NVMe: Use CMB for the IO SQes if available

Some controllers have a controller-side memory buffer available for use
for submissions, completions, lists, or data.

If a CMB is available, the entire CMB will be ioremapped and it will
attempt to map the IO SQes onto the CMB. The queues will be shrunk as
needed. The CMB will not be used if the queue depth is shrunk below some
threshold where it may have reduced performance over a larger queue
in system memory.

Signed-off-by: Jon Derrick <jonathan.derrick@intel.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@fb.com>
diff --git a/include/linux/nvme.h b/include/linux/nvme.h
index c0d94ed..fa3fe16 100644
--- a/include/linux/nvme.h
+++ b/include/linux/nvme.h
@@ -32,6 +32,8 @@
 	__u32			aqa;	/* Admin Queue Attributes */
 	__u64			asq;	/* Admin SQ Base Address */
 	__u64			acq;	/* Admin CQ Base Address */
+	__u32			cmbloc; /* Controller Memory Buffer Location */
+	__u32			cmbsz;  /* Controller Memory Buffer Size */
 };
 
 #define NVME_CAP_MQES(cap)	((cap) & 0xffff)
@@ -40,6 +42,17 @@
 #define NVME_CAP_MPSMIN(cap)	(((cap) >> 48) & 0xf)
 #define NVME_CAP_MPSMAX(cap)	(((cap) >> 52) & 0xf)
 
+#define NVME_CMB_BIR(cmbloc)	((cmbloc) & 0x7)
+#define NVME_CMB_OFST(cmbloc)	(((cmbloc) >> 12) & 0xfffff)
+#define NVME_CMB_SZ(cmbsz)	(((cmbsz) >> 12) & 0xfffff)
+#define NVME_CMB_SZU(cmbsz)	(((cmbsz) >> 8) & 0xf)
+
+#define NVME_CMB_WDS(cmbsz)	((cmbsz) & 0x10)
+#define NVME_CMB_RDS(cmbsz)	((cmbsz) & 0x8)
+#define NVME_CMB_LISTS(cmbsz)	((cmbsz) & 0x4)
+#define NVME_CMB_CQS(cmbsz)	((cmbsz) & 0x2)
+#define NVME_CMB_SQS(cmbsz)	((cmbsz) & 0x1)
+
 enum {
 	NVME_CC_ENABLE		= 1 << 0,
 	NVME_CC_CSS_NVM		= 0 << 4,
@@ -100,6 +113,10 @@
 	u32 max_hw_sectors;
 	u32 stripe_size;
 	u32 page_size;
+	void __iomem *cmb;
+	dma_addr_t cmb_dma_addr;
+	u64 cmb_size;
+	u32 cmbsz;
 	u16 oncs;
 	u16 abort_limit;
 	u8 event_limit;