E) ipt_MASQUERADE(E) nf_nat_masquerade_ipv4(E) iptable_nat(E) nf_conntrack_ipv4(E) nf_defrag_ipv4(E) nf_nat_ipv4(E) iptable_filter(E) ip_tables(E) x_tables(E) nf_nat(E) nf_conntrack(E) bridge(E) stp(E) llc(E) dm_thin_pool(E) dm_persistent_data(E) dm_bio_prison(E) dm_bufio(E) ppdev(E) vmw_balloon(E) coretemp(E) crc32_pclmul(E) ghash_clmulni_intel(E) aesni_intel(E) aes_x86_64(E) glue_helper(E) lrw(E) gf128mul(E) ablk_helper(E) cryptd(E) psmouse(E) serio_raw(E) vmw_vmci(E) i2c_piix4(E) shpchp(E) parport_pc(E) parport(E) mac_hid(E) nfsd(E) auth_rpcgss(E) nfs_acl(E) lockd(E) grace(E) sunrpc(E) xfs(E) libcrc32c(E) hid_generic(E) usbhid(E) hid(E) e1000(E) mptspi(E) [34560.045106] mptscsih(E) mptbase(E) vmwgfx(E) drm_kms_helper(E) ttm(E) drm(E) autofs4(E) [last unloaded: fscache] [34560.045897] CPU: 0 PID: 130543 Comm: bash Tainted: G OE 4.2.0-rc5-dp-00057-gf993a93 #11 [34560.046699] Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 05/20/2014 [34560.047525] task: ffff880031b0a980 ti: ffff880045fec000 task.ti: ffff880045fec000 [34560.048264] RIP: 0010:[] [] nfs_init_commit+0x26/0x130 [nfs] [34560.049000] RSP: 0018:ffff880045fefc18 EFLAGS: 00010246 [34560.049717] RAX: 0000000000000000 RBX: ffff8800208fbc80 RCX: ffff880045fefd50 [34560.050396] RDX: ffff880031c19ec0 RSI: ffff880045fefc88 RDI: ffff8800208fbc80 [34560.051041] RBP: ffff880045fefc28 R08: ffff8800208fbe68 R09: ffff880045fefc88 [34560.051666] R10: 0000000000000000 R11: 0000000000000000 R12: ffff880045fefc78 [34560.052247] R13: ffff880045fefc88 R14: ffff880045fefa90 R15: ffff880045fefd50 [34560.052825] FS: 00007fa02d58c740(0000) GS:ffff88006d600000(0000) knlGS:0000000000000000 [34560.053410] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [34560.053992] CR2: 0000000000000068 CR3: 000000003b37a000 CR4: 00000000001406f0 [34560.054615] Stack: [34560.055200] ffff8800208fbc80 ffff8800208fbc80 ffff880045fefcc8 ffffffffa05c1a5b [34560.055800] ffff880045fefcc8 ffff880045fefd50 0000000045fefcb8 ffff880045fefd40 [34560.056418] ffff8800420608e0 ffffffffa04f3910 0000000100000001 ffff880045fefd50 [34560.057013] Call Trace: [34560.057672] [] pnfs_generic_commit_pagelist+0x1cb/0x300 [nfsv4] [34560.058277] [] ? ff_layout_commit_pagelist+0x20/0x20 [nfs_layout_flexfiles] [34560.058907] [] ff_layout_commit_pagelist+0x15/0x20 [nfs_layout_flexfiles] [34560.059557] [] nfs_generic_commit_list+0xb1/0xf0 [nfs] [34560.060214] [] ? nfs_scan_commit+0x37/0xa0 [nfs] [34560.060825] [] nfs_commit_inode+0x81/0x150 [nfs] [34560.061432] [] nfs_wb_all+0x1ae/0x400 [nfs] [34560.062035] [] nfs_getattr+0x33d/0x510 [nfs] [34560.062630] [] vfs_getattr_nosec+0x2c/0x40 [34560.063223] [] vfs_getattr+0x26/0x30 [34560.063818] [] vfs_fstatat+0x65/0xa0 [34560.064413] [] SYSC_newstat+0x1f/0x40 [34560.065016] [] ? do_audit_syscall_entry+0x66/0x70 [34560.065626] [] ? syscall_trace_enter_phase1+0x113/0x170 [34560.066245] [] ? trace_hardirqs_on_thunk+0x17/0x19 [34560.066868] [] SyS_newstat+0xe/0x10 [34560.067533] [] entry_SYSCALL_64_fastpath+0x16/0x7a [34560.068173] Code: 0f 1f 44 00 00 0f 1f 44 00 00 55 4c 8d 87 e8 01 00 00 48 89 e5 53 48 89 fb 48 83 ec 08 4c 8b 0e 49 8b 41 18 4c 39 ce 48 8b 40 40 <4c> 8b 50 68 74 24 48 8b 87 e8 01 00 00 48 8b 7e 08 4d 89 41 08 [34560.069609] RIP [] nfs_init_commit+0x26/0x130 [nfs] [34560.070295] RSP [34560.071008] CR2: 0000000000000068 [34560.073207] ---[ end trace f85f873260977406 ]--- [fixes 27571297a7e(pNFS: Tighten up locking around DS commit buckets)] Signed-off-by: Peng Tao Signed-off-by: Trond Myklebust --- fs/nfs/pnfs_nfs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/nfs/pnfs_nfs.c b/fs/nfs/pnfs_nfs.c index bbd407b6bc1f..24655b807d44 100644 --- a/fs/nfs/pnfs_nfs.c +++ b/fs/nfs/pnfs_nfs.c @@ -239,7 +239,7 @@ void pnfs_fetch_commit_bucket_list(struct list_head *pages, bucket = &cinfo->ds->buckets[data->ds_commit_index]; spin_lock(cinfo->lock); - list_splice_init(pages, &bucket->committing); + list_splice_init(&bucket->committing, pages); data->lseg = bucket->clseg; bucket->clseg = NULL; spin_unlock(cinfo->lock); -- cgit v1.2.3 From e76d28dd9ca84fa96dc0054bab34ccec34a7662a Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 20 Aug 2015 13:12:51 -0500 Subject: NFSv4.1/pnfs: Ensure the flexfiles layoutstats timers are consistent We want to ensure that the stopwatches for the busy timer and the aggregate timer are consistent. This means that they need to use the same start/stop times. Signed-off-by: Trond Myklebust --- fs/nfs/flexfilelayout/flexfilelayout.c | 51 ++++++++++++++++------------------ 1 file changed, 24 insertions(+), 27 deletions(-) diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index 4bd3cff94190..0fcb8670f25f 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -419,42 +419,35 @@ ff_layout_get_lseg_count(struct nfs4_ff_layout_segment *fls) } static void -nfs4_ff_start_busy_timer(struct nfs4_ff_busy_timer *timer) +nfs4_ff_start_busy_timer(struct nfs4_ff_busy_timer *timer, ktime_t now) { /* first IO request? */ if (atomic_inc_return(&timer->n_ops) == 1) { - timer->start_time = ktime_get(); + timer->start_time = now; } } static ktime_t -nfs4_ff_end_busy_timer(struct nfs4_ff_busy_timer *timer) +nfs4_ff_end_busy_timer(struct nfs4_ff_busy_timer *timer, ktime_t now) { - ktime_t start, now; + ktime_t start; if (atomic_dec_return(&timer->n_ops) < 0) WARN_ON_ONCE(1); - now = ktime_get(); start = timer->start_time; timer->start_time = now; return ktime_sub(now, start); } -static ktime_t -nfs4_ff_layout_calc_completion_time(struct rpc_task *task) -{ - return ktime_sub(ktime_get(), task->tk_start); -} - static bool nfs4_ff_layoutstat_start_io(struct nfs4_ff_layout_mirror *mirror, - struct nfs4_ff_layoutstat *layoutstat) + struct nfs4_ff_layoutstat *layoutstat, + ktime_t now) { static const ktime_t notime = {0}; - ktime_t now = ktime_get(); - nfs4_ff_start_busy_timer(&layoutstat->busy_timer); + nfs4_ff_start_busy_timer(&layoutstat->busy_timer, now); if (ktime_equal(mirror->start_time, notime)) mirror->start_time = now; if (ktime_equal(mirror->last_report_time, notime)) @@ -482,30 +475,33 @@ static void nfs4_ff_layout_stat_io_update_completed(struct nfs4_ff_layoutstat *layoutstat, __u64 requested, __u64 completed, - ktime_t time_completed) + ktime_t time_completed, + ktime_t time_started) { struct nfs4_ff_io_stat *iostat = &layoutstat->io_stat; + ktime_t completion_time = ktime_sub(time_completed, time_started); ktime_t timer; iostat->ops_completed++; iostat->bytes_completed += completed; iostat->bytes_not_delivered += requested - completed; - timer = nfs4_ff_end_busy_timer(&layoutstat->busy_timer); + timer = nfs4_ff_end_busy_timer(&layoutstat->busy_timer, time_completed); iostat->total_busy_time = ktime_add(iostat->total_busy_time, timer); iostat->aggregate_completion_time = - ktime_add(iostat->aggregate_completion_time, time_completed); + ktime_add(iostat->aggregate_completion_time, + completion_time); } static void nfs4_ff_layout_stat_io_start_read(struct nfs4_ff_layout_mirror *mirror, - __u64 requested) + __u64 requested, ktime_t now) { bool report; spin_lock(&mirror->lock); - report = nfs4_ff_layoutstat_start_io(mirror, &mirror->read_stat); + report = nfs4_ff_layoutstat_start_io(mirror, &mirror->read_stat, now); nfs4_ff_layout_stat_io_update_requested(&mirror->read_stat, requested); spin_unlock(&mirror->lock); @@ -523,18 +519,18 @@ nfs4_ff_layout_stat_io_end_read(struct rpc_task *task, spin_lock(&mirror->lock); nfs4_ff_layout_stat_io_update_completed(&mirror->read_stat, requested, completed, - nfs4_ff_layout_calc_completion_time(task)); + ktime_get(), task->tk_start); spin_unlock(&mirror->lock); } static void nfs4_ff_layout_stat_io_start_write(struct nfs4_ff_layout_mirror *mirror, - __u64 requested) + __u64 requested, ktime_t now) { bool report; spin_lock(&mirror->lock); - report = nfs4_ff_layoutstat_start_io(mirror , &mirror->write_stat); + report = nfs4_ff_layoutstat_start_io(mirror , &mirror->write_stat, now); nfs4_ff_layout_stat_io_update_requested(&mirror->write_stat, requested); spin_unlock(&mirror->lock); @@ -555,8 +551,7 @@ nfs4_ff_layout_stat_io_end_write(struct rpc_task *task, spin_lock(&mirror->lock); nfs4_ff_layout_stat_io_update_completed(&mirror->write_stat, - requested, completed, - nfs4_ff_layout_calc_completion_time(task)); + requested, completed, ktime_get(), task->tk_start); spin_unlock(&mirror->lock); } @@ -1063,7 +1058,8 @@ static int ff_layout_read_prepare_common(struct rpc_task *task, { nfs4_ff_layout_stat_io_start_read( FF_LAYOUT_COMP(hdr->lseg, hdr->pgio_mirror_idx), - hdr->args.count); + hdr->args.count, + task->tk_start); if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags))) { rpc_exit(task, -EIO); @@ -1249,7 +1245,8 @@ static int ff_layout_write_prepare_common(struct rpc_task *task, { nfs4_ff_layout_stat_io_start_write( FF_LAYOUT_COMP(hdr->lseg, hdr->pgio_mirror_idx), - hdr->args.count); + hdr->args.count, + task->tk_start); if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags))) { rpc_exit(task, -EIO); @@ -1330,7 +1327,7 @@ static void ff_layout_commit_prepare_common(struct rpc_task *task, { nfs4_ff_layout_stat_io_start_write( FF_LAYOUT_COMP(cdata->lseg, cdata->ds_commit_index), - 0); + 0, task->tk_start); } static void ff_layout_commit_prepare_v3(struct rpc_task *task, void *data) -- cgit v1.2.3 From dd52128afddecade1dcd0767aec124c13c8a40bb Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 20 Aug 2015 17:59:49 -0500 Subject: NFSv4.1/pnfs Ensure flexfiles reports all connection related errors Make sure that we also handle RPC level connection and protocol negotiation errors. Reported-by: Tom Haynes Signed-off-by: Trond Myklebust --- fs/nfs/flexfilelayout/flexfilelayout.c | 48 +++++++++++++++++++++++++--------- 1 file changed, 35 insertions(+), 13 deletions(-) diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index 0fcb8670f25f..7e4cbdfb3713 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -967,11 +967,36 @@ static int ff_layout_async_handle_error(struct rpc_task *task, static void ff_layout_io_track_ds_error(struct pnfs_layout_segment *lseg, int idx, u64 offset, u64 length, - u32 status, int opnum) + u32 status, int opnum, int error) { struct nfs4_ff_layout_mirror *mirror; int err; + if (status == 0) { + switch (error) { + case -ETIMEDOUT: + case -EPFNOSUPPORT: + case -EPROTONOSUPPORT: + case -EOPNOTSUPP: + case -ECONNREFUSED: + case -ECONNRESET: + case -EHOSTDOWN: + case -EHOSTUNREACH: + case -ENETUNREACH: + case -EADDRINUSE: + case -ENOBUFS: + case -EPIPE: + case -EPERM: + status = NFS4ERR_NXIO; + break; + case -EACCES: + status = NFS4ERR_ACCESS; + break; + default: + return; + } + } + mirror = FF_LAYOUT_COMP(lseg, idx); err = ff_layout_track_ds_error(FF_LAYOUT_FROM_HDR(lseg->pls_layout), mirror, offset, length, status, opnum, @@ -988,12 +1013,11 @@ static int ff_layout_read_done_cb(struct rpc_task *task, int err; trace_nfs4_pnfs_read(hdr, task->tk_status); - if (task->tk_status == -ETIMEDOUT && !hdr->res.op_status) - hdr->res.op_status = NFS4ERR_NXIO; - if (task->tk_status < 0 && hdr->res.op_status) + if (task->tk_status < 0) ff_layout_io_track_ds_error(hdr->lseg, hdr->pgio_mirror_idx, hdr->args.offset, hdr->args.count, - hdr->res.op_status, OP_READ); + hdr->res.op_status, OP_READ, + task->tk_status); err = ff_layout_async_handle_error(task, hdr->args.context->state, hdr->ds_clp, hdr->lseg, hdr->pgio_mirror_idx); @@ -1163,12 +1187,11 @@ static int ff_layout_write_done_cb(struct rpc_task *task, int err; trace_nfs4_pnfs_write(hdr, task->tk_status); - if (task->tk_status == -ETIMEDOUT && !hdr->res.op_status) - hdr->res.op_status = NFS4ERR_NXIO; - if (task->tk_status < 0 && hdr->res.op_status) + if (task->tk_status < 0) ff_layout_io_track_ds_error(hdr->lseg, hdr->pgio_mirror_idx, hdr->args.offset, hdr->args.count, - hdr->res.op_status, OP_WRITE); + hdr->res.op_status, OP_WRITE, + task->tk_status); err = ff_layout_async_handle_error(task, hdr->args.context->state, hdr->ds_clp, hdr->lseg, hdr->pgio_mirror_idx); @@ -1208,12 +1231,11 @@ static int ff_layout_commit_done_cb(struct rpc_task *task, int err; trace_nfs4_pnfs_commit_ds(data, task->tk_status); - if (task->tk_status == -ETIMEDOUT && !data->res.op_status) - data->res.op_status = NFS4ERR_NXIO; - if (task->tk_status < 0 && data->res.op_status) + if (task->tk_status < 0) ff_layout_io_track_ds_error(data->lseg, data->ds_commit_index, data->args.offset, data->args.count, - data->res.op_status, OP_COMMIT); + data->res.op_status, OP_COMMIT, + task->tk_status); err = ff_layout_async_handle_error(task, NULL, data->ds_clp, data->lseg, data->ds_commit_index); -- cgit v1.2.3 From aaae3f00d3f67f681a1f3cb7af999e976e8a24ce Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 20 Aug 2015 18:56:07 -0500 Subject: NFSv4: Force a post-op attribute update when holding a delegation If the ctime or mtime or change attribute have changed because of an operation we initiated, we should make sure that we force an attribute update. However we do not want to mark the page cache for revalidation. Signed-off-by: Trond Myklebust Cc: stable@vger.kernel.org # v4.0+ --- fs/nfs/inode.c | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 2744d48bbbfe..e2cc0031decb 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -1477,6 +1477,13 @@ static int nfs_post_op_update_inode_locked(struct inode *inode, struct nfs_fattr { unsigned long invalid = NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE; + /* + * Don't revalidate the pagecache if we hold a delegation, but do + * force an attribute update + */ + if (NFS_PROTO(inode)->have_delegation(inode, FMODE_READ)) + invalid = NFS_INO_INVALID_ATTR|NFS_INO_REVAL_FORCED; + if (S_ISDIR(inode->i_mode)) invalid |= NFS_INO_INVALID_DATA; nfs_set_cache_invalid(inode, invalid); -- cgit v1.2.3 From 6a463beb9a433d91f3eaf85c2f0ca0aeddf3e0ab Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 20 Aug 2015 15:40:47 -0500 Subject: NFSv4.1/pnfs: Add a tracepoint for return-on-close events Allow tracing of return-on-close. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4trace.h | 1 + fs/nfs/pnfs.c | 1 + 2 files changed, 2 insertions(+) diff --git a/fs/nfs/nfs4trace.h b/fs/nfs/nfs4trace.h index 470af1a78bec..0b67f7825d00 100644 --- a/fs/nfs/nfs4trace.h +++ b/fs/nfs/nfs4trace.h @@ -1136,6 +1136,7 @@ TRACE_EVENT(nfs4_layoutget, DEFINE_NFS4_INODE_EVENT(nfs4_layoutcommit); DEFINE_NFS4_INODE_EVENT(nfs4_layoutreturn); +DEFINE_NFS4_INODE_EVENT(nfs4_layoutreturn_on_close); #endif /* CONFIG_NFS_V4_1 */ diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 8a3f30b695e2..cf90eeda9d71 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1144,6 +1144,7 @@ void pnfs_roc_set_barrier(struct inode *ino, u32 barrier) if (pnfs_seqid_is_newer(barrier, lo->plh_barrier)) lo->plh_barrier = barrier; spin_unlock(&ino->i_lock); + trace_nfs4_layoutreturn_on_close(ino, 0); } void pnfs_roc_get_barrier(struct inode *ino, u32 *barrier) -- cgit v1.2.3 From 7cd148610af9312aa6454395fe174ebfe9496aa1 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 20 Aug 2015 20:07:54 -0500 Subject: NFSv4: Add a tracepoint for CB_GETATTR Signed-off-by: Trond Myklebust --- fs/nfs/callback_proc.c | 6 ++++- fs/nfs/nfs4trace.h | 59 ++++++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 64 insertions(+), 1 deletion(-) diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c index 624bef79ba7c..07258d269e5a 100644 --- a/fs/nfs/callback_proc.c +++ b/fs/nfs/callback_proc.c @@ -40,8 +40,11 @@ __be32 nfs4_callback_getattr(struct cb_getattrargs *args, rpc_peeraddr2str(cps->clp->cl_rpcclient, RPC_DISPLAY_ADDR)); inode = nfs_delegation_find_inode(cps->clp, &args->fh); - if (inode == NULL) + if (inode == NULL) { + trace_nfs4_cb_getattr(cps->clp, &args->fh, inode, + -ntohl(res->status)); goto out; + } nfsi = NFS_I(inode); rcu_read_lock(); delegation = rcu_dereference(nfsi->delegation); @@ -60,6 +63,7 @@ __be32 nfs4_callback_getattr(struct cb_getattrargs *args, res->status = 0; out_iput: rcu_read_unlock(); + trace_nfs4_cb_getattr(cps->clp, &args->fh, inode, -ntohl(res->status)); iput(inode); out: dprintk("%s: exit with status = %d\n", __func__, ntohl(res->status)); diff --git a/fs/nfs/nfs4trace.h b/fs/nfs/nfs4trace.h index 0b67f7825d00..1aa096b665dc 100644 --- a/fs/nfs/nfs4trace.h +++ b/fs/nfs/nfs4trace.h @@ -884,6 +884,65 @@ DEFINE_NFS4_GETATTR_EVENT(nfs4_getattr); DEFINE_NFS4_GETATTR_EVENT(nfs4_lookup_root); DEFINE_NFS4_GETATTR_EVENT(nfs4_fsinfo); +DECLARE_EVENT_CLASS(nfs4_inode_callback_event, + TP_PROTO( + const struct nfs_client *clp, + const struct nfs_fh *fhandle, + const struct inode *inode, + int error + ), + + TP_ARGS(clp, fhandle, inode, error), + + TP_STRUCT__entry( + __field(int, error) + __field(dev_t, dev) + __field(u32, fhandle) + __field(u64, fileid) + __string(dstaddr, clp ? + rpc_peeraddr2str(clp->cl_rpcclient, + RPC_DISPLAY_ADDR) : "unknown") + ), + + TP_fast_assign( + __entry->error = error; + __entry->fhandle = nfs_fhandle_hash(fhandle); + if (inode != NULL) { + __entry->fileid = NFS_FILEID(inode); + __entry->dev = inode->i_sb->s_dev; + } else { + __entry->fileid = 0; + __entry->dev = 0; + } + __assign_str(dstaddr, clp ? + rpc_peeraddr2str(clp->cl_rpcclient, + RPC_DISPLAY_ADDR) : "unknown") + ), + + TP_printk( + "error=%d (%s) fileid=%02x:%02x:%llu fhandle=0x%08x " + "dstaddr=%s", + __entry->error, + show_nfsv4_errors(__entry->error), + MAJOR(__entry->dev), MINOR(__entry->dev), + (unsigned long long)__entry->fileid, + __entry->fhandle, + __get_str(dstaddr) + ) +); + +#define DEFINE_NFS4_INODE_CALLBACK_EVENT(name) \ + DEFINE_EVENT(nfs4_inode_callback_event, name, \ + TP_PROTO( \ + const struct nfs_client *clp, \ + const struct nfs_fh *fhandle, \ + const struct inode *inode, \ + int error \ + ), \ + TP_ARGS(clp, fhandle, inode, error)) +DEFINE_NFS4_INODE_CALLBACK_EVENT(nfs4_cb_getattr); + + DECLARE_EVENT_CLASS(nfs4_idmap_event, TP_PROTO( const char *name, -- cgit v1.2.3 From 249b2eef647f97164b8bb61f2d9282f227a17992 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 20 Aug 2015 20:43:14 -0500 Subject: NFSv4: Add a tracepoint for CB_LAYOUTRECALL Only support for single file layoutrecall for now. Signed-off-by: Trond Myklebust --- fs/nfs/callback_proc.c | 3 ++- fs/nfs/nfs4trace.h | 1 + 2 files changed, 3 insertions(+), 1 deletion(-) diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c index 07258d269e5a..b85cf7a30232 100644 --- a/fs/nfs/callback_proc.c +++ b/fs/nfs/callback_proc.c @@ -41,7 +41,7 @@ __be32 nfs4_callback_getattr(struct cb_getattrargs *args, inode = nfs_delegation_find_inode(cps->clp, &args->fh); if (inode == NULL) { - trace_nfs4_cb_getattr(cps->clp, &args->fh, inode, + trace_nfs4_cb_getattr(cps->clp, &args->fh, NULL, -ntohl(res->status)); goto out; } @@ -198,6 +198,7 @@ unlock: spin_unlock(&ino->i_lock); pnfs_free_lseg_list(&free_me_list); pnfs_put_layout_hdr(lo); + trace_nfs4_cb_layoutrecall_inode(clp, &args->cbl_fh, ino, -rv); iput(ino); out: return rv; diff --git a/fs/nfs/nfs4trace.h b/fs/nfs/nfs4trace.h index 1aa096b665dc..28df12e525ba 100644 --- a/fs/nfs/nfs4trace.h +++ b/fs/nfs/nfs4trace.h @@ -941,6 +941,7 @@ DECLARE_EVENT_CLASS(nfs4_inode_callback_event, ), \ TP_ARGS(clp, fhandle, inode, error)) DEFINE_NFS4_INODE_CALLBACK_EVENT(nfs4_cb_getattr); +DEFINE_NFS4_INODE_CALLBACK_EVENT(nfs4_cb_layoutrecall_inode); DECLARE_EVENT_CLASS(nfs4_idmap_event, -- cgit v1.2.3 From 3976143b063e3c42fe5471d87860f6ae118e0eee Mon Sep 17 00:00:00 2001 From: Peng Tao Date: Fri, 21 Aug 2015 12:49:44 +0800 Subject: NFS41: remove NFS_LAYOUT_ROC flag If we return delegation before closing, we fail to do roc check during close because NFS_LAYOUT_ROC is cleared by delegreturn and it causes layouts to be still hanging around after delegreturn + close, which is a voilation against protocol. Signed-off-by: Peng Tao Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 7 ++----- fs/nfs/pnfs.h | 1 - 2 files changed, 2 insertions(+), 6 deletions(-) diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index cf90eeda9d71..247c5a5d2d6b 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1068,8 +1068,7 @@ bool pnfs_roc(struct inode *ino) spin_lock(&ino->i_lock); lo = nfsi->layout; - if (!lo || !test_and_clear_bit(NFS_LAYOUT_ROC, &lo->plh_flags) || - test_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags)) + if (!lo || test_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags)) goto out_noroc; /* no roc if we hold a delegation */ @@ -1617,10 +1616,8 @@ pnfs_layout_process(struct nfs4_layoutget *lgp) pnfs_get_lseg(lseg); pnfs_layout_insert_lseg(lo, lseg); - if (res->return_on_close) { + if (res->return_on_close) set_bit(NFS_LSEG_ROC, &lseg->pls_flags); - set_bit(NFS_LAYOUT_ROC, &lo->plh_flags); - } spin_unlock(&ino->i_lock); pnfs_free_lseg_list(&free_me); diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h index a3d57a8fac76..02c27f93caf1 100644 --- a/fs/nfs/pnfs.h +++ b/fs/nfs/pnfs.h @@ -94,7 +94,6 @@ enum { NFS_LAYOUT_RO_FAILED = 0, /* get ro layout failed stop trying */ NFS_LAYOUT_RW_FAILED, /* get rw layout failed stop trying */ NFS_LAYOUT_BULK_RECALL, /* bulk recall affecting layout */ - NFS_LAYOUT_ROC, /* some lseg had roc bit set */ NFS_LAYOUT_RETURN, /* Return this layout ASAP */ NFS_LAYOUT_RETURN_BEFORE_CLOSE, /* Return this layout before close */ NFS_LAYOUT_INVALID_STID, /* layout stateid id is invalid */ -- cgit v1.2.3 From 5420401079e152ff68a8024f6a375804b1c21505 Mon Sep 17 00:00:00 2001 From: Peng Tao Date: Sat, 22 Aug 2015 06:40:00 +0800 Subject: NFS41/flexfiles: zero out DS write wcc We do not want to update inode attributes with DS values. Cc: stable@vger.kernel.org # v4.0+ Signed-off-by: Peng Tao Signed-off-by: Trond Myklebust --- fs/nfs/flexfilelayout/flexfilelayout.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index 7e4cbdfb3713..cc2c5f7f2bc1 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -1218,6 +1218,8 @@ static int ff_layout_write_done_cb(struct rpc_task *task, hdr->res.verf->committed == NFS_DATA_SYNC) ff_layout_set_layoutcommit(hdr); + /* zero out fattr since we don't care DS attr at all */ + hdr->fattr.valid = 0; if (task->tk_status >= 0) nfs_writeback_update_inode(hdr); -- cgit v1.2.3 From a6970317f2d5f596302d957ae2f5b0d7c8676371 Mon Sep 17 00:00:00 2001 From: Oleg Drokin Date: Tue, 18 Aug 2015 21:04:35 -0400 Subject: staging/lustre/o2iblnd: remove references to ib_reg_phsy_mr() Removed references to ib_reg_phsy_mr() and PMR which was added to deal with some Chelsio specific scenario, but no longer needed now. Reported-by: Stephen Rothwell Signed-off-by: Amir Shehata Signed-off-by: Oleg Drokin Signed-off-by: Greg Kroah-Hartman Signed-off-by: Trond Myklebust --- .../staging/lustre/lnet/klnds/o2iblnd/o2iblnd.c | 251 +-------------------- .../staging/lustre/lnet/klnds/o2iblnd/o2iblnd.h | 32 +-- .../staging/lustre/lnet/klnds/o2iblnd/o2iblnd_cb.c | 55 +---- .../lustre/lnet/klnds/o2iblnd/o2iblnd_modparams.c | 6 - 4 files changed, 19 insertions(+), 325 deletions(-) diff --git a/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd.c b/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd.c index 4eb24a11b02a..3c3e79a51a0f 100644 --- a/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd.c +++ b/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd.c @@ -1789,140 +1789,6 @@ struct list_head *kiblnd_pool_alloc_node(kib_poolset_t *ps) goto again; } -void kiblnd_pmr_pool_unmap(kib_phys_mr_t *pmr) -{ - kib_pmr_pool_t *ppo = pmr->pmr_pool; - struct ib_mr *mr = pmr->pmr_mr; - - pmr->pmr_mr = NULL; - kiblnd_pool_free_node(&ppo->ppo_pool, &pmr->pmr_list); - if (mr != NULL) - ib_dereg_mr(mr); -} - -int kiblnd_pmr_pool_map(kib_pmr_poolset_t *pps, kib_hca_dev_t *hdev, - kib_rdma_desc_t *rd, __u64 *iova, kib_phys_mr_t **pp_pmr) -{ - kib_phys_mr_t *pmr; - struct list_head *node; - int rc; - int i; - - node = kiblnd_pool_alloc_node(&pps->pps_poolset); - if (node == NULL) { - CERROR("Failed to allocate PMR descriptor\n"); - return -ENOMEM; - } - - pmr = container_of(node, kib_phys_mr_t, pmr_list); - if (pmr->pmr_pool->ppo_hdev != hdev) { - kiblnd_pool_free_node(&pmr->pmr_pool->ppo_pool, node); - return -EAGAIN; - } - - for (i = 0; i < rd->rd_nfrags; i++) { - pmr->pmr_ipb[i].addr = rd->rd_frags[i].rf_addr; - pmr->pmr_ipb[i].size = rd->rd_frags[i].rf_nob; - } - - pmr->pmr_mr = ib_reg_phys_mr(hdev->ibh_pd, - pmr->pmr_ipb, rd->rd_nfrags, - IB_ACCESS_LOCAL_WRITE | - IB_ACCESS_REMOTE_WRITE, - iova); - if (!IS_ERR(pmr->pmr_mr)) { - pmr->pmr_iova = *iova; - *pp_pmr = pmr; - return 0; - } - - rc = PTR_ERR(pmr->pmr_mr); - CERROR("Failed ib_reg_phys_mr: %d\n", rc); - - pmr->pmr_mr = NULL; - kiblnd_pool_free_node(&pmr->pmr_pool->ppo_pool, node); - - return rc; -} - -static void kiblnd_destroy_pmr_pool(kib_pool_t *pool) -{ - kib_pmr_pool_t *ppo = container_of(pool, kib_pmr_pool_t, ppo_pool); - kib_phys_mr_t *pmr; - kib_phys_mr_t *tmp; - - LASSERT(pool->po_allocated == 0); - - list_for_each_entry_safe(pmr, tmp, &pool->po_free_list, pmr_list) { - LASSERT(pmr->pmr_mr == NULL); - list_del(&pmr->pmr_list); - - if (pmr->pmr_ipb != NULL) { - LIBCFS_FREE(pmr->pmr_ipb, - IBLND_MAX_RDMA_FRAGS * - sizeof(struct ib_phys_buf)); - } - - LIBCFS_FREE(pmr, sizeof(kib_phys_mr_t)); - } - - kiblnd_fini_pool(pool); - if (ppo->ppo_hdev != NULL) - kiblnd_hdev_decref(ppo->ppo_hdev); - - LIBCFS_FREE(ppo, sizeof(kib_pmr_pool_t)); -} - -static inline int kiblnd_pmr_pool_size(int ncpts) -{ - int size = *kiblnd_tunables.kib_pmr_pool_size / ncpts; - - return max(IBLND_PMR_POOL, size); -} - -static int kiblnd_create_pmr_pool(kib_poolset_t *ps, int size, - kib_pool_t **pp_po) -{ - struct kib_pmr_pool *ppo; - struct kib_pool *pool; - kib_phys_mr_t *pmr; - int i; - - LIBCFS_CPT_ALLOC(ppo, lnet_cpt_table(), - ps->ps_cpt, sizeof(kib_pmr_pool_t)); - if (ppo == NULL) { - CERROR("Failed to allocate PMR pool\n"); - return -ENOMEM; - } - - pool = &ppo->ppo_pool; - kiblnd_init_pool(ps, pool, size); - - for (i = 0; i < size; i++) { - LIBCFS_CPT_ALLOC(pmr, lnet_cpt_table(), - ps->ps_cpt, sizeof(kib_phys_mr_t)); - if (pmr == NULL) - break; - - pmr->pmr_pool = ppo; - LIBCFS_CPT_ALLOC(pmr->pmr_ipb, lnet_cpt_table(), ps->ps_cpt, - IBLND_MAX_RDMA_FRAGS * sizeof(*pmr->pmr_ipb)); - if (pmr->pmr_ipb == NULL) - break; - - list_add(&pmr->pmr_list, &pool->po_free_list); - } - - if (i < size) { - ps->ps_pool_destroy(pool); - return -ENOMEM; - } - - ppo->ppo_hdev = kiblnd_current_hdev(ps->ps_net->ibn_dev); - *pp_po = pool; - return 0; -} - static void kiblnd_destroy_tx_pool(kib_pool_t *pool) { kib_tx_pool_t *tpo = container_of(pool, kib_tx_pool_t, tpo_pool); @@ -2078,7 +1944,6 @@ static void kiblnd_net_fini_pools(kib_net_t *net) cfs_cpt_for_each(i, lnet_cpt_table()) { kib_tx_poolset_t *tps; kib_fmr_poolset_t *fps; - kib_pmr_poolset_t *pps; if (net->ibn_tx_ps != NULL) { tps = net->ibn_tx_ps[i]; @@ -2089,11 +1954,6 @@ static void kiblnd_net_fini_pools(kib_net_t *net) fps = net->ibn_fmr_ps[i]; kiblnd_fini_fmr_poolset(fps); } - - if (net->ibn_pmr_ps != NULL) { - pps = net->ibn_pmr_ps[i]; - kiblnd_fini_poolset(&pps->pps_poolset); - } } if (net->ibn_tx_ps != NULL) { @@ -2105,18 +1965,13 @@ static void kiblnd_net_fini_pools(kib_net_t *net) cfs_percpt_free(net->ibn_fmr_ps); net->ibn_fmr_ps = NULL; } - - if (net->ibn_pmr_ps != NULL) { - cfs_percpt_free(net->ibn_pmr_ps); - net->ibn_pmr_ps = NULL; - } } static int kiblnd_net_init_pools(kib_net_t *net, __u32 *cpts, int ncpts) { unsigned long flags; int cpt; - int rc; + int rc = 0; int i; read_lock_irqsave(&kiblnd_data.kib_global_lock, flags); @@ -2137,12 +1992,16 @@ static int kiblnd_net_init_pools(kib_net_t *net, __u32 *cpts, int ncpts) goto failed; } - /* TX pool must be created later than FMR/PMR, see LU-2268 - * for details */ + /* + * TX pool must be created later than FMR, see LU-2268 + * for details + */ LASSERT(net->ibn_tx_ps == NULL); - /* premapping can fail if ibd_nmr > 1, so we always create - * FMR/PMR pool and map-on-demand if premapping failed */ + /* + * premapping can fail if ibd_nmr > 1, so we always create + * FMR pool and map-on-demand if premapping failed + */ net->ibn_fmr_ps = cfs_percpt_alloc(lnet_cpt_table(), sizeof(kib_fmr_poolset_t)); @@ -2158,7 +2017,7 @@ static int kiblnd_net_init_pools(kib_net_t *net, __u32 *cpts, int ncpts) kiblnd_fmr_pool_size(ncpts), kiblnd_fmr_flush_trigger(ncpts)); if (rc == -ENOSYS && i == 0) /* no FMR */ - break; /* create PMR pool */ + break; if (rc != 0) { /* a real error */ CERROR("Can't initialize FMR pool for CPT %d: %d\n", @@ -2175,38 +2034,8 @@ static int kiblnd_net_init_pools(kib_net_t *net, __u32 *cpts, int ncpts) cfs_percpt_free(net->ibn_fmr_ps); net->ibn_fmr_ps = NULL; - CWARN("Device does not support FMR, failing back to PMR\n"); - - if (*kiblnd_tunables.kib_pmr_pool_size < - *kiblnd_tunables.kib_ntx / 4) { - CERROR("Can't set pmr pool size (%d) < ntx / 4(%d)\n", - *kiblnd_tunables.kib_pmr_pool_size, - *kiblnd_tunables.kib_ntx / 4); - rc = -EINVAL; - goto failed; - } - - net->ibn_pmr_ps = cfs_percpt_alloc(lnet_cpt_table(), - sizeof(kib_pmr_poolset_t)); - if (net->ibn_pmr_ps == NULL) { - CERROR("Failed to allocate PMR pool array\n"); - rc = -ENOMEM; - goto failed; - } - - for (i = 0; i < ncpts; i++) { - cpt = (cpts == NULL) ? i : cpts[i]; - rc = kiblnd_init_poolset(&net->ibn_pmr_ps[cpt]->pps_poolset, - cpt, net, "PMR", - kiblnd_pmr_pool_size(ncpts), - kiblnd_create_pmr_pool, - kiblnd_destroy_pmr_pool, NULL, NULL); - if (rc != 0) { - CERROR("Can't initialize PMR pool for CPT %d: %d\n", - cpt, rc); + CWARN("Device does not support FMR\n"); goto failed; - } - } create_tx_pool: net->ibn_tx_ps = cfs_percpt_alloc(lnet_cpt_table(), @@ -2318,17 +2147,13 @@ void kiblnd_hdev_destroy(kib_hca_dev_t *hdev) static int kiblnd_hdev_setup_mrs(kib_hca_dev_t *hdev) { struct ib_mr *mr; - int i; int rc; - __u64 mm_size; - __u64 mr_size; int acflags = IB_ACCESS_LOCAL_WRITE | IB_ACCESS_REMOTE_WRITE; rc = kiblnd_hdev_get_attr(hdev); if (rc != 0) return rc; - if (hdev->ibh_mr_shift == 64) { LIBCFS_ALLOC(hdev->ibh_mrs, 1 * sizeof(*hdev->ibh_mrs)); if (hdev->ibh_mrs == NULL) { CERROR("Failed to allocate MRs table\n"); @@ -2347,53 +2172,6 @@ static int kiblnd_hdev_setup_mrs(kib_hca_dev_t *hdev) hdev->ibh_mrs[0] = mr; - goto out; - } - - mr_size = 1ULL << hdev->ibh_mr_shift; - mm_size = (unsigned long)high_memory - PAGE_OFFSET; - - hdev->ibh_nmrs = (int)((mm_size + mr_size - 1) >> hdev->ibh_mr_shift); - - if (hdev->ibh_mr_shift < 32 || hdev->ibh_nmrs > 1024) { - /* it's 4T..., assume we will re-code at that time */ - CERROR("Can't support memory size: x%#llx with MR size: x%#llx\n", - mm_size, mr_size); - return -EINVAL; - } - - /* create an array of MRs to cover all memory */ - LIBCFS_ALLOC(hdev->ibh_mrs, sizeof(*hdev->ibh_mrs) * hdev->ibh_nmrs); - if (hdev->ibh_mrs == NULL) { - CERROR("Failed to allocate MRs' table\n"); - return -ENOMEM; - } - - for (i = 0; i < hdev->ibh_nmrs; i++) { - struct ib_phys_buf ipb; - __u64 iova; - - ipb.size = hdev->ibh_mr_size; - ipb.addr = i * mr_size; - iova = ipb.addr; - - mr = ib_reg_phys_mr(hdev->ibh_pd, &ipb, 1, acflags, &iova); - if (IS_ERR(mr)) { - CERROR("Failed ib_reg_phys_mr addr %#llx size %#llx : %ld\n", - ipb.addr, ipb.size, PTR_ERR(mr)); - kiblnd_hdev_cleanup_mrs(hdev); - return PTR_ERR(mr); - } - - LASSERT(iova == ipb.addr); - - hdev->ibh_mrs[i] = mr; - } - -out: - if (hdev->ibh_mr_size != ~0ULL || hdev->ibh_nmrs != 1) - LCONSOLE_INFO("Register global MR array, MR size: %#llx, array size: %d\n", - hdev->ibh_mr_size, hdev->ibh_nmrs); return 0; } @@ -2564,14 +2342,9 @@ int kiblnd_dev_failover(kib_dev_t *dev) kiblnd_fail_poolset(&net->ibn_tx_ps[i]->tps_poolset, &zombie_tpo); - if (net->ibn_fmr_ps != NULL) { + if (net->ibn_fmr_ps) kiblnd_fail_fmr_poolset(net->ibn_fmr_ps[i], &zombie_fpo); - - } else if (net->ibn_pmr_ps != NULL) { - kiblnd_fail_poolset(&net->ibn_pmr_ps[i]-> - pps_poolset, &zombie_ppo); - } } } diff --git a/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd.h b/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd.h index f5d1d9f8f1ed..2687ae39fcbb 100644 --- a/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd.h +++ b/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd.h @@ -104,7 +104,6 @@ typedef struct { int *kib_map_on_demand; /* map-on-demand if RD has more * fragments than this value, 0 * disable map-on-demand */ - int *kib_pmr_pool_size; /* # physical MR in pool */ int *kib_fmr_pool_size; /* # FMRs in pool */ int *kib_fmr_flush_trigger; /* When to trigger FMR flush */ int *kib_fmr_cache; /* enable FMR pool cache? */ @@ -163,7 +162,6 @@ kiblnd_concurrent_sends_v1(void) /* Pools (shared by connections on each CPT) */ /* These pools can grow at runtime, so don't need give a very large value */ #define IBLND_TX_POOL 256 -#define IBLND_PMR_POOL 256 #define IBLND_FMR_POOL 256 #define IBLND_FMR_POOL_FLUSH 192 @@ -232,17 +230,6 @@ typedef struct { struct page *ibp_pages[0]; /* page array */ } kib_pages_t; -struct kib_pmr_pool; - -typedef struct { - struct list_head pmr_list; /* chain node */ - struct ib_phys_buf *pmr_ipb; /* physical buffer */ - struct ib_mr *pmr_mr; /* IB MR */ - struct kib_pmr_pool *pmr_pool; /* owner of this MR */ - __u64 pmr_iova; /* Virtual I/O address */ - int pmr_refcount; /* reference count */ -} kib_phys_mr_t; - struct kib_pool; struct kib_poolset; @@ -298,15 +285,6 @@ typedef struct { kib_pages_t *tpo_tx_pages; /* premapped tx msg pages */ } kib_tx_pool_t; -typedef struct { - kib_poolset_t pps_poolset; /* pool-set */ -} kib_pmr_poolset_t; - -typedef struct kib_pmr_pool { - struct kib_hca_dev *ppo_hdev; /* device for this pool */ - kib_pool_t ppo_pool; /* pool */ -} kib_pmr_pool_t; - typedef struct { spinlock_t fps_lock; /* serialize */ struct kib_net *fps_net; /* IB network */ @@ -347,7 +325,6 @@ typedef struct kib_net { kib_tx_poolset_t **ibn_tx_ps; /* tx pool-set */ kib_fmr_poolset_t **ibn_fmr_ps; /* fmr pool-set */ - kib_pmr_poolset_t **ibn_pmr_ps; /* pmr pool-set */ kib_dev_t *ibn_dev; /* underlying IB device */ } kib_net_t; @@ -554,10 +531,7 @@ typedef struct kib_tx /* transmit message */ int tx_nfrags; /* # entries in... */ struct scatterlist *tx_frags; /* dma_map_sg descriptor */ __u64 *tx_pages; /* rdma phys page addrs */ - union { - kib_phys_mr_t *pmr; /* MR for physical buffer */ - kib_fmr_t fmr; /* FMR */ - } tx_u; + kib_fmr_t fmr; /* FMR */ int tx_dmadir; /* dma direction */ } kib_tx_t; @@ -978,10 +952,6 @@ int kiblnd_fmr_pool_map(kib_fmr_poolset_t *fps, __u64 *pages, int npages, __u64 iov, kib_fmr_t *fmr); void kiblnd_fmr_pool_unmap(kib_fmr_t *fmr, int status); -int kiblnd_pmr_pool_map(kib_pmr_poolset_t *pps, kib_hca_dev_t *hdev, - kib_rdma_desc_t *rd, __u64 *iova, kib_phys_mr_t **pp_pmr); -void kiblnd_pmr_pool_unmap(kib_phys_mr_t *pmr); - int kiblnd_startup (lnet_ni_t *ni); void kiblnd_shutdown (lnet_ni_t *ni); int kiblnd_ctl (lnet_ni_t *ni, unsigned int cmd, void *arg); diff --git a/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd_cb.c b/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd_cb.c index 477aa8b76f32..a131024e4932 100644 --- a/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd_cb.c +++ b/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd_cb.c @@ -121,7 +121,6 @@ kiblnd_get_idle_tx(lnet_ni_t *ni, lnet_nid_t target) LASSERT(tx->tx_conn == NULL); LASSERT(tx->tx_lntmsg[0] == NULL); LASSERT(tx->tx_lntmsg[1] == NULL); - LASSERT(tx->tx_u.pmr == NULL); LASSERT(tx->tx_nfrags == 0); return tx; @@ -575,7 +574,7 @@ kiblnd_fmr_map_tx(kib_net_t *net, kib_tx_t *tx, kib_rdma_desc_t *rd, int nob) cpt = tx->tx_pool->tpo_pool.po_owner->ps_cpt; fps = net->ibn_fmr_ps[cpt]; - rc = kiblnd_fmr_pool_map(fps, pages, npages, 0, &tx->tx_u.fmr); + rc = kiblnd_fmr_pool_map(fps, pages, npages, 0, &tx->fmr); if (rc != 0) { CERROR("Can't map %d pages: %d\n", npages, rc); return rc; @@ -583,8 +582,8 @@ kiblnd_fmr_map_tx(kib_net_t *net, kib_tx_t *tx, kib_rdma_desc_t *rd, int nob) /* If rd is not tx_rd, it's going to get sent to a peer, who will need * the rkey */ - rd->rd_key = (rd != tx->tx_rd) ? tx->tx_u.fmr.fmr_pfmr->fmr->rkey : - tx->tx_u.fmr.fmr_pfmr->fmr->lkey; + rd->rd_key = (rd != tx->tx_rd) ? tx->fmr.fmr_pfmr->fmr->rkey : + tx->fmr.fmr_pfmr->fmr->lkey; rd->rd_frags[0].rf_addr &= ~hdev->ibh_page_mask; rd->rd_frags[0].rf_nob = nob; rd->rd_nfrags = 1; @@ -592,42 +591,6 @@ kiblnd_fmr_map_tx(kib_net_t *net, kib_tx_t *tx, kib_rdma_desc_t *rd, int nob) return 0; } -static int -kiblnd_pmr_map_tx(kib_net_t *net, kib_tx_t *tx, kib_rdma_desc_t *rd, int nob) -{ - kib_hca_dev_t *hdev; - kib_pmr_poolset_t *pps; - __u64 iova; - int cpt; - int rc; - - LASSERT(tx->tx_pool != NULL); - LASSERT(tx->tx_pool->tpo_pool.po_owner != NULL); - - hdev = tx->tx_pool->tpo_hdev; - - iova = rd->rd_frags[0].rf_addr & ~hdev->ibh_page_mask; - - cpt = tx->tx_pool->tpo_pool.po_owner->ps_cpt; - - pps = net->ibn_pmr_ps[cpt]; - rc = kiblnd_pmr_pool_map(pps, hdev, rd, &iova, &tx->tx_u.pmr); - if (rc != 0) { - CERROR("Failed to create MR by phybuf: %d\n", rc); - return rc; - } - - /* If rd is not tx_rd, it's going to get sent to a peer, who will need - * the rkey */ - rd->rd_key = (rd != tx->tx_rd) ? tx->tx_u.pmr->pmr_mr->rkey : - tx->tx_u.pmr->pmr_mr->lkey; - rd->rd_nfrags = 1; - rd->rd_frags[0].rf_addr = iova; - rd->rd_frags[0].rf_nob = nob; - - return 0; -} - void kiblnd_unmap_tx(lnet_ni_t *ni, kib_tx_t *tx) { @@ -635,13 +598,9 @@ kiblnd_unmap_tx(lnet_ni_t *ni, kib_tx_t *tx) LASSERT(net != NULL); - if (net->ibn_fmr_ps != NULL && tx->tx_u.fmr.fmr_pfmr != NULL) { - kiblnd_fmr_pool_unmap(&tx->tx_u.fmr, tx->tx_status); - tx->tx_u.fmr.fmr_pfmr = NULL; - - } else if (net->ibn_pmr_ps != NULL && tx->tx_u.pmr != NULL) { - kiblnd_pmr_pool_unmap(tx->tx_u.pmr); - tx->tx_u.pmr = NULL; + if (net->ibn_fmr_ps && tx->fmr.fmr_pfmr) { + kiblnd_fmr_pool_unmap(&tx->fmr, tx->tx_status); + tx->fmr.fmr_pfmr = NULL; } if (tx->tx_nfrags != 0) { @@ -687,8 +646,6 @@ kiblnd_map_tx(lnet_ni_t *ni, kib_tx_t *tx, if (net->ibn_fmr_ps != NULL) return kiblnd_fmr_map_tx(net, tx, rd, nob); - else if (net->ibn_pmr_ps != NULL) - return kiblnd_pmr_map_tx(net, tx, rd, nob); return -EINVAL; } diff --git a/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd_modparams.c b/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd_modparams.c index b0e00361cfce..b3d1b5d627cb 100644 --- a/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd_modparams.c +++ b/drivers/staging/lustre/lnet/klnds/o2iblnd/o2iblnd_modparams.c @@ -126,11 +126,6 @@ static int fmr_cache = 1; module_param(fmr_cache, int, 0444); MODULE_PARM_DESC(fmr_cache, "non-zero to enable FMR caching"); -/* NB: this value is shared by all CPTs, it can grow at runtime */ -static int pmr_pool_size = 512; -module_param(pmr_pool_size, int, 0444); -MODULE_PARM_DESC(pmr_pool_size, "size of MR cache pmr pool on each CPT"); - /* * 0: disable failover * 1: enable failover if necessary @@ -170,7 +165,6 @@ kib_tunables_t kiblnd_tunables = { .kib_fmr_pool_size = &fmr_pool_size, .kib_fmr_flush_trigger = &fmr_flush_trigger, .kib_fmr_cache = &fmr_cache, - .kib_pmr_pool_size = &pmr_pool_size, .kib_require_priv_port = &require_privileged_port, .kib_use_priv_port = &use_privileged_port, .kib_nscheds = &nscheds -- cgit v1.2.3 From 28a0d72c6867af307c000e068a6da93ae98bcd0c Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 24 Aug 2015 18:08:30 -0400 Subject: NFSv4.1/flexfiles: Add refcounting to struct nfs4_ff_layout_mirror We do want to share mirrors between layout segments, so add a refcount to enable that. Signed-off-by: Trond Myklebust --- fs/nfs/flexfilelayout/flexfilelayout.c | 36 +++++++++++++++++++++++++--------- fs/nfs/flexfilelayout/flexfilelayout.h | 1 + 2 files changed, 28 insertions(+), 9 deletions(-) diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index cc2c5f7f2bc1..62de0b8038c8 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -135,6 +135,31 @@ decode_name(struct xdr_stream *xdr, u32 *id) return 0; } +static struct nfs4_ff_layout_mirror *ff_layout_alloc_mirror(gfp_t gfp_flags) +{ + struct nfs4_ff_layout_mirror *mirror; + + mirror = kzalloc(sizeof(*mirror), gfp_flags); + if (mirror != NULL) { + spin_lock_init(&mirror->lock); + atomic_set(&mirror->ref, 1); + } + return mirror; +} + +static void ff_layout_free_mirror(struct nfs4_ff_layout_mirror *mirror) +{ + kfree(mirror->fh_versions); + nfs4_ff_layout_put_deviceid(mirror->mirror_ds); + kfree(mirror); +} + +static void ff_layout_put_mirror(struct nfs4_ff_layout_mirror *mirror) +{ + if (mirror != NULL && atomic_dec_and_test(&mirror->ref)) + ff_layout_free_mirror(mirror); +} + static void ff_layout_free_mirror_array(struct nfs4_ff_layout_segment *fls) { int i; @@ -144,11 +169,7 @@ static void ff_layout_free_mirror_array(struct nfs4_ff_layout_segment *fls) /* normally mirror_ds is freed in * .free_deviceid_node but we still do it here * for .alloc_lseg error path */ - if (fls->mirror_array[i]) { - kfree(fls->mirror_array[i]->fh_versions); - nfs4_ff_layout_put_deviceid(fls->mirror_array[i]->mirror_ds); - kfree(fls->mirror_array[i]); - } + ff_layout_put_mirror(fls->mirror_array[i]); } kfree(fls->mirror_array); fls->mirror_array = NULL; @@ -262,15 +283,12 @@ ff_layout_alloc_lseg(struct pnfs_layout_hdr *lh, if (ds_count != 1) goto out_err_free; - fls->mirror_array[i] = - kzalloc(sizeof(struct nfs4_ff_layout_mirror), - gfp_flags); + fls->mirror_array[i] = ff_layout_alloc_mirror(gfp_flags); if (fls->mirror_array[i] == NULL) { rc = -ENOMEM; goto out_err_free; } - spin_lock_init(&fls->mirror_array[i]->lock); fls->mirror_array[i]->ds_count = ds_count; fls->mirror_array[i]->lseg = &fls->generic_hdr; diff --git a/fs/nfs/flexfilelayout/flexfilelayout.h b/fs/nfs/flexfilelayout/flexfilelayout.h index f92f9a0a856b..26b8258e256f 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.h +++ b/fs/nfs/flexfilelayout/flexfilelayout.h @@ -77,6 +77,7 @@ struct nfs4_ff_layout_mirror { u32 uid; u32 gid; struct rpc_cred *cred; + atomic_t ref; spinlock_t lock; struct nfs4_ff_layoutstat read_stat; struct nfs4_ff_layoutstat write_stat; -- cgit v1.2.3 From 0b7baf9433d4e0a0c83eecbadc1f035d5370c6dc Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 24 Aug 2015 18:22:28 -0400 Subject: NFSv4.1/flexfiles: Remove mirror backpointer to lseg. When we start sharing mirrors between several lsegs, we won't be able to keep it. Signed-off-by: Trond Myklebust --- fs/nfs/flexfilelayout/flexfilelayout.c | 25 ++++++++++++------------- fs/nfs/flexfilelayout/flexfilelayout.h | 1 - 2 files changed, 12 insertions(+), 14 deletions(-) diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index 62de0b8038c8..f3efff640989 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -290,7 +290,6 @@ ff_layout_alloc_lseg(struct pnfs_layout_hdr *lh, } fls->mirror_array[i]->ds_count = ds_count; - fls->mirror_array[i]->lseg = &fls->generic_hdr; /* deviceid */ rc = decode_deviceid(&stream, &devid); @@ -513,7 +512,8 @@ nfs4_ff_layout_stat_io_update_completed(struct nfs4_ff_layoutstat *layoutstat, } static void -nfs4_ff_layout_stat_io_start_read(struct nfs4_ff_layout_mirror *mirror, +nfs4_ff_layout_stat_io_start_read(struct inode *inode, + struct nfs4_ff_layout_mirror *mirror, __u64 requested, ktime_t now) { bool report; @@ -524,8 +524,7 @@ nfs4_ff_layout_stat_io_start_read(struct nfs4_ff_layout_mirror *mirror, spin_unlock(&mirror->lock); if (report) - pnfs_report_layoutstat(mirror->lseg->pls_layout->plh_inode, - GFP_KERNEL); + pnfs_report_layoutstat(inode, GFP_KERNEL); } static void @@ -542,7 +541,8 @@ nfs4_ff_layout_stat_io_end_read(struct rpc_task *task, } static void -nfs4_ff_layout_stat_io_start_write(struct nfs4_ff_layout_mirror *mirror, +nfs4_ff_layout_stat_io_start_write(struct inode *inode, + struct nfs4_ff_layout_mirror *mirror, __u64 requested, ktime_t now) { bool report; @@ -553,8 +553,7 @@ nfs4_ff_layout_stat_io_start_write(struct nfs4_ff_layout_mirror *mirror, spin_unlock(&mirror->lock); if (report) - pnfs_report_layoutstat(mirror->lseg->pls_layout->plh_inode, - GFP_NOIO); + pnfs_report_layoutstat(inode, GFP_NOIO); } static void @@ -1098,7 +1097,7 @@ ff_layout_reset_to_mds(struct pnfs_layout_segment *lseg, int idx) static int ff_layout_read_prepare_common(struct rpc_task *task, struct nfs_pgio_header *hdr) { - nfs4_ff_layout_stat_io_start_read( + nfs4_ff_layout_stat_io_start_read(hdr->inode, FF_LAYOUT_COMP(hdr->lseg, hdr->pgio_mirror_idx), hdr->args.count, task->tk_start); @@ -1285,7 +1284,7 @@ static int ff_layout_commit_done_cb(struct rpc_task *task, static int ff_layout_write_prepare_common(struct rpc_task *task, struct nfs_pgio_header *hdr) { - nfs4_ff_layout_stat_io_start_write( + nfs4_ff_layout_stat_io_start_write(hdr->inode, FF_LAYOUT_COMP(hdr->lseg, hdr->pgio_mirror_idx), hdr->args.count, task->tk_start); @@ -1367,7 +1366,7 @@ static void ff_layout_write_count_stats(struct rpc_task *task, void *data) static void ff_layout_commit_prepare_common(struct rpc_task *task, struct nfs_commit_data *cdata) { - nfs4_ff_layout_stat_io_start_write( + nfs4_ff_layout_stat_io_start_write(cdata->inode, FF_LAYOUT_COMP(cdata->lseg, cdata->ds_commit_index), 0, task->tk_start); } @@ -1912,8 +1911,8 @@ ff_layout_mirror_prepare_stats(struct nfs42_layoutstat_args *args, devinfo->layout_type = LAYOUT_FLEX_FILES; devinfo->layoutstats_encode = ff_layout_encode_layoutstats; devinfo->layout_private = mirror; - /* lseg refcount put in cleanup_layoutstats */ - pnfs_get_lseg(pls); + /* mirror refcount put in cleanup_layoutstats */ + atomic_inc(&mirror->ref); ++(*dev_count); } @@ -1965,7 +1964,7 @@ ff_layout_cleanup_layoutstats(struct nfs42_layoutstat_data *data) for (i = 0; i < data->args.num_dev; i++) { mirror = data->args.devinfo[i].layout_private; data->args.devinfo[i].layout_private = NULL; - pnfs_put_lseg(mirror->lseg); + ff_layout_put_mirror(mirror); } } diff --git a/fs/nfs/flexfilelayout/flexfilelayout.h b/fs/nfs/flexfilelayout/flexfilelayout.h index 26b8258e256f..fe9d3ff7cf85 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.h +++ b/fs/nfs/flexfilelayout/flexfilelayout.h @@ -67,7 +67,6 @@ struct nfs4_ff_layoutstat { }; struct nfs4_ff_layout_mirror { - struct pnfs_layout_segment *lseg; /* back pointer */ u32 ds_count; u32 efficiency; struct nfs4_ff_layout_ds *mirror_ds; -- cgit v1.2.3 From 266d12d42ebd2587d3d0e8c4a4e54bfb5ed29543 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 24 Aug 2015 20:03:17 -0400 Subject: NFSv4.1/flexfile: Ensure uniqueness of mirrors across layout segments Keep the full list of mirrors in the struct nfs4_ff_layout_mirror so that they can be shared among the layout segments that use them. Also ensure that we send out only one copy of the layoutstats per mirror. Signed-off-by: Trond Myklebust --- fs/nfs/flexfilelayout/flexfilelayout.c | 125 +++++++++++++++++++++++++-------- fs/nfs/flexfilelayout/flexfilelayout.h | 3 + 2 files changed, 99 insertions(+), 29 deletions(-) diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index f3efff640989..0fbf37de2a41 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -34,6 +34,7 @@ ff_layout_alloc_layout_hdr(struct inode *inode, gfp_t gfp_flags) ffl = kzalloc(sizeof(*ffl), gfp_flags); if (ffl) { INIT_LIST_HEAD(&ffl->error_list); + INIT_LIST_HEAD(&ffl->mirrors); return &ffl->generic_hdr; } else return NULL; @@ -135,6 +136,66 @@ decode_name(struct xdr_stream *xdr, u32 *id) return 0; } +static bool ff_mirror_match_fh(const struct nfs4_ff_layout_mirror *m1, + const struct nfs4_ff_layout_mirror *m2) +{ + int i, j; + + if (m1->fh_versions_cnt != m2->fh_versions_cnt) + return false; + for (i = 0; i < m1->fh_versions_cnt; i++) { + bool found_fh = false; + for (j = 0; j < m2->fh_versions_cnt; i++) { + if (nfs_compare_fh(&m1->fh_versions[i], + &m2->fh_versions[j]) == 0) { + found_fh = true; + break; + } + } + if (!found_fh) + return false; + } + return true; +} + +static struct nfs4_ff_layout_mirror * +ff_layout_add_mirror(struct pnfs_layout_hdr *lo, + struct nfs4_ff_layout_mirror *mirror) +{ + struct nfs4_flexfile_layout *ff_layout = FF_LAYOUT_FROM_HDR(lo); + struct nfs4_ff_layout_mirror *pos; + struct inode *inode = lo->plh_inode; + + spin_lock(&inode->i_lock); + list_for_each_entry(pos, &ff_layout->mirrors, mirrors) { + if (mirror->mirror_ds != pos->mirror_ds) + continue; + if (!ff_mirror_match_fh(mirror, pos)) + continue; + if (atomic_inc_not_zero(&pos->ref)) { + spin_unlock(&inode->i_lock); + return pos; + } + } + list_add(&mirror->mirrors, &ff_layout->mirrors); + mirror->layout = lo; + spin_unlock(&inode->i_lock); + return mirror; +} + +void +ff_layout_remove_mirror(struct nfs4_ff_layout_mirror *mirror) +{ + struct inode *inode; + if (mirror->layout == NULL) + return; + inode = mirror->layout->plh_inode; + spin_lock(&inode->i_lock); + list_del(&mirror->mirrors); + spin_unlock(&inode->i_lock); + mirror->layout = NULL; +} + static struct nfs4_ff_layout_mirror *ff_layout_alloc_mirror(gfp_t gfp_flags) { struct nfs4_ff_layout_mirror *mirror; @@ -143,12 +204,14 @@ static struct nfs4_ff_layout_mirror *ff_layout_alloc_mirror(gfp_t gfp_flags) if (mirror != NULL) { spin_lock_init(&mirror->lock); atomic_set(&mirror->ref, 1); + INIT_LIST_HEAD(&mirror->mirrors); } return mirror; } static void ff_layout_free_mirror(struct nfs4_ff_layout_mirror *mirror) { + ff_layout_remove_mirror(mirror); kfree(mirror->fh_versions); nfs4_ff_layout_put_deviceid(mirror->mirror_ds); kfree(mirror); @@ -267,6 +330,7 @@ ff_layout_alloc_lseg(struct pnfs_layout_hdr *lh, goto out_err_free; for (i = 0; i < fls->mirror_array_cnt; i++) { + struct nfs4_ff_layout_mirror *mirror; struct nfs4_deviceid devid; struct nfs4_deviceid_node *idnode; u32 ds_count; @@ -355,6 +419,12 @@ ff_layout_alloc_lseg(struct pnfs_layout_hdr *lh, if (rc) goto out_err_free; + mirror = ff_layout_add_mirror(lh, fls->mirror_array[i]); + if (mirror != fls->mirror_array[i]) { + ff_layout_free_mirror(fls->mirror_array[i]); + fls->mirror_array[i] = mirror; + } + dprintk("%s: uid %d gid %d\n", __func__, fls->mirror_array[i]->uid, fls->mirror_array[i]->gid); @@ -1883,27 +1953,30 @@ ff_layout_encode_layoutstats(struct xdr_stream *xdr, *start = cpu_to_be32((xdr->p - start - 1) * 4); } -static bool +static int ff_layout_mirror_prepare_stats(struct nfs42_layoutstat_args *args, - struct pnfs_layout_segment *pls, - int *dev_count, int dev_limit) + struct pnfs_layout_hdr *lo, + int dev_limit) { + struct nfs4_flexfile_layout *ff_layout = FF_LAYOUT_FROM_HDR(lo); struct nfs4_ff_layout_mirror *mirror; struct nfs4_deviceid_node *dev; struct nfs42_layoutstat_devinfo *devinfo; - int i; + int i = 0; - for (i = 0; i < FF_LAYOUT_MIRROR_COUNT(pls); i++) { - if (*dev_count >= dev_limit) + list_for_each_entry(mirror, &ff_layout->mirrors, mirrors) { + if (i >= dev_limit) break; - mirror = FF_LAYOUT_COMP(pls, i); - if (!mirror || !mirror->mirror_ds) + if (!mirror->mirror_ds) continue; - dev = FF_LAYOUT_DEVID_NODE(pls, i); - devinfo = &args->devinfo[*dev_count]; + /* mirror refcount put in cleanup_layoutstats */ + if (!atomic_inc_not_zero(&mirror->ref)) + continue; + dev = &mirror->mirror_ds->id_node; + devinfo = &args->devinfo[i]; memcpy(&devinfo->dev_id, &dev->deviceid, NFS4_DEVICEID4_SIZE); - devinfo->offset = pls->pls_range.offset; - devinfo->length = pls->pls_range.length; + devinfo->offset = 0; + devinfo->length = NFS4_MAX_UINT64; devinfo->read_count = mirror->read_stat.io_stat.ops_completed; devinfo->read_bytes = mirror->read_stat.io_stat.bytes_completed; devinfo->write_count = mirror->write_stat.io_stat.ops_completed; @@ -1911,24 +1984,24 @@ ff_layout_mirror_prepare_stats(struct nfs42_layoutstat_args *args, devinfo->layout_type = LAYOUT_FLEX_FILES; devinfo->layoutstats_encode = ff_layout_encode_layoutstats; devinfo->layout_private = mirror; - /* mirror refcount put in cleanup_layoutstats */ - atomic_inc(&mirror->ref); - ++(*dev_count); + i++; } - - return *dev_count < dev_limit; + return i; } static int ff_layout_prepare_layoutstats(struct nfs42_layoutstat_args *args) { - struct pnfs_layout_segment *pls; + struct nfs4_flexfile_layout *ff_layout; + struct nfs4_ff_layout_mirror *mirror; int dev_count = 0; spin_lock(&args->inode->i_lock); - list_for_each_entry(pls, &NFS_I(args->inode)->layout->plh_segs, pls_list) { - dev_count += FF_LAYOUT_MIRROR_COUNT(pls); + ff_layout = FF_LAYOUT_FROM_HDR(NFS_I(args->inode)->layout); + list_for_each_entry(mirror, &ff_layout->mirrors, mirrors) { + if (atomic_read(&mirror->ref) != 0) + dev_count ++; } spin_unlock(&args->inode->i_lock); /* For now, send at most PNFS_LAYOUTSTATS_MAXDEV statistics */ @@ -1937,20 +2010,14 @@ ff_layout_prepare_layoutstats(struct nfs42_layoutstat_args *args) __func__, dev_count, PNFS_LAYOUTSTATS_MAXDEV); dev_count = PNFS_LAYOUTSTATS_MAXDEV; } - args->devinfo = kmalloc(dev_count * sizeof(*args->devinfo), GFP_KERNEL); + args->devinfo = kmalloc_array(dev_count, sizeof(*args->devinfo), GFP_NOIO); if (!args->devinfo) return -ENOMEM; - dev_count = 0; spin_lock(&args->inode->i_lock); - list_for_each_entry(pls, &NFS_I(args->inode)->layout->plh_segs, pls_list) { - if (!ff_layout_mirror_prepare_stats(args, pls, &dev_count, - PNFS_LAYOUTSTATS_MAXDEV)) { - break; - } - } + args->num_dev = ff_layout_mirror_prepare_stats(args, + &ff_layout->generic_hdr, dev_count); spin_unlock(&args->inode->i_lock); - args->num_dev = dev_count; return 0; } diff --git a/fs/nfs/flexfilelayout/flexfilelayout.h b/fs/nfs/flexfilelayout/flexfilelayout.h index fe9d3ff7cf85..68cc0d9828f9 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.h +++ b/fs/nfs/flexfilelayout/flexfilelayout.h @@ -67,6 +67,8 @@ struct nfs4_ff_layoutstat { }; struct nfs4_ff_layout_mirror { + struct pnfs_layout_hdr *layout; + struct list_head mirrors; u32 ds_count; u32 efficiency; struct nfs4_ff_layout_ds *mirror_ds; @@ -95,6 +97,7 @@ struct nfs4_ff_layout_segment { struct nfs4_flexfile_layout { struct pnfs_layout_hdr generic_hdr; struct pnfs_ds_commit_info commit_info; + struct list_head mirrors; struct list_head error_list; /* nfs4_ff_layout_ds_err */ }; -- cgit v1.2.3 From bbf58bf3488e41f346536aa89d62bdf2fe771128 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 24 Aug 2015 20:39:18 -0400 Subject: NFSv4.2/pnfs: Make the layoutstats timer configurable Allow advanced users to set the layoutstats timer in order to lengthen or shorten the period between layoutstat transmissions to the server. Signed-off-by: Trond Myklebust --- Documentation/kernel-parameters.txt | 9 +++++++++ fs/nfs/flexfilelayout/flexfilelayout.c | 5 ++++- fs/nfs/pnfs.c | 4 ++++ fs/nfs/pnfs.h | 3 +++ 4 files changed, 20 insertions(+), 1 deletion(-) diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt index 1d6f0459cd7b..30d78b561574 100644 --- a/Documentation/kernel-parameters.txt +++ b/Documentation/kernel-parameters.txt @@ -2279,6 +2279,15 @@ bytes respectively. Such letter suffixes can also be entirely omitted. The default parameter value of '0' causes the kernel not to attempt recovery of lost locks. + nfs4.layoutstats_timer = + [NFSv4.2] Change the rate at which the kernel sends + layoutstats to the pNFS metadata server. + + Setting this to value to 0 causes the kernel to use + whatever value is the default set by the layout + driver. A non-zero value sets the minimum interval + in seconds between layoutstats transmissions. + nfsd.nfs4_disable_idmapping= [NFSv4] When set to the default of '1', the NFSv4 server will return only numeric uids and gids to diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index 0fbf37de2a41..9f6fb8876b3f 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -533,14 +533,17 @@ nfs4_ff_layoutstat_start_io(struct nfs4_ff_layout_mirror *mirror, ktime_t now) { static const ktime_t notime = {0}; + s64 report_interval = FF_LAYOUTSTATS_REPORT_INTERVAL; nfs4_ff_start_busy_timer(&layoutstat->busy_timer, now); if (ktime_equal(mirror->start_time, notime)) mirror->start_time = now; if (ktime_equal(mirror->last_report_time, notime)) mirror->last_report_time = now; + if (layoutstats_timer != 0) + report_interval = (s64)layoutstats_timer * 1000LL; if (ktime_to_ms(ktime_sub(now, mirror->last_report_time)) >= - FF_LAYOUTSTATS_REPORT_INTERVAL) { + report_interval) { mirror->last_report_time = now; return true; } diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 247c5a5d2d6b..3530bb703214 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -2285,3 +2285,7 @@ out_put: } EXPORT_SYMBOL_GPL(pnfs_report_layoutstat); #endif + +unsigned int layoutstats_timer; +module_param(layoutstats_timer, uint, 0644); +EXPORT_SYMBOL_GPL(layoutstats_timer); diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h index 02c27f93caf1..d3979dd1037a 100644 --- a/fs/nfs/pnfs.h +++ b/fs/nfs/pnfs.h @@ -528,12 +528,15 @@ pnfs_use_threshold(struct nfs4_threshold **dst, struct nfs4_threshold *src, nfss->pnfs_curr_ld->id == src->l_type); } +extern unsigned int layoutstats_timer; + #ifdef NFS_DEBUG void nfs4_print_deviceid(const struct nfs4_deviceid *dev_id); #else static inline void nfs4_print_deviceid(const struct nfs4_deviceid *dev_id) { } + #endif /* NFS_DEBUG */ #else /* CONFIG_NFS_V4_1 */ -- cgit v1.2.3 From e3b1df2dbd7bcda807a94db131fda6c2bbd1480a Mon Sep 17 00:00:00 2001 From: kbuild test robot Date: Tue, 25 Aug 2015 11:19:25 +0800 Subject: NFSv4.1/flexfile: ff_layout_remove_mirror can be static Signed-off-by: Fengguang Wu Signed-off-by: Trond Myklebust --- fs/nfs/flexfilelayout/flexfilelayout.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index 9f6fb8876b3f..7fefa8ad9578 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -183,7 +183,7 @@ ff_layout_add_mirror(struct pnfs_layout_hdr *lo, return mirror; } -void +static void ff_layout_remove_mirror(struct nfs4_ff_layout_mirror *mirror) { struct inode *inode; -- cgit v1.2.3 From 82714bd1424a88e4bb43813c8a78fbe8f6c5feab Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 25 Aug 2015 08:41:24 -0400 Subject: NFSv4.1/pnfs Improve the packing of struct pnfs_layout_hdr Eliminate a couple of holes in the structure, and move the 2 atomics into the same cacheline. Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.h | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h index d3979dd1037a..4df87ef3dccc 100644 --- a/fs/nfs/pnfs.h +++ b/fs/nfs/pnfs.h @@ -183,15 +183,15 @@ struct pnfs_layoutdriver_type { struct pnfs_layout_hdr { atomic_t plh_refcount; + atomic_t plh_outstanding; /* number of RPCs out */ struct list_head plh_layouts; /* other client layouts */ struct list_head plh_bulk_destroy; struct list_head plh_segs; /* layout segments list */ - nfs4_stateid plh_stateid; - atomic_t plh_outstanding; /* number of RPCs out */ unsigned long plh_block_lgets; /* block LAYOUTGET if >0 */ - u32 plh_barrier; /* ignore lower seqids */ unsigned long plh_retry_timestamp; unsigned long plh_flags; + nfs4_stateid plh_stateid; + u32 plh_barrier; /* ignore lower seqids */ enum pnfs_iomode plh_return_iomode; loff_t plh_lwb; /* last write byte for layoutcommit */ struct rpc_cred *plh_lc_cred; /* layoutcommit cred */ -- cgit v1.2.3 From 540d9864e1c5f8d3ca2ecc919e7d8a47e713ec3f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 25 Aug 2015 11:16:13 -0400 Subject: NFSv4.1/pnfs: Add sanity check for the layout range returned by the server Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 25 ++++++++++++++++++++++++- 1 file changed, 24 insertions(+), 1 deletion(-) diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 3530bb703214..68cc4b169769 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1561,6 +1561,26 @@ out_unlock: } EXPORT_SYMBOL_GPL(pnfs_update_layout); +static bool +pnfs_sanity_check_layout_range(struct pnfs_layout_range *range) +{ + switch (range->iomode) { + case IOMODE_READ: + case IOMODE_RW: + break; + default: + return false; + } + if (range->offset == NFS4_MAX_UINT64) + return false; + if (range->length == 0) + return false; + if (range->length != NFS4_MAX_UINT64 && + range->length > NFS4_MAX_UINT64 - range->offset) + return false; + return true; +} + struct pnfs_layout_segment * pnfs_layout_process(struct nfs4_layoutget *lgp) { @@ -1569,7 +1589,10 @@ pnfs_layout_process(struct nfs4_layoutget *lgp) struct pnfs_layout_segment *lseg; struct inode *ino = lo->plh_inode; LIST_HEAD(free_me); - int status = 0; + int status = -EINVAL; + + if (!pnfs_sanity_check_layout_range(&res->range)) + goto out; /* Inject layout blob into I/O device driver */ lseg = NFS_SERVER(ino)->pnfs_curr_ld->alloc_lseg(lo, res, lgp->gfp_flags); -- cgit v1.2.3 From 03772d2f00313bb22d91b2019cb8e6e91b415653 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 25 Aug 2015 08:54:17 -0400 Subject: NFSv4.1/pnfs: Allow pNFS device drivers to customise layout segment insertion This is needed in order to allow merging of contiguous layout segments, and also to correct the ordering of layouts for those device drivers that don't necessarily want to place the read-write layouts first. Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 59 ++++++++++++++++++++++++++++++++++++++++++++++++++--------- fs/nfs/pnfs.h | 11 +++++++++++ 2 files changed, 61 insertions(+), 9 deletions(-) diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 68cc4b169769..914c1daf08df 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1188,16 +1188,41 @@ pnfs_lseg_range_cmp(const struct pnfs_layout_range *l1, return (int)(l1->iomode == IOMODE_READ) - (int)(l2->iomode == IOMODE_READ); } -static void -pnfs_layout_insert_lseg(struct pnfs_layout_hdr *lo, - struct pnfs_layout_segment *lseg) +static bool +pnfs_lseg_range_is_after(const struct pnfs_layout_range *l1, + const struct pnfs_layout_range *l2) +{ + return pnfs_lseg_range_cmp(l1, l2) > 0; +} + +static bool +pnfs_lseg_no_merge(struct pnfs_layout_segment *lseg, + struct pnfs_layout_segment *old) +{ + return false; +} + +void +pnfs_generic_layout_insert_lseg(struct pnfs_layout_hdr *lo, + struct pnfs_layout_segment *lseg, + bool (*is_after)(const struct pnfs_layout_range *, + const struct pnfs_layout_range *), + bool (*do_merge)(struct pnfs_layout_segment *, + struct pnfs_layout_segment *), + struct list_head *free_me) { - struct pnfs_layout_segment *lp; + struct pnfs_layout_segment *lp, *tmp; dprintk("%s:Begin\n", __func__); - list_for_each_entry(lp, &lo->plh_segs, pls_list) { - if (pnfs_lseg_range_cmp(&lseg->pls_range, &lp->pls_range) > 0) + list_for_each_entry_safe(lp, tmp, &lo->plh_segs, pls_list) { + if (test_bit(NFS_LSEG_VALID, &lp->pls_flags) == 0) + continue; + if (do_merge(lseg, lp)) { + mark_lseg_invalid(lp, free_me); + continue; + } + if (is_after(&lseg->pls_range, &lp->pls_range)) continue; list_add_tail(&lseg->pls_list, &lp->pls_list); dprintk("%s: inserted lseg %p " @@ -1219,6 +1244,24 @@ out: dprintk("%s:Return\n", __func__); } +EXPORT_SYMBOL_GPL(pnfs_generic_layout_insert_lseg); + +static void +pnfs_layout_insert_lseg(struct pnfs_layout_hdr *lo, + struct pnfs_layout_segment *lseg, + struct list_head *free_me) +{ + struct inode *inode = lo->plh_inode; + struct pnfs_layoutdriver_type *ld = NFS_SERVER(inode)->pnfs_curr_ld; + + if (ld->add_lseg != NULL) + ld->add_lseg(lo, lseg, free_me); + else + pnfs_generic_layout_insert_lseg(lo, lseg, + pnfs_lseg_range_is_after, + pnfs_lseg_no_merge, + free_me); +} static struct pnfs_layout_hdr * alloc_init_layout_hdr(struct inode *ino, @@ -1311,8 +1354,6 @@ pnfs_find_lseg(struct pnfs_layout_hdr *lo, ret = pnfs_get_lseg(lseg); break; } - if (lseg->pls_range.offset > range->offset) - break; } dprintk("%s:Return lseg %p ref %d\n", @@ -1637,7 +1678,7 @@ pnfs_layout_process(struct nfs4_layoutget *lgp) clear_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags); pnfs_get_lseg(lseg); - pnfs_layout_insert_lseg(lo, lseg); + pnfs_layout_insert_lseg(lo, lseg, &free_me); if (res->return_on_close) set_bit(NFS_LSEG_ROC, &lseg->pls_flags); diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h index 4df87ef3dccc..869069d8b996 100644 --- a/fs/nfs/pnfs.h +++ b/fs/nfs/pnfs.h @@ -128,6 +128,9 @@ struct pnfs_layoutdriver_type { struct pnfs_layout_segment * (*alloc_lseg) (struct pnfs_layout_hdr *layoutid, struct nfs4_layoutget_res *lgr, gfp_t gfp_flags); void (*free_lseg) (struct pnfs_layout_segment *lseg); + void (*add_lseg) (struct pnfs_layout_hdr *layoutid, + struct pnfs_layout_segment *lseg, + struct list_head *free_me); void (*return_range) (struct pnfs_layout_hdr *lo, struct pnfs_layout_range *range); @@ -285,6 +288,14 @@ struct pnfs_layout_segment *pnfs_update_layout(struct inode *ino, gfp_t gfp_flags); void pnfs_clear_layoutreturn_waitbit(struct pnfs_layout_hdr *lo); +void pnfs_generic_layout_insert_lseg(struct pnfs_layout_hdr *lo, + struct pnfs_layout_segment *lseg, + bool (*is_after)(const struct pnfs_layout_range *lseg_range, + const struct pnfs_layout_range *old), + bool (*do_merge)(struct pnfs_layout_segment *lseg, + struct pnfs_layout_segment *old), + struct list_head *free_me); + void nfs4_deviceid_mark_client_invalid(struct nfs_client *clp); int pnfs_read_done_resend_to_mds(struct nfs_pgio_header *); int pnfs_write_done_resend_to_mds(struct nfs_pgio_header *); -- cgit v1.2.3 From 0762ed2ced40bfe648378ef0e1635cf4d3a6fb76 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 25 Aug 2015 17:38:25 -0400 Subject: NFSv4.1/flexfiles: Allow coalescing of new layout segments and existing ones In order to ensure atomicity of updates, we merge the old layout segments into the new ones, and then invalidate the old ones. Also ensure that we order the list of layout segments so that RO segments are preferred over RW. Signed-off-by: Trond Myklebust --- fs/nfs/flexfilelayout/flexfilelayout.c | 60 ++++++++++++++++++++++++++++++++++ fs/nfs/pnfs.h | 16 +++++++++ 2 files changed, 76 insertions(+) diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index 7fefa8ad9578..4ec624cfcf8b 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -265,6 +265,65 @@ static void _ff_layout_free_lseg(struct nfs4_ff_layout_segment *fls) } } +static bool +ff_lseg_range_is_after(const struct pnfs_layout_range *l1, + const struct pnfs_layout_range *l2) +{ + u64 end1, end2; + + if (l1->iomode != l2->iomode) + return l1->iomode != IOMODE_READ; + end1 = pnfs_calc_offset_end(l1->offset, l1->length); + end2 = pnfs_calc_offset_end(l2->offset, l2->length); + if (end1 < l2->offset) + return false; + if (end2 < l1->offset) + return true; + return l2->offset <= l1->offset; +} + +static bool +ff_lseg_merge(struct pnfs_layout_segment *new, + struct pnfs_layout_segment *old) +{ + u64 new_end, old_end; + + if (new->pls_range.iomode != old->pls_range.iomode) + return false; + old_end = pnfs_calc_offset_end(old->pls_range.offset, + old->pls_range.length); + if (old_end < new->pls_range.offset) + return false; + new_end = pnfs_calc_offset_end(new->pls_range.offset, + new->pls_range.length); + if (new_end < old->pls_range.offset) + return false; + + /* Mergeable: copy info from 'old' to 'new' */ + if (new_end < old_end) + new_end = old_end; + if (new->pls_range.offset < old->pls_range.offset) + new->pls_range.offset = old->pls_range.offset; + new->pls_range.length = pnfs_calc_offset_length(new->pls_range.offset, + new_end); + if (test_bit(NFS_LSEG_ROC, &old->pls_flags)) + set_bit(NFS_LSEG_ROC, &new->pls_flags); + if (test_bit(NFS_LSEG_LAYOUTRETURN, &old->pls_flags)) + set_bit(NFS_LSEG_LAYOUTRETURN, &new->pls_flags); + return true; +} + +static void +ff_layout_add_lseg(struct pnfs_layout_hdr *lo, + struct pnfs_layout_segment *lseg, + struct list_head *free_me) +{ + pnfs_generic_layout_insert_lseg(lo, lseg, + ff_lseg_range_is_after, + ff_lseg_merge, + free_me); +} + static void ff_layout_sort_mirrors(struct nfs4_ff_layout_segment *fls) { int i, j; @@ -2046,6 +2105,7 @@ static struct pnfs_layoutdriver_type flexfilelayout_type = { .free_layout_hdr = ff_layout_free_layout_hdr, .alloc_lseg = ff_layout_alloc_lseg, .free_lseg = ff_layout_free_lseg, + .add_lseg = ff_layout_add_lseg, .pg_read_ops = &ff_layout_pg_read_ops, .pg_write_ops = &ff_layout_pg_write_ops, .get_ds_info = ff_layout_get_ds_info, diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h index 869069d8b996..78c9351ff117 100644 --- a/fs/nfs/pnfs.h +++ b/fs/nfs/pnfs.h @@ -539,6 +539,22 @@ pnfs_use_threshold(struct nfs4_threshold **dst, struct nfs4_threshold *src, nfss->pnfs_curr_ld->id == src->l_type); } +static inline u64 +pnfs_calc_offset_end(u64 offset, u64 len) +{ + if (len == NFS4_MAX_UINT64 || len >= NFS4_MAX_UINT64 - offset) + return NFS4_MAX_UINT64; + return offset + len - 1; +} + +static inline u64 +pnfs_calc_offset_length(u64 offset, u64 end) +{ + if (end == NFS4_MAX_UINT64 || end <= offset) + return NFS4_MAX_UINT64; + return 1 + end - offset; +} + extern unsigned int layoutstats_timer; #ifdef NFS_DEBUG -- cgit v1.2.3 From 19cf6335134d82be792831e14aae9d037d0cb30b Mon Sep 17 00:00:00 2001 From: Peng Tao Date: Wed, 26 Aug 2015 00:13:15 +0800 Subject: nfs42: decode_layoutstats does not need res parameter Reviewed-by: Christoph Hellwig Signed-off-by: Peng Tao Signed-off-by: Trond Myklebust --- fs/nfs/nfs42xdr.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) diff --git a/fs/nfs/nfs42xdr.c b/fs/nfs/nfs42xdr.c index a6bd27da6286..0eb29e14070d 100644 --- a/fs/nfs/nfs42xdr.c +++ b/fs/nfs/nfs42xdr.c @@ -238,8 +238,7 @@ out_overflow: return -EIO; } -static int decode_layoutstats(struct xdr_stream *xdr, - struct nfs42_layoutstat_res *res) +static int decode_layoutstats(struct xdr_stream *xdr) { return decode_op_hdr(xdr, OP_LAYOUTSTATS); } @@ -343,7 +342,7 @@ static int nfs4_xdr_dec_layoutstats(struct rpc_rqst *rqstp, goto out; WARN_ON(res->num_dev > PNFS_LAYOUTSTATS_MAXDEV); for (i = 0; i < res->num_dev; i++) { - status = decode_layoutstats(xdr, res); + status = decode_layoutstats(xdr); if (status) goto out; } -- cgit v1.2.3 From 1090c3bf81ef12eb22383d4b6627d372ce5115d9 Mon Sep 17 00:00:00 2001 From: Peng Tao Date: Wed, 26 Aug 2015 00:13:16 +0800 Subject: nfs42: remove unused declaration Reviewed-by: Christoph Hellwig Signed-off-by: Peng Tao Signed-off-by: Trond Myklebust --- fs/nfs/nfs42.h | 2 -- 1 file changed, 2 deletions(-) diff --git a/fs/nfs/nfs42.h b/fs/nfs/nfs42.h index ff66ae700b89..814c1255f1d2 100644 --- a/fs/nfs/nfs42.h +++ b/fs/nfs/nfs42.h @@ -17,7 +17,5 @@ int nfs42_proc_deallocate(struct file *, loff_t, loff_t); loff_t nfs42_proc_llseek(struct file *, loff_t, int); int nfs42_proc_layoutstats_generic(struct nfs_server *, struct nfs42_layoutstat_data *); -/* nfs4.2xdr.h */ -extern struct rpc_procinfo nfs4_2_procedures[]; #endif /* __LINUX_FS_NFS_NFS4_2_H */ -- cgit v1.2.3 From 0bdb8fa6ecb4a356f26c6874db51b5488706e088 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 27 Aug 2015 19:17:33 -0400 Subject: NFSv4.1/pNFS: pnfs_mark_matching_lsegs_return must notify of layout return It's not sufficient to just mark the layout segment for layout return. We also need to set the NFS_LAYOUT_RETURN_BEFORE_CLOSE flag in the layout header. Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 914c1daf08df..4eec540de9ea 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1717,6 +1717,8 @@ pnfs_mark_matching_lsegs_return(struct pnfs_layout_hdr *lo, lseg->pls_range.length); set_bit(NFS_LSEG_LAYOUTRETURN, &lseg->pls_flags); mark_lseg_invalid(lseg, tmp_list); + set_bit(NFS_LAYOUT_RETURN_BEFORE_CLOSE, + &lo->plh_flags); } } -- cgit v1.2.3 From ae57ca0f4fce219ef34c28f0edc210598c465a4d Mon Sep 17 00:00:00 2001 From: Kinglong Mee Date: Wed, 26 Aug 2015 21:10:55 +0800 Subject: NFS: Check size by inode_newsize_ok in nfs_setattr Set rlimit for NFS's files is useless right now. For local process's rlimit, it should be checked by nfs client. The same, CIFS also call inode_change_ok checking rlimit at its client in cifs_setattr_nounix() and cifs_setattr_unix(). v3, fix bad using of error Signed-off-by: Kinglong Mee Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 18 ++++++++++-------- 1 file changed, 10 insertions(+), 8 deletions(-) diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index e2cc0031decb..99a68bd9c178 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -504,7 +504,7 @@ nfs_setattr(struct dentry *dentry, struct iattr *attr) { struct inode *inode = d_inode(dentry); struct nfs_fattr *fattr; - int error = -ENOMEM; + int error = 0; nfs_inc_stats(inode, NFSIOS_VFSSETATTR); @@ -513,15 +513,14 @@ nfs_setattr(struct dentry *dentry, struct iattr *attr) attr->ia_valid &= ~ATTR_MODE; if (attr->ia_valid & ATTR_SIZE) { - loff_t i_size; - BUG_ON(!S_ISREG(inode->i_mode)); - i_size = i_size_read(inode); - if (attr->ia_size == i_size) + error = inode_newsize_ok(inode, attr->ia_size); + if (error) + return error; + + if (attr->ia_size == i_size_read(inode)) attr->ia_valid &= ~ATTR_SIZE; - else if (attr->ia_size < i_size && IS_SWAPFILE(inode)) - return -ETXTBSY; } /* Optimization: if the end result is no change, don't RPC */ @@ -536,8 +535,11 @@ nfs_setattr(struct dentry *dentry, struct iattr *attr) nfs_sync_inode(inode); fattr = nfs_alloc_fattr(); - if (fattr == NULL) + if (fattr == NULL) { + error = -ENOMEM; goto out; + } + /* * Return any delegations if we're going to change ACLs */ -- cgit v1.2.3 From c5c3fb5f975a8bcc42cd039b83d9a4729ce489bb Mon Sep 17 00:00:00 2001 From: Kinglong Mee Date: Wed, 26 Aug 2015 21:11:39 +0800 Subject: NFS: Make opened as optional argument in _nfs4_do_open Check opened, only update it when non-NULL. It's not needs define an unused value for the opened when calling _nfs4_do_open. v3, same as v2. Signed-off-by: Kinglong Mee Signed-off-by: Trond Myklebust --- fs/nfs/nfs4file.c | 3 +-- fs/nfs/nfs4proc.c | 5 ++--- 2 files changed, 3 insertions(+), 5 deletions(-) diff --git a/fs/nfs/nfs4file.c b/fs/nfs/nfs4file.c index dcd39d4e2efe..43f1590b9240 100644 --- a/fs/nfs/nfs4file.c +++ b/fs/nfs/nfs4file.c @@ -27,7 +27,6 @@ nfs4_file_open(struct inode *inode, struct file *filp) struct inode *dir; unsigned openflags = filp->f_flags; struct iattr attr; - int opened = 0; int err; /* @@ -66,7 +65,7 @@ nfs4_file_open(struct inode *inode, struct file *filp) nfs_sync_inode(inode); } - inode = NFS_PROTO(dir)->open_context(dir, ctx, openflags, &attr, &opened); + inode = NFS_PROTO(dir)->open_context(dir, ctx, openflags, &attr, NULL); if (IS_ERR(inode)) { err = PTR_ERR(inode); switch (err) { diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 6e988fd92f69..4687661bfbdc 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -2452,7 +2452,7 @@ static int _nfs4_do_open(struct inode *dir, nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel); } } - if (opendata->file_created) + if (opened && opendata->file_created) *opened |= FILE_CREATED; if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) { @@ -3562,7 +3562,6 @@ nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr, struct nfs4_label l, *ilabel = NULL; struct nfs_open_context *ctx; struct nfs4_state *state; - int opened = 0; int status = 0; ctx = alloc_nfs_open_context(dentry, FMODE_READ); @@ -3572,7 +3571,7 @@ nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr, ilabel = nfs4_label_init_security(dir, dentry, sattr, &l); sattr->ia_mode &= ~current_umask(); - state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, &opened); + state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL); if (IS_ERR(state)) { status = PTR_ERR(state); goto out; -- cgit v1.2.3 From 5153aacfb8e2744af68e7b84ccd3f02aeefe4f48 Mon Sep 17 00:00:00 2001 From: Kinglong Mee Date: Wed, 26 Aug 2015 21:12:15 +0800 Subject: NFS: Update NFS4_BITMAP_SIZE v4.1/v4.2 have define attributes at word2, nfs client also support security label now. v3, same as v2. Signed-off-by: Kinglong Mee Signed-off-by: Trond Myklebust --- include/uapi/linux/nfs4.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/include/uapi/linux/nfs4.h b/include/uapi/linux/nfs4.h index 2119c7c274d7..2b871e0858d9 100644 --- a/include/uapi/linux/nfs4.h +++ b/include/uapi/linux/nfs4.h @@ -15,7 +15,7 @@ #include -#define NFS4_BITMAP_SIZE 2 +#define NFS4_BITMAP_SIZE 3 #define NFS4_VERIFIER_SIZE 8 #define NFS4_STATEID_SEQID_SIZE 4 #define NFS4_STATEID_OTHER_SIZE 12 -- cgit v1.2.3 From 8c61282ff61c28d5a12bb53f0eaa221d30fd3ae1 Mon Sep 17 00:00:00 2001 From: Kinglong Mee Date: Wed, 26 Aug 2015 21:12:58 +0800 Subject: NFS: Get suppattr_exclcreat when getting server capabilities Create file with attributs as NFS4_CREATE_EXCLUSIVE4_1 mode depends on suppattr_exclcreat attribut. v3, same as v2. Signed-off-by: Kinglong Mee Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 14 +++++++++++++- fs/nfs/nfs4xdr.c | 26 +++++++++++++++++++++----- include/linux/nfs_fs_sb.h | 5 +++++ include/linux/nfs_xdr.h | 2 ++ 4 files changed, 41 insertions(+), 6 deletions(-) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 4687661bfbdc..a6a28d45cca4 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -2893,8 +2893,10 @@ static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync) static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle) { + u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion; struct nfs4_server_caps_arg args = { .fhandle = fhandle, + .bitmask = bitmask, }; struct nfs4_server_caps_res res = {}; struct rpc_message msg = { @@ -2904,10 +2906,18 @@ static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *f }; int status; + bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS | + FATTR4_WORD0_FH_EXPIRE_TYPE | + FATTR4_WORD0_LINK_SUPPORT | + FATTR4_WORD0_SYMLINK_SUPPORT | + FATTR4_WORD0_ACLSUPPORT; + if (minorversion) + bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT; + status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0); if (status == 0) { /* Sanity check the server answers */ - switch (server->nfs_client->cl_minorversion) { + switch (minorversion) { case 0: res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK; res.attr_bitmask[2] = 0; @@ -2960,6 +2970,8 @@ static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *f server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE; server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY; server->cache_consistency_bitmask[2] = 0; + memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask, + sizeof(server->exclcreat_bitmask)); server->acl_bitmask = res.acl_bitmask; server->fh_expire_type = res.fh_expire_type; } diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index c42459e45f62..ad8dde12f23b 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -2582,6 +2582,7 @@ static void nfs4_xdr_enc_server_caps(struct rpc_rqst *req, struct xdr_stream *xdr, struct nfs4_server_caps_arg *args) { + const u32 *bitmask = args->bitmask; struct compound_hdr hdr = { .minorversion = nfs4_xdr_minorversion(&args->seq_args), }; @@ -2589,11 +2590,7 @@ static void nfs4_xdr_enc_server_caps(struct rpc_rqst *req, encode_compound_hdr(xdr, req, &hdr); encode_sequence(xdr, &args->seq_args, &hdr); encode_putfh(xdr, args->fhandle, &hdr); - encode_getattr_one(xdr, FATTR4_WORD0_SUPPORTED_ATTRS| - FATTR4_WORD0_FH_EXPIRE_TYPE| - FATTR4_WORD0_LINK_SUPPORT| - FATTR4_WORD0_SYMLINK_SUPPORT| - FATTR4_WORD0_ACLSUPPORT, &hdr); + encode_getattr_three(xdr, bitmask[0], bitmask[1], bitmask[2], &hdr); encode_nops(&hdr); } @@ -3370,6 +3367,22 @@ out_overflow: return -EIO; } +static int decode_attr_exclcreat_supported(struct xdr_stream *xdr, + uint32_t *bitmap, uint32_t *bitmask) +{ + if (likely(bitmap[2] & FATTR4_WORD2_SUPPATTR_EXCLCREAT)) { + int ret; + ret = decode_attr_bitmap(xdr, bitmask); + if (unlikely(ret < 0)) + return ret; + bitmap[2] &= ~FATTR4_WORD2_SUPPATTR_EXCLCREAT; + } else + bitmask[0] = bitmask[1] = bitmask[2] = 0; + dprintk("%s: bitmask=%08x:%08x:%08x\n", __func__, + bitmask[0], bitmask[1], bitmask[2]); + return 0; +} + static int decode_attr_filehandle(struct xdr_stream *xdr, uint32_t *bitmap, struct nfs_fh *fh) { __be32 *p; @@ -4323,6 +4336,9 @@ static int decode_server_caps(struct xdr_stream *xdr, struct nfs4_server_caps_re goto xdr_error; if ((status = decode_attr_aclsupport(xdr, bitmap, &res->acl_bitmask)) != 0) goto xdr_error; + if ((status = decode_attr_exclcreat_supported(xdr, bitmap, + res->exclcreat_bitmask)) != 0) + goto xdr_error; status = verify_attr_len(xdr, savep, attrlen); xdr_error: dprintk("%s: xdr returned %d!\n", __func__, -status); diff --git a/include/linux/nfs_fs_sb.h b/include/linux/nfs_fs_sb.h index 20bc8e51b161..570a7df2775b 100644 --- a/include/linux/nfs_fs_sb.h +++ b/include/linux/nfs_fs_sb.h @@ -173,6 +173,11 @@ struct nfs_server { set of attributes supported on this filesystem excluding the label support bit. */ + u32 exclcreat_bitmask[3]; + /* V4 bitmask representing the + set of attributes supported + on this filesystem for the + exclusive create. */ u32 cache_consistency_bitmask[3]; /* V4 bitmask representing the subset of change attribute, size, ctime diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index b9b530409ff7..0d7c832ec415 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -1057,11 +1057,13 @@ struct nfs4_statfs_res { struct nfs4_server_caps_arg { struct nfs4_sequence_args seq_args; struct nfs_fh *fhandle; + const u32 * bitmask; }; struct nfs4_server_caps_res { struct nfs4_sequence_res seq_res; u32 attr_bitmask[3]; + u32 exclcreat_bitmask[3]; u32 acl_bitmask; u32 has_links; u32 has_symlinks; -- cgit v1.2.3 From 5334c5bdac926c5f8d89729beccb46fe88eda9e7 Mon Sep 17 00:00:00 2001 From: Kinglong Mee Date: Wed, 26 Aug 2015 21:13:37 +0800 Subject: NFS: Send attributes in OPEN request for NFS4_CREATE_EXCLUSIVE4_1 Client sends a SETATTR request after OPEN for updating attributes. For create file with S_ISGID is set, the S_ISGID in SETATTR will be ignored at nfs server as chmod of no PERMISSION. v3, same as v2. Signed-off-by: Kinglong Mee Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 18 ++++++++++++++---- fs/nfs/nfs4xdr.c | 26 ++++++++++++++++++-------- include/linux/nfs_xdr.h | 2 +- 3 files changed, 33 insertions(+), 13 deletions(-) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index a6a28d45cca4..2923abf2fc0c 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -2307,15 +2307,25 @@ static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *st * fields corresponding to attributes that were used to store the verifier. * Make sure we clobber those fields in the later setattr call */ -static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, struct iattr *sattr) +static inline void nfs4_exclusive_attrset(struct nfs4_opendata *opendata, + struct iattr *sattr, struct nfs4_label **label) { - if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_ACCESS) && + const u32 *attrset = opendata->o_res.attrset; + + if ((attrset[1] & FATTR4_WORD1_TIME_ACCESS) && !(sattr->ia_valid & ATTR_ATIME_SET)) sattr->ia_valid |= ATTR_ATIME; - if ((opendata->o_res.attrset[1] & FATTR4_WORD1_TIME_MODIFY) && + if ((attrset[1] & FATTR4_WORD1_TIME_MODIFY) && !(sattr->ia_valid & ATTR_MTIME_SET)) sattr->ia_valid |= ATTR_MTIME; + + /* Except MODE, it seems harmless of setting twice. */ + if ((attrset[1] & FATTR4_WORD1_MODE)) + sattr->ia_valid &= ~ATTR_MODE; + + if (attrset[2] & FATTR4_WORD2_SECURITY_LABEL) + *label = NULL; } static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata, @@ -2440,7 +2450,7 @@ static int _nfs4_do_open(struct inode *dir, if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) && (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) { - nfs4_exclusive_attrset(opendata, sattr); + nfs4_exclusive_attrset(opendata, sattr, &label); nfs_fattr_init(opendata->o_res.f_attr); status = nfs4_do_setattr(state->inode, cred, diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index ad8dde12f23b..a7be571c1666 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -1001,7 +1001,8 @@ static void encode_nfs4_verifier(struct xdr_stream *xdr, const nfs4_verifier *ve static void encode_attrs(struct xdr_stream *xdr, const struct iattr *iap, const struct nfs4_label *label, - const struct nfs_server *server) + const struct nfs_server *server, + bool excl_check) { char owner_name[IDMAP_NAMESZ]; char owner_group[IDMAP_NAMESZ]; @@ -1067,6 +1068,17 @@ static void encode_attrs(struct xdr_stream *xdr, const struct iattr *iap, bmval[1] |= FATTR4_WORD1_TIME_MODIFY_SET; len += 4; } + + if (excl_check) { + const u32 *excl_bmval = server->exclcreat_bitmask; + bmval[0] &= excl_bmval[0]; + bmval[1] &= excl_bmval[1]; + bmval[2] &= excl_bmval[2]; + + if (!(excl_bmval[2] & FATTR4_WORD2_SECURITY_LABEL)) + label = NULL; + } + if (label) { len += 4 + 4 + 4 + (XDR_QUADLEN(label->len) << 2); bmval[2] |= FATTR4_WORD2_SECURITY_LABEL; @@ -1170,7 +1182,7 @@ static void encode_create(struct xdr_stream *xdr, const struct nfs4_create_arg * } encode_string(xdr, create->name->len, create->name->name); - encode_attrs(xdr, create->attrs, create->label, create->server); + encode_attrs(xdr, create->attrs, create->label, create->server, false); } static void encode_getattr_one(struct xdr_stream *xdr, uint32_t bitmap, struct compound_hdr *hdr) @@ -1384,18 +1396,17 @@ static inline void encode_openhdr(struct xdr_stream *xdr, const struct nfs_opena static inline void encode_createmode(struct xdr_stream *xdr, const struct nfs_openargs *arg) { - struct iattr dummy; __be32 *p; p = reserve_space(xdr, 4); switch(arg->createmode) { case NFS4_CREATE_UNCHECKED: *p = cpu_to_be32(NFS4_CREATE_UNCHECKED); - encode_attrs(xdr, arg->u.attrs, arg->label, arg->server); + encode_attrs(xdr, arg->u.attrs, arg->label, arg->server, false); break; case NFS4_CREATE_GUARDED: *p = cpu_to_be32(NFS4_CREATE_GUARDED); - encode_attrs(xdr, arg->u.attrs, arg->label, arg->server); + encode_attrs(xdr, arg->u.attrs, arg->label, arg->server, false); break; case NFS4_CREATE_EXCLUSIVE: *p = cpu_to_be32(NFS4_CREATE_EXCLUSIVE); @@ -1404,8 +1415,7 @@ static inline void encode_createmode(struct xdr_stream *xdr, const struct nfs_op case NFS4_CREATE_EXCLUSIVE4_1: *p = cpu_to_be32(NFS4_CREATE_EXCLUSIVE4_1); encode_nfs4_verifier(xdr, &arg->u.verifier); - dummy.ia_valid = 0; - encode_attrs(xdr, &dummy, arg->label, arg->server); + encode_attrs(xdr, arg->u.attrs, arg->label, arg->server, true); } } @@ -1661,7 +1671,7 @@ static void encode_setattr(struct xdr_stream *xdr, const struct nfs_setattrargs { encode_op_hdr(xdr, OP_SETATTR, decode_setattr_maxsz, hdr); encode_nfs4_stateid(xdr, &arg->stateid); - encode_attrs(xdr, arg->iap, arg->label, server); + encode_attrs(xdr, arg->iap, arg->label, server, false); } static void encode_setclientid(struct xdr_stream *xdr, const struct nfs4_setclientid *setclientid, struct compound_hdr *hdr) diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index 0d7c832ec415..b4392d86d157 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -379,7 +379,7 @@ struct nfs_openargs { struct stateowner_id id; union { struct { - struct iattr * attrs; /* UNCHECKED, GUARDED */ + struct iattr * attrs; /* UNCHECKED, GUARDED, EXCLUSIVE4_1 */ nfs4_verifier verifier; /* EXCLUSIVE */ }; nfs4_stateid delegation; /* CLAIM_DELEGATE_CUR */ -- cgit v1.2.3 From d13549074cf066d6d5bb29903d044beffea342d3 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 27 Aug 2015 20:37:39 -0400 Subject: NFSv4.1/flexfiles: Fix a protocol error in layoutreturn According to the flexfiles protocol, the layoutreturn should specify an array of errors in the following format: struct ff_ioerr4 { offset4 ffie_offset; length4 ffie_length; stateid4 ffie_stateid; device_error4 ffie_errors<>; }; This patch fixes up the code to ensure that our ffie_errors is indeed encoded as an array (albeit with only a single entry). Reported-by: Tom Haynes Cc: stable@vger.kernel.org Signed-off-by: Trond Myklebust --- fs/nfs/flexfilelayout/flexfilelayoutdev.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) diff --git a/fs/nfs/flexfilelayout/flexfilelayoutdev.c b/fs/nfs/flexfilelayout/flexfilelayoutdev.c index f13e1969eedd..b28fa4cbea52 100644 --- a/fs/nfs/flexfilelayout/flexfilelayoutdev.c +++ b/fs/nfs/flexfilelayout/flexfilelayoutdev.c @@ -500,16 +500,19 @@ int ff_layout_encode_ds_ioerr(struct nfs4_flexfile_layout *flo, range->offset, range->length)) continue; /* offset(8) + length(8) + stateid(NFS4_STATEID_SIZE) - * + deviceid(NFS4_DEVICEID4_SIZE) + status(4) + opnum(4) + * + array length + deviceid(NFS4_DEVICEID4_SIZE) + * + status(4) + opnum(4) */ p = xdr_reserve_space(xdr, - 24 + NFS4_STATEID_SIZE + NFS4_DEVICEID4_SIZE); + 28 + NFS4_STATEID_SIZE + NFS4_DEVICEID4_SIZE); if (unlikely(!p)) return -ENOBUFS; p = xdr_encode_hyper(p, err->offset); p = xdr_encode_hyper(p, err->length); p = xdr_encode_opaque_fixed(p, &err->stateid, NFS4_STATEID_SIZE); + /* Encode 1 error */ + *p++ = cpu_to_be32(1); p = xdr_encode_opaque_fixed(p, &err->deviceid, NFS4_DEVICEID4_SIZE); *p++ = cpu_to_be32(err->status); -- cgit v1.2.3 From 6669cb8bed02ec1b60e80f2e1e317afc28544207 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 27 Aug 2015 20:43:20 -0400 Subject: NFSv4.1/pnfs: Ensure layoutreturn reserves space for the opaque payload The "FIXME" is outdated. Flexfiles does add a payload. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4xdr.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index a7be571c1666..ff4784c54e04 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -400,7 +400,8 @@ static int nfs4_stat_to_errno(int); #define decode_layoutcommit_maxsz (op_decode_hdr_maxsz + 3) #define encode_layoutreturn_maxsz (8 + op_encode_hdr_maxsz + \ encode_stateid_maxsz + \ - 1 /* FIXME: opaque lrf_body always empty at the moment */) + 1 + \ + XDR_QUADLEN(NFS4_OPAQUE_LIMIT)) #define decode_layoutreturn_maxsz (op_decode_hdr_maxsz + \ 1 + decode_stateid_maxsz) #define encode_secinfo_no_name_maxsz (op_encode_hdr_maxsz + 1) -- cgit v1.2.3 From 0c78789e3a030615c6650fde89546cadf40ec2cc Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 29 Aug 2015 13:36:30 -0700 Subject: SUNRPC: xs_reset_transport must mark the connection as disconnected In case the reconnection attempt fails. Cc: stable@vger.kernel.org Signed-off-by: Trond Myklebust --- net/sunrpc/xprtsock.c | 1 + 1 file changed, 1 insertion(+) diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c index 723df7cb096d..214ca9dfb14e 100644 --- a/net/sunrpc/xprtsock.c +++ b/net/sunrpc/xprtsock.c @@ -829,6 +829,7 @@ static void xs_reset_transport(struct sock_xprt *transport) sk->sk_user_data = NULL; xs_restore_old_callbacks(transport, sk); + xprt_clear_connected(xprt); write_unlock_bh(&sk->sk_callback_lock); xs_sock_reset_connection_flags(xprt); -- cgit v1.2.3 From 099392048cd443349c50310f7fdc96070e40f4e7 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 29 Aug 2015 19:11:21 -0700 Subject: SUNRPC: Prevent SYN+SYNACK+RST storms Add a shutdown() call before we release the socket in order to ensure the reset is sent before we try to reconnect. Signed-off-by: Trond Myklebust --- net/sunrpc/xprtsock.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c index 214ca9dfb14e..7be90bc1a7c2 100644 --- a/net/sunrpc/xprtsock.c +++ b/net/sunrpc/xprtsock.c @@ -822,6 +822,8 @@ static void xs_reset_transport(struct sock_xprt *transport) if (atomic_read(&transport->xprt.swapper)) sk_clear_memalloc(sk); + kernel_sock_shutdown(sock, SHUT_RDWR); + write_lock_bh(&sk->sk_callback_lock); transport->inet = NULL; transport->sock = NULL; -- cgit v1.2.3 From 90816d1ddacfb5a8b783f67e2c1a1bc77dc50ff4 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sun, 30 Aug 2015 09:53:06 -0700 Subject: NFSv4.1/flexfiles: Don't mark the entire deviceid as bad for file errors If the file was fenced and/or has been deleted on the DS, then we want to retry pNFS after a layoutreturn with error report. If the server cannot fix the problem, then we rely on it to tell us so in the response to the LAYOUTGET. Signed-off-by: Trond Myklebust --- fs/nfs/flexfilelayout/flexfilelayout.c | 24 ++++++++++++++++-------- 1 file changed, 16 insertions(+), 8 deletions(-) diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index 4ec624cfcf8b..61ccf1122494 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -1075,18 +1075,26 @@ static int ff_layout_async_handle_error_v3(struct rpc_task *task, if (task->tk_status >= 0) return 0; - if (task->tk_status != -EJUKEBOX) { + switch (task->tk_status) { + /* File access problems. Don't mark the device as unavailable */ + case -EACCES: + case -ESTALE: + case -EISDIR: + case -EBADHANDLE: + case -ELOOP: + case -ENOSPC: + break; + case -EJUKEBOX: + nfs_inc_stats(lseg->pls_layout->plh_inode, NFSIOS_DELAY); + goto out_retry; + default: dprintk("%s DS connection error %d\n", __func__, task->tk_status); nfs4_mark_deviceid_unavailable(devid); - if (ff_layout_has_available_ds(lseg)) - return -NFS4ERR_RESET_TO_PNFS; - else - return -NFS4ERR_RESET_TO_MDS; } - - if (task->tk_status == -EJUKEBOX) - nfs_inc_stats(lseg->pls_layout->plh_inode, NFSIOS_DELAY); + /* FIXME: Need to prevent infinite looping here. */ + return -NFS4ERR_RESET_TO_PNFS; +out_retry: task->tk_status = 0; rpc_restart_call(task); rpc_delay(task, NFS_JUKEBOX_RETRY_TIME); -- cgit v1.2.3 From 4a1e2feb9d246775dee0f78ed5b18826bae2b1c5 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sun, 30 Aug 2015 18:37:59 -0700 Subject: NFSv4.1: Fix a protocol issue with CLOSE stateids According to RFC5661 Section 18.2.4, CLOSE is supposed to return the zero stateid. This means that nfs_clear_open_stateid_locked() cannot assume that the result stateid will always match the 'other' field of the existing open stateid when trying to determine a race with a parallel OPEN. Instead, we look at the argument, and check for matches. Cc: stable@vger.kernel.org # v4.0+ Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 15 ++++++++++----- 1 file changed, 10 insertions(+), 5 deletions(-) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 2923abf2fc0c..366b81c185f6 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -1231,6 +1231,7 @@ static void nfs_resync_open_stateid_locked(struct nfs4_state *state) } static void nfs_clear_open_stateid_locked(struct nfs4_state *state, + nfs4_stateid *arg_stateid, nfs4_stateid *stateid, fmode_t fmode) { clear_bit(NFS_O_RDWR_STATE, &state->flags); @@ -1249,8 +1250,9 @@ static void nfs_clear_open_stateid_locked(struct nfs4_state *state, if (stateid == NULL) return; /* Handle races with OPEN */ - if (!nfs4_stateid_match_other(stateid, &state->open_stateid) || - !nfs4_stateid_is_newer(stateid, &state->open_stateid)) { + if (!nfs4_stateid_match_other(arg_stateid, &state->open_stateid) || + (nfs4_stateid_match_other(stateid, &state->open_stateid) && + !nfs4_stateid_is_newer(stateid, &state->open_stateid))) { nfs_resync_open_stateid_locked(state); return; } @@ -1259,10 +1261,12 @@ static void nfs_clear_open_stateid_locked(struct nfs4_state *state, nfs4_stateid_copy(&state->open_stateid, stateid); } -static void nfs_clear_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, fmode_t fmode) +static void nfs_clear_open_stateid(struct nfs4_state *state, + nfs4_stateid *arg_stateid, + nfs4_stateid *stateid, fmode_t fmode) { write_seqlock(&state->seqlock); - nfs_clear_open_stateid_locked(state, stateid, fmode); + nfs_clear_open_stateid_locked(state, arg_stateid, stateid, fmode); write_sequnlock(&state->seqlock); if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags)) nfs4_schedule_state_manager(state->owner->so_server->nfs_client); @@ -2707,7 +2711,8 @@ static void nfs4_close_done(struct rpc_task *task, void *data) goto out_release; } } - nfs_clear_open_stateid(state, res_stateid, calldata->arg.fmode); + nfs_clear_open_stateid(state, &calldata->arg.stateid, + res_stateid, calldata->arg.fmode); out_release: nfs_release_seqid(calldata->arg.seqid); nfs_refresh_inode(calldata->inode, calldata->res.fattr); -- cgit v1.2.3 From 4ae93560b11cb2b57ea5732d442458694ab0c168 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 31 Aug 2015 01:25:11 -0700 Subject: NFSv4.1/pnfs: Don't ask for a read layout for an empty file. Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 4eec540de9ea..c4f918eca3d2 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1494,6 +1494,9 @@ pnfs_update_layout(struct inode *ino, if (!pnfs_enabled_sb(NFS_SERVER(ino))) goto out; + if (iomode == IOMODE_READ && i_size_read(ino) == 0) + goto out; + if (pnfs_within_mdsthreshold(ctx, ino, iomode)) goto out; -- cgit v1.2.3 From 21b874c873b5019db8bb4b4f6aa929c4bac0a398 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 31 Aug 2015 01:19:22 -0700 Subject: NFSv4.1/pnfs: Handle LAYOUTGET return values correctly According to RFC5661 section 18.43.3, if the server cannot satisfy the loga_minlength argument to LAYOUTGET, there are 2 cases: 1) If loga_minlength == 0, it returns NFS4ERR_LAYOUTTRYLATER 2) If loga_minlength != 0, it returns NFS4ERR_BADLAYOUT Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 15 ++++++++++++++- 1 file changed, 14 insertions(+), 1 deletion(-) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 366b81c185f6..51c7164abd1a 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -7780,11 +7780,20 @@ static void nfs4_layoutget_done(struct rpc_task *task, void *calldata) switch (task->tk_status) { case 0: goto out; + /* + * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of + * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3). + */ + case -NFS4ERR_BADLAYOUT: + goto out_overflow; /* * NFS4ERR_LAYOUTTRYLATER is a conflict with another client - * (or clients) writing to the same RAID stripe + * (or clients) writing to the same RAID stripe except when + * the minlength argument is 0 (see RFC5661 section 18.43.3). */ case -NFS4ERR_LAYOUTTRYLATER: + if (lgp->args.minlength == 0) + goto out_overflow; /* * NFS4ERR_RECALLCONFLICT is when conflict with self (must recall * existing layout before getting a new one). @@ -7840,6 +7849,10 @@ static void nfs4_layoutget_done(struct rpc_task *task, void *calldata) rpc_restart_call_prepare(task); out: dprintk("<-- %s\n", __func__); + return; +out_overflow: + task->tk_status = -EOVERFLOW; + goto out; } static size_t max_response_pages(struct nfs_server *server) -- cgit v1.2.3 From 2d89a1d3c9ff8ceb115f001e66cff9788338ca47 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 31 Aug 2015 02:05:47 -0700 Subject: NFSv4.1/pNFS: Don't request a minimal read layout beyond the end of file If we have a read layout, then sanity check the minimal layout length so that it does not extend beyond the end of file. Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index c4f918eca3d2..ba1246433794 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -868,6 +868,7 @@ send_layoutget(struct pnfs_layout_hdr *lo, struct nfs_server *server = NFS_SERVER(ino); struct nfs4_layoutget *lgp; struct pnfs_layout_segment *lseg; + loff_t i_size; dprintk("--> %s\n", __func__); @@ -875,9 +876,17 @@ send_layoutget(struct pnfs_layout_hdr *lo, if (lgp == NULL) return NULL; + i_size = i_size_read(ino); + lgp->args.minlength = PAGE_CACHE_SIZE; if (lgp->args.minlength > range->length) lgp->args.minlength = range->length; + if (range->iomode == IOMODE_READ) { + if (range->offset >= i_size) + lgp->args.minlength = 0; + else if (i_size - range->offset < lgp->args.minlength) + lgp->args.minlength = i_size - range->offset; + } lgp->args.maxcount = PNFS_LAYOUT_MAXSIZE; lgp->args.range = *range; lgp->args.type = server->pnfs_curr_ld->id; -- cgit v1.2.3 From 972398fa0a5f47c6ee0bde4d6d24b29f90ec888d Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 1 Sep 2015 12:03:56 -0700 Subject: NFSv4.1/flexfiles: Fix freeing of mirrors Mirrors are now shared objects, so we should not be freeing them directly inside ff_layout_free_lseg(). We should already be doing the right thing in _ff_layout_free_lseg(), so just let it handle things. Also ensure that ff_layout_free_mirror() frees the RPC credential if it is set. Fixes: 28a0d72c6867 ("Add refcounting to struct nfs4_ff_layout_mirror") Signed-off-by: Trond Myklebust --- fs/nfs/flexfilelayout/flexfilelayout.c | 14 ++------------ 1 file changed, 2 insertions(+), 12 deletions(-) diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index 61ccf1122494..ee8e7013454f 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -213,6 +213,8 @@ static void ff_layout_free_mirror(struct nfs4_ff_layout_mirror *mirror) { ff_layout_remove_mirror(mirror); kfree(mirror->fh_versions); + if (mirror->cred) + put_rpccred(mirror->cred); nfs4_ff_layout_put_deviceid(mirror->mirror_ds); kfree(mirror); } @@ -525,21 +527,9 @@ static void ff_layout_free_lseg(struct pnfs_layout_segment *lseg) { struct nfs4_ff_layout_segment *fls = FF_LAYOUT_LSEG(lseg); - int i; dprintk("--> %s\n", __func__); - for (i = 0; i < fls->mirror_array_cnt; i++) { - if (fls->mirror_array[i]) { - nfs4_ff_layout_put_deviceid(fls->mirror_array[i]->mirror_ds); - fls->mirror_array[i]->mirror_ds = NULL; - if (fls->mirror_array[i]->cred) { - put_rpccred(fls->mirror_array[i]->cred); - fls->mirror_array[i]->cred = NULL; - } - } - } - if (lseg->pls_range.iomode == IOMODE_RW) { struct nfs4_flexfile_layout *ffl; struct inode *inode; -- cgit v1.2.3 From 388ef16640cefd202daa723fba02e7c0266f8454 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 1 Sep 2015 03:31:33 -0700 Subject: NFSv4.1/flexfiles: Fix incorrect usage of pnfs_generic_mark_devid_invalid() Unlike the files layout, flexfiles does not test for the NFS_DEVICEID_INVALID flag. Instead it relies on NFS_DEVICEID_UNAVAILABLE. Fix is to replace with nfs4_mark_deviceid_unavailable(). Signed-off-by: Trond Myklebust --- fs/nfs/flexfilelayout/flexfilelayoutdev.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/fs/nfs/flexfilelayout/flexfilelayoutdev.c b/fs/nfs/flexfilelayout/flexfilelayoutdev.c index b28fa4cbea52..883d35c86778 100644 --- a/fs/nfs/flexfilelayout/flexfilelayoutdev.c +++ b/fs/nfs/flexfilelayout/flexfilelayoutdev.c @@ -344,7 +344,7 @@ nfs4_ff_layout_select_ds_fh(struct pnfs_layout_segment *lseg, u32 mirror_idx) __func__, mirror_idx); if (mirror && mirror->mirror_ds) { devid = &mirror->mirror_ds->id_node; - pnfs_generic_mark_devid_invalid(devid); + nfs4_mark_deviceid_unavailable(devid); } goto out; } @@ -374,7 +374,7 @@ nfs4_ff_layout_prepare_ds(struct pnfs_layout_segment *lseg, u32 ds_idx, __func__, ds_idx); if (mirror && mirror->mirror_ds) { devid = &mirror->mirror_ds->id_node; - pnfs_generic_mark_devid_invalid(devid); + nfs4_mark_deviceid_unavailable(devid); } goto out; } -- cgit v1.2.3 From 81d6dc8b3431d298abaef11d8bc64646fc691618 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 1 Sep 2015 02:49:44 -0700 Subject: NFSv4.1/flexfiles: RW layouts are valid only if all mirrors are valid Unlike read layouts, the writeable layout cannot fall back to using only one of the mirrors. It need to write to all of them. Signed-off-by: Trond Myklebust --- fs/nfs/flexfilelayout/flexfilelayoutdev.c | 30 ++++++++++++++++++++++++++++-- 1 file changed, 28 insertions(+), 2 deletions(-) diff --git a/fs/nfs/flexfilelayout/flexfilelayoutdev.c b/fs/nfs/flexfilelayout/flexfilelayoutdev.c index 883d35c86778..b6c21e9fa002 100644 --- a/fs/nfs/flexfilelayout/flexfilelayoutdev.c +++ b/fs/nfs/flexfilelayout/flexfilelayoutdev.c @@ -528,11 +528,11 @@ int ff_layout_encode_ds_ioerr(struct nfs4_flexfile_layout *flo, return 0; } -bool ff_layout_has_available_ds(struct pnfs_layout_segment *lseg) +static bool ff_read_layout_has_available_ds(struct pnfs_layout_segment *lseg) { struct nfs4_ff_layout_mirror *mirror; struct nfs4_deviceid_node *devid; - int idx; + u32 idx; for (idx = 0; idx < FF_LAYOUT_MIRROR_COUNT(lseg); idx++) { mirror = FF_LAYOUT_COMP(lseg, idx); @@ -546,6 +546,32 @@ bool ff_layout_has_available_ds(struct pnfs_layout_segment *lseg) return false; } +static bool ff_rw_layout_has_available_ds(struct pnfs_layout_segment *lseg) +{ + struct nfs4_ff_layout_mirror *mirror; + struct nfs4_deviceid_node *devid; + u32 idx; + + for (idx = 0; idx < FF_LAYOUT_MIRROR_COUNT(lseg); idx++) { + mirror = FF_LAYOUT_COMP(lseg, idx); + if (!mirror || !mirror->mirror_ds) + return false; + devid = &mirror->mirror_ds->id_node; + if (ff_layout_test_devid_unavailable(devid)) + return false; + } + + return FF_LAYOUT_MIRROR_COUNT(lseg) != 0; +} + +bool ff_layout_has_available_ds(struct pnfs_layout_segment *lseg) +{ + if (lseg->pls_range.iomode == IOMODE_READ) + return ff_read_layout_has_available_ds(lseg); + /* Note: RW layout needs all mirrors available */ + return ff_rw_layout_has_available_ds(lseg); +} + module_param(dataserver_retrans, uint, 0644); MODULE_PARM_DESC(dataserver_retrans, "The number of times the NFSv4.1 client " "retries a request before it attempts further " -- cgit v1.2.3 From 889d94d49a3cbcbc8c8d07208549fa614d33da76 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 1 Sep 2015 00:58:24 -0700 Subject: NFSv4.1/flexfiles: Mark layout for return if the mirrors are invalid If a read-write layout has an invalid mirror, then we should mark it as invalid, and return it. If a read-only layout has an invalid mirror, then mark it as invalid and check if there is still at least one valid mirror before we return it. Note: Also fix incorrect use of pnfs_generic_mark_devid_invalid(). We really want nfs4_mark_deviceid_unavailable(). Signed-off-by: Trond Myklebust --- fs/nfs/flexfilelayout/flexfilelayoutdev.c | 45 ++++++++++++++++++++----------- 1 file changed, 30 insertions(+), 15 deletions(-) diff --git a/fs/nfs/flexfilelayout/flexfilelayoutdev.c b/fs/nfs/flexfilelayout/flexfilelayoutdev.c index b6c21e9fa002..e125e55de86d 100644 --- a/fs/nfs/flexfilelayout/flexfilelayoutdev.c +++ b/fs/nfs/flexfilelayout/flexfilelayoutdev.c @@ -172,6 +172,32 @@ out_err: return NULL; } +static void ff_layout_mark_devid_invalid(struct pnfs_layout_segment *lseg, + struct nfs4_deviceid_node *devid) +{ + nfs4_mark_deviceid_unavailable(devid); + if (!ff_layout_has_available_ds(lseg)) + pnfs_error_mark_layout_for_return(lseg->pls_layout->plh_inode, + lseg); +} + +static bool ff_layout_mirror_valid(struct pnfs_layout_segment *lseg, + struct nfs4_ff_layout_mirror *mirror) +{ + if (mirror == NULL || mirror->mirror_ds == NULL) { + pnfs_error_mark_layout_for_return(lseg->pls_layout->plh_inode, + lseg); + return false; + } + if (mirror->mirror_ds->ds == NULL) { + struct nfs4_deviceid_node *devid; + devid = &mirror->mirror_ds->id_node; + ff_layout_mark_devid_invalid(lseg, devid); + return false; + } + return true; +} + static u64 end_offset(u64 start, u64 len) { @@ -336,16 +362,10 @@ nfs4_ff_layout_select_ds_fh(struct pnfs_layout_segment *lseg, u32 mirror_idx) { struct nfs4_ff_layout_mirror *mirror = FF_LAYOUT_COMP(lseg, mirror_idx); struct nfs_fh *fh = NULL; - struct nfs4_deviceid_node *devid; - if (mirror == NULL || mirror->mirror_ds == NULL || - mirror->mirror_ds->ds == NULL) { - printk(KERN_ERR "NFS: %s: No data server for mirror offset index %d\n", + if (!ff_layout_mirror_valid(lseg, mirror)) { + pr_err_ratelimited("NFS: %s: No data server for mirror offset index %d\n", __func__, mirror_idx); - if (mirror && mirror->mirror_ds) { - devid = &mirror->mirror_ds->id_node; - nfs4_mark_deviceid_unavailable(devid); - } goto out; } @@ -368,14 +388,9 @@ nfs4_ff_layout_prepare_ds(struct pnfs_layout_segment *lseg, u32 ds_idx, unsigned int max_payload; rpc_authflavor_t flavor; - if (mirror == NULL || mirror->mirror_ds == NULL || - mirror->mirror_ds->ds == NULL) { - printk(KERN_ERR "NFS: %s: No data server for offset index %d\n", + if (!ff_layout_mirror_valid(lseg, mirror)) { + pr_err_ratelimited("NFS: %s: No data server for offset index %d\n", __func__, ds_idx); - if (mirror && mirror->mirror_ds) { - devid = &mirror->mirror_ds->id_node; - nfs4_mark_deviceid_unavailable(devid); - } goto out; } -- cgit v1.2.3 From 4a70316caef7d158445e672e146eb9f1b8c1aeee Mon Sep 17 00:00:00 2001 From: Kinglong Mee Date: Mon, 31 Aug 2015 10:53:33 +0800 Subject: nfs: Fix truncated client owner id without proto type The length of "Linux NFSv4.0 " is 14, not 10. Without this patch, I get a truncated client owner id as, "Linux NFSv4.0 ::1/::1" With this patch, "Linux NFSv4.0 ::1/::1 tcp" Fixes: a319268891 ("nfs: make nfs4_init_nonuniform_client_string use a dynamically allocated buffer") Signed-off-by: Kinglong Mee Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 51c7164abd1a..3f73539579e5 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -5020,7 +5020,7 @@ nfs4_init_nonuniform_client_string(struct nfs_client *clp) return 0; retry: rcu_read_lock(); - len = 10 + strlen(clp->cl_ipaddr) + 1 + + len = 14 + strlen(clp->cl_ipaddr) + 1 + strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) + 1 + strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)) + -- cgit v1.2.3 From 4a3e5779cf6c6d557682b499c2190ad04c80c6fd Mon Sep 17 00:00:00 2001 From: Kinglong Mee Date: Mon, 31 Aug 2015 10:53:43 +0800 Subject: nfs: Remove unneeded checking of the return value from scnprintf The return value from scnprintf always less than the buffer length. So, result >= len always false. This patch removes those checking. int vscnprintf(char *buf, size_t size, const char *fmt, va_list args) { int i; i = vsnprintf(buf, size, fmt, args); if (likely(i < size)) return i; if (size != 0) return size - 1; return 0; } Signed-off-by: Kinglong Mee Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 19 +------------------ 1 file changed, 1 insertion(+), 18 deletions(-) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 3f73539579e5..693b903b48bd 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -5014,11 +5014,10 @@ nfs4_init_nonuniform_client_string(struct nfs_client *clp) int result; size_t len; char *str; - bool retried = false; if (clp->cl_owner_id != NULL) return 0; -retry: + rcu_read_lock(); len = 14 + strlen(clp->cl_ipaddr) + 1 + strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) + @@ -5046,14 +5045,6 @@ retry: rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO)); rcu_read_unlock(); - /* Did something change? */ - if (result >= len) { - kfree(str); - if (retried) - return -EINVAL; - retried = true; - goto retry; - } clp->cl_owner_id = str; return 0; } @@ -5085,10 +5076,6 @@ nfs4_init_uniquifier_client_string(struct nfs_client *clp) clp->rpc_ops->version, clp->cl_minorversion, nfs4_client_id_uniquifier, clp->cl_rpcclient->cl_nodename); - if (result >= len) { - kfree(str); - return -EINVAL; - } clp->cl_owner_id = str; return 0; } @@ -5124,10 +5111,6 @@ nfs4_init_uniform_client_string(struct nfs_client *clp) result = scnprintf(str, len, "Linux NFSv%u.%u %s", clp->rpc_ops->version, clp->cl_minorversion, clp->cl_rpcclient->cl_nodename); - if (result >= len) { - kfree(str); - return -EINVAL; - } clp->cl_owner_id = str; return 0; } -- cgit v1.2.3 From f95c03b2d5fb6d56c0d7ec21970329b49b657fed Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 2 Sep 2015 15:15:11 -0700 Subject: NFSv4.1/flexfiles: Mark the layout for return in ff_layout_io_track_ds_error() When I/O cannot complete due to a fatal error on the DS, ensure that we invalidate the corresponding layout segment and return it. Signed-off-by: Trond Myklebust --- fs/nfs/flexfilelayout/flexfilelayout.c | 10 +--------- 1 file changed, 1 insertion(+), 9 deletions(-) diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index ee8e7013454f..3f073a7de870 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -1148,6 +1148,7 @@ static void ff_layout_io_track_ds_error(struct pnfs_layout_segment *lseg, err = ff_layout_track_ds_error(FF_LAYOUT_FROM_HDR(lseg->pls_layout), mirror, offset, length, status, opnum, GFP_NOIO); + pnfs_error_mark_layout_for_return(lseg->pls_layout->plh_inode, lseg); dprintk("%s: err %d op %d status %u\n", __func__, err, opnum, status); } @@ -1156,7 +1157,6 @@ static void ff_layout_io_track_ds_error(struct pnfs_layout_segment *lseg, static int ff_layout_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr) { - struct inode *inode; int err; trace_nfs4_pnfs_read(hdr, task->tk_status); @@ -1176,8 +1176,6 @@ static int ff_layout_read_done_cb(struct rpc_task *task, pnfs_read_resend_pnfs(hdr); return task->tk_status; case -NFS4ERR_RESET_TO_MDS: - inode = hdr->lseg->pls_layout->plh_inode; - pnfs_error_mark_layout_for_return(inode, hdr->lseg); ff_layout_reset_read(hdr); return task->tk_status; case -EAGAIN: @@ -1330,7 +1328,6 @@ static void ff_layout_read_count_stats(struct rpc_task *task, void *data) static int ff_layout_write_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr) { - struct inode *inode; int err; trace_nfs4_pnfs_write(hdr, task->tk_status); @@ -1346,8 +1343,6 @@ static int ff_layout_write_done_cb(struct rpc_task *task, switch (err) { case -NFS4ERR_RESET_TO_PNFS: case -NFS4ERR_RESET_TO_MDS: - inode = hdr->lseg->pls_layout->plh_inode; - pnfs_error_mark_layout_for_return(inode, hdr->lseg); if (err == -NFS4ERR_RESET_TO_PNFS) { pnfs_set_retry_layoutget(hdr->lseg->pls_layout); ff_layout_reset_write(hdr, true); @@ -1376,7 +1371,6 @@ static int ff_layout_write_done_cb(struct rpc_task *task, static int ff_layout_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data) { - struct inode *inode; int err; trace_nfs4_pnfs_commit_ds(data, task->tk_status); @@ -1391,8 +1385,6 @@ static int ff_layout_commit_done_cb(struct rpc_task *task, switch (err) { case -NFS4ERR_RESET_TO_PNFS: case -NFS4ERR_RESET_TO_MDS: - inode = data->lseg->pls_layout->plh_inode; - pnfs_error_mark_layout_for_return(inode, data->lseg); if (err == -NFS4ERR_RESET_TO_PNFS) pnfs_set_retry_layoutget(data->lseg->pls_layout); else -- cgit v1.2.3 From 7cc8c5cde0a5872f5d013f82978b73c011d8f8f1 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 2 Sep 2015 15:22:48 -0700 Subject: NFSv4.1/flexfiles: Clean up ff_layout_write_done_cb/ff_layout_commit_done_cb Signed-off-by: Trond Myklebust --- fs/nfs/flexfilelayout/flexfilelayout.c | 20 +++++++++----------- 1 file changed, 9 insertions(+), 11 deletions(-) diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index 3f073a7de870..fbc5a56de875 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -1342,14 +1342,12 @@ static int ff_layout_write_done_cb(struct rpc_task *task, switch (err) { case -NFS4ERR_RESET_TO_PNFS: + pnfs_set_retry_layoutget(hdr->lseg->pls_layout); + ff_layout_reset_write(hdr, true); + return task->tk_status; case -NFS4ERR_RESET_TO_MDS: - if (err == -NFS4ERR_RESET_TO_PNFS) { - pnfs_set_retry_layoutget(hdr->lseg->pls_layout); - ff_layout_reset_write(hdr, true); - } else { - pnfs_clear_retry_layoutget(hdr->lseg->pls_layout); - ff_layout_reset_write(hdr, false); - } + pnfs_clear_retry_layoutget(hdr->lseg->pls_layout); + ff_layout_reset_write(hdr, false); return task->tk_status; case -EAGAIN: rpc_restart_call_prepare(task); @@ -1384,11 +1382,11 @@ static int ff_layout_commit_done_cb(struct rpc_task *task, switch (err) { case -NFS4ERR_RESET_TO_PNFS: + pnfs_set_retry_layoutget(data->lseg->pls_layout); + pnfs_generic_prepare_to_resend_writes(data); + return -EAGAIN; case -NFS4ERR_RESET_TO_MDS: - if (err == -NFS4ERR_RESET_TO_PNFS) - pnfs_set_retry_layoutget(data->lseg->pls_layout); - else - pnfs_clear_retry_layoutget(data->lseg->pls_layout); + pnfs_clear_retry_layoutget(data->lseg->pls_layout); pnfs_generic_prepare_to_resend_writes(data); return -EAGAIN; case -EAGAIN: -- cgit v1.2.3 From 5cf9d70659594e1a75b34d18619d0bb6e0cbbafa Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 4 Sep 2015 15:07:37 -0400 Subject: NFS: Optimise away the close-to-open getattr if there is no cached data If there is no cached data, then there is no need to track the file change attribute on close. Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-) diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 99a68bd9c178..6307d8de103d 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -761,11 +761,13 @@ EXPORT_SYMBOL_GPL(nfs_put_lock_context); * @ctx: pointer to context * @is_sync: is this a synchronous close * - * always ensure that the attributes are up to date if we're mounted - * with close-to-open semantics + * Ensure that the attributes are up to date if we're mounted + * with close-to-open semantics and we have cached data that will + * need to be revalidated on open. */ void nfs_close_context(struct nfs_open_context *ctx, int is_sync) { + struct nfs_inode *nfsi; struct inode *inode; struct nfs_server *server; @@ -774,7 +776,12 @@ void nfs_close_context(struct nfs_open_context *ctx, int is_sync) if (!is_sync) return; inode = d_inode(ctx->dentry); - if (!list_empty(&NFS_I(inode)->open_files)) + nfsi = NFS_I(inode); + if (inode->i_mapping->nrpages == 0) + return; + if (nfsi->cache_validity & NFS_INO_INVALID_DATA) + return; + if (!list_empty(&nfsi->open_files)) return; server = NFS_SERVER(inode); if (server->flags & NFS_MOUNT_NOCTO) -- cgit v1.2.3 From 4eae50143bcbfda819c650b7ed6739f3b6338ffc Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 4 Sep 2015 15:17:53 -0400 Subject: Revert "NFS: Make close(2) asynchronous when closing NFS O_DIRECT files" This reverts commit f895c53f8ace3c3e49ebf9def90e63fc6d46d2bf. This commit causes a NFSv4 regression in that close()+unlink() can end up failing. The reason is that we no longer have a guarantee that the CLOSE has completed on the server, meaning that the subsequent call to REMOVE may fail with NFS4ERR_FILE_OPEN if the server implements Windows unlink() semantics. Reported-by: Cc: Chuck Lever Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 6307d8de103d..326d9e10d833 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -853,6 +853,11 @@ void put_nfs_open_context(struct nfs_open_context *ctx) } EXPORT_SYMBOL_GPL(put_nfs_open_context); +static void put_nfs_open_context_sync(struct nfs_open_context *ctx) +{ + __put_nfs_open_context(ctx, 1); +} + /* * Ensure that mmap has a recent RPC credential for use when writing out * shared pages @@ -908,7 +913,7 @@ void nfs_file_clear_open_context(struct file *filp) spin_lock(&inode->i_lock); list_move_tail(&ctx->list, &NFS_I(inode)->open_files); spin_unlock(&inode->i_lock); - __put_nfs_open_context(ctx, filp->f_flags & O_DIRECT ? 0 : 1); + put_nfs_open_context_sync(ctx); } } -- cgit v1.2.3 From 7d160a6c462c2c690e074c173b43aad7204049ad Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 5 Sep 2015 19:06:57 -0400 Subject: NFSv4: Express delegation limit in units of pages Since we're tracking modifications to the page cache on a per-page basis, it makes sense to express the limit to how much we may cache in units of pages. Signed-off-by: Trond Myklebust --- fs/nfs/delegation.c | 4 ++-- fs/nfs/delegation.h | 2 +- fs/nfs/nfs4xdr.c | 16 ++++++++++------ include/linux/nfs_xdr.h | 2 +- 4 files changed, 14 insertions(+), 10 deletions(-) diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c index 029d688a969f..cd503cc2251c 100644 --- a/fs/nfs/delegation.c +++ b/fs/nfs/delegation.c @@ -175,7 +175,7 @@ void nfs_inode_reclaim_delegation(struct inode *inode, struct rpc_cred *cred, if (delegation->inode != NULL) { nfs4_stateid_copy(&delegation->stateid, &res->delegation); delegation->type = res->delegation_type; - delegation->maxsize = res->maxsize; + delegation->pagemod_limit = res->pagemod_limit; oldcred = delegation->cred; delegation->cred = get_rpccred(cred); clear_bit(NFS_DELEGATION_NEED_RECLAIM, @@ -337,7 +337,7 @@ int nfs_inode_set_delegation(struct inode *inode, struct rpc_cred *cred, struct return -ENOMEM; nfs4_stateid_copy(&delegation->stateid, &res->delegation); delegation->type = res->delegation_type; - delegation->maxsize = res->maxsize; + delegation->pagemod_limit = res->pagemod_limit; delegation->change_attr = inode->i_version; delegation->cred = get_rpccred(cred); delegation->inode = inode; diff --git a/fs/nfs/delegation.h b/fs/nfs/delegation.h index e3c20a3ccc93..554178a17376 100644 --- a/fs/nfs/delegation.h +++ b/fs/nfs/delegation.h @@ -18,7 +18,7 @@ struct nfs_delegation { struct inode *inode; nfs4_stateid stateid; fmode_t type; - loff_t maxsize; + unsigned long pagemod_limit; __u64 change_attr; unsigned long flags; spinlock_t lock; diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index ff4784c54e04..788adf3897c7 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -4932,24 +4932,28 @@ static int decode_lookup(struct xdr_stream *xdr) } /* This is too sick! */ -static int decode_space_limit(struct xdr_stream *xdr, u64 *maxsize) +static int decode_space_limit(struct xdr_stream *xdr, + unsigned long *pagemod_limit) { __be32 *p; uint32_t limit_type, nblocks, blocksize; + u64 maxsize = 0; p = xdr_inline_decode(xdr, 12); if (unlikely(!p)) goto out_overflow; limit_type = be32_to_cpup(p++); switch (limit_type) { - case 1: - xdr_decode_hyper(p, maxsize); + case NFS4_LIMIT_SIZE: + xdr_decode_hyper(p, &maxsize); break; - case 2: + case NFS4_LIMIT_BLOCKS: nblocks = be32_to_cpup(p++); blocksize = be32_to_cpup(p); - *maxsize = (uint64_t)nblocks * (uint64_t)blocksize; + maxsize = (uint64_t)nblocks * (uint64_t)blocksize; } + maxsize >>= PAGE_CACHE_SHIFT; + *pagemod_limit = min_t(u64, maxsize, ULONG_MAX); return 0; out_overflow: print_overflow_msg(__func__, xdr); @@ -4977,7 +4981,7 @@ static int decode_rw_delegation(struct xdr_stream *xdr, break; case NFS4_OPEN_DELEGATE_WRITE: res->delegation_type = FMODE_WRITE|FMODE_READ; - if (decode_space_limit(xdr, &res->maxsize) < 0) + if (decode_space_limit(xdr, &res->pagemod_limit) < 0) return -EIO; } return decode_ace(xdr, NULL, res->server->nfs_client); diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index b4392d86d157..52faf7e96c65 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -406,8 +406,8 @@ struct nfs_openres { const struct nfs_server *server; fmode_t delegation_type; nfs4_stateid delegation; + unsigned long pagemod_limit; __u32 do_recall; - __u64 maxsize; __u32 attrset[NFS4_BITMAP_SIZE]; struct nfs4_string *owner; struct nfs4_string *group_owner; -- cgit v1.2.3 From 5445b1fbd123420bffed5e629a420aa2a16bf849 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 5 Sep 2015 19:06:58 -0400 Subject: NFSv4: Respect the server imposed limit on how many changes we may cache The NFSv4 delegation spec allows the server to tell a client to limit how much data it cache after the file is closed. In return, the server guarantees enough free space to avoid ENOSPC situations, etc. Prior to this patch, we assumed we could always cache aggressively after close. Unfortunately, this causes problems with servers that set the limit to 0 and therefore do not offer any ENOSPC guarantees. Signed-off-by: Trond Myklebust --- fs/nfs/delegation.c | 25 +++++++++++++++++++++++++ fs/nfs/delegation.h | 1 + fs/nfs/file.c | 10 +--------- fs/nfs/internal.h | 1 - fs/nfs/nfs4file.c | 29 ++++++++++++++++++++++++++++- 5 files changed, 55 insertions(+), 11 deletions(-) diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c index cd503cc2251c..2714ef835bdd 100644 --- a/fs/nfs/delegation.c +++ b/fs/nfs/delegation.c @@ -900,3 +900,28 @@ bool nfs4_copy_delegation_stateid(nfs4_stateid *dst, struct inode *inode, rcu_read_unlock(); return ret; } + +/** + * nfs4_delegation_flush_on_close - Check if we must flush file on close + * @inode: inode to check + * + * This function checks the number of outstanding writes to the file + * against the delegation 'space_limit' field to see if + * the spec requires us to flush the file on close. + */ +bool nfs4_delegation_flush_on_close(const struct inode *inode) +{ + struct nfs_inode *nfsi = NFS_I(inode); + struct nfs_delegation *delegation; + bool ret = true; + + rcu_read_lock(); + delegation = rcu_dereference(nfsi->delegation); + if (delegation == NULL || !(delegation->type & FMODE_WRITE)) + goto out; + if (nfsi->nrequests < delegation->pagemod_limit) + ret = false; +out: + rcu_read_unlock(); + return ret; +} diff --git a/fs/nfs/delegation.h b/fs/nfs/delegation.h index 554178a17376..a44829173e57 100644 --- a/fs/nfs/delegation.h +++ b/fs/nfs/delegation.h @@ -61,6 +61,7 @@ bool nfs4_copy_delegation_stateid(nfs4_stateid *dst, struct inode *inode, fmode_ void nfs_mark_delegation_referenced(struct nfs_delegation *delegation); int nfs4_have_delegation(struct inode *inode, fmode_t flags); int nfs4_check_delegation(struct inode *inode, fmode_t flags); +bool nfs4_delegation_flush_on_close(const struct inode *inode); #endif diff --git a/fs/nfs/file.c b/fs/nfs/file.c index 526a2681d975..c0f9b1ed12b9 100644 --- a/fs/nfs/file.c +++ b/fs/nfs/file.c @@ -142,7 +142,7 @@ EXPORT_SYMBOL_GPL(nfs_file_llseek); /* * Flush all dirty pages, and check for write errors. */ -int +static int nfs_file_flush(struct file *file, fl_owner_t id) { struct inode *inode = file_inode(file); @@ -153,17 +153,9 @@ nfs_file_flush(struct file *file, fl_owner_t id) if ((file->f_mode & FMODE_WRITE) == 0) return 0; - /* - * If we're holding a write delegation, then just start the i/o - * but don't wait for completion (or send a commit). - */ - if (NFS_PROTO(inode)->have_delegation(inode, FMODE_WRITE)) - return filemap_fdatawrite(file->f_mapping); - /* Flush writes to the server and return any errors */ return vfs_fsync(file, 0); } -EXPORT_SYMBOL_GPL(nfs_file_flush); ssize_t nfs_file_read(struct kiocb *iocb, struct iov_iter *to) diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h index 9ab3b1c21bb4..56cfde26fb9c 100644 --- a/fs/nfs/internal.h +++ b/fs/nfs/internal.h @@ -360,7 +360,6 @@ int nfs_rename(struct inode *, struct dentry *, struct inode *, struct dentry *) /* file.c */ int nfs_file_fsync_commit(struct file *, loff_t, loff_t, int); loff_t nfs_file_llseek(struct file *, loff_t, int); -int nfs_file_flush(struct file *, fl_owner_t); ssize_t nfs_file_read(struct kiocb *, struct iov_iter *); ssize_t nfs_file_splice_read(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int); diff --git a/fs/nfs/nfs4file.c b/fs/nfs/nfs4file.c index 43f1590b9240..b0dbe0abed53 100644 --- a/fs/nfs/nfs4file.c +++ b/fs/nfs/nfs4file.c @@ -6,7 +6,9 @@ #include #include #include +#include "delegation.h" #include "internal.h" +#include "iostat.h" #include "fscache.h" #include "pnfs.h" @@ -99,6 +101,31 @@ out_drop: goto out_put_ctx; } +/* + * Flush all dirty pages, and check for write errors. + */ +static int +nfs4_file_flush(struct file *file, fl_owner_t id) +{ + struct inode *inode = file_inode(file); + + dprintk("NFS: flush(%pD2)\n", file); + + nfs_inc_stats(inode, NFSIOS_VFSFLUSH); + if ((file->f_mode & FMODE_WRITE) == 0) + return 0; + + /* + * If we're holding a write delegation, then check if we're required + * to flush the i/o on close. If not, then just start the i/o now. + */ + if (!nfs4_delegation_flush_on_close(inode)) + return filemap_fdatawrite(file->f_mapping); + + /* Flush writes to the server and return any errors */ + return vfs_fsync(file, 0); +} + static int nfs4_file_fsync(struct file *file, loff_t start, loff_t end, int datasync) { @@ -177,7 +204,7 @@ const struct file_operations nfs4_file_operations = { .write_iter = nfs_file_write, .mmap = nfs_file_mmap, .open = nfs4_file_open, - .flush = nfs_file_flush, + .flush = nfs4_file_flush, .release = nfs_file_release, .fsync = nfs4_file_fsync, .lock = nfs_lock, -- cgit v1.2.3