From 7bb46a6734a7e1ad4beaecc11cae7ed3ff81d30f Mon Sep 17 00:00:00 2001 From: "npiggin@suse.de" Date: Thu, 27 May 2010 01:05:33 +1000 Subject: fs: introduce new truncate sequence Introduce a new truncate calling sequence into fs/mm subsystems. Rather than setattr > vmtruncate > truncate, have filesystems call their truncate sequence from ->setattr if filesystem specific operations are required. vmtruncate is deprecated, and truncate_pagecache and inode_newsize_ok helpers introduced previously should be used. simple_setattr is introduced for simple in-ram filesystems to implement the new truncate sequence. Eventually all filesystems should be converted to implement a setattr, and the default code in notify_change should go away. simple_setsize is also introduced to perform just the ATTR_SIZE portion of simple_setattr (ie. changing i_size and trimming pagecache). To implement the new truncate sequence: - filesystem specific manipulations (eg freeing blocks) must be done in the setattr method rather than ->truncate. - vmtruncate can not be used by core code to trim blocks past i_size in the event of write failure after allocation, so this must be performed in the fs code. - convert usage of helpers block_write_begin, nobh_write_begin, cont_write_begin, and *blockdev_direct_IO* to use _newtrunc postfixed variants. These avoid calling vmtruncate to trim blocks (see previous). - inode_setattr should not be used. generic_setattr is a new function to be used to copy simple attributes into the generic inode. - make use of the better opportunity to handle errors with the new sequence. Big problem with the previous calling sequence: the filesystem is not called until i_size has already changed. This means it is not allowed to fail the call, and also it does not know what the previous i_size was. Also, generic code calling vmtruncate to truncate allocated blocks in case of error had no good way to return a meaningful error (or, for example, atomically handle block deallocation). Cc: Christoph Hellwig Acked-by: Jan Kara Signed-off-by: Nick Piggin Signed-off-by: Al Viro --- fs/attr.c | 50 ++++++++++++++++++----- fs/buffer.c | 123 +++++++++++++++++++++++++++++++++++++++++++++------------ fs/direct-io.c | 61 ++++++++++++++++++---------- fs/libfs.c | 76 +++++++++++++++++++++++++++++++++++ 4 files changed, 254 insertions(+), 56 deletions(-) (limited to 'fs') diff --git a/fs/attr.c b/fs/attr.c index 0815e93bb487..b4fa3b0aa596 100644 --- a/fs/attr.c +++ b/fs/attr.c @@ -67,14 +67,14 @@ EXPORT_SYMBOL(inode_change_ok); * @offset: the new size to assign to the inode * @Returns: 0 on success, -ve errno on failure * + * inode_newsize_ok must be called with i_mutex held. + * * inode_newsize_ok will check filesystem limits and ulimits to check that the * new inode size is within limits. inode_newsize_ok will also send SIGXFSZ * when necessary. Caller must not proceed with inode size change if failure is * returned. @inode must be a file (not directory), with appropriate * permissions to allow truncate (inode_newsize_ok does NOT check these * conditions). - * - * inode_newsize_ok must be called with i_mutex held. */ int inode_newsize_ok(const struct inode *inode, loff_t offset) { @@ -104,17 +104,25 @@ out_big: } EXPORT_SYMBOL(inode_newsize_ok); -int inode_setattr(struct inode * inode, struct iattr * attr) +/** + * generic_setattr - copy simple metadata updates into the generic inode + * @inode: the inode to be updated + * @attr: the new attributes + * + * generic_setattr must be called with i_mutex held. + * + * generic_setattr updates the inode's metadata with that specified + * in attr. Noticably missing is inode size update, which is more complex + * as it requires pagecache updates. See simple_setsize. + * + * The inode is not marked as dirty after this operation. The rationale is + * that for "simple" filesystems, the struct inode is the inode storage. + * The caller is free to mark the inode dirty afterwards if needed. + */ +void generic_setattr(struct inode *inode, const struct iattr *attr) { unsigned int ia_valid = attr->ia_valid; - if (ia_valid & ATTR_SIZE && - attr->ia_size != i_size_read(inode)) { - int error = vmtruncate(inode, attr->ia_size); - if (error) - return error; - } - if (ia_valid & ATTR_UID) inode->i_uid = attr->ia_uid; if (ia_valid & ATTR_GID) @@ -135,6 +143,28 @@ int inode_setattr(struct inode * inode, struct iattr * attr) mode &= ~S_ISGID; inode->i_mode = mode; } +} +EXPORT_SYMBOL(generic_setattr); + +/* + * note this function is deprecated, the new truncate sequence should be + * used instead -- see eg. simple_setsize, generic_setattr. + */ +int inode_setattr(struct inode *inode, const struct iattr *attr) +{ + unsigned int ia_valid = attr->ia_valid; + + if (ia_valid & ATTR_SIZE && + attr->ia_size != i_size_read(inode)) { + int error; + + error = vmtruncate(inode, attr->ia_size); + if (error) + return error; + } + + generic_setattr(inode, attr); + mark_inode_dirty(inode); return 0; diff --git a/fs/buffer.c b/fs/buffer.c index e8aa7081d25c..d54812b198e9 100644 --- a/fs/buffer.c +++ b/fs/buffer.c @@ -1949,14 +1949,11 @@ static int __block_commit_write(struct inode *inode, struct page *page, } /* - * block_write_begin takes care of the basic task of block allocation and - * bringing partial write blocks uptodate first. - * - * If *pagep is not NULL, then block_write_begin uses the locked page - * at *pagep rather than allocating its own. In this case, the page will - * not be unlocked or deallocated on failure. + * Filesystems implementing the new truncate sequence should use the + * _newtrunc postfix variant which won't incorrectly call vmtruncate. + * The filesystem needs to handle block truncation upon failure. */ -int block_write_begin(struct file *file, struct address_space *mapping, +int block_write_begin_newtrunc(struct file *file, struct address_space *mapping, loff_t pos, unsigned len, unsigned flags, struct page **pagep, void **fsdata, get_block_t *get_block) @@ -1992,20 +1989,50 @@ int block_write_begin(struct file *file, struct address_space *mapping, unlock_page(page); page_cache_release(page); *pagep = NULL; - - /* - * prepare_write() may have instantiated a few blocks - * outside i_size. Trim these off again. Don't need - * i_size_read because we hold i_mutex. - */ - if (pos + len > inode->i_size) - vmtruncate(inode, inode->i_size); } } out: return status; } +EXPORT_SYMBOL(block_write_begin_newtrunc); + +/* + * block_write_begin takes care of the basic task of block allocation and + * bringing partial write blocks uptodate first. + * + * If *pagep is not NULL, then block_write_begin uses the locked page + * at *pagep rather than allocating its own. In this case, the page will + * not be unlocked or deallocated on failure. + */ +int block_write_begin(struct file *file, struct address_space *mapping, + loff_t pos, unsigned len, unsigned flags, + struct page **pagep, void **fsdata, + get_block_t *get_block) +{ + int ret; + + ret = block_write_begin_newtrunc(file, mapping, pos, len, flags, + pagep, fsdata, get_block); + + /* + * prepare_write() may have instantiated a few blocks + * outside i_size. Trim these off again. Don't need + * i_size_read because we hold i_mutex. + * + * Filesystems which pass down their own page also cannot + * call into vmtruncate here because it would lead to lock + * inversion problems (*pagep is locked). This is a further + * example of where the old truncate sequence is inadequate. + */ + if (unlikely(ret) && *pagep == NULL) { + loff_t isize = mapping->host->i_size; + if (pos + len > isize) + vmtruncate(mapping->host, isize); + } + + return ret; +} EXPORT_SYMBOL(block_write_begin); int block_write_end(struct file *file, struct address_space *mapping, @@ -2324,7 +2351,7 @@ out: * For moronic filesystems that do not allow holes in file. * We may have to extend the file. */ -int cont_write_begin(struct file *file, struct address_space *mapping, +int cont_write_begin_newtrunc(struct file *file, struct address_space *mapping, loff_t pos, unsigned len, unsigned flags, struct page **pagep, void **fsdata, get_block_t *get_block, loff_t *bytes) @@ -2345,11 +2372,30 @@ int cont_write_begin(struct file *file, struct address_space *mapping, } *pagep = NULL; - err = block_write_begin(file, mapping, pos, len, + err = block_write_begin_newtrunc(file, mapping, pos, len, flags, pagep, fsdata, get_block); out: return err; } +EXPORT_SYMBOL(cont_write_begin_newtrunc); + +int cont_write_begin(struct file *file, struct address_space *mapping, + loff_t pos, unsigned len, unsigned flags, + struct page **pagep, void **fsdata, + get_block_t *get_block, loff_t *bytes) +{ + int ret; + + ret = cont_write_begin_newtrunc(file, mapping, pos, len, flags, + pagep, fsdata, get_block, bytes); + if (unlikely(ret)) { + loff_t isize = mapping->host->i_size; + if (pos + len > isize) + vmtruncate(mapping->host, isize); + } + + return ret; +} EXPORT_SYMBOL(cont_write_begin); int block_prepare_write(struct page *page, unsigned from, unsigned to, @@ -2381,7 +2427,7 @@ EXPORT_SYMBOL(block_commit_write); * * We are not allowed to take the i_mutex here so we have to play games to * protect against truncate races as the page could now be beyond EOF. Because - * vmtruncate() writes the inode size before removing pages, once we have the + * truncate writes the inode size before removing pages, once we have the * page lock we can determine safely if the page is beyond EOF. If it is not * beyond EOF, then the page is guaranteed safe against truncation until we * unlock the page. @@ -2464,10 +2510,11 @@ static void attach_nobh_buffers(struct page *page, struct buffer_head *head) } /* - * On entry, the page is fully not uptodate. - * On exit the page is fully uptodate in the areas outside (from,to) + * Filesystems implementing the new truncate sequence should use the + * _newtrunc postfix variant which won't incorrectly call vmtruncate. + * The filesystem needs to handle block truncation upon failure. */ -int nobh_write_begin(struct file *file, struct address_space *mapping, +int nobh_write_begin_newtrunc(struct file *file, struct address_space *mapping, loff_t pos, unsigned len, unsigned flags, struct page **pagep, void **fsdata, get_block_t *get_block) @@ -2500,8 +2547,8 @@ int nobh_write_begin(struct file *file, struct address_space *mapping, unlock_page(page); page_cache_release(page); *pagep = NULL; - return block_write_begin(file, mapping, pos, len, flags, pagep, - fsdata, get_block); + return block_write_begin_newtrunc(file, mapping, pos, len, + flags, pagep, fsdata, get_block); } if (PageMappedToDisk(page)) @@ -2605,8 +2652,34 @@ out_release: page_cache_release(page); *pagep = NULL; - if (pos + len > inode->i_size) - vmtruncate(inode, inode->i_size); + return ret; +} +EXPORT_SYMBOL(nobh_write_begin_newtrunc); + +/* + * On entry, the page is fully not uptodate. + * On exit the page is fully uptodate in the areas outside (from,to) + */ +int nobh_write_begin(struct file *file, struct address_space *mapping, + loff_t pos, unsigned len, unsigned flags, + struct page **pagep, void **fsdata, + get_block_t *get_block) +{ + int ret; + + ret = nobh_write_begin_newtrunc(file, mapping, pos, len, flags, + pagep, fsdata, get_block); + + /* + * prepare_write() may have instantiated a few blocks + * outside i_size. Trim these off again. Don't need + * i_size_read because we hold i_mutex. + */ + if (unlikely(ret)) { + loff_t isize = mapping->host->i_size; + if (pos + len > isize) + vmtruncate(mapping->host, isize); + } return ret; } diff --git a/fs/direct-io.c b/fs/direct-io.c index da111aacb46e..7600aacf531d 100644 --- a/fs/direct-io.c +++ b/fs/direct-io.c @@ -1134,27 +1134,8 @@ direct_io_worker(int rw, struct kiocb *iocb, struct inode *inode, return ret; } -/* - * This is a library function for use by filesystem drivers. - * - * The locking rules are governed by the flags parameter: - * - if the flags value contains DIO_LOCKING we use a fancy locking - * scheme for dumb filesystems. - * For writes this function is called under i_mutex and returns with - * i_mutex held, for reads, i_mutex is not held on entry, but it is - * taken and dropped again before returning. - * For reads and writes i_alloc_sem is taken in shared mode and released - * on I/O completion (which may happen asynchronously after returning to - * the caller). - * - * - if the flags value does NOT contain DIO_LOCKING we don't use any - * internal locking but rather rely on the filesystem to synchronize - * direct I/O reads/writes versus each other and truncate. - * For reads and writes both i_mutex and i_alloc_sem are not held on - * entry and are never taken. - */ ssize_t -__blockdev_direct_IO(int rw, struct kiocb *iocb, struct inode *inode, +__blockdev_direct_IO_newtrunc(int rw, struct kiocb *iocb, struct inode *inode, struct block_device *bdev, const struct iovec *iov, loff_t offset, unsigned long nr_segs, get_block_t get_block, dio_iodone_t end_io, dio_submit_t submit_io, int flags) @@ -1247,9 +1228,46 @@ __blockdev_direct_IO(int rw, struct kiocb *iocb, struct inode *inode, nr_segs, blkbits, get_block, end_io, submit_io, dio); +out: + return retval; +} +EXPORT_SYMBOL(__blockdev_direct_IO_newtrunc); + +/* + * This is a library function for use by filesystem drivers. + * + * The locking rules are governed by the flags parameter: + * - if the flags value contains DIO_LOCKING we use a fancy locking + * scheme for dumb filesystems. + * For writes this function is called under i_mutex and returns with + * i_mutex held, for reads, i_mutex is not held on entry, but it is + * taken and dropped again before returning. + * For reads and writes i_alloc_sem is taken in shared mode and released + * on I/O completion (which may happen asynchronously after returning to + * the caller). + * + * - if the flags value does NOT contain DIO_LOCKING we don't use any + * internal locking but rather rely on the filesystem to synchronize + * direct I/O reads/writes versus each other and truncate. + * For reads and writes both i_mutex and i_alloc_sem are not held on + * entry and are never taken. + */ +ssize_t +__blockdev_direct_IO(int rw, struct kiocb *iocb, struct inode *inode, + struct block_device *bdev, const struct iovec *iov, loff_t offset, + unsigned long nr_segs, get_block_t get_block, dio_iodone_t end_io, + dio_submit_t submit_io, int flags) +{ + ssize_t retval; + + retval = __blockdev_direct_IO_newtrunc(rw, iocb, inode, bdev, iov, + offset, nr_segs, get_block, end_io, submit_io, flags); /* * In case of error extending write may have instantiated a few * blocks outside i_size. Trim these off again for DIO_LOCKING. + * NOTE: DIO_NO_LOCK/DIO_OWN_LOCK callers have to handle this in + * their own manner. This is a further example of where the old + * truncate sequence is inadequate. * * NOTE: filesystems with their own locking have to handle this * on their own. @@ -1257,12 +1275,13 @@ __blockdev_direct_IO(int rw, struct kiocb *iocb, struct inode *inode, if (flags & DIO_LOCKING) { if (unlikely((rw & WRITE) && retval < 0)) { loff_t isize = i_size_read(inode); + loff_t end = offset + iov_length(iov, nr_segs); + if (end > isize) vmtruncate(inode, isize); } } -out: return retval; } EXPORT_SYMBOL(__blockdev_direct_IO); diff --git a/fs/libfs.c b/fs/libfs.c index b84d0a7a2204..09e1016eb774 100644 --- a/fs/libfs.c +++ b/fs/libfs.c @@ -8,6 +8,7 @@ #include #include #include +#include #include #include #include @@ -325,6 +326,81 @@ int simple_rename(struct inode *old_dir, struct dentry *old_dentry, return 0; } +/** + * simple_setsize - handle core mm and vfs requirements for file size change + * @inode: inode + * @newsize: new file size + * + * Returns 0 on success, -error on failure. + * + * simple_setsize must be called with inode_mutex held. + * + * simple_setsize will check that the requested new size is OK (see + * inode_newsize_ok), and then will perform the necessary i_size update + * and pagecache truncation (if necessary). It will be typically be called + * from the filesystem's setattr function when ATTR_SIZE is passed in. + * + * The inode itself must have correct permissions and attributes to allow + * i_size to be changed, this function then just checks that the new size + * requested is valid. + * + * In the case of simple in-memory filesystems with inodes stored solely + * in the inode cache, and file data in the pagecache, nothing more needs + * to be done to satisfy a truncate request. Filesystems with on-disk + * blocks for example will need to free them in the case of truncate, in + * that case it may be easier not to use simple_setsize (but each of its + * components will likely be required at some point to update pagecache + * and inode etc). + */ +int simple_setsize(struct inode *inode, loff_t newsize) +{ + loff_t oldsize; + int error; + + error = inode_newsize_ok(inode, newsize); + if (error) + return error; + + oldsize = inode->i_size; + i_size_write(inode, newsize); + truncate_pagecache(inode, oldsize, newsize); + + return error; +} +EXPORT_SYMBOL(simple_setsize); + +/** + * simple_setattr - setattr for simple in-memory filesystem + * @dentry: dentry + * @iattr: iattr structure + * + * Returns 0 on success, -error on failure. + * + * simple_setattr implements setattr for an in-memory filesystem which + * does not store its own file data or metadata (eg. uses the page cache + * and inode cache as its data store). + */ +int simple_setattr(struct dentry *dentry, struct iattr *iattr) +{ + struct inode *inode = dentry->d_inode; + int error; + + error = inode_change_ok(inode, iattr); + if (error) + return error; + + if (iattr->ia_valid & ATTR_SIZE) { + error = simple_setsize(inode, iattr->ia_size); + if (error) + return error; + } + + generic_setattr(inode, iattr); + + return error; +} +EXPORT_SYMBOL(simple_setattr); + int simple_readpage(struct file *file, struct page *page) { clear_highpage(page); -- cgit v1.2.3