Release 4.7 fs/direct-io.c
/*
* fs/direct-io.c
*
* Copyright (C) 2002, Linus Torvalds.
*
* O_DIRECT
*
* 04Jul2002 Andrew Morton
* Initial version
* 11Sep2002 janetinc@us.ibm.com
* added readv/writev support.
* 29Oct2002 Andrew Morton
* rewrote bio_add_page() support.
* 30Oct2002 pbadari@us.ibm.com
* added support for non-aligned IO.
* 06Nov2002 pbadari@us.ibm.com
* added asynchronous IO support.
* 21Jul2003 nathans@sgi.com
* added IO completion notifier.
*/
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/types.h>
#include <linux/fs.h>
#include <linux/mm.h>
#include <linux/slab.h>
#include <linux/highmem.h>
#include <linux/pagemap.h>
#include <linux/task_io_accounting_ops.h>
#include <linux/bio.h>
#include <linux/wait.h>
#include <linux/err.h>
#include <linux/blkdev.h>
#include <linux/buffer_head.h>
#include <linux/rwsem.h>
#include <linux/uio.h>
#include <linux/atomic.h>
#include <linux/prefetch.h>
/*
* How many user pages to map in one call to get_user_pages(). This determines
* the size of a structure in the slab cache
*/
#define DIO_PAGES 64
/*
* This code generally works in units of "dio_blocks". A dio_block is
* somewhere between the hard sector size and the filesystem block size. it
* is determined on a per-invocation basis. When talking to the filesystem
* we need to convert dio_blocks to fs_blocks by scaling the dio_block quantity
* down by dio->blkfactor. Similarly, fs-blocksize quantities are converted
* to bio_block quantities by shifting left by blkfactor.
*
* If blkfactor is zero then the user's request was aligned to the filesystem's
* blocksize.
*/
/* dio_state only used in the submission path */
struct dio_submit {
struct bio *bio; /* bio under assembly */
unsigned blkbits; /* doesn't change */
unsigned blkfactor; /* When we're using an alignment which
is finer than the filesystem's soft
blocksize, this specifies how much
finer. blkfactor=2 means 1/4-block
alignment. Does not change */
unsigned start_zero_done; /* flag: sub-blocksize zeroing has
been performed at the start of a
write */
int pages_in_io; /* approximate total IO pages */
sector_t block_in_file; /* Current offset into the underlying
file in dio_block units. */
unsigned blocks_available; /* At block_in_file. changes */
int reap_counter; /* rate limit reaping */
sector_t final_block_in_request;/* doesn't change */
int boundary; /* prev block is at a boundary */
get_block_t *get_block; /* block mapping function */
dio_submit_t *submit_io; /* IO submition function */
loff_t logical_offset_in_bio; /* current first logical block in bio */
sector_t final_block_in_bio; /* current final block in bio + 1 */
sector_t next_block_for_io; /* next block to be put under IO,
in dio_blocks units */
/*
* Deferred addition of a page to the dio. These variables are
* private to dio_send_cur_page(), submit_page_section() and
* dio_bio_add_page().
*/
struct page *cur_page; /* The page */
unsigned cur_page_offset; /* Offset into it, in bytes */
unsigned cur_page_len; /* Nr of bytes at cur_page_offset */
sector_t cur_page_block; /* Where it starts */
loff_t cur_page_fs_offset; /* Offset in file */
struct iov_iter *iter;
/*
* Page queue. These variables belong to dio_refill_pages() and
* dio_get_page().
*/
unsigned head; /* next page to process */
unsigned tail; /* last valid page + 1 */
size_t from, to;
};
/* dio_state communicated between submission path and end_io */
struct dio {
int flags; /* doesn't change */
int rw;
blk_qc_t bio_cookie;
struct block_device *bio_bdev;
struct inode *inode;
loff_t i_size; /* i_size when submitted */
dio_iodone_t *end_io; /* IO completion function */
void *private; /* copy from map_bh.b_private */
/* BIO completion state */
spinlock_t bio_lock; /* protects BIO fields below */
int page_errors; /* errno from get_user_pages() */
int is_async; /* is IO async ? */
bool defer_completion; /* defer AIO completion to workqueue? */
bool should_dirty; /* if pages should be dirtied */
int io_error; /* IO error in completion path */
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 */
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.
*/
union {
struct page *pages[DIO_PAGES]; /* page buffer */
struct work_struct complete_work;/* deferred AIO completion */
};
}
____cacheline_aligned_in_smp;
static struct kmem_cache *dio_cache __read_mostly;
/*
* How many pages are in the queue?
*/
static inline unsigned dio_pages_present(struct dio_submit *sdio)
{
return sdio->tail - sdio->head;
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
andrew morton | andrew morton | 17 | 80.95% | 2 | 66.67% |
andi kleen | andi kleen | 4 | 19.05% | 1 | 33.33% |
| Total | 21 | 100.00% | 3 | 100.00% |
/*
* Go grab and pin some userspace pages. Typically we'll get 64 at a time.
*/
static inline int dio_refill_pages(struct dio *dio, struct dio_submit *sdio)
{
ssize_t ret;
ret = iov_iter_get_pages(sdio->iter, dio->pages, LONG_MAX, DIO_PAGES,
&sdio->from);
if (ret < 0 && sdio->blocks_available && (dio->rw & WRITE)) {
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
* leaking stale data in the file.
*/
if (dio->page_errors == 0)
dio->page_errors = ret;
get_page(page);
dio->pages[0] = page;
sdio->head = 0;
sdio->tail = 1;
sdio->from = 0;
sdio->to = PAGE_SIZE;
return 0;
}
if (ret >= 0) {
iov_iter_advance(sdio->iter, ret);
ret += sdio->from;
sdio->head = 0;
sdio->tail = (ret + PAGE_SIZE - 1) / PAGE_SIZE;
sdio->to = ((ret - 1) & (PAGE_SIZE - 1)) + 1;
return 0;
}
return ret;
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
andrew morton | andrew morton | 107 | 54.31% | 3 | 27.27% |
al viro | al viro | 57 | 28.93% | 1 | 9.09% |
nick piggin | nick piggin | 15 | 7.61% | 2 | 18.18% |
andi kleen | andi kleen | 14 | 7.11% | 2 | 18.18% |
miklos szeredi | miklos szeredi | 2 | 1.02% | 1 | 9.09% |
kirill a. shutemov | kirill a. shutemov | 1 | 0.51% | 1 | 9.09% |
jens axboe | jens axboe | 1 | 0.51% | 1 | 9.09% |
| Total | 197 | 100.00% | 11 | 100.00% |
/*
* Get another userspace page. Returns an ERR_PTR on error. Pages are
* buffered inside the dio so that we can call get_user_pages() against a
* decent number of pages, less frequently. To provide nicer use of the
* L1 cache.
*/
static inline struct page *dio_get_page(struct dio *dio,
struct dio_submit *sdio)
{
if (dio_pages_present(sdio) == 0) {
int ret;
ret = dio_refill_pages(dio, sdio);
if (ret)
return ERR_PTR(ret);
BUG_ON(dio_pages_present(sdio) == 0);
}
return dio->pages[sdio->head];
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
andrew morton | andrew morton | 59 | 81.94% | 1 | 25.00% |
andi kleen | andi kleen | 10 | 13.89% | 2 | 50.00% |
boaz harrosh | boaz harrosh | 3 | 4.17% | 1 | 25.00% |
| Total | 72 | 100.00% | 4 | 100.00% |
/**
* dio_complete() - called when all DIO BIO I/O has been completed
* @offset: the byte offset in the file of the completed operation
*
* This drops i_dio_count, lets interested parties know that a DIO operation
* has completed, and calculates the resulting return code for the operation.
*
* It lets the filesystem know if it registered an interest earlier via
* get_block. Pass the private field of the map buffer_head so that
* filesystems can use it to hold additional state between get_block calls and
* dio_complete.
*/
static ssize_t dio_complete(struct dio *dio, ssize_t ret, bool is_async)
{
loff_t offset = dio->iocb->ki_pos;
ssize_t transferred = 0;
/*
* AIO submission can race with bio completion to get here while
* expecting to have the last io completed by bio completion.
* In that case -EIOCBQUEUED is in fact not an error we want
* to preserve through this call.
*/
if (ret == -EIOCBQUEUED)
ret = 0;
if (dio->result) {
transferred = dio->result;
/* Check for short read case */
if ((dio->rw == READ) && ((offset + transferred) > dio->i_size))
transferred = dio->i_size - offset;
}
if (ret == 0)
ret = dio->page_errors;
if (ret == 0)
ret = dio->io_error;
if (ret == 0)
ret = transferred;
if (dio->end_io) {
int err;
// XXX: ki_pos??
err = dio->end_io(dio->iocb, offset, ret, dio->private);
if (err)
ret = err;
}
if (!(dio->flags & DIO_SKIP_DIO_COUNT))
inode_dio_end(dio->inode);
if (is_async) {
/*
* generic_write_sync expects ki_pos to have been updated
* already, but the submission path only does this for
* synchronous I/O.
*/
dio->iocb->ki_pos += transferred;
if (dio->rw & WRITE)
ret = generic_write_sync(dio->iocb, transferred);
dio->iocb->ki_complete(dio->iocb, ret, 0);
}
kmem_cache_free(dio_cache, dio);
return ret;
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
christoph hellwig | christoph hellwig | 120 | 50.42% | 8 | 53.33% |
zach brown | zach brown | 66 | 27.73% | 2 | 13.33% |
stephen lord | stephen lord | 31 | 13.03% | 1 | 6.67% |
jens axboe | jens axboe | 12 | 5.04% | 1 | 6.67% |
jan kara | jan kara | 6 | 2.52% | 1 | 6.67% |
edward shishkin | edward shishkin | 2 | 0.84% | 1 | 6.67% |
andi kleen | andi kleen | 1 | 0.42% | 1 | 6.67% |
| Total | 238 | 100.00% | 15 | 100.00% |
static void dio_aio_complete_work(struct work_struct *work)
{
struct dio *dio = container_of(work, struct dio, complete_work);
dio_complete(dio, 0, true);
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
christoph hellwig | christoph hellwig | 35 | 100.00% | 1 | 100.00% |
| Total | 35 | 100.00% | 1 | 100.00% |
static int dio_bio_complete(struct dio *dio, struct bio *bio);
/*
* Asynchronous IO callback.
*/
static void dio_bio_end_aio(struct bio *bio)
{
struct dio *dio = bio->bi_private;
unsigned long remaining;
unsigned long flags;
/* cleanup the bio */
dio_bio_complete(dio, bio);
spin_lock_irqsave(&dio->bio_lock, flags);
remaining = --dio->refcount;
if (remaining == 1 && dio->waiter)
wake_up_process(dio->waiter);
spin_unlock_irqrestore(&dio->bio_lock, flags);
if (remaining == 0) {
if (dio->result && dio->defer_completion) {
INIT_WORK(&dio->complete_work, dio_aio_complete_work);
queue_work(dio->inode->i_sb->s_dio_done_wq,
&dio->complete_work);
} else {
dio_complete(dio, 0, true);
}
}
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
zach brown | zach brown | 68 | 49.28% | 4 | 50.00% |
christoph hellwig | christoph hellwig | 42 | 30.43% | 2 | 25.00% |
andrew morton | andrew morton | 27 | 19.57% | 1 | 12.50% |
neil brown | neil brown | 1 | 0.72% | 1 | 12.50% |
| Total | 138 | 100.00% | 8 | 100.00% |
/*
* The BIO completion handler simply queues the BIO up for the process-context
* handler.
*
* 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 void dio_bio_end_io(struct bio *bio)
{
struct dio *dio = bio->bi_private;
unsigned long flags;
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);
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
andrew morton | andrew morton | 72 | 92.31% | 4 | 57.14% |
zach brown | zach brown | 5 | 6.41% | 2 | 28.57% |
neil brown | neil brown | 1 | 1.28% | 1 | 14.29% |
| Total | 78 | 100.00% | 7 | 100.00% |
/**
* dio_end_io - handle the end io action for the given bio
* @bio: The direct io bio thats being completed
* @error: Error if there was one
*
* This is meant to be called by any filesystem that uses their own dio_submit_t
* so that the DIO specific endio actions are dealt with after the filesystem
* has done it's completion work.
*/
void dio_end_io(struct bio *bio, int error)
{
struct dio *dio = bio->bi_private;
if (dio->is_async)
dio_bio_end_aio(bio);
else
dio_bio_end_io(bio);
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
josef bacik | josef bacik | 39 | 100.00% | 1 | 100.00% |
| Total | 39 | 100.00% | 1 | 100.00% |
EXPORT_SYMBOL_GPL(dio_end_io);
static inline void
dio_bio_alloc(struct dio *dio, struct dio_submit *sdio,
struct block_device *bdev,
sector_t first_sector, int nr_vecs)
{
struct bio *bio;
/*
* bio_alloc() is guaranteed to return a bio when called with
* __GFP_RECLAIM and we request a valid number of vectors.
*/
bio = bio_alloc(GFP_KERNEL, nr_vecs);
bio->bi_bdev = bdev;
bio->bi_iter.bi_sector = first_sector;
if (dio->is_async)
bio->bi_end_io = dio_bio_end_aio;
else
bio->bi_end_io = dio_bio_end_io;
sdio->bio = bio;
sdio->logical_offset_in_bio = sdio->cur_page_fs_offset;
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
andrew morton | andrew morton | 71 | 78.89% | 2 | 25.00% |
andi kleen | andi kleen | 9 | 10.00% | 2 | 25.00% |
josef bacik | josef bacik | 6 | 6.67% | 1 | 12.50% |
kent overstreet | kent overstreet | 2 | 2.22% | 1 | 12.50% |
mel gorman | mel gorman | 1 | 1.11% | 1 | 12.50% |
david dillow | david dillow | 1 | 1.11% | 1 | 12.50% |
| Total | 90 | 100.00% | 8 | 100.00% |
/*
* In the AIO read case we speculatively dirty the pages before starting IO.
* During IO completion, any of these pages which happen to have been written
* back will be redirtied by bio_check_pages_dirty().
*
* bios hold a dio reference between submit_bio and ->end_io.
*/
static inline void dio_bio_submit(struct dio *dio, struct dio_submit *sdio)
{
struct bio *bio = sdio->bio;
unsigned long flags;
bio->bi_private = dio;
spin_lock_irqsave(&dio->bio_lock, flags);
dio->refcount++;
spin_unlock_irqrestore(&dio->bio_lock, flags);
if (dio->is_async && dio->rw == READ && dio->should_dirty)
bio_set_pages_dirty(bio);
dio->bio_bdev = bio->bi_bdev;
if (sdio->submit_io) {
sdio->submit_io(dio->rw, bio, dio->inode,
sdio->logical_offset_in_bio);
dio->bio_cookie = BLK_QC_T_NONE;
} else
dio->bio_cookie = submit_bio(dio->rw, bio);
sdio->bio = NULL;
sdio->boundary = 0;
sdio->logical_offset_in_bio = 0;
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
andrew morton | andrew morton | 67 | 43.23% | 4 | 33.33% |
josef bacik | josef bacik | 28 | 18.06% | 1 | 8.33% |
zach brown | zach brown | 23 | 14.84% | 2 | 16.67% |
jens axboe | jens axboe | 20 | 12.90% | 2 | 16.67% |
andi kleen | andi kleen | 13 | 8.39% | 2 | 16.67% |
ming lei | ming lei | 4 | 2.58% | 1 | 8.33% |
| Total | 155 | 100.00% | 12 | 100.00% |
/*
* Release any resources in case of a failure
*/
static inline void dio_cleanup(struct dio *dio, struct dio_submit *sdio)
{
while (sdio->head < sdio->tail)
put_page(dio->pages[sdio->head++]);
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
andrew morton | andrew morton | 18 | 45.00% | 1 | 20.00% |
al viro | al viro | 13 | 32.50% | 1 | 20.00% |
andi kleen | andi kleen | 8 | 20.00% | 2 | 40.00% |
kirill a. shutemov | kirill a. shutemov | 1 | 2.50% | 1 | 20.00% |
| Total | 40 | 100.00% | 5 | 100.00% |
/*
* Wait for the next BIO to complete. Remove it and return it. NULL is
* returned once all BIOs have been completed. This must only be called once
* all bios have been issued so that dio->refcount can only decrease. This
* requires that that the caller hold a reference on the dio.
*/
static struct bio *dio_await_one(struct dio *dio)
{
unsigned long flags;
struct bio *bio = NULL;
spin_lock_irqsave(&dio->bio_lock, flags);
/*
* Wait as long as the list is empty and there are bios in flight. bio
* completion drops the count, maybe adds to the list, and wakes while
* holding the bio_lock so we don't need set_current_state()'s barrier
* and can call it after testing our condition.
*/
while (dio->refcount > 1 && dio->bio_list == NULL) {
__set_current_state(TASK_UNINTERRUPTIBLE);
dio->waiter = current;
spin_unlock_irqrestore(&dio->bio_lock, flags);
if (!(dio->iocb->ki_flags & IOCB_HIPRI) ||
!blk_poll(bdev_get_queue(dio->bio_bdev), dio->bio_cookie))
io_schedule();
/* wake up sets us TASK_RUNNING */
spin_lock_irqsave(&dio->bio_lock, flags);
dio->waiter = NULL;
}
if (dio->bio_list) {
bio = dio->bio_list;
dio->bio_list = bio->bi_private;
}
spin_unlock_irqrestore(&dio->bio_lock, flags);
return bio;
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
andrew morton | andrew morton | 104 | 67.10% | 4 | 50.00% |
zach brown | zach brown | 23 | 14.84% | 2 | 25.00% |
jens axboe | jens axboe | 17 | 10.97% | 1 | 12.50% |
christoph hellwig | christoph hellwig | 11 | 7.10% | 1 | 12.50% |
| Total | 155 | 100.00% | 8 | 100.00% |
/*
* Process one completed BIO. No locks are held.
*/
static int dio_bio_complete(struct dio *dio, struct bio *bio)
{
struct bio_vec *bvec;
unsigned i;
int err;
if (bio->bi_error)
dio->io_error = -EIO;
if (dio->is_async && dio->rw == READ && dio->should_dirty) {
err = bio->bi_error;
bio_check_pages_dirty(bio); /* transfers ownership */
} else {
bio_for_each_segment_all(bvec, bio, i) {
struct page *page = bvec->bv_page;
if (dio->rw == READ && !PageCompound(page) &&
dio->should_dirty)
set_page_dirty_lock(page);
put_page(page);
}
err = bio->bi_error;
bio_put(bio);
}
return err;
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
andrew morton | andrew morton | 82 | 61.19% | 3 | 27.27% |
sasha levin | sasha levin | 16 | 11.94% | 1 | 9.09% |
kent overstreet | kent overstreet | 11 | 8.21% | 1 | 9.09% |
ming lei | ming lei | 8 | 5.97% | 1 | 9.09% |
kenneth w. chen | kenneth w. chen | 7 | 5.22% | 2 | 18.18% |
mike krinkin | mike krinkin | 6 | 4.48% | 1 | 9.09% |
christoph hellwig | christoph hellwig | 3 | 2.24% | 1 | 9.09% |
kirill a. shutemov | kirill a. shutemov | 1 | 0.75% | 1 | 9.09% |
| Total | 134 | 100.00% | 11 | 100.00% |
/*
* Wait on and process all in-flight BIOs. This must only be called once
* all bios have been issued so that the refcount can only decrease.
* This just waits for all bios to make it through dio_bio_complete. IO
* errors are propagated through dio->io_error and should be propagated via
* dio_complete().
*/
static void dio_await_completion(struct dio *dio)
{
struct bio *bio;
do {
bio = dio_await_one(dio);
if (bio)
dio_bio_complete(dio, bio);
} while (bio);
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
andrew morton | andrew morton | 26 | 61.90% | 1 | 33.33% |
zach brown | zach brown | 16 | 38.10% | 2 | 66.67% |
| Total | 42 | 100.00% | 3 | 100.00% |
/*
* A really large O_DIRECT read or write can generate a lot of BIOs. So
* to keep the memory consumption sane we periodically reap any completed BIOs
* during the BIO generation phase.
*
* This also helps to limit the peak amount of pinned userspace memory.
*/
static inline int dio_bio_reap(struct dio *dio, struct dio_submit *sdio)
{
int ret = 0;
if (sdio->reap_counter++ >= 64) {
while (dio->bio_list) {
unsigned long flags;
struct bio *bio;
int ret2;
spin_lock_irqsave(&dio->bio_lock, flags);
bio = dio->bio_list;
dio->bio_list = bio->bi_private;
spin_unlock_irqrestore(&dio->bio_lock, flags);
ret2 = dio_bio_complete(dio, bio);
if (ret == 0)
ret = ret2;
}
sdio->reap_counter = 0;
}
return ret;
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
andrew morton | andrew morton | 107 | 93.04% | 3 | 60.00% |
andi kleen | andi kleen | 8 | 6.96% | 2 | 40.00% |
| Total | 115 | 100.00% | 5 | 100.00% |
/*
* Create workqueue for deferred direct IO completions. We allocate the
* workqueue when it's first needed. This avoids creating workqueue for
* filesystems that don't need it and also allows us to create the workqueue
* late enough so the we can include s_id in the name of the workqueue.
*/
static int sb_init_dio_done_wq(struct super_block *sb)
{
struct workqueue_struct *old;
struct workqueue_struct *wq = alloc_workqueue("dio/%s",
WQ_MEM_RECLAIM, 0,
sb->s_id);
if (!wq)
return -ENOMEM;
/*
* This has to be atomic as more DIOs can race to create the workqueue
*/
old = cmpxchg(&sb->s_dio_done_wq, NULL, wq);
/* Someone created workqueue before us? Free ours... */
if (old)
destroy_workqueue(wq);
return 0;
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
christoph hellwig | christoph hellwig | 63 | 88.73% | 1 | 50.00% |
olof johansson | olof johansson | 8 | 11.27% | 1 | 50.00% |
| Total | 71 | 100.00% | 2 | 100.00% |
static int dio_set_defer_completion(struct dio *dio)
{
struct super_block *sb = dio->inode->i_sb;
if (dio->defer_completion)
return 0;
dio->defer_completion = true;
if (!sb->s_dio_done_wq)
return sb_init_dio_done_wq(sb);
return 0;
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
christoph hellwig | christoph hellwig | 53 | 100.00% | 1 | 100.00% |
| Total | 53 | 100.00% | 1 | 100.00% |
/*
* Call into the fs to map some more disk blocks. We record the current number
* of available blocks at sdio->blocks_available. These are in units of the
* fs blocksize, (1 << inode->i_blkbits).
*
* The fs is allowed to map lots of blocks at once. If it wants to do that,
* it uses the passed inode-relative block number as the file offset, as usual.
*
* get_block() is passed the number of i_blkbits-sized blocks which direct_io
* has remaining to do. The fs should not map more than this number of blocks.
*
* If the fs has mapped a lot of blocks, it should populate bh->b_size to
* indicate how much contiguous disk space has been made available at
* bh->b_blocknr.
*
* If *any* of the mapped blocks are new, then the fs must set buffer_new().
* This isn't very efficient...
*
* In the case of filesystem holes: the fs may return an arbitrarily-large
* hole by returning an appropriate value in b_size and by clearing
* buffer_mapped(). However the direct-io code will only process holes one
* block at a time - it will repeatedly call get_block() as it walks the hole.
*/
static int get_more_blocks(struct dio *dio, struct dio_submit *sdio,
struct buffer_head *map_bh)
{
int ret;
sector_t fs_startblk; /* Into file, in filesystem-sized blocks */
sector_t fs_endblk; /* Into file, in filesystem-sized blocks */
unsigned long fs_count; /* Number of filesystem-sized blocks */
int create;
unsigned int i_blkbits = sdio->blkbits + sdio->blkfactor;
/*
* If there was a memory error and we've overwritten all the
* mapped blocks then we can now return that memory error
*/
ret = dio->page_errors;
if (ret == 0) {
BUG_ON(sdio->block_in_file >= sdio->final_block_in_request);
fs_startblk = sdio->block_in_file >> sdio->blkfactor;
fs_endblk = (sdio->final_block_in_request - 1) >>
sdio->blkfactor;
fs_count = fs_endblk - fs_startblk + 1;
map_bh->b_state = 0;
map_bh->b_size = fs_count << i_blkbits;
/*
* For writes that could fill holes 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->flags & DIO_SKIP_HOLES) {
if (fs_startblk <= ((i_size_read(dio->inode) - 1) >>
i_blkbits))
create = 0;
}
ret = (*sdio->get_block)(dio->inode, fs_startblk,
map_bh, create);
/* Store for completion */
dio->private = map_bh->b_private;
if (ret == 0 && buffer_defer_completion(map_bh))
ret = dio_set_defer_completion(dio);
}
return ret;
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
andrew morton | andrew morton | 108 | 50.00% | 5 | 31.25% |
nathan scott | nathan scott | 27 | 12.50% | 2 | 12.50% |
andi kleen | andi kleen | 26 | 12.04% | 2 | 12.50% |
christoph hellwig | christoph hellwig | 21 | 9.72% | 2 | 12.50% |
linus torvalds | linus torvalds | 12 | 5.56% | 1 | 6.25% |
tao ma | tao ma | 12 | 5.56% | 1 | 6.25% |
eryu guan | eryu guan | 8 | 3.70% | 1 | 6.25% |
badari pulavarty | badari pulavarty | 1 | 0.46% | 1 | 6.25% |
jens axboe | jens axboe | 1 | 0.46% | 1 | 6.25% |
| Total | 216 | 100.00% | 16 | 100.00% |
/*
* There is no bio. Make one now.
*/
static inline int dio_new_bio(struct dio *dio, struct dio_submit *sdio,
sector_t start_sector, struct buffer_head *map_bh)
{
sector_t sector;
int ret, nr_pages;
ret = dio_bio_reap(dio, sdio);
if (ret)
goto out;
sector = start_sector << (sdio->blkbits - 9);
nr_pages = min(sdio->pages_in_io, BIO_MAX_PAGES);
BUG_ON(nr_pages <= 0);
dio_bio_alloc(dio, sdio, map_bh->b_bdev, sector, nr_pages);
sdio->boundary = 0;
out:
return ret;
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
andrew morton | andrew morton | 85 | 80.95% | 6 | 60.00% |
andi kleen | andi kleen | 19 | 18.10% | 3 | 30.00% |
kent overstreet | kent overstreet | 1 | 0.95% | 1 | 10.00% |
| Total | 105 | 100.00% | 10 | 100.00% |
/*
* Attempt to put the current chunk of 'cur_page' into the current BIO. If
* that was successful then update final_block_in_bio and take a ref against
* the just-added page.
*
* Return zero on success. Non-zero means the caller needs to start a new BIO.
*/
static inline int dio_bio_add_page(struct dio_submit *sdio)
{
int ret;
ret = bio_add_page(sdio->bio, sdio->cur_page,
sdio->cur_page_len, sdio->cur_page_offset);
if (ret == sdio->cur_page_len) {
/*
* Decrement count only, if we are done with this page
*/
if ((sdio->cur_page_len + sdio->cur_page_offset) == PAGE_SIZE)
sdio->pages_in_io--;
get_page(sdio->cur_page);
sdio->final_block_in_bio = sdio->cur_page_block +
(sdio->cur_page_len >> sdio->blkbits);
ret = 0;
} else {
ret = 1;
}
return ret;
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
andrew morton | andrew morton | 75 | 71.43% | 3 | 42.86% |
andi kleen | andi kleen | 16 | 15.24% | 2 | 28.57% |
badari pulavarty | badari pulavarty | 13 | 12.38% | 1 | 14.29% |
kirill a. shutemov | kirill a. shutemov | 1 | 0.95% | 1 | 14.29% |
| Total | 105 | 100.00% | 7 | 100.00% |
/*
* Put cur_page under IO. The section of cur_page which is described by
* cur_page_offset,cur_page_len is put into a BIO. The section of cur_page
* starts on-disk at cur_page_block.
*
* We take a ref against the page here (on behalf of its presence in the bio).
*
* The caller of this function is responsible for removing cur_page from the
* dio, and for dropping the refcount which came from that presence.
*/
static inline int dio_send_cur_page(struct dio *dio, struct dio_submit *sdio,
struct buffer_head *map_bh)
{
int ret = 0;
if (sdio->bio) {
loff_t cur_offset = sdio->cur_page_fs_offset;
loff_t bio_next_offset = sdio->logical_offset_in_bio +
sdio->bio->bi_iter.bi_size;
/*
* See whether this new request is contiguous with the old.
*
* Btrfs cannot handle having logically non-contiguous requests
* submitted. For example if you have
*
* Logical: [0-4095][HOLE][8192-12287]
* Physical: [0-4095] [4096-8191]
*
* We cannot submit those pages together as one BIO. So if our
* current logical offset in the file does not equal what would
* be the next logical offset in the bio, submit the bio we
* have.
*/
if (sdio->final_block_in_bio != sdio->cur_page_block ||
cur_offset != bio_next_offset)
dio_bio_submit(dio, sdio);
}
if (sdio->bio == NULL) {
ret = dio_new_bio(dio, sdio, sdio->cur_page_block, map_bh);
if (ret)
goto out;
}
if (dio_bio_add_page(sdio) != 0) {
dio_bio_submit(dio, sdio);
ret = dio_new_bio(dio, sdio, sdio->cur_page_block, map_bh);
if (ret == 0) {
ret = dio_bio_add_page(sdio);
BUG_ON(ret != 0);
}
}
out:
return ret;
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
andrew morton | andrew morton | 113 | 66.08% | 3 | 30.00% |
andi kleen | andi kleen | 34 | 19.88% | 3 | 30.00% |
josef bacik | josef bacik | 20 | 11.70% | 1 | 10.00% |
kent overstreet | kent overstreet | 2 | 1.17% | 1 | 10.00% |
jeff moyer | jeff moyer | 1 | 0.58% | 1 | 10.00% |
namhyung kim | namhyung kim | 1 | 0.58% | 1 | 10.00% |
| Total | 171 | 100.00% | 10 | 100.00% |
/*
* An autonomous function to put a chunk of a page under deferred IO.
*
* The caller doesn't actually know (or care) whether this piece of page is in
* a BIO, or is under IO or whatever. We just take care of all possible
* situations here. The separation between the logic of do_direct_IO() and
* that of submit_page_section() is important for clarity. Please don't break.
*
* The chunk of page starts on-disk at blocknr.
*
* We perform deferred IO, by recording the last-submitted page inside our
* private part of the dio structure. If possible, we just expand the IO
* across that page here.
*
* If that doesn't work out then we put the old page into the bio and add this
* page to the dio instead.
*/
static inline int
submit_page_section(struct dio *dio, struct dio_submit *sdio, struct page *page,
unsigned offset, unsigned len, sector_t blocknr,
struct buffer_head *map_bh)
{
int ret = 0;
if (dio->rw & WRITE) {
/*
* Read accounting is performed in submit_bio()
*/
task_io_account_write(len);
}
/*
* Can we just grow the current page's presence in the dio?
*/
if (sdio->cur_page == page &&
sdio->cur_page_offset + sdio->cur_page_len == offset &&
sdio->cur_page_block +
(sdio->cur_page_len >> sdio->blkbits) == blocknr) {
sdio->cur_page_len += len;
goto out;
}
/*
* If there's a deferred page already there then send it.
*/
if (sdio->cur_page) {
ret = dio_send_cur_page(dio, sdio, map_bh);
put_page(sdio->cur_page);
sdio->cur_page = NULL;
if (ret)
return ret;
}
get_page(page); /* It is in dio */
sdio->cur_page = page;
sdio->cur_page_offset = offset;
sdio->cur_page_len = len;
sdio->cur_page_block = blocknr;
sdio->cur_page_fs_offset = sdio->block_in_file << sdio->blkbits;
out:
/*
* If sdio->boundary then we want to schedule the IO now to
* avoid metadata seeks.
*/
if (sdio->boundary) {
ret = dio_send_cur_page(dio, sdio, map_bh);
dio_bio_submit(dio, sdio);
put_page(sdio->cur_page);
sdio->cur_page