*
* O_DIRECT
*
- * 04Jul2002 akpm@zip.com.au
+ * 04Jul2002 Andrew Morton
* Initial version
* 11Sep2002 janetinc@us.ibm.com
* added readv/writev support.
- * 29Oct2002 akpm@zip.com.au
+ * 29Oct2002 Andrew Morton
* rewrote bio_add_page() support.
* 30Oct2002 pbadari@us.ibm.com
* added support for non-aligned IO.
*
* If blkfactor is zero then the user's request was aligned to the filesystem's
* blocksize.
- *
- * lock_type is DIO_LOCKING for regular files on direct-IO-naive filesystems.
- * This determines whether we need to do the fancy locking which prevents
- * direct-IO from being able to read uninitialised disk blocks. If its zero
- * (blockdev) this locking is not done, and if it is DIO_OWN_LOCKING i_mutex is
- * not held for the entire direct write (taken briefly, initially, during a
- * direct read though, but its never held for the duration of a direct-IO).
*/
struct dio {
struct inode *inode;
int rw;
loff_t i_size; /* i_size when submitted */
- int lock_type; /* doesn't change */
+ int flags; /* doesn't change */
unsigned blkbits; /* doesn't change */
unsigned blkfactor; /* When we're using an alignment which
is finer than the filesystem's soft
unsigned cur_page_len; /* Nr of bytes at cur_page_offset */
sector_t cur_page_block; /* Where it starts */
+ /* BIO completion state */
+ spinlock_t bio_lock; /* protects BIO fields below */
+ unsigned long refcount; /* direct_io_worker() and bios */
+ struct bio *bio_list; /* singly linked via bi_private */
+ struct task_struct *waiter; /* waiting task (NULL if none) */
+
+ /* AIO related stuff */
+ struct kiocb *iocb; /* kiocb */
+ int is_async; /* is IO async ? */
+ int io_error; /* IO error in completion path */
+ ssize_t result; /* IO result */
+
/*
* Page fetching state. These variables belong to dio_refill_pages().
*/
* Page queue. These variables belong to dio_refill_pages() and
* dio_get_page().
*/
- struct page *pages[DIO_PAGES]; /* page buffer */
unsigned head; /* next page to process */
unsigned tail; /* last valid page + 1 */
int page_errors; /* errno from get_user_pages() */
- /* BIO completion state */
- spinlock_t bio_lock; /* protects BIO fields below */
- unsigned long refcount; /* direct_io_worker() and bios */
- struct bio *bio_list; /* singly linked via bi_private */
- struct task_struct *waiter; /* waiting task (NULL if none) */
-
- /* AIO related stuff */
- struct kiocb *iocb; /* kiocb */
- int is_async; /* is IO async ? */
- int io_error; /* IO error in completion path */
- ssize_t result; /* IO result */
+ /*
+ * pages[] (and any fields placed after it) are not zeroed out at
+ * allocation time. Don't add new fields after pages[] unless you
+ * wish that they not be zeroed.
+ */
+ struct page *pages[DIO_PAGES]; /* page buffer */
};
/*
int nr_pages;
nr_pages = min(dio->total_pages - dio->curr_page, DIO_PAGES);
- down_read(¤t->mm->mmap_sem);
- ret = get_user_pages(
- current, /* Task for fault acounting */
- current->mm, /* whose pages? */
+ ret = get_user_pages_fast(
dio->curr_user_address, /* Where from? */
nr_pages, /* How many pages? */
dio->rw == READ, /* Write to memory? */
- 0, /* force (?) */
- &dio->pages[0],
- NULL); /* vmas */
- up_read(¤t->mm->mmap_sem);
+ &dio->pages[0]); /* Put results here */
if (ret < 0 && dio->blocks_available && (dio->rw & WRITE)) {
- struct page *page = ZERO_PAGE(dio->curr_user_address);
+ struct page *page = ZERO_PAGE(0);
/*
* A memory fault, but the filesystem has some outstanding
* mapped blocks. We need to use those blocks up to avoid
if (dio->end_io && dio->result)
dio->end_io(dio->iocb, offset, transferred,
dio->map_bh.b_private);
- if (dio->lock_type == DIO_LOCKING)
+
+ if (dio->flags & DIO_LOCKING)
/* lockdep: non-owner release */
up_read_non_owner(&dio->inode->i_alloc_sem);
/*
* Asynchronous IO callback.
*/
-static int dio_bio_end_aio(struct bio *bio, unsigned int bytes_done, int error)
+static void dio_bio_end_aio(struct bio *bio, int error)
{
struct dio *dio = bio->bi_private;
unsigned long remaining;
unsigned long flags;
- if (bio->bi_size)
- return 1;
-
/* cleanup the bio */
dio_bio_complete(dio, bio);
aio_complete(dio->iocb, ret, 0);
kfree(dio);
}
-
- return 0;
}
/*
* During I/O bi_private points at the dio. After I/O, bi_private is used to
* implement a singly-linked list of completed BIOs, at dio->bio_list.
*/
-static int dio_bio_end_io(struct bio *bio, unsigned int bytes_done, int error)
+static void dio_bio_end_io(struct bio *bio, int error)
{
struct dio *dio = bio->bi_private;
unsigned long flags;
- if (bio->bi_size)
- return 1;
-
spin_lock_irqsave(&dio->bio_lock, flags);
bio->bi_private = dio->bio_list;
dio->bio_list = bio;
if (--dio->refcount == 1 && dio->waiter)
wake_up_process(dio->waiter);
spin_unlock_irqrestore(&dio->bio_lock, flags);
- return 0;
}
static int
struct bio *bio;
bio = bio_alloc(GFP_KERNEL, nr_vecs);
- if (bio == NULL)
- return -ENOMEM;
bio->bi_bdev = bdev;
bio->bi_sector = first_sector;
map_bh->b_state = 0;
map_bh->b_size = fs_count << dio->inode->i_blkbits;
+ /*
+ * For writes inside i_size on a DIO_SKIP_HOLES filesystem we
+ * forbid block creations: only overwrites are permitted.
+ * We will return early to the caller once we see an
+ * unmapped buffer head returned, and the caller will fall
+ * back to buffered I/O.
+ *
+ * Otherwise the decision is left to the get_blocks method,
+ * which may decide to handle it or also return an unmapped
+ * buffer head.
+ */
create = dio->rw & WRITE;
- if (dio->lock_type == DIO_LOCKING) {
+ if (dio->flags & DIO_SKIP_HOLES) {
if (dio->block_in_file < (i_size_read(dio->inode) >>
dio->blkbits))
create = 0;
- } else if (dio->lock_type == DIO_NO_LOCKING) {
- create = 0;
}
- /*
- * For writes inside i_size we forbid block creations: only
- * overwrites are permitted. We fall back to buffered writes
- * at a higher level for inside-i_size block-instantiating
- * writes.
- */
ret = (*dio->get_block)(dio->inode, fs_startblk,
map_bh, create);
}
this_chunk_bytes = this_chunk_blocks << dio->blkbits;
- page = ZERO_PAGE(dio->curr_user_address);
+ page = ZERO_PAGE(0);
if (submit_page_section(dio, page, 0, this_chunk_bytes,
dio->next_block_for_io))
return;
page_cache_release(page);
goto out;
}
- zero_user_page(page, block_in_page << blkbits,
- 1 << blkbits, KM_USER0);
+ zero_user(page, block_in_page << blkbits,
+ 1 << blkbits);
dio->block_in_file++;
block_in_page++;
goto next_block;
ssize_t ret2;
size_t bytes;
- dio->bio = NULL;
dio->inode = inode;
dio->rw = rw;
dio->blkbits = blkbits;
dio->blkfactor = inode->i_blkbits - blkbits;
- dio->start_zero_done = 0;
- dio->size = 0;
dio->block_in_file = offset >> blkbits;
- dio->blocks_available = 0;
- dio->cur_page = NULL;
- dio->boundary = 0;
- dio->reap_counter = 0;
dio->get_block = get_block;
dio->end_io = end_io;
- dio->map_bh.b_private = NULL;
- dio->map_bh.b_state = 0;
dio->final_block_in_bio = -1;
dio->next_block_for_io = -1;
- dio->page_errors = 0;
- dio->io_error = 0;
- dio->result = 0;
dio->iocb = iocb;
dio->i_size = i_size_read(inode);
spin_lock_init(&dio->bio_lock);
dio->refcount = 1;
- dio->bio_list = NULL;
- dio->waiter = NULL;
/*
* In case of non-aligned buffers, we may need 2 more
*/
if (unlikely(dio->blkfactor))
dio->pages_in_io = 2;
- else
- dio->pages_in_io = 0;
for (seg = 0; seg < nr_segs; seg++) {
user_addr = (unsigned long)iov[seg].iov_base;
if (dio->bio)
dio_bio_submit(dio);
- /* All IO is now issued, send it on its way */
- blk_run_address_space(inode->i_mapping);
-
/*
* It is possible that, we return short IO due to end of file.
* In that case, we need to release all the pages we got hold on.
* we can let i_mutex go now that its achieved its purpose
* of protecting us from looking up uninitialized blocks.
*/
- if ((rw == READ) && (dio->lock_type == DIO_LOCKING))
+ if (rw == READ && (dio->flags & DIO_LOCKING))
mutex_unlock(&dio->inode->i_mutex);
/*
((rw & READ) || (dio->result == dio->size)))
ret = -EIOCBQUEUED;
- if (ret != -EIOCBQUEUED)
+ if (ret != -EIOCBQUEUED) {
+ /* All IO is now issued, send it on its way */
+ blk_run_address_space(inode->i_mapping);
dio_await_completion(dio);
+ }
/*
* Sync will always be dropping the final ref and completing the
/*
* This is a library function for use by filesystem drivers.
- * The locking rules are governed by the dio_lock_type parameter.
- *
- * DIO_NO_LOCKING (no locking, for raw block device access)
- * For writes, i_mutex is not held on entry; it is never taken.
- *
- * DIO_LOCKING (simple locking for regular files)
- * For writes we are called under i_mutex and return with i_mutex held, even
- * though it is internally dropped.
- * For reads, i_mutex is not held on entry, but it is taken and dropped before
- * returning.
*
- * DIO_OWN_LOCKING (filesystem provides synchronisation and handling of
- * uninitialised data, allowing parallel direct readers and writers)
- * For writes we are called without i_mutex, return without it, never touch it.
- * For reads we are called under i_mutex and return with i_mutex held, even
- * though it may be internally dropped.
+ * 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).
*
- * Additional i_alloc_sem locking requirements described inline below.
+ * - 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,
- int dio_lock_type)
+ int flags)
{
int seg;
size_t size;
ssize_t retval = -EINVAL;
loff_t end = offset;
struct dio *dio;
- int release_i_mutex = 0;
- int acquire_i_mutex = 0;
if (rw & WRITE)
- rw = WRITE_SYNC;
+ rw = WRITE_ODIRECT_PLUG;
if (bdev)
- bdev_blkbits = blksize_bits(bdev_hardsect_size(bdev));
+ bdev_blkbits = blksize_bits(bdev_logical_block_size(bdev));
if (offset & blocksize_mask) {
if (bdev)
retval = -ENOMEM;
if (!dio)
goto out;
-
/*
- * For block device access DIO_NO_LOCKING is used,
- * neither readers nor writers do any locking at all
- * For regular files using DIO_LOCKING,
- * readers need to grab i_mutex and i_alloc_sem
- * writers need to grab i_alloc_sem only (i_mutex is already held)
- * For regular files using DIO_OWN_LOCKING,
- * neither readers nor writers take any locks here
+ * Believe it or not, zeroing out the page array caused a .5%
+ * performance regression in a database benchmark. So, we take
+ * care to only zero out what's needed.
*/
- dio->lock_type = dio_lock_type;
- if (dio_lock_type != DIO_NO_LOCKING) {
+ memset(dio, 0, offsetof(struct dio, pages));
+
+ dio->flags = flags;
+ if (dio->flags & DIO_LOCKING) {
/* watch out for a 0 len io from a tricksy fs */
if (rw == READ && end > offset) {
- struct address_space *mapping;
+ struct address_space *mapping =
+ iocb->ki_filp->f_mapping;
- mapping = iocb->ki_filp->f_mapping;
- if (dio_lock_type != DIO_OWN_LOCKING) {
- mutex_lock(&inode->i_mutex);
- release_i_mutex = 1;
- }
+ /* will be released by direct_io_worker */
+ mutex_lock(&inode->i_mutex);
retval = filemap_write_and_wait_range(mapping, offset,
end - 1);
if (retval) {
+ mutex_unlock(&inode->i_mutex);
kfree(dio);
goto out;
}
-
- if (dio_lock_type == DIO_OWN_LOCKING) {
- mutex_unlock(&inode->i_mutex);
- acquire_i_mutex = 1;
- }
}
- if (dio_lock_type == DIO_LOCKING)
- /* lockdep: not the owner will release it */
- down_read_non_owner(&inode->i_alloc_sem);
+ /*
+ * Will be released at I/O completion, possibly in a
+ * different thread.
+ */
+ down_read_non_owner(&inode->i_alloc_sem);
}
/*
retval = direct_io_worker(rw, iocb, inode, iov, offset,
nr_segs, blkbits, get_block, end_io, dio);
- if (rw == READ && dio_lock_type == DIO_LOCKING)
- release_i_mutex = 0;
+ /*
+ * In case of error extending write may have instantiated a few
+ * blocks outside i_size. Trim these off again for DIO_LOCKING.
+ *
+ * NOTE: filesystems with their own locking have to handle this
+ * on their own.
+ */
+ if (flags & DIO_LOCKING) {
+ if (unlikely((rw & WRITE) && retval < 0)) {
+ loff_t isize = i_size_read(inode);
+ if (end > isize)
+ vmtruncate(inode, isize);
+ }
+ }
out:
- if (release_i_mutex)
- mutex_unlock(&inode->i_mutex);
- else if (acquire_i_mutex)
- mutex_lock(&inode->i_mutex);
return retval;
}
EXPORT_SYMBOL(__blockdev_direct_IO);