ext4: avoid ext4_error()'s caused by ENOMEM in the truncate path

PATCH Applied Anna Pendleton 3 participants 3 messages May 7, 2020 – May 21, 2020 ext4 file system

lore

Share & tools

lore

Anna Pendleton

author

2020-05-07 17:50 UTC #0 lore raw

PATCH 1/1 ext4: avoid ext4_error()'s caused by ENOMEM in the truncate path

From: Theodore Ts'o <[email protected]>

We can't fail in the truncate path without requiring an fsck. Add work around for this by using a combination of retry loops and the __GFP_NOFAIL flag.

From: Theodore Ts'o <[email protected]>

Signed-off-by Theodore Ts'o <[email protected]>

Signed-off-by Anna Pendleton <[email protected]>

Reviewed-by Harshad Shirwadkar <[email protected]>

---
 fs/ext4/ext4.h    |  1 +
 fs/ext4/extents.c | 43 +++++++++++++++++++++++++++++++++----------
 2 files changed, 34 insertions(+), 10 deletions(-)

fs/ext4/ext4.h +1 -0

index 91eb4381cae5..af010f5009e4 100644
@@ -632,6 +632,7 @@ enum {
  */
 #define EXT4_EX_NOCACHE				0x40000000
 #define EXT4_EX_FORCE_CACHE			0x20000000
+#define EXT4_EX_NOFAIL				0x10000000
 
 /*
  * Flags used by ext4_free_blocks

fs/ext4/extents.c +33 ~~-10~~

116 lines not on this page · raw message

Signature · 1 line

2.26.2.526.g744177e7f7-goog

kbuild test robot

build bot

2020-05-12 06:27 UTC #1 lore raw

I love your patch! Perhaps something to improve:

Hi Anna,

I love your patch! Perhaps something to improve:

[auto build test WARNING on ext4/dev]
[also build test WARNING on tytso-fscrypt/master v5.7-rc4 next-20200508] [if your patch is applied to the wrong git tree, please drop us a note to help improve the system. BTW, we also suggest to use '--base' option to specify the base tree in git format-patch, please see https://stackoverflow.com/a/37406982]

url:    https://github.com/0day-ci/linux/commits/Anna-Pendleton/ext4-avoid-ext4_error-s-caused-by-ENOMEM-in-the-truncate-path/20200508-062229
base:   https://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4.git dev
reproduce:
        # apt-get install sparse
        # sparse version: 
        make ARCH=x86_64 allmodconfig
        make C=1 CF='-fdiagnostic-prefix -D__CHECK_ENDIAN__'
:::::: branch date: 22 hours ago
:::::: commit date: 22 hours ago

If you fix the issue, kindly add following tag as appropriate

Reported-by kbuild test robot <[email protected]>

sparse warnings: (new ones prefixed by >>)

fs/ext4/extents.c:493:67: sparse: warning: incorrect type in initializer (different base types)

>> fs/ext4/extents.c:493:67: sparse:    expected int gfp_flags
>> fs/ext4/extents.c:493:67: sparse:    got restricted gfp_t
   fs/ext4/extents.c:496:27: sparse: warning: invalid assignment: |=
>> fs/ext4/extents.c:496:27: sparse:    left side has type int
>> fs/ext4/extents.c:496:27: sparse:    right side has type restricted gfp_t
   fs/ext4/extents.c:498:47: sparse: warning: incorrect type in argument 3 (different base types)
>> fs/ext4/extents.c:498:47: sparse:    expected restricted gfp_t [usertype] gfp
>> fs/ext4/extents.c:498:47: sparse:    got int gfp_flags
   fs/ext4/extents.c:848:25: sparse: warning: incorrect type in initializer (different base types)
   fs/ext4/extents.c:848:25: sparse:    expected int gfp_flags
   fs/ext4/extents.c:848:25: sparse:    got restricted gfp_t
   fs/ext4/extents.c:851:27: sparse: warning: invalid assignment: |=
   fs/ext4/extents.c:851:27: sparse:    left side has type int
   fs/ext4/extents.c:851:27: sparse:    right side has type restricted gfp_t
   fs/ext4/extents.c:872:33: sparse: warning: incorrect type in argument 3 (different base types)
>> fs/ext4/extents.c:872:33: sparse:    expected restricted gfp_t [usertype] flags

fs/ext4/extents.c:872:33: sparse: got int gfp_flags
fs/ext4/extents.c:1022:25: sparse: warning: incorrect type in initializer (different base types)
fs/ext4/extents.c:1022:25: sparse: expected int gfp_flags
fs/ext4/extents.c:1022:25: sparse: got restricted gfp_t
fs/ext4/extents.c:1027:27: sparse: warning: invalid assignment: |=
fs/ext4/extents.c:1027:27: sparse: left side has type int
fs/ext4/extents.c:1027:27: sparse: right side has type restricted gfp_t
fs/ext4/extents.c:1062:56: sparse: warning: incorrect type in argument 3 (different base types)
fs/ext4/extents.c:1062:56: sparse: expected restricted gfp_t [usertype] flags
fs/ext4/extents.c:1062:56: sparse: got int gfp_flags

# https://github.com/0day-ci/linux/commit/7d07f7d72800b68a55822a5705514d3c3bc82b63
git remote add linux-review https://github.com/0day-ci/linux
git remote update linux-review
git checkout 7d07f7d72800b68a55822a5705514d3c3bc82b63
vim +493 fs/ext4/extents.c

4068664e3cd231 Dmitry Monakhov 2019-11-06  485  
7d7ea89e756ea1 Theodore Ts'o   2013-08-16  486  static struct buffer_head *
7d7ea89e756ea1 Theodore Ts'o   2013-08-16  487  __read_extent_tree_block(const char *function, unsigned int line,
107a7bd31ac003 Theodore Ts'o   2013-08-16  488  			 struct inode *inode, ext4_fsblk_t pblk, int depth,
107a7bd31ac003 Theodore Ts'o   2013-08-16  489  			 int flags)
f84891289e62a7 Darrick J. Wong 2012-04-29  490  {
7d7ea89e756ea1 Theodore Ts'o   2013-08-16  491  	struct buffer_head		*bh;
7d7ea89e756ea1 Theodore Ts'o   2013-08-16  492  	int				err;
7d07f7d72800b6 Theodore Ts'o   2020-05-07 @493  	int				gfp_flags = __GFP_MOVABLE | GFP_NOFS;
f84891289e62a7 Darrick J. Wong 2012-04-29  494  
7d07f7d72800b6 Theodore Ts'o   2020-05-07  495  	if (flags & EXT4_EX_NOFAIL)
7d07f7d72800b6 Theodore Ts'o   2020-05-07 @496  		gfp_flags |= __GFP_NOFAIL;
7d07f7d72800b6 Theodore Ts'o   2020-05-07  497  
7d07f7d72800b6 Theodore Ts'o   2020-05-07 @498  	bh = sb_getblk_gfp(inode->i_sb, pblk, gfp_flags);
7d7ea89e756ea1 Theodore Ts'o   2013-08-16  499  	if (unlikely(!bh))
7d7ea89e756ea1 Theodore Ts'o   2013-08-16  500  		return ERR_PTR(-ENOMEM);
7d7ea89e756ea1 Theodore Ts'o   2013-08-16  501  
7d7ea89e756ea1 Theodore Ts'o   2013-08-16  502  	if (!bh_uptodate_or_lock(bh)) {
7d7ea89e756ea1 Theodore Ts'o   2013-08-16  503  		trace_ext4_ext_load_extent(inode, pblk, _RET_IP_);
7d7ea89e756ea1 Theodore Ts'o   2013-08-16  504  		err = bh_submit_read(bh);
7d7ea89e756ea1 Theodore Ts'o   2013-08-16  505  		if (err < 0)
7d7ea89e756ea1 Theodore Ts'o   2013-08-16  506  			goto errout;
7d7ea89e756ea1 Theodore Ts'o   2013-08-16  507  	}
7869a4a6c5caa7 Theodore Ts'o   2013-08-16  508  	if (buffer_verified(bh) && !(flags & EXT4_EX_FORCE_CACHE))
7d7ea89e756ea1 Theodore Ts'o   2013-08-16  509  		return bh;
0a944e8a6c66ca Theodore Ts'o   2019-05-22  510  	if (!ext4_has_feature_journal(inode->i_sb) ||
0a944e8a6c66ca Theodore Ts'o   2019-05-22  511  	    (inode->i_ino !=
0a944e8a6c66ca Theodore Ts'o   2019-05-22  512  	     le32_to_cpu(EXT4_SB(inode->i_sb)->s_es->s_journal_inum))) {
7d7ea89e756ea1 Theodore Ts'o   2013-08-16  513  		err = __ext4_ext_check(function, line, inode,
c349179b4808f7 Theodore Ts'o   2013-08-16  514  				       ext_block_hdr(bh), depth, pblk);
7d7ea89e756ea1 Theodore Ts'o   2013-08-16  515  		if (err)
7d7ea89e756ea1 Theodore Ts'o   2013-08-16  516  			goto errout;
0a944e8a6c66ca Theodore Ts'o   2019-05-22  517  	}
f84891289e62a7 Darrick J. Wong 2012-04-29  518  	set_buffer_verified(bh);
107a7bd31ac003 Theodore Ts'o   2013-08-16  519  	/*
107a7bd31ac003 Theodore Ts'o   2013-08-16  520  	 * If this is a leaf block, cache all of its entries
107a7bd31ac003 Theodore Ts'o   2013-08-16  521  	 */
107a7bd31ac003 Theodore Ts'o   2013-08-16  522  	if (!(flags & EXT4_EX_NOCACHE) && depth == 0) {
107a7bd31ac003 Theodore Ts'o   2013-08-16  523  		struct ext4_extent_header *eh = ext_block_hdr(bh);
4068664e3cd231 Dmitry Monakhov 2019-11-06  524  		ext4_cache_extents(inode, eh);
107a7bd31ac003 Theodore Ts'o   2013-08-16  525  	}
7d7ea89e756ea1 Theodore Ts'o   2013-08-16  526  	return bh;
7d7ea89e756ea1 Theodore Ts'o   2013-08-16  527  errout:
7d7ea89e756ea1 Theodore Ts'o   2013-08-16  528  	put_bh(bh);
7d7ea89e756ea1 Theodore Ts'o   2013-08-16  529  	return ERR_PTR(err);
7d7ea89e756ea1 Theodore Ts'o   2013-08-16  530

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/[email protected]

Theodore Y. Ts'o

maintainer Applied

2020-05-21 16:55 UTC #2 lore raw

In reply to Anna Pendleton

Quoted 10 lines · Anna Pendleton

From: Theodore Ts'o <[email protected]>
We can't fail in the truncate path without requiring an fsck. Add work around for this by using a combination of retry loops and the __GFP_NOFAIL flag.
From: Theodore Ts'o <[email protected]>
Signed-off-by Theodore Ts'o <[email protected]>
Signed-off-by Anna Pendleton <[email protected]>
Reviewed-by Harshad Shirwadkar <[email protected]>

Thanks, applied.

Per the feedback from the kbuild test robot, I changed "int gfp_flags =..." to "gfp_t gfp_flags=..." in three places in the patches.

Cheers,

- Ted