udf: Fix race between file type conversion and writeback
Subsystems (2)
Jan Kara
reviewerv4 cover udf: Fix race between file type conversion and writeback
Hello,
here is the third version of patches to fix a race between conversion of UDF file from inline format to the standard format and writeback resulting in possible memory corruption. As Christoph didn't like opencoding folio_prepare_writeback() in UDF either, this version provides mpage_writepages() variant with a per-folio callback instead. Review/Acks are welcome in particular for the first patch. I'd like to push the fix to Linus soon as it's kind of nasty issue.
Changes since v3:
* Create mpage_writepages() variant with a callback instead of opencoding.
Changes since v2:
* Opencoded folio_prepare_writeback() in udf instead of exporting it
Changes since v1:
* Fix crash in udf_writepages() when file had no folio in the mapping
Honza
Previous versions:
Jan Kara
reviewerPATCH v3 1/2 mpage: Provide variant of mpage_writepages() with own optional folio handler
Some filesystems need to treat some folios specially (for example for inodes with inline data). Doing the handling in their .writepages method in a race-free manner results in duplicating some of the writeback internals. So provide generalized version of mpage_writepages() that allows filesystem to provide a handler called for each folio which can handle the folio in a special way.
--- fs/mpage.c | 21 +++++++++++++++++---- include/linux/mpage.h | 11 +++++++++-- 2 files changed, 26 insertions(+), 6 deletions(-)
fs/mpage.c +17 -4
@@ -655,8 +655,10 @@ static int mpage_write_folio(struct writeback_control *wbc, struct folio *folio,* address_space_operation. */ int -mpage_writepages (struct address_space *mapping, -struct writeback_control *wbc,get_block_t get_block) +__mpage_writepages (struct address_space *mapping, +struct writeback_control *wbc,get_block_t get_block, +int (*write_folio)(struct folio *folio, +struct writeback_control *wbc)) {struct mpage_data mpd = { .get_block = get_block,
@@ -666,11 +668,22 @@ mpage_writepages(struct address_space *mapping,int error;blk_start_plug (&plug); -while ((folio =writeback_iter (mapping, wbc, folio, &error))) +while ((folio =writeback_iter (mapping, wbc, folio, &error))) { +if (write_folio) { + error =write_folio (folio, wbc); +/* +* == 0 means folio is handled, < 0 means error. In +* both cases hand back control to writeback_iter() +*/ +if (error <=0 ) +continue ; +/* Let mpage_write_folio() handle the folio. */ + } error =mpage_write_folio (wbc, folio, &mpd); + }if (mpd.bio)mpage_bio_submit_write (mpd.bio);blk_finish_plug (&plug);return error; } -EXPORT_SYMBOL (mpage_writepages); +EXPORT_SYMBOL (__mpage_writepages);
include/linux/mpage.h +9 -2
@@ -17,7 +17,14 @@ struct readahead_control;void mpage_readahead (struct readahead_control *,get_block_t get_block);int mpage_read_folio (struct folio *folio,get_block_t get_block); -int mpage_writepages (struct address_space *mapping, -struct writeback_control *wbc,get_block_t get_block); +int __mpage_writepages (struct address_space *mapping, +struct writeback_control *wbc,get_block_t get_block, +int (*write_folio)(struct folio *folio, +struct writeback_control *wbc)); +static inline int mpage_writepages (struct address_space *mapping, +struct writeback_control *wbc,get_block_t get_block) +{ +return __mpage_writepages (mapping, wbc, get_block,NULL ); +}#endif
Signature · 1 line
2.51.0
Re: [PATCH v3 1/2] mpage: Provide variant of mpage_writepages() with own optional folio handler
Looks good:
Jan Kara
reviewerPATCH v3 2/2 udf: Fix race between file type conversion and writeback
In reply to Jan Karaudf_setsize() can race with udf_writepages() as follows:
udf_setsize() udf_writepages()
if (iinfo->i_alloc_type == ICBTAG_FLAG_AD_IN_ICB)
err = udf_expand_file_adinicb(inode);
err = udf_extend_file(inode, newsize);
udf_adinicb_writepages()
memcpy_from_file_folio() - crash
because inode size is too big.
Fix the problem by rechecking file type under folio lock in udf_writepages() which properly serializes with udf_expand_file_adinicb(). Since it is quite difficult to implement this locking with current writeback_iter() logic, let's just opencode the logic necessary to prepare (the only) folio the inode can have for writeback.
--- fs/udf/inode.c | 33 +++++++++++++++------------------ 1 file changed, 15 insertions(+), 18 deletions(-)
fs/udf/inode.c +15 -18
@@ -181,22 +181,23 @@ static void udf_write_failed(struct address_space *mapping, loff_t to) } } -static int udf_adinicb_writepages (struct address_space *mapping, -struct writeback_control *wbc) +static int udf_handle_page_wb (struct folio *folio, +struct writeback_control *wbc) { -struct inode *inode = mapping->host; +struct inode *inode = folio->mapping->host;struct udf_inode_info *iinfo =UDF_I (inode); -struct folio *folio =NULL ; -int error =0 ; -while ((folio =writeback_iter (mapping, wbc, folio, &error))) { -BUG_ON (!folio_test_locked (folio)); -BUG_ON (folio->index !=0 ); -memcpy_from_file_folio (iinfo->i_data + iinfo->i_lenEAttr, folio, -0 ,i_size_read (inode)); -folio_unlock (folio); - } +/* +* Inodes in the normal format are handled by the generic code. This +* check is race-free as the folio lock protects us from inode type +* conversion. +*/ +if (iinfo->i_alloc_type !=ICBTAG_FLAG_AD_IN_ICB ) +return 1 ; +memcpy_from_file_folio (iinfo->i_data + iinfo->i_lenEAttr, folio, +0 ,i_size_read (inode)); +folio_unlock (folio);mark_inode_dirty (inode);return 0 ; }
@@ -204,12 +205,8 @@ static int udf_adinicb_writepages(struct address_space *mapping,static int udf_writepages (struct address_space *mapping,struct writeback_control *wbc) { -struct inode *inode = mapping->host; -struct udf_inode_info *iinfo =UDF_I (inode); - -if (iinfo->i_alloc_type ==ICBTAG_FLAG_AD_IN_ICB ) -return udf_adinicb_writepages (mapping, wbc); -return mpage_writepages (mapping, wbc, udf_get_block_wb); +return __mpage_writepages (mapping, wbc, udf_get_block_wb, + udf_handle_page_wb); }static void udf_adinicb_read_folio (struct folio *folio)
Signature · 1 line
2.51.0
Looks good: