July 2024 - Linux-stable-mirror

[PATCH v2 4.19 4/4] ext4: drop unnecessary journal handle in delalloc write

by WangYuli

From: Zhang Yi <yi.zhang(a)huawei.com> commit cc883236b79297f6266ca6f4e7f24f3fd3c736c1 upstream After we factor out the inline data write procedure from ext4_da_write_end(), we don't need to start journal handle for the cases of both buffer overwrite and append-write. If we need to update i_disksize, mark_inode_dirty() do start handle and update inode buffer. So we could just remove all the journal handle codes in the delalloc write procedure. After this patch, we could get a lot of performance improvement. Below is the Unixbench comparison data test on my machine with 'Intel Xeon Gold 5120' CPU and nvme SSD backend. Test cmd: ./Run -c 56 -i 3 fstime fsbuffer fsdisk Before this patch: System Benchmarks Partial Index BASELINE RESULT INDEX File Copy 1024 bufsize 2000 maxblocks 3960.0 422965.0 1068.1 File Copy 256 bufsize 500 maxblocks 1655.0 105077.0 634.9 File Copy 4096 bufsize 8000 maxblocks 5800.0 1429092.0 2464.0 ====== System Benchmarks Index Score (Partial Only) 1186.6 After this patch: System Benchmarks Partial Index BASELINE RESULT INDEX File Copy 1024 bufsize 2000 maxblocks 3960.0 732716.0 1850.3 File Copy 256 bufsize 500 maxblocks 1655.0 184940.0 1117.5 File Copy 4096 bufsize 8000 maxblocks 5800.0 2427152.0 4184.7 ====== System Benchmarks Index Score (Partial Only) 2053.0 Cc: stable(a)vger.kernel.org Signed-off-by: Zhang Yi <yi.zhang(a)huawei.com> Reviewed-by: Jan Kara <jack(a)suse.cz> Signed-off-by: Theodore Ts'o <tytso(a)mit.edu> Link: https://lore.kernel.org/r/20210716122024.1105856-5-yi.zhang@huawei.com Reviewed-by: Cheng Nie <niecheng1(a)uniontech.com> Signed-off-by: Dandan Zhang <zhangdandan(a)uniontech.com> Signed-off-by: WangYuli <wangyuli(a)uniontech.com> --- fs/ext4/inode.c | 60 +++++-------------------------------------------- 1 file changed, 5 insertions(+), 55 deletions(-) diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index 1e6753084a00..597c2f49c889 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -3032,19 +3032,6 @@ static int ext4_nonda_switch(struct super_block *sb) return 0; } -/* We always reserve for an inode update; the superblock could be there too */ -static int ext4_da_write_credits(struct inode *inode, loff_t pos, unsigned len) -{ - if (likely(ext4_has_feature_large_file(inode->i_sb))) - return 1; - - if (pos + len <= 0x7fffffffULL) - return 1; - - /* We might need to update the superblock to set LARGE_FILE */ - return 2; -} - static int ext4_da_write_begin(struct file *file, struct address_space *mapping, loff_t pos, unsigned len, unsigned flags, struct page **pagep, void **fsdata) @@ -3053,7 +3040,6 @@ static int ext4_da_write_begin(struct file *file, struct address_space *mapping, struct page *page; pgoff_t index; struct inode *inode = mapping->host; - handle_t *handle; if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) return -EIO; @@ -3079,41 +3065,11 @@ static int ext4_da_write_begin(struct file *file, struct address_space *mapping, return 0; } - /* - * grab_cache_page_write_begin() can take a long time if the - * system is thrashing due to memory pressure, or if the page - * is being written back. So grab it first before we start - * the transaction handle. This also allows us to allocate - * the page (if needed) without using GFP_NOFS. - */ -retry_grab: +retry: page = grab_cache_page_write_begin(mapping, index, flags); if (!page) return -ENOMEM; - unlock_page(page); - - /* - * With delayed allocation, we don't log the i_disksize update - * if there is delayed block allocation. But we still need - * to journalling the i_disksize update if writes to the end - * of file which has an already mapped buffer. - */ -retry_journal: - handle = ext4_journal_start(inode, EXT4_HT_WRITE_PAGE, - ext4_da_write_credits(inode, pos, len)); - if (IS_ERR(handle)) { - put_page(page); - return PTR_ERR(handle); - } - lock_page(page); - if (page->mapping != mapping) { - /* The page got truncated from under us */ - unlock_page(page); - put_page(page); - ext4_journal_stop(handle); - goto retry_grab; - } /* In case writeback began while the page was unlocked */ wait_for_stable_page(page); @@ -3125,20 +3081,18 @@ static int ext4_da_write_begin(struct file *file, struct address_space *mapping, #endif if (ret < 0) { unlock_page(page); - ext4_journal_stop(handle); + put_page(page); /* * block_write_begin may have instantiated a few blocks * outside i_size. Trim these off again. Don't need - * i_size_read because we hold i_mutex. + * i_size_read because we hold inode lock. */ if (pos + len > inode->i_size) ext4_truncate_failed_write(inode); if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries)) - goto retry_journal; - - put_page(page); + goto retry; return ret; } @@ -3175,8 +3129,6 @@ static int ext4_da_write_end(struct file *file, struct page *page, void *fsdata) { struct inode *inode = mapping->host; - int ret; - handle_t *handle = ext4_journal_current_handle(); loff_t new_i_size; unsigned long start, end; int write_mode = (int)(unsigned long)fsdata; @@ -3215,9 +3167,7 @@ static int ext4_da_write_end(struct file *file, ext4_da_should_update_i_disksize(page, end)) ext4_update_i_disksize(inode, new_i_size); - copied = generic_write_end(file, mapping, pos, len, copied, page, fsdata); - ret = ext4_journal_stop(handle); - return ret ? ret : copied; + return generic_write_end(file, mapping, pos, len, copied, page, fsdata); } static void ext4_da_invalidatepage(struct page *page, unsigned int offset, -- 2.43.4

9 months, 1 week

1
0
0 0

[PATCH v2 4.19 3/4] ext4: factor out write end code of inline file

by WangYuli

From: Zhang Yi <yi.zhang(a)huawei.com> commit 6984aef59814fb5c47b0e30c56e101186b5ebf8c upstream Now that the inline_data file write end procedure are falled into the common write end functions, it is not clear. Factor them out and do some cleanup. This patch also drop ext4_da_write_inline_data_end() and switch to use ext4_write_inline_data_end() instead because we also need to do the same error processing if we failed to write data into inline entry. Cc: stable(a)vger.kernel.org Signed-off-by: Zhang Yi <yi.zhang(a)huawei.com> Reviewed-by: Jan Kara <jack(a)suse.cz> Signed-off-by: Theodore Ts'o <tytso(a)mit.edu> Link: https://lore.kernel.org/r/20210716122024.1105856-4-yi.zhang@huawei.com Reviewed-by: Cheng Nie <niecheng1(a)uniontech.com> Signed-off-by: Dandan Zhang <zhangdandan(a)uniontech.com> Signed-off-by: WangYuli <wangyuli(a)uniontech.com> --- fs/ext4/ext4.h | 3 -- fs/ext4/inline.c | 127 ++++++++++++++++++++++++----------------------- fs/ext4/inode.c | 74 +++++++++------------------ 3 files changed, 89 insertions(+), 115 deletions(-) diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h index fa2579abea7d..6394c8a3803c 100644 --- a/fs/ext4/ext4.h +++ b/fs/ext4/ext4.h @@ -3083,9 +3083,6 @@ extern int ext4_da_write_inline_data_begin(struct address_space *mapping, unsigned flags, struct page **pagep, void **fsdata); -extern int ext4_da_write_inline_data_end(struct inode *inode, loff_t pos, - unsigned len, unsigned copied, - struct page *page); extern int ext4_try_add_inline_entry(handle_t *handle, struct ext4_filename *fname, struct inode *dir, struct inode *inode); diff --git a/fs/ext4/inline.c b/fs/ext4/inline.c index de04bd5fb551..eeb696c85a61 100644 --- a/fs/ext4/inline.c +++ b/fs/ext4/inline.c @@ -741,40 +741,82 @@ int ext4_try_to_write_inline_data(struct address_space *mapping, int ext4_write_inline_data_end(struct inode *inode, loff_t pos, unsigned len, unsigned copied, struct page *page) { - int ret, no_expand; + handle_t *handle = ext4_journal_current_handle(); + int no_expand; void *kaddr; struct ext4_iloc iloc; + int ret = 0, ret2; if (unlikely(copied < len) && !PageUptodate(page)) - return 0; + copied = 0; - ret = ext4_get_inode_loc(inode, &iloc); - if (ret) { - ext4_std_error(inode->i_sb, ret); - return ret; - } + if (likely(copied)) { + ret = ext4_get_inode_loc(inode, &iloc); + if (ret) { + unlock_page(page); + put_page(page); + ext4_std_error(inode->i_sb, ret); + goto out; + } + ext4_write_lock_xattr(inode, &no_expand); + BUG_ON(!ext4_has_inline_data(inode)); - ext4_write_lock_xattr(inode, &no_expand); - BUG_ON(!ext4_has_inline_data(inode)); + kaddr = kmap_atomic(page); + ext4_write_inline_data(inode, &iloc, kaddr, pos, copied); + kunmap_atomic(kaddr); + SetPageUptodate(page); + /* clear page dirty so that writepages wouldn't work for us. */ + ClearPageDirty(page); - /* - * ei->i_inline_off may have changed since ext4_write_begin() - * called ext4_try_to_write_inline_data() - */ - (void) ext4_find_inline_data_nolock(inode); + /* + * ei->i_inline_off may have changed since ext4_write_begin() + * called ext4_try_to_write_inline_data() + */ + (void) ext4_find_inline_data_nolock(inode); - kaddr = kmap_atomic(page); - ext4_write_inline_data(inode, &iloc, kaddr, pos, copied); - kunmap_atomic(kaddr); - SetPageUptodate(page); - /* clear page dirty so that writepages wouldn't work for us. */ - ClearPageDirty(page); + ext4_write_unlock_xattr(inode, &no_expand); + brelse(iloc.bh); - ext4_write_unlock_xattr(inode, &no_expand); - brelse(iloc.bh); - mark_inode_dirty(inode); + /* + * It's important to update i_size while still holding page + * lock: page writeout could otherwise come in and zero + * beyond i_size. + */ + ext4_update_inode_size(inode, pos + copied); + } + unlock_page(page); + put_page(page); - return copied; + /* + * Don't mark the inode dirty under page lock. First, it unnecessarily + * makes the holding time of page lock longer. Second, it forces lock + * ordering of page lock and transaction start for journaling + * filesystems. + */ + if (likely(copied)) + mark_inode_dirty(inode); +out: + /* + * If we didn't copy as much data as expected, we need to trim back + * size of xattr containing inline data. + */ + if (pos + len > inode->i_size && ext4_can_truncate(inode)) + ext4_orphan_add(handle, inode); + + ret2 = ext4_journal_stop(handle); + if (!ret) + ret = ret2; + if (pos + len > inode->i_size) { + ext4_truncate_failed_write(inode); + /* + * If truncate failed early the inode might still be + * on the orphan list; we need to make sure the inode + * is removed from the orphan list in that case. + */ + if (inode->i_nlink) + ext4_orphan_del(NULL, inode); + } + return ret ? ret : copied; } struct buffer_head * @@ -955,43 +997,6 @@ int ext4_da_write_inline_data_begin(struct address_space *mapping, return ret; } -int ext4_da_write_inline_data_end(struct inode *inode, loff_t pos, - unsigned len, unsigned copied, - struct page *page) -{ - int ret; - - ret = ext4_write_inline_data_end(inode, pos, len, copied, page); - if (ret < 0) { - unlock_page(page); - put_page(page); - return ret; - } - copied = ret; - - /* - * No need to use i_size_read() here, the i_size - * cannot change under us because we hold i_mutex. - * - * But it's important to update i_size while still holding page lock: - * page writeout could otherwise come in and zero beyond i_size. - */ - if (pos+copied > inode->i_size) - i_size_write(inode, pos+copied); - unlock_page(page); - put_page(page); - - /* - * Don't mark the inode dirty under page lock. First, it unnecessarily - * makes the holding time of page lock longer. Second, it forces lock - * ordering of page lock and transaction start for journaling - * filesystems. - */ - mark_inode_dirty(inode); - - return copied; -} - #ifdef INLINE_DIR_DEBUG void ext4_show_inline_dir(struct inode *dir, struct buffer_head *bh, void *inline_start, int inline_size) diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index 44a715e6aae1..1e6753084a00 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -1415,23 +1415,13 @@ static int ext4_write_end(struct file *file, loff_t old_size = inode->i_size; int ret = 0, ret2; int i_size_changed = 0; - int inline_data = ext4_has_inline_data(inode); trace_ext4_write_end(inode, pos, len, copied); - if (inline_data && - ext4_test_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA)) { - ret = ext4_write_inline_data_end(inode, pos, len, - copied, page); - if (ret < 0) { - unlock_page(page); - put_page(page); - goto errout; - } - copied = ret; - ret = 0; - } else - copied = block_write_end(file, mapping, pos, - len, copied, page, fsdata); + + if (ext4_has_inline_data(inode)) + return ext4_write_inline_data_end(inode, pos, len, copied, page); + + copied = block_write_end(file, mapping, pos, len, copied, page, fsdata); /* * it's important to update i_size while still holding page lock: * page writeout could otherwise come in and zero beyond i_size. @@ -1448,10 +1438,9 @@ static int ext4_write_end(struct file *file, * ordering of page lock and transaction start for journaling * filesystems. */ - if (i_size_changed || inline_data) + if (i_size_changed) ext4_mark_inode_dirty(handle, inode); -errout: if (pos + len > inode->i_size && ext4_can_truncate(inode)) /* if we have allocated more blocks and copied * less. We will have blocks allocated outside @@ -1523,7 +1512,6 @@ static int ext4_journalled_write_end(struct file *file, int partial = 0; unsigned from, to; int size_changed = 0; - int inline_data = ext4_has_inline_data(inode); trace_ext4_journalled_write_end(inode, pos, len, copied); from = pos & (PAGE_SIZE - 1); @@ -1531,17 +1519,10 @@ static int ext4_journalled_write_end(struct file *file, BUG_ON(!ext4_handle_valid(handle)); - if (inline_data) { - ret = ext4_write_inline_data_end(inode, pos, len, - copied, page); - if (ret < 0) { - unlock_page(page); - put_page(page); - goto errout; - } - copied = ret; - ret = 0; - } else if (unlikely(copied < len) && !PageUptodate(page)) { + if (ext4_has_inline_data(inode)) + return ext4_write_inline_data_end(inode, pos, len, copied, page); + + if (unlikely(copied < len) && !PageUptodate(page)) { copied = 0; ext4_journalled_zero_new_buffers(handle, page, from, to); } else { @@ -1563,13 +1544,12 @@ static int ext4_journalled_write_end(struct file *file, if (old_size < pos) pagecache_isize_extended(inode, old_size, pos); - if (size_changed || inline_data) { + if (size_changed) { ret2 = ext4_mark_inode_dirty(handle, inode); if (!ret) ret = ret2; } -errout: if (pos + len > inode->i_size && ext4_can_truncate(inode)) /* if we have allocated more blocks and copied * less. We will have blocks allocated outside @@ -3195,7 +3175,7 @@ static int ext4_da_write_end(struct file *file, struct page *page, void *fsdata) { struct inode *inode = mapping->host; - int ret = 0, ret2; + int ret; handle_t *handle = ext4_journal_current_handle(); loff_t new_i_size; unsigned long start, end; @@ -3206,6 +3186,12 @@ static int ext4_da_write_end(struct file *file, len, copied, page, fsdata); trace_ext4_da_write_end(inode, pos, len, copied); + + if (write_mode != CONVERT_INLINE_DATA && + ext4_test_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA) && + ext4_has_inline_data(inode)) + return ext4_write_inline_data_end(inode, pos, len, copied, page); + start = pos & (PAGE_SIZE - 1); end = start + copied - 1; @@ -3225,26 +3211,12 @@ static int ext4_da_write_end(struct file *file, * ext4_da_write_inline_data_end(). */ new_i_size = pos + copied; - if (copied && new_i_size > inode->i_size) { - if (ext4_has_inline_data(inode) || - ext4_da_should_update_i_disksize(page, end)) - ext4_update_i_disksize(inode, new_i_size); - } - - if (write_mode != CONVERT_INLINE_DATA && - ext4_test_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA) && - ext4_has_inline_data(inode)) - ret = ext4_da_write_inline_data_end(inode, pos, len, copied, - page); - else - ret = generic_write_end(file, mapping, pos, len, copied, - page, fsdata); - - copied = ret; - ret2 = ext4_journal_stop(handle); - if (!ret) - ret = ret2; + if (copied && new_i_size > inode->i_size && + ext4_da_should_update_i_disksize(page, end)) + ext4_update_i_disksize(inode, new_i_size); + copied = generic_write_end(file, mapping, pos, len, copied, page, fsdata); + ret = ext4_journal_stop(handle); return ret ? ret : copied; } -- 2.43.4

9 months, 1 week

1
0
0 0

[PATCH v2 4.19 2/4] ext4: correct the error path of ext4_write_inline_data_end()

by WangYuli

From: Zhang Yi <yi.zhang(a)huawei.com> commit 55ce2f649b9e88111270333a8127e23f4f8f42d7 upstream Current error path of ext4_write_inline_data_end() is not correct. Firstly, it should pass out the error value if ext4_get_inode_loc() return fail, or else it could trigger infinite loop if we inject error here. And then it's better to add inode to orphan list if it return fail in ext4_journal_stop(), otherwise we could not restore inline xattr entry after power failure. Finally, we need to reset the 'ret' value if ext4_write_inline_data_end() return success in ext4_write_end() and ext4_journalled_write_end(), otherwise we could not get the error return value of ext4_journal_stop(). Cc: stable(a)vger.kernel.org Signed-off-by: Zhang Yi <yi.zhang(a)huawei.com> Reviewed-by: Jan Kara <jack(a)suse.cz> Signed-off-by: Theodore Ts'o <tytso(a)mit.edu> Link: https://lore.kernel.org/r/20210716122024.1105856-3-yi.zhang@huawei.com Reviewed-by: Cheng Nie <niecheng1(a)uniontech.com> Signed-off-by: Dandan Zhang <zhangdandan(a)uniontech.com> Signed-off-by: WangYuli <wangyuli(a)uniontech.com> --- fs/ext4/inline.c | 15 +++++---------- fs/ext4/inode.c | 7 +++++-- 2 files changed, 10 insertions(+), 12 deletions(-) diff --git a/fs/ext4/inline.c b/fs/ext4/inline.c index 71bb3cfc5933..de04bd5fb551 100644 --- a/fs/ext4/inline.c +++ b/fs/ext4/inline.c @@ -745,18 +745,13 @@ int ext4_write_inline_data_end(struct inode *inode, loff_t pos, unsigned len, void *kaddr; struct ext4_iloc iloc; - if (unlikely(copied < len)) { - if (!PageUptodate(page)) { - copied = 0; - goto out; - } - } + if (unlikely(copied < len) && !PageUptodate(page)) + return 0; ret = ext4_get_inode_loc(inode, &iloc); if (ret) { ext4_std_error(inode->i_sb, ret); - copied = 0; - goto out; + return ret; } ext4_write_lock_xattr(inode, &no_expand); @@ -769,7 +764,7 @@ int ext4_write_inline_data_end(struct inode *inode, loff_t pos, unsigned len, (void) ext4_find_inline_data_nolock(inode); kaddr = kmap_atomic(page); - ext4_write_inline_data(inode, &iloc, kaddr, pos, len); + ext4_write_inline_data(inode, &iloc, kaddr, pos, copied); kunmap_atomic(kaddr); SetPageUptodate(page); /* clear page dirty so that writepages wouldn't work for us. */ @@ -778,7 +773,7 @@ int ext4_write_inline_data_end(struct inode *inode, loff_t pos, unsigned len, ext4_write_unlock_xattr(inode, &no_expand); brelse(iloc.bh); mark_inode_dirty(inode); -out: + return copied; } diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index d8a8e4ee5ff8..44a715e6aae1 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -1428,6 +1428,7 @@ static int ext4_write_end(struct file *file, goto errout; } copied = ret; + ret = 0; } else copied = block_write_end(file, mapping, pos, len, copied, page, fsdata); @@ -1450,13 +1451,14 @@ static int ext4_write_end(struct file *file, if (i_size_changed || inline_data) ext4_mark_inode_dirty(handle, inode); +errout: if (pos + len > inode->i_size && ext4_can_truncate(inode)) /* if we have allocated more blocks and copied * less. We will have blocks allocated outside * inode->i_size. So truncate them */ ext4_orphan_add(handle, inode); -errout: + ret2 = ext4_journal_stop(handle); if (!ret) ret = ret2; @@ -1538,6 +1540,7 @@ static int ext4_journalled_write_end(struct file *file, goto errout; } copied = ret; + ret = 0; } else if (unlikely(copied < len) && !PageUptodate(page)) { copied = 0; ext4_journalled_zero_new_buffers(handle, page, from, to); @@ -1566,6 +1569,7 @@ static int ext4_journalled_write_end(struct file *file, ret = ret2; } +errout: if (pos + len > inode->i_size && ext4_can_truncate(inode)) /* if we have allocated more blocks and copied * less. We will have blocks allocated outside @@ -1573,7 +1577,6 @@ static int ext4_journalled_write_end(struct file *file, */ ext4_orphan_add(handle, inode); -errout: ret2 = ext4_journal_stop(handle); if (!ret) ret = ret2; -- 2.43.4

9 months, 1 week

1
0
0 0

[PATCH 1/4] ext4: check and update i_disksize properly

by WangYuli

From: Zhang Yi <yi.zhang(a)huawei.com> After commit 3da40c7b0898 ("ext4: only call ext4_truncate when size <= isize"), i_disksize could always be updated to i_size in ext4_setattr(), and we could sure that i_disksize <= i_size since holding inode lock and if i_disksize < i_size there are delalloc writes pending in the range upto i_size. If the end of the current write is <= i_size, there's no need to touch i_disksize since writeback will push i_disksize upto i_size eventually. So we can switch to check i_size instead of i_disksize in ext4_da_write_end() when write to the end of the file. we also could remove ext4_mark_inode_dirty() together because we defer inode dirtying to generic_write_end() or ext4_da_write_inline_data_end(). Signed-off-by: Zhang Yi <yi.zhang(a)huawei.com> Reviewed-by: Jan Kara <jack(a)suse.cz> Signed-off-by: Theodore Ts'o <tytso(a)mit.edu> Link: https://lore.kernel.org/r/20210716122024.1105856-2-yi.zhang@huawei.com Reviewed-by: Cheng Nie <niecheng1(a)uniontech.com> Signed-off-by: Dandan Zhang <zhangdandan(a)uniontech.com> Signed-off-by: WangYuli <wangyuli(a)uniontech.com> --- fs/ext4/inode.c | 34 ++++++++++++++++++---------------- 1 file changed, 18 insertions(+), 16 deletions(-) diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index 646285fbc9fc..d8a8e4ee5ff8 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -3207,35 +3207,37 @@ static int ext4_da_write_end(struct file *file, end = start + copied - 1; /* - * generic_write_end() will run mark_inode_dirty() if i_size - * changes. So let's piggyback the i_disksize mark_inode_dirty - * into that. + * Since we are holding inode lock, we are sure i_disksize <= + * i_size. We also know that if i_disksize < i_size, there are + * delalloc writes pending in the range upto i_size. If the end of + * the current write is <= i_size, there's no need to touch + * i_disksize since writeback will push i_disksize upto i_size + * eventually. If the end of the current write is > i_size and + * inside an allocated block (ext4_da_should_update_i_disksize() + * check), we need to update i_disksize here as neither + * ext4_writepage() nor certain ext4_writepages() paths not + * allocating blocks update i_disksize. + * + * Note that we defer inode dirtying to generic_write_end() / + * ext4_da_write_inline_data_end(). */ new_i_size = pos + copied; - if (copied && new_i_size > EXT4_I(inode)->i_disksize) { + if (copied && new_i_size > inode->i_size) { if (ext4_has_inline_data(inode) || - ext4_da_should_update_i_disksize(page, end)) { + ext4_da_should_update_i_disksize(page, end)) ext4_update_i_disksize(inode, new_i_size); - /* We need to mark inode dirty even if - * new_i_size is less that inode->i_size - * bu greater than i_disksize.(hint delalloc) - */ - ext4_mark_inode_dirty(handle, inode); - } } if (write_mode != CONVERT_INLINE_DATA && ext4_test_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA) && ext4_has_inline_data(inode)) - ret2 = ext4_da_write_inline_data_end(inode, pos, len, copied, + ret = ext4_da_write_inline_data_end(inode, pos, len, copied, page); else - ret2 = generic_write_end(file, mapping, pos, len, copied, + ret = generic_write_end(file, mapping, pos, len, copied, page, fsdata); - copied = ret2; - if (ret2 < 0) - ret = ret2; + copied = ret; ret2 = ext4_journal_stop(handle); if (!ret) ret = ret2; -- 2.43.4

9 months, 1 week

2
1
0 0

[PATCH 4/4] ext4: drop unnecessary journal handle in delalloc write

by WangYuli

From: Zhang Yi <yi.zhang(a)huawei.com> After we factor out the inline data write procedure from ext4_da_write_end(), we don't need to start journal handle for the cases of both buffer overwrite and append-write. If we need to update i_disksize, mark_inode_dirty() do start handle and update inode buffer. So we could just remove all the journal handle codes in the delalloc write procedure. After this patch, we could get a lot of performance improvement. Below is the Unixbench comparison data test on my machine with 'Intel Xeon Gold 5120' CPU and nvme SSD backend. Test cmd: ./Run -c 56 -i 3 fstime fsbuffer fsdisk Before this patch: System Benchmarks Partial Index BASELINE RESULT INDEX File Copy 1024 bufsize 2000 maxblocks 3960.0 422965.0 1068.1 File Copy 256 bufsize 500 maxblocks 1655.0 105077.0 634.9 File Copy 4096 bufsize 8000 maxblocks 5800.0 1429092.0 2464.0 ====== System Benchmarks Index Score (Partial Only) 1186.6 After this patch: System Benchmarks Partial Index BASELINE RESULT INDEX File Copy 1024 bufsize 2000 maxblocks 3960.0 732716.0 1850.3 File Copy 256 bufsize 500 maxblocks 1655.0 184940.0 1117.5 File Copy 4096 bufsize 8000 maxblocks 5800.0 2427152.0 4184.7 ====== System Benchmarks Index Score (Partial Only) 2053.0 Signed-off-by: Zhang Yi <yi.zhang(a)huawei.com> Reviewed-by: Jan Kara <jack(a)suse.cz> Signed-off-by: Theodore Ts'o <tytso(a)mit.edu> Link: https://lore.kernel.org/r/20210716122024.1105856-5-yi.zhang@huawei.com Reviewed-by: Cheng Nie <niecheng1(a)uniontech.com> Signed-off-by: Dandan Zhang <zhangdandan(a)uniontech.com> Signed-off-by: WangYuli <wangyuli(a)uniontech.com> --- fs/ext4/inode.c | 60 +++++-------------------------------------------- 1 file changed, 5 insertions(+), 55 deletions(-) diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index 1e6753084a00..597c2f49c889 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -3032,19 +3032,6 @@ static int ext4_nonda_switch(struct super_block *sb) return 0; } -/* We always reserve for an inode update; the superblock could be there too */ -static int ext4_da_write_credits(struct inode *inode, loff_t pos, unsigned len) -{ - if (likely(ext4_has_feature_large_file(inode->i_sb))) - return 1; - - if (pos + len <= 0x7fffffffULL) - return 1; - - /* We might need to update the superblock to set LARGE_FILE */ - return 2; -} - static int ext4_da_write_begin(struct file *file, struct address_space *mapping, loff_t pos, unsigned len, unsigned flags, struct page **pagep, void **fsdata) @@ -3053,7 +3040,6 @@ static int ext4_da_write_begin(struct file *file, struct address_space *mapping, struct page *page; pgoff_t index; struct inode *inode = mapping->host; - handle_t *handle; if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) return -EIO; @@ -3079,41 +3065,11 @@ static int ext4_da_write_begin(struct file *file, struct address_space *mapping, return 0; } - /* - * grab_cache_page_write_begin() can take a long time if the - * system is thrashing due to memory pressure, or if the page - * is being written back. So grab it first before we start - * the transaction handle. This also allows us to allocate - * the page (if needed) without using GFP_NOFS. - */ -retry_grab: +retry: page = grab_cache_page_write_begin(mapping, index, flags); if (!page) return -ENOMEM; - unlock_page(page); - - /* - * With delayed allocation, we don't log the i_disksize update - * if there is delayed block allocation. But we still need - * to journalling the i_disksize update if writes to the end - * of file which has an already mapped buffer. - */ -retry_journal: - handle = ext4_journal_start(inode, EXT4_HT_WRITE_PAGE, - ext4_da_write_credits(inode, pos, len)); - if (IS_ERR(handle)) { - put_page(page); - return PTR_ERR(handle); - } - lock_page(page); - if (page->mapping != mapping) { - /* The page got truncated from under us */ - unlock_page(page); - put_page(page); - ext4_journal_stop(handle); - goto retry_grab; - } /* In case writeback began while the page was unlocked */ wait_for_stable_page(page); @@ -3125,20 +3081,18 @@ static int ext4_da_write_begin(struct file *file, struct address_space *mapping, #endif if (ret < 0) { unlock_page(page); - ext4_journal_stop(handle); + put_page(page); /* * block_write_begin may have instantiated a few blocks * outside i_size. Trim these off again. Don't need - * i_size_read because we hold i_mutex. + * i_size_read because we hold inode lock. */ if (pos + len > inode->i_size) ext4_truncate_failed_write(inode); if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries)) - goto retry_journal; - - put_page(page); + goto retry; return ret; } @@ -3175,8 +3129,6 @@ static int ext4_da_write_end(struct file *file, struct page *page, void *fsdata) { struct inode *inode = mapping->host; - int ret; - handle_t *handle = ext4_journal_current_handle(); loff_t new_i_size; unsigned long start, end; int write_mode = (int)(unsigned long)fsdata; @@ -3215,9 +3167,7 @@ static int ext4_da_write_end(struct file *file, ext4_da_should_update_i_disksize(page, end)) ext4_update_i_disksize(inode, new_i_size); - copied = generic_write_end(file, mapping, pos, len, copied, page, fsdata); - ret = ext4_journal_stop(handle); - return ret ? ret : copied; + return generic_write_end(file, mapping, pos, len, copied, page, fsdata); } static void ext4_da_invalidatepage(struct page *page, unsigned int offset, -- 2.43.4

9 months, 1 week

1
0
0 0

[PATCH 2/4] ext4: correct the error path of ext4_write_inline_data_end()

by WangYuli

From: Zhang Yi <yi.zhang(a)huawei.com> Current error path of ext4_write_inline_data_end() is not correct. Firstly, it should pass out the error value if ext4_get_inode_loc() return fail, or else it could trigger infinite loop if we inject error here. And then it's better to add inode to orphan list if it return fail in ext4_journal_stop(), otherwise we could not restore inline xattr entry after power failure. Finally, we need to reset the 'ret' value if ext4_write_inline_data_end() return success in ext4_write_end() and ext4_journalled_write_end(), otherwise we could not get the error return value of ext4_journal_stop(). Signed-off-by: Zhang Yi <yi.zhang(a)huawei.com> Reviewed-by: Jan Kara <jack(a)suse.cz> Signed-off-by: Theodore Ts'o <tytso(a)mit.edu> Link: https://lore.kernel.org/r/20210716122024.1105856-3-yi.zhang@huawei.com Reviewed-by: Cheng Nie <niecheng1(a)uniontech.com> Signed-off-by: Dandan Zhang <zhangdandan(a)uniontech.com> Signed-off-by: WangYuli <wangyuli(a)uniontech.com> --- fs/ext4/inline.c | 15 +++++---------- fs/ext4/inode.c | 7 +++++-- 2 files changed, 10 insertions(+), 12 deletions(-) diff --git a/fs/ext4/inline.c b/fs/ext4/inline.c index 71bb3cfc5933..de04bd5fb551 100644 --- a/fs/ext4/inline.c +++ b/fs/ext4/inline.c @@ -745,18 +745,13 @@ int ext4_write_inline_data_end(struct inode *inode, loff_t pos, unsigned len, void *kaddr; struct ext4_iloc iloc; - if (unlikely(copied < len)) { - if (!PageUptodate(page)) { - copied = 0; - goto out; - } - } + if (unlikely(copied < len) && !PageUptodate(page)) + return 0; ret = ext4_get_inode_loc(inode, &iloc); if (ret) { ext4_std_error(inode->i_sb, ret); - copied = 0; - goto out; + return ret; } ext4_write_lock_xattr(inode, &no_expand); @@ -769,7 +764,7 @@ int ext4_write_inline_data_end(struct inode *inode, loff_t pos, unsigned len, (void) ext4_find_inline_data_nolock(inode); kaddr = kmap_atomic(page); - ext4_write_inline_data(inode, &iloc, kaddr, pos, len); + ext4_write_inline_data(inode, &iloc, kaddr, pos, copied); kunmap_atomic(kaddr); SetPageUptodate(page); /* clear page dirty so that writepages wouldn't work for us. */ @@ -778,7 +773,7 @@ int ext4_write_inline_data_end(struct inode *inode, loff_t pos, unsigned len, ext4_write_unlock_xattr(inode, &no_expand); brelse(iloc.bh); mark_inode_dirty(inode); -out: + return copied; } diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index d8a8e4ee5ff8..44a715e6aae1 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -1428,6 +1428,7 @@ static int ext4_write_end(struct file *file, goto errout; } copied = ret; + ret = 0; } else copied = block_write_end(file, mapping, pos, len, copied, page, fsdata); @@ -1450,13 +1451,14 @@ static int ext4_write_end(struct file *file, if (i_size_changed || inline_data) ext4_mark_inode_dirty(handle, inode); +errout: if (pos + len > inode->i_size && ext4_can_truncate(inode)) /* if we have allocated more blocks and copied * less. We will have blocks allocated outside * inode->i_size. So truncate them */ ext4_orphan_add(handle, inode); -errout: + ret2 = ext4_journal_stop(handle); if (!ret) ret = ret2; @@ -1538,6 +1540,7 @@ static int ext4_journalled_write_end(struct file *file, goto errout; } copied = ret; + ret = 0; } else if (unlikely(copied < len) && !PageUptodate(page)) { copied = 0; ext4_journalled_zero_new_buffers(handle, page, from, to); @@ -1566,6 +1569,7 @@ static int ext4_journalled_write_end(struct file *file, ret = ret2; } +errout: if (pos + len > inode->i_size && ext4_can_truncate(inode)) /* if we have allocated more blocks and copied * less. We will have blocks allocated outside @@ -1573,7 +1577,6 @@ static int ext4_journalled_write_end(struct file *file, */ ext4_orphan_add(handle, inode); -errout: ret2 = ext4_journal_stop(handle); if (!ret) ret = ret2; -- 2.43.4

9 months, 1 week

1
0
0 0

[PATCH 4.19 0/4] ext4: improve delalloc buffer write performance

by WangYuli

A patchset from linux-5.15 should be backported to 4.19 that can significantly improve ext4 fs read and write performance. Unixbench test results for linux-4.19.318 on Phytium D2000 CPU are shown below. Test cmd: (Phytium D2000 only has 8 cores) ./Run fs -c 8 Before this patch set: File Copy 1024 bufsize 2000 maxblocks 1124181 File Copy 256 bufsize 500 maxblocks 281885 File Copy 4096 bufsize 8000 maxblocks 3383785 File Read 1024 bufsize 2000 maxblocks 8702173 File Read 256 bufsize 500 maxblocks 3869384 File Read 4096 bufsize 8000 maxblocks 13043151 File Write 1024 bufsize 2000 maxblocks 1107185 File Write 256 bufsize 500 maxblocks 270493 File Write 4096 bufsize 8000 maxblocks 4018084 After this patch set: File Copy 1024 bufsize 2000 maxblocks 2026206 File Copy 256 bufsize 500 maxblocks 829534 File Copy 4096 bufsize 8000 maxblocks 4066659 File Read 1024 bufsize 2000 maxblocks 8877219 File Read 256 bufsize 500 maxblocks 3997445 File Read 4096 bufsize 8000 maxblocks 13179885 File Write 1024 bufsize 2000 maxblocks 4256929 File Write 256 bufsize 500 maxblocks 1305320 File Write 4096 bufsize 8000 maxblocks 10721052 We can observe a quantum leap in the test results as a consequence of applying this patchset Link: https://lore.kernel.org/all/20210716122024.1105856-1-yi.zhang@huawei.com/ Original description: This patchset address to improve buffer write performance with delalloc. The first patch reduce the unnecessary update i_disksize, the second two patch refactor the inline data write procedure and also do some small fix, the last patch do improve by remove all unnecessary journal handle in the delalloc write procedure. After this patch set, we could get a lot of performance improvement. Below is the Unixbench comparison data test on my machine with 'Intel Xeon Gold 5120' CPU and nvme SSD backend. Test cmd: ./Run -c 56 -i 3 fstime fsbuffer fsdisk Before this patch set: System Benchmarks Partial Index BASELINE RESULT INDEX File Copy 1024 bufsize 2000 maxblocks 3960.0 422965.0 1068.1 File Copy 256 bufsize 500 maxblocks 1655.0 105077.0 634.9 File Copy 4096 bufsize 8000 maxblocks 5800.0 1429092.0 2464.0 ======== System Benchmarks Index Score (Partial Only) 1186.6 After this patch set: System Benchmarks Partial Index BASELINE RESULT INDEX File Copy 1024 bufsize 2000 maxblocks 3960.0 732716.0 1850.3 File Copy 256 bufsize 500 maxblocks 1655.0 184940.0 1117.5 File Copy 4096 bufsize 8000 maxblocks 5800.0 2427152.0 4184.7 ======== System Benchmarks Index Score (Partial Only) 2053.0 Zhang Yi (4): ext4: check and update i_disksize properly ext4: correct the error path of ext4_write_inline_data_end() ext4: factor out write end code of inline file ext4: drop unnecessary journal handle in delalloc write fs/ext4/ext4.h | 3 - fs/ext4/inline.c | 120 ++++++++++++++++++------------------- fs/ext4/inode.c | 150 ++++++++++++----------------------------------- 3 files changed, 99 insertions(+), 174 deletions(-) -- 2.31.1

9 months, 1 week

1
0
0 0

Re:

by Jackson Jikarjo

Hello, I sent you a message a few hours ago but no reply yet, or you didn't receive it? Kindly read my letter and reply back. I want to make an inquiry Thanks. Dr.Allen Cheng Human Resource Manager | Product Research Assistant FC Industrial Laboratories Ltd

9 months, 1 week

1
0
0 0

[PATCH] mm: huge_memory: use !CONFIG_64BIT to relax huge page alignment on 32 bit machines

by Yang Shi

Yves-Alexis Perez reported commit 4ef9ad19e176 ("mm: huge_memory: don't force huge page alignment on 32 bit") didn't work for x86_32 [1]. It is because x86_32 uses CONFIG_X86_32 instead of CONFIG_32BIT. !CONFIG_64BIT should cover all 32 bit machines. [1] https://lore.kernel.org/linux-mm/CAHbLzkr1LwH3pcTgM+aGQ31ip2bKqiqEQ8=FQB+t2… Fixes: 4ef9ad19e176 ("mm: huge_memory: don't force huge page alignment on 32 bit") Reported-by: Yves-Alexis Perez <corsac(a)debian.org> Tested-By: Yves-Alexis Perez <corsac(a)debian.org> Cc: <stable(a)vger.kernel.org> [6.8+] Signed-off-by: Yang Shi <yang(a)os.amperecomputing.com> --- mm/huge_memory.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/mm/huge_memory.c b/mm/huge_memory.c index 2120f7478e55..64f00aedf9af 100644 --- a/mm/huge_memory.c +++ b/mm/huge_memory.c @@ -857,7 +857,7 @@ static unsigned long __thp_get_unmapped_area(struct file *filp, loff_t off_align = round_up(off, size); unsigned long len_pad, ret, off_sub; - if (IS_ENABLED(CONFIG_32BIT) || in_compat_syscall()) + if (!IS_ENABLED(CONFIG_64BIT) || in_compat_syscall()) return 0; if (off_end <= off_align || (off_end - off_align) < size) -- 2.41.0

9 months, 1 week

4
3
0 0

[PATCH v5 3/7] RISC-V: Check scalar unaligned access on all CPUs

by Jesse Taube

Originally, the check_unaligned_access_emulated_all_cpus function only checked the boot hart. This fixes the function to check all harts. Fixes: 71c54b3d169d ("riscv: report misaligned accesses emulation to hwprobe") Signed-off-by: Jesse Taube <jesse(a)rivosinc.com> Reviewed-by: Charlie Jenkins <charlie(a)rivosinc.com> Cc: stable(a)vger.kernel.org --- V1 -> V2: - New patch V2 -> V3: - Split patch V3 -> V4: - Re-add check for a system where a heterogeneous CPU is hotplugged into a previously homogenous system. V4 -> V5: - Change work_struct *unused to work_struct *work __always_unused --- arch/riscv/kernel/traps_misaligned.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) diff --git a/arch/riscv/kernel/traps_misaligned.c b/arch/riscv/kernel/traps_misaligned.c index b62d5a2f4541..9a1e94383d6d 100644 --- a/arch/riscv/kernel/traps_misaligned.c +++ b/arch/riscv/kernel/traps_misaligned.c @@ -526,11 +526,11 @@ int handle_misaligned_store(struct pt_regs *regs) return 0; } -static bool check_unaligned_access_emulated(int cpu) +static void check_unaligned_access_emulated(struct work_struct *work __always_unused) { + int cpu = smp_processor_id(); long *mas_ptr = per_cpu_ptr(&misaligned_access_speed, cpu); unsigned long tmp_var, tmp_val; - bool misaligned_emu_detected; *mas_ptr = RISCV_HWPROBE_MISALIGNED_UNKNOWN; @@ -538,19 +538,16 @@ static bool check_unaligned_access_emulated(int cpu) " "REG_L" %[tmp], 1(%[ptr])\n" : [tmp] "=r" (tmp_val) : [ptr] "r" (&tmp_var) : "memory"); - misaligned_emu_detected = (*mas_ptr == RISCV_HWPROBE_MISALIGNED_EMULATED); /* * If unaligned_ctl is already set, this means that we detected that all * CPUS uses emulated misaligned access at boot time. If that changed * when hotplugging the new cpu, this is something we don't handle. */ - if (unlikely(unaligned_ctl && !misaligned_emu_detected)) { + if (unlikely(unaligned_ctl && (*mas_ptr != RISCV_HWPROBE_MISALIGNED_EMULATED))) { pr_crit("CPU misaligned accesses non homogeneous (expected all emulated)\n"); while (true) cpu_relax(); } - - return misaligned_emu_detected; } bool check_unaligned_access_emulated_all_cpus(void) @@ -562,8 +559,11 @@ bool check_unaligned_access_emulated_all_cpus(void) * accesses emulated since tasks requesting such control can run on any * CPU. */ + schedule_on_each_cpu(check_unaligned_access_emulated); + for_each_online_cpu(cpu) - if (!check_unaligned_access_emulated(cpu)) + if (per_cpu(misaligned_access_speed, cpu) + != RISCV_HWPROBE_MISALIGNED_EMULATED) return false; unaligned_ctl = true; -- 2.45.2

9 months, 1 week

2
1
0 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

Linux-stable-mirror July 2024