f2fs: allow writeback on error status filesystem
PATCH v2 1/1 f2fs: allow writeback on error status filesystem
71.07% 0.01% kworker/u256:1+ [kernel.kallsyms] [k] wb_writeback
|
--71.06%--wb_writeback
| |--68.96%--__writeback_inodes_wb
| |
| --68.95%--writeback_sb_inodes
| |
| |--65.08%--__writeback_single_inode
| | |
| | --64.35%--do_writepages
| | |
| | |--59.83%--f2fs_write_node_pages
| | | |
| | | --59.74%--f2fs_sync_node_pages
| | | |
| | | |--27.91%--pagevec_lookup_range_tag
| | | | |
| | | | --27.90%--find_get_pages_range_tag
If filesystem was injected checkpoint errror, before umount, kworker will always hold one core in order to writeback a large number of node pages, that looks not reasonable, to avoid that, we can allow data/node write in such case, since we can force all data/node writes with OPU mode, and clear recovery flag on node, and checkpoint is not allowed as well, so we don't need to worry about writeback's effect on data/node in previous checkpoint, then with this way, it can decrease memory footprint cost by node/data pages and avoid looping into data/node writeback process.
--- fs/f2fs/checkpoint.c | 7 +++++++ fs/f2fs/data.c | 22 +++++++++++++++------- fs/f2fs/node.c | 7 +++++-- 3 files changed, 27 insertions(+), 9 deletions(-)
fs/f2fs/checkpoint.c +7 -0
@@ -1238,6 +1238,13 @@ static int block_operations(struct f2fs_sb_info *sbi) goto retry_flush_nodes; } +if (unlikely (f2fs_cp_error (sbi))) { +up_write (&sbi->node_write); +up_write (&sbi->node_change); +f2fs_unlock_all (sbi); +return -EIO ; + } +/* * sbi->node_change is used only for AIO write_begin path which produces * dirty node blocks and some checkpoint values by block allocation.
fs/f2fs/data.c +15 -7
@@ -2508,6 +2508,8 @@ bool f2fs_should_update_outplace(struct inode *inode, struct f2fs_io_info *fio) {struct f2fs_sb_info *sbi =F2FS_I_SB (inode); +if (unlikely (f2fs_cp_error (sbi))) +return true;if (f2fs_lfs_mode (sbi))return true;if (S_ISDIR (inode->i_mode))
@@ -2691,13 +2693,19 @@ int f2fs_write_single_data_page(struct page *page, int *submitted,/* we should bypass data pages to proceed the kworkder jobs */ if (unlikely (f2fs_cp_error (sbi))) {mapping_set_error (page->mapping, -EIO ); -/* -* don't drop any dirty dentry pages for keeping lastest -* directory structure. -*/ -if (S_ISDIR (inode->i_mode)) - goto redirty_out; - goto out; + +if (is_sbi_flag_set (sbi,SBI_IS_CLOSE )) + goto out; + +if (has_not_enough_free_secs (sbi,0 ,0 )) { +/* +* don't drop any dirty dentry pages for keeping lastest +* directory structure. +*/ +if (S_ISDIR (inode->i_mode)) + goto redirty_out; + goto out; + } }if (unlikely (is_sbi_flag_set (sbi,SBI_POR_DOING )))
fs/f2fs/node.c +5 -2
@@ -1527,7 +1527,10 @@ static int __write_node_page(struct page *page, bool atomic, bool *submitted,unlock_page (page);return 0 ; } - goto redirty_out; +if (has_not_enough_free_secs (sbi,0 ,0 )) + goto redirty_out; +set_fsync_mark (page,0 ); +set_dentry_mark (page,0 ); }if (unlikely (is_sbi_flag_set (sbi,SBI_POR_DOING )))
@@ -1568,7 +1571,7 @@ static int __write_node_page(struct page *page, bool atomic, bool *submitted, goto redirty_out; } -if (atomic && !test_opt (sbi,NOBARRIER )) +if (atomic && !test_opt (sbi,NOBARRIER ) && !f2fs_cp_error (sbi)) fio.op_flags |=REQ_PREFLUSH |REQ_FUA ;/* should add to global list before clearing PAGECACHE status */
Signature · 1 line
2.18.0.rc1
Jaegeuk,
Let's consider this patch in next Linux version.
BTW, do you still encounter consistency problem w/o this version?
Quoted 120 lines · Chao Yu
71.07% 0.01% kworker/u256:1+ [kernel.kallsyms] [k]
wb_writeback
|
--71.06%--wb_writeback
| |--68.96%--__writeback_inodes_wb
| |
| --68.95%--writeback_sb_inodes
| |
| |--65.08%--__writeback_single_inode
| | |
| | --64.35%--do_writepages
| | |
| | |--59.83%--f2fs_write_node_pages
| | | |
| | | --59.74%--f2fs_sync_node_pages
| | | |
| | | |--27.91%--pagevec_lookup_range_tag
| | | | |
| | | | --27.90%--find_get_pages_range_tagIf filesystem was injected checkpoint errror, before umount, kworker will always hold one core in order to writeback a large number of node pages, that looks not reasonable, to avoid that, we can allow data/node write in such case, since we can force all data/node writes with OPU mode, and clear recovery flag on node, and checkpoint is not allowed as well, so we don't need to worry about writeback's effect on data/node in previous checkpoint, then with this way, it can decrease memory footprint cost by node/data pages and avoid looping into data/node writeback process.
Signed-off-by Chao Yu <[email protected]>--- fs/f2fs/checkpoint.c | 7 +++++++ fs/f2fs/data.c | 22 +++++++++++++++------- fs/f2fs/node.c | 7 +++++-- 3 files changed, 27 insertions(+), 9 deletions(-)fs/f2fs/checkpoint.c +7
-0@@ -1238,6 +1238,13 @@ static int block_operations(struct f2fs_sb_info *sbi) goto retry_flush_nodes; } +if (unlikely (f2fs_cp_error (sbi))) { +up_write (&sbi->node_write); +up_write (&sbi->node_change); +f2fs_unlock_all (sbi); +return -EIO ; + } +/* * sbi->node_change is used only for AIO write_begin path which produces * dirty node blocks and some checkpoint values by block allocation. fs/f2fs/data.c +15
-7@@ -2508,6 +2508,8 @@ bool f2fs_should_update_outplace(struct inode *inode, struct f2fs_io_info *fio) {struct f2fs_sb_info *sbi =F2FS_I_SB (inode); +if (unlikely (f2fs_cp_error (sbi))) +return true;if (f2fs_lfs_mode (sbi))return true;if (S_ISDIR (inode->i_mode))@@ -2691,13 +2693,19 @@ int f2fs_write_single_data_page(struct page *page, int *submitted,/* we should bypass data pages to proceed the kworkder jobs */ if (unlikely (f2fs_cp_error (sbi))) {mapping_set_error (page->mapping, -EIO ); -/* -* don't drop any dirty dentry pages for keeping lastest -* directory structure. -*/ -if (S_ISDIR (inode->i_mode)) - goto redirty_out; - goto out; + +if (is_sbi_flag_set (sbi,SBI_IS_CLOSE )) + goto out; + +if (has_not_enough_free_secs (sbi,0 ,0 )) { +/* +* don't drop any dirty dentry pages for keeping lastest +* directory structure. +*/ +if (S_ISDIR (inode->i_mode)) + goto redirty_out; + goto out; + } }if (unlikely (is_sbi_flag_set (sbi,SBI_POR_DOING )))fs/f2fs/node.c +5
-2@@ -1527,7 +1527,10 @@ static int __write_node_page(struct page *page, bool atomic, bool *submitted,unlock_page (page);return 0 ; } - goto redirty_out; +if (has_not_enough_free_secs (sbi,0 ,0 )) + goto redirty_out; +set_fsync_mark (page,0 ); +set_dentry_mark (page,0 ); }if (unlikely (is_sbi_flag_set (sbi,SBI_POR_DOING )))@@ -1568,7 +1571,7 @@ static int __write_node_page(struct page *page, bool atomic, bool *submitted, goto redirty_out; } -if (atomic && !test_opt (sbi,NOBARRIER )) +if (atomic && !test_opt (sbi,NOBARRIER ) && !f2fs_cp_error (sbi)) fio.op_flags |=REQ_PREFLUSH |REQ_FUA ;/* should add to global list before clearing PAGECACHE status */