niansa-misc/linux
1
0
Fork 0
mirror of synced 2025-03-06 20:59:54 +01:00
Code Issues Projects Releases Packages Wiki Activity

btrfs: convert btrfs_buffered_write() to use folios

Browse source 
The buffered write path is still heavily utilizing the page interface.
Since we have converted it to do a page-by-page copying, it's much easier
to convert all involved functions to folio interface, this involves:

- btrfs_copy_from_user()
- btrfs_drop_folio()
- prepare_uptodate_page()
- prepare_one_page()
- lock_and_cleanup_extent_if_need()
- btrfs_dirty_page()

All function are changed to accept a folio parameter, and if the word
"page" is in the function name, change that to "folio" too.

The function btrfs_dirty_page() is exported for v1 space cache, convert
v1 cache call site to convert its page to folio for the new interface.

And there is a small enhancement for prepare_one_folio(), instead of
manually waiting for the page writeback, let __filemap_get_folio() to
handle that by using FGP_WRITEBEGIN, which implies
(FGP_LOCK | FGP_WRITE | FGP_CREAT | FGP_STABLE).

Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This commit is contained in:
Qu Wenruo 2024-10-10 15:16:13 +10:30 • committed by David Sterba
parent c87c299776
commit e820dbeb6a
3 changed files with 66 additions and 80 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

138
fs/btrfs/file.c
View file

@ -42,7 +42,7 @@
* calls into generic code. * calls into generic code.
*/ */
static noinline int btrfs_copy_from_user(loff_t pos, size_t write_bytes, static noinline int btrfs_copy_from_user(loff_t pos, size_t write_bytes,
struct page *page, struct iov_iter *i) struct folio *folio, struct iov_iter *i)
{ {
size_t copied = 0; size_t copied = 0;
size_t total_copied = 0; size_t total_copied = 0;
@ -53,10 +53,10 @@ static noinline int btrfs_copy_from_user(loff_t pos, size_t write_bytes,
/* /*
* Copy data from userspace to the current page * Copy data from userspace to the current page
*/ */
copied = copy_page_from_iter_atomic(page, offset, count, i); copied = copy_folio_from_iter_atomic(folio, offset, count, i);
/* Flush processor's dcache for this page */ /* Flush processor's dcache for this page */
flush_dcache_page(page); flush_dcache_folio(folio);
/* /*
* if we get a partial write, we can end up with * if we get a partial write, we can end up with
@ -68,7 +68,7 @@ static noinline int btrfs_copy_from_user(loff_t pos, size_t write_bytes,
* back to page at a time copies after we return 0. * back to page at a time copies after we return 0.
*/ */
if (unlikely(copied < count)) { if (unlikely(copied < count)) {
if (!PageUptodate(page)) { if (!folio_test_uptodate(folio)) {
iov_iter_revert(i, copied); iov_iter_revert(i, copied);
copied = 0; copied = 0;
} }
@ -84,37 +84,36 @@ static noinline int btrfs_copy_from_user(loff_t pos, size_t write_bytes,
} }
/* /*
* unlocks pages after btrfs_file_write is done with them * Unlock folio after btrfs_file_write() is done with it.
*/ */
static void btrfs_drop_page(struct btrfs_fs_info *fs_info, struct page *page, static void btrfs_drop_folio(struct btrfs_fs_info *fs_info, struct folio *folio,
u64 pos, u64 copied) u64 pos, u64 copied)
{ {
u64 block_start = round_down(pos, fs_info->sectorsize); u64 block_start = round_down(pos, fs_info->sectorsize);
u64 block_len = round_up(pos + copied, fs_info->sectorsize) - block_start; u64 block_len = round_up(pos + copied, fs_info->sectorsize) - block_start;
ASSERT(block_len <= U32_MAX); ASSERT(block_len <= U32_MAX);
/* /*
* Page checked is some magic around finding pages that have been * Folio checked is some magic around finding folios that have been
* modified without going through btrfs_set_page_dirty clear it here. * modified without going through btrfs_dirty_folio(). Clear it here.
* There should be no need to mark the pages accessed as * There should be no need to mark the pages accessed as
* prepare_one_page() should have marked them accessed in * prepare_one_folio() should have marked them accessed in
* prepare_one_page() via find_or_create_page() * prepare_one_folio() via find_or_create_page()
*/ */
btrfs_folio_clamp_clear_checked(fs_info, page_folio(page), block_start, btrfs_folio_clamp_clear_checked(fs_info, folio, block_start, block_len);
block_len); folio_unlock(folio);
unlock_page(page); folio_put(folio);
put_page(page);
} }
/* /*
* After btrfs_copy_from_user(), update the following things for delalloc: * After btrfs_copy_from_user(), update the following things for delalloc:
* - Mark newly dirtied pages as DELALLOC in the io tree. * - Mark newly dirtied folio as DELALLOC in the io tree.
* Used to advise which range is to be written back. * Used to advise which range is to be written back.
* - Mark modified pages as Uptodate/Dirty and not needing COW fixup * - Mark modified folio as Uptodate/Dirty and not needing COW fixup
* - Update inode size for past EOF write * - Update inode size for past EOF write
*/ */
int btrfs_dirty_page(struct btrfs_inode *inode, struct page *page, loff_t pos, int btrfs_dirty_folio(struct btrfs_inode *inode, struct folio *folio, loff_t pos,
size_t write_bytes, struct extent_state **cached, bool noreserve) size_t write_bytes, struct extent_state **cached, bool noreserve)
{ {
struct btrfs_fs_info *fs_info = inode->root->fs_info; struct btrfs_fs_info *fs_info = inode->root->fs_info;
int ret = 0; int ret = 0;
@ -122,7 +121,6 @@ int btrfs_dirty_page(struct btrfs_inode *inode, struct page *page, loff_t pos,
u64 start_pos; u64 start_pos;
u64 end_of_last_block; u64 end_of_last_block;
u64 end_pos = pos + write_bytes; u64 end_pos = pos + write_bytes;
struct folio *folio = page_folio(page);
loff_t isize = i_size_read(&inode->vfs_inode); loff_t isize = i_size_read(&inode->vfs_inode);
unsigned int extra_bits = 0; unsigned int extra_bits = 0;
@ -835,14 +833,12 @@ out:
} }
/* /*
* on error we return an unlocked page and the error value * On error return an unlocked folio and the error value
* on success we return a locked page and 0 * On success return a locked folio and 0
*/ */
static int prepare_uptodate_page(struct inode *inode, static int prepare_uptodate_folio(struct inode *inode, struct folio *folio, u64 pos,
struct page *page, u64 pos, u64 len, bool force_uptodate)
u64 len, bool force_uptodate)
{ {
struct folio *folio = page_folio(page);
u64 clamp_start = max_t(u64, pos, folio_pos(folio)); u64 clamp_start = max_t(u64, pos, folio_pos(folio));
u64 clamp_end = min_t(u64, pos + len, folio_pos(folio) + folio_size(folio)); u64 clamp_end = min_t(u64, pos + len, folio_pos(folio) + folio_size(folio));
int ret = 0; int ret = 0;
@ -873,23 +869,13 @@ static int prepare_uptodate_page(struct inode *inode,
* The private flag check is essential for subpage as we need to store * The private flag check is essential for subpage as we need to store
* extra bitmap using folio private. * extra bitmap using folio private.
*/ */
if (page->mapping != inode->i_mapping || !folio_test_private(folio)) { if (folio->mapping != inode->i_mapping || !folio_test_private(folio)) {
folio_unlock(folio); folio_unlock(folio);
return -EAGAIN; return -EAGAIN;
} }
return 0; return 0;
} }
static fgf_t get_prepare_fgp_flags(bool nowait)
{
fgf_t fgp_flags = FGP_LOCK | FGP_ACCESSED | FGP_CREAT;
if (nowait)
fgp_flags |= FGP_NOWAIT;
return fgp_flags;
}
static gfp_t get_prepare_gfp_flags(struct inode *inode, bool nowait) static gfp_t get_prepare_gfp_flags(struct inode *inode, bool nowait)
{ {
gfp_t gfp; gfp_t gfp;
@ -904,60 +890,60 @@ static gfp_t get_prepare_gfp_flags(struct inode *inode, bool nowait)
} }
/* /*
* this just gets page into the page cache and locks them down. * Get folio into the page cache and lock it.
*/ */
static noinline int prepare_one_page(struct inode *inode, struct page **page_ret, static noinline int prepare_one_folio(struct inode *inode, struct folio **folio_ret,
loff_t pos, size_t write_bytes, loff_t pos, size_t write_bytes,
bool force_uptodate, bool nowait) bool force_uptodate, bool nowait)
{ {
unsigned long index = pos >> PAGE_SHIFT; unsigned long index = pos >> PAGE_SHIFT;
gfp_t mask = get_prepare_gfp_flags(inode, nowait); gfp_t mask = get_prepare_gfp_flags(inode, nowait);
fgf_t fgp_flags = get_prepare_fgp_flags(nowait); fgf_t fgp_flags = (nowait ? FGP_WRITEBEGIN | FGP_NOWAIT : FGP_WRITEBEGIN);
struct page *page; struct folio *folio;
int ret = 0; int ret = 0;
again: again:
page = pagecache_get_page(inode->i_mapping, index, fgp_flags, folio = __filemap_get_folio(inode->i_mapping, index, fgp_flags, mask);
mask | __GFP_WRITE); if (IS_ERR(folio)) {
if (!page) {
if (nowait) if (nowait)
ret = -EAGAIN; ret = -EAGAIN;
else else
ret = -ENOMEM; ret = PTR_ERR(folio);
return ret; return ret;
} }
ret = set_page_extent_mapped(page); /* Only support page sized folio yet. */
ASSERT(folio_order(folio) == 0);
ret = set_folio_extent_mapped(folio);
if (ret < 0) { if (ret < 0) {
unlock_page(page); folio_unlock(folio);
put_page(page); folio_put(folio);
return ret; return ret;
} }
ret = prepare_uptodate_page(inode, page, pos, write_bytes, force_uptodate); ret = prepare_uptodate_folio(inode, folio, pos, write_bytes, force_uptodate);
if (ret) { if (ret) {
/* The page is already unlocked. */ /* The folio is already unlocked. */
put_page(page); folio_put(folio);
if (!nowait && ret == -EAGAIN) { if (!nowait && ret == -EAGAIN) {
ret = 0; ret = 0;
goto again; goto again;
} }
return ret; return ret;
} }
wait_on_page_writeback(page); *folio_ret = folio;
*page_ret = page;
return 0; return 0;
} }
/* /*
* This function locks the extent and properly waits for data=ordered extents * Locks the extent and properly waits for data=ordered extents to finish
* to finish before allowing the pages to be modified if need. * before allowing the folios to be modified if need.
* *
* The return value: * Return:
* 1 - the extent is locked * 1 - the extent is locked
* 0 - the extent is not locked, and everything is OK * 0 - the extent is not locked, and everything is OK
* -EAGAIN - need re-prepare the pages * -EAGAIN - need to prepare the folios again
*/ */
static noinline int static noinline int
lock_and_cleanup_extent_if_need(struct btrfs_inode *inode, struct page *page, lock_and_cleanup_extent_if_need(struct btrfs_inode *inode, struct folio *folio,
loff_t pos, size_t write_bytes, loff_t pos, size_t write_bytes,
u64 *lockstart, u64 *lockend, bool nowait, u64 *lockstart, u64 *lockend, bool nowait,
struct extent_state **cached_state) struct extent_state **cached_state)
@ -976,8 +962,8 @@ lock_and_cleanup_extent_if_need(struct btrfs_inode *inode, struct page *page,
if (nowait) { if (nowait) {
if (!try_lock_extent(&inode->io_tree, start_pos, last_pos, if (!try_lock_extent(&inode->io_tree, start_pos, last_pos,
cached_state)) { cached_state)) {
unlock_page(page); folio_unlock(folio);
put_page(page); folio_put(folio);
return -EAGAIN; return -EAGAIN;
} }
} else { } else {
@ -991,8 +977,8 @@ lock_and_cleanup_extent_if_need(struct btrfs_inode *inode, struct page *page,
ordered->file_offset <= last_pos) { ordered->file_offset <= last_pos) {
unlock_extent(&inode->io_tree, start_pos, last_pos, unlock_extent(&inode->io_tree, start_pos, last_pos,
cached_state); cached_state);
unlock_page(page); folio_unlock(folio);
put_page(page); folio_put(folio);
btrfs_start_ordered_extent(ordered); btrfs_start_ordered_extent(ordered);
btrfs_put_ordered_extent(ordered); btrfs_put_ordered_extent(ordered);
return -EAGAIN; return -EAGAIN;
@ -1006,10 +992,10 @@ lock_and_cleanup_extent_if_need(struct btrfs_inode *inode, struct page *page,
} }
/* /*
* We should be called after prepare_one_page() which should have locked * We should be called after prepare_one_folio() which should have locked
* all pages in the range. * all pages in the range.
*/ */
WARN_ON(!PageLocked(page)); WARN_ON(!folio_test_locked(folio));
return ret; return ret;
} }
@ -1190,12 +1176,12 @@ ssize_t btrfs_buffered_write(struct kiocb *iocb, struct iov_iter *i)
size_t copied; size_t copied;
size_t dirty_sectors; size_t dirty_sectors;
size_t num_sectors; size_t num_sectors;
struct page *page = NULL; struct folio *folio = NULL;
int extents_locked; int extents_locked;
bool force_page_uptodate = false; bool force_page_uptodate = false;
/* /*
* Fault pages before locking them in prepare_one_page() * Fault pages before locking them in prepare_one_folio()
* to avoid recursive lock * to avoid recursive lock
*/ */
if (unlikely(fault_in_iov_iter_readable(i, write_bytes))) { if (unlikely(fault_in_iov_iter_readable(i, write_bytes))) {
@ -1261,8 +1247,8 @@ again:
break; break;
} }
ret = prepare_one_page(inode, &page, pos, write_bytes, ret = prepare_one_folio(inode, &folio, pos, write_bytes,
force_page_uptodate, false); force_page_uptodate, false);
if (ret) { if (ret) {
btrfs_delalloc_release_extents(BTRFS_I(inode), btrfs_delalloc_release_extents(BTRFS_I(inode),
reserve_bytes); reserve_bytes);
@ -1270,7 +1256,7 @@ again:
} }
extents_locked = lock_and_cleanup_extent_if_need(BTRFS_I(inode), extents_locked = lock_and_cleanup_extent_if_need(BTRFS_I(inode),
page, pos, write_bytes, &lockstart, folio, pos, write_bytes, &lockstart,
&lockend, nowait, &cached_state); &lockend, nowait, &cached_state);
if (extents_locked < 0) { if (extents_locked < 0) {
if (!nowait && extents_locked == -EAGAIN) if (!nowait && extents_locked == -EAGAIN)
@ -1282,7 +1268,7 @@ again:
break; break;
} }
copied = btrfs_copy_from_user(pos, write_bytes, page, i); copied = btrfs_copy_from_user(pos, write_bytes, folio, i);
num_sectors = BTRFS_BYTES_TO_BLKS(fs_info, reserve_bytes); num_sectors = BTRFS_BYTES_TO_BLKS(fs_info, reserve_bytes);
dirty_sectors = round_up(copied + sector_offset, dirty_sectors = round_up(copied + sector_offset,
@ -1314,8 +1300,8 @@ again:
release_bytes = round_up(copied + sector_offset, release_bytes = round_up(copied + sector_offset,
fs_info->sectorsize); fs_info->sectorsize);
ret = btrfs_dirty_page(BTRFS_I(inode), page, pos, copied, ret = btrfs_dirty_folio(BTRFS_I(inode), folio, pos, copied,
&cached_state, only_release_metadata); &cached_state, only_release_metadata);
/* /*
* If we have not locked the extent range, because the range's * If we have not locked the extent range, because the range's
@ -1332,7 +1318,7 @@ again:
btrfs_delalloc_release_extents(BTRFS_I(inode), reserve_bytes); btrfs_delalloc_release_extents(BTRFS_I(inode), reserve_bytes);
if (ret) { if (ret) {
btrfs_drop_page(fs_info, page, pos, copied); btrfs_drop_folio(fs_info, folio, pos, copied);
break; break;
} }
@ -1340,7 +1326,7 @@ again:
if (only_release_metadata) if (only_release_metadata)
btrfs_check_nocow_unlock(BTRFS_I(inode)); btrfs_check_nocow_unlock(BTRFS_I(inode));
btrfs_drop_page(fs_info, page, pos, copied); btrfs_drop_folio(fs_info, folio, pos, copied);
cond_resched(); cond_resched();

4
fs/btrfs/file.h
View file

@ -34,8 +34,8 @@ int btrfs_mark_extent_written(struct btrfs_trans_handle *trans,
ssize_t btrfs_do_write_iter(struct kiocb *iocb, struct iov_iter *from, ssize_t btrfs_do_write_iter(struct kiocb *iocb, struct iov_iter *from,
const struct btrfs_ioctl_encoded_io_args *encoded); const struct btrfs_ioctl_encoded_io_args *encoded);
int btrfs_release_file(struct inode *inode, struct file *file); int btrfs_release_file(struct inode *inode, struct file *file);
int btrfs_dirty_page(struct btrfs_inode *inode, struct page *page, loff_t pos, int btrfs_dirty_folio(struct btrfs_inode *inode, struct folio *folio, loff_t pos,
size_t write_bytes, struct extent_state **cached, bool noreserve); size_t write_bytes, struct extent_state **cached, bool noreserve);
int btrfs_fdatawrite_range(struct btrfs_inode *inode, loff_t start, loff_t end); int btrfs_fdatawrite_range(struct btrfs_inode *inode, loff_t start, loff_t end);
int btrfs_check_nocow_lock(struct btrfs_inode *inode, loff_t pos, int btrfs_check_nocow_lock(struct btrfs_inode *inode, loff_t pos,
size_t *write_bytes, bool nowait); size_t *write_bytes, bool nowait);

4
fs/btrfs/free-space-cache.c
View file

@ -1464,8 +1464,8 @@ static int __btrfs_write_out_cache(struct inode *inode,
u64 dirty_start = i * PAGE_SIZE; u64 dirty_start = i * PAGE_SIZE;
u64 dirty_len = min_t(u64, dirty_start + PAGE_SIZE, i_size) - dirty_start; u64 dirty_len = min_t(u64, dirty_start + PAGE_SIZE, i_size) - dirty_start;
ret = btrfs_dirty_page(BTRFS_I(inode), io_ctl->pages[i], ret = btrfs_dirty_folio(BTRFS_I(inode), page_folio(io_ctl->pages[i]),
dirty_start, dirty_len, &cached_state, false); dirty_start, dirty_len, &cached_state, false);
if (ret < 0) if (ret < 0)
goto out_nospc; goto out_nospc;
} }