Message ID | HK2PR06MB045211675B43EED794E597B6D56E0@HK2PR06MB0452.apcprd06.prod.outlook.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
Hi Changwei, On 2018/5/28 22:40, Changwei Ge wrote: > From: Changwei Ge <ge.changwei@h3c.com> > > Somehow, file system metadata was corrupted, which causes > ocfs2_check_dir_entry() to fail in function ocfs2_dir_foreach_blk_el(). > > According to the original design intention, if above happens we should > skip the problematic block and continue to retrieve dir entry. But there > is obviouse misuse of brelse around related code. > > After failure of ocfs2_check_dir_entry(), currunt code just moves to next > position and uses the problematic buffer head again and again during > which the problematic buffer head is released for multiple times. I > suppose, this a serious issue which is long-lived in ocfs2. This may > cause other file systems which is also used in a the same host insane. > > So we should also consider about bakcporting this patch into > linux -stable. > > Suggested-by: Changkuo Shi <shi.changkuo@h3c.com> > Cc: stable@vger.kernel.org > Signed-off-by: Changwei Ge <ge.changwei@h3c.com> > --- > fs/ocfs2/dir.c | 3 +-- > 1 file changed, 1 insertion(+), 2 deletions(-) > > diff --git a/fs/ocfs2/dir.c b/fs/ocfs2/dir.c > index b048d4f..c121abb 100644 > --- a/fs/ocfs2/dir.c > +++ b/fs/ocfs2/dir.c > @@ -1897,8 +1897,7 @@ static int ocfs2_dir_foreach_blk_el(struct inode *inode, > /* On error, skip the f_pos to the > next block. */ > ctx->pos = (ctx->pos | (sb->s_blocksize - 1)) + 1; > - brelse(bh); > - continue; > + break; I guess return is more appropriate than break here as it will cause double free buffer: " ctx->pos = (ctx->pos | (sb->s_blocksize - 1)) + 1; brelse(bh); break; " " brelse(bh); bh = NULL; if (!persist && stored) break; " Thanks, Jun > } > if (le64_to_cpu(de->inode)) { > unsigned char d_type = DT_UNKNOWN; >
Hi Jun, Thanks for looking into this, my code from Linux mainline is not looked like yours... On 2018/11/10 14:36, piaojun wrote: > Hi Changwei, > > On 2018/5/28 22:40, Changwei Ge wrote: >> From: Changwei Ge <ge.changwei@h3c.com> >> >> Somehow, file system metadata was corrupted, which causes >> ocfs2_check_dir_entry() to fail in function ocfs2_dir_foreach_blk_el(). >> >> According to the original design intention, if above happens we should >> skip the problematic block and continue to retrieve dir entry. But there >> is obviouse misuse of brelse around related code. >> >> After failure of ocfs2_check_dir_entry(), currunt code just moves to next >> position and uses the problematic buffer head again and again during >> which the problematic buffer head is released for multiple times. I >> suppose, this a serious issue which is long-lived in ocfs2. This may >> cause other file systems which is also used in a the same host insane. >> >> So we should also consider about bakcporting this patch into >> linux -stable. >> >> Suggested-by: Changkuo Shi <shi.changkuo@h3c.com> >> Cc: stable@vger.kernel.org >> Signed-off-by: Changwei Ge <ge.changwei@h3c.com> >> --- >> fs/ocfs2/dir.c | 3 +-- >> 1 file changed, 1 insertion(+), 2 deletions(-) >> >> diff --git a/fs/ocfs2/dir.c b/fs/ocfs2/dir.c >> index b048d4f..c121abb 100644 >> --- a/fs/ocfs2/dir.c >> +++ b/fs/ocfs2/dir.c >> @@ -1897,8 +1897,7 @@ static int ocfs2_dir_foreach_blk_el(struct inode *inode, >> /* On error, skip the f_pos to the >> next block. */ >> ctx->pos = (ctx->pos | (sb->s_blocksize - 1)) + 1; >> - brelse(bh); >> - continue; >> + break; > > I guess return is more appropriate than break here as it will cause double > free buffer: > > " > ctx->pos = (ctx->pos | (sb->s_blocksize - 1)) + 1; > brelse(bh); > break; > " Linux mainline should not have 'brelse(bh)', could your please rebase your tree? ''' ccd979bdb (Mark Fasheh 2005-12-15 14:31:24 -0800 1897) /* On error, skip the f_pos to the ccd979bdb (Mark Fasheh 2005-12-15 14:31:24 -0800 1898) next block. */ 3704412bd (Al Viro 2013-05-22 21:06:00 -0400 1899) ctx->pos = (ctx->pos | (sb->s_blocksize - 1)) + 1; 29aa30167 (Changwei Ge 2018-11-02 15:48:15 -0700 1900) break; ''' > > " > brelse(bh); > bh = NULL; > if (!persist && stored) > break; > " > > Thanks, > Jun > >> } >> if (le64_to_cpu(de->inode)) { >> unsigned char d_type = DT_UNKNOWN; >> > > _______________________________________________ > Ocfs2-devel mailing list > Ocfs2-devel@oss.oracle.com > https://oss.oracle.com/mailman/listinfo/ocfs2-devel >
Hi Changwei, Sorry for the oversight for your patch. There is nothing wrong. Thanks, Jun On 2018/11/10 15:58, Changwei Ge wrote: > Hi Jun, > > Thanks for looking into this, my code from Linux mainline is not looked like yours... > > On 2018/11/10 14:36, piaojun wrote: >> Hi Changwei, >> >> On 2018/5/28 22:40, Changwei Ge wrote: >>> From: Changwei Ge <ge.changwei@h3c.com> >>> >>> Somehow, file system metadata was corrupted, which causes >>> ocfs2_check_dir_entry() to fail in function ocfs2_dir_foreach_blk_el(). >>> >>> According to the original design intention, if above happens we should >>> skip the problematic block and continue to retrieve dir entry. But there >>> is obviouse misuse of brelse around related code. >>> >>> After failure of ocfs2_check_dir_entry(), currunt code just moves to next >>> position and uses the problematic buffer head again and again during >>> which the problematic buffer head is released for multiple times. I >>> suppose, this a serious issue which is long-lived in ocfs2. This may >>> cause other file systems which is also used in a the same host insane. >>> >>> So we should also consider about bakcporting this patch into >>> linux -stable. >>> >>> Suggested-by: Changkuo Shi <shi.changkuo@h3c.com> >>> Cc: stable@vger.kernel.org >>> Signed-off-by: Changwei Ge <ge.changwei@h3c.com> >>> --- >>> fs/ocfs2/dir.c | 3 +-- >>> 1 file changed, 1 insertion(+), 2 deletions(-) >>> >>> diff --git a/fs/ocfs2/dir.c b/fs/ocfs2/dir.c >>> index b048d4f..c121abb 100644 >>> --- a/fs/ocfs2/dir.c >>> +++ b/fs/ocfs2/dir.c >>> @@ -1897,8 +1897,7 @@ static int ocfs2_dir_foreach_blk_el(struct inode *inode, >>> /* On error, skip the f_pos to the >>> next block. */ >>> ctx->pos = (ctx->pos | (sb->s_blocksize - 1)) + 1; >>> - brelse(bh); >>> - continue; >>> + break; >> >> I guess return is more appropriate than break here as it will cause double >> free buffer: >> >> " >> ctx->pos = (ctx->pos | (sb->s_blocksize - 1)) + 1; >> brelse(bh); >> break; >> " > > Linux mainline should not have 'brelse(bh)', could your please rebase your tree? > > ''' > ccd979bdb (Mark Fasheh 2005-12-15 14:31:24 -0800 1897) /* On error, skip the f_pos to the > ccd979bdb (Mark Fasheh 2005-12-15 14:31:24 -0800 1898) next block. */ > 3704412bd (Al Viro 2013-05-22 21:06:00 -0400 1899) ctx->pos = (ctx->pos | (sb->s_blocksize - 1)) + 1; > 29aa30167 (Changwei Ge 2018-11-02 15:48:15 -0700 1900) break; > ''' >> >> " >> brelse(bh); >> bh = NULL; >> if (!persist && stored) >> break; >> " >> >> Thanks, >> Jun >> >>> } >>> if (le64_to_cpu(de->inode)) { >>> unsigned char d_type = DT_UNKNOWN; >>> >> >> _______________________________________________ >> Ocfs2-devel mailing list >> Ocfs2-devel@oss.oracle.com >> https://oss.oracle.com/mailman/listinfo/ocfs2-devel >> > . >
diff --git a/fs/ocfs2/dir.c b/fs/ocfs2/dir.c index b048d4f..c121abb 100644 --- a/fs/ocfs2/dir.c +++ b/fs/ocfs2/dir.c @@ -1897,8 +1897,7 @@ static int ocfs2_dir_foreach_blk_el(struct inode *inode, /* On error, skip the f_pos to the next block. */ ctx->pos = (ctx->pos | (sb->s_blocksize - 1)) + 1; - brelse(bh); - continue; + break; } if (le64_to_cpu(de->inode)) { unsigned char d_type = DT_UNKNOWN;