| Message ID | 20190911093650.35329-1-yuchao0@huawei.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | quota: fix wrong condition in is_quota_modification() | expand |
On Wed 11-09-19 17:36:50, Chao Yu wrote: > Quoted from > commit 3da40c7b0898 ("ext4: only call ext4_truncate when size <= isize") > > " At LSF we decided that if we truncate up from isize we shouldn't trim > fallocated blocks that were fallocated with KEEP_SIZE and are past the > new i_size. This patch fixes ext4 to do this. " > > And generic/092 of fstest have covered this case for long time, however > is_quota_modification() didn't adjust based on that rule, so that in > below condition, we will lose to quota block change: > - fallocate blocks beyond EOF > - remount > - truncate(file_path, file_size) > > Fix it. > > Signed-off-by: Chao Yu <yuchao0@huawei.com> > --- > include/linux/quotaops.h | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/include/linux/quotaops.h b/include/linux/quotaops.h > index dc905a4ff8d7..bd30acad3a7f 100644 > --- a/include/linux/quotaops.h > +++ b/include/linux/quotaops.h > @@ -22,7 +22,7 @@ static inline struct quota_info *sb_dqopt(struct super_block *sb) > /* i_mutex must being held */ > static inline bool is_quota_modification(struct inode *inode, struct iattr *ia) > { > - return (ia->ia_valid & ATTR_SIZE && ia->ia_size != inode->i_size) || > + return (ia->ia_valid & ATTR_SIZE && ia->ia_size <= inode->i_size) || > (ia->ia_valid & ATTR_UID && !uid_eq(ia->ia_uid, inode->i_uid)) || > (ia->ia_valid & ATTR_GID && !gid_eq(ia->ia_gid, inode->i_gid)); > } OK, but your change makes i_size extension not to be quota modification which is IMO wrong. So I think the condition should just be: return (ia->ia_valid & ATTR_SIZE) || ... I'll fix the patch up and pull it into my tree. Honza
On 2019/9/12 18:06, Jan Kara wrote: > On Wed 11-09-19 17:36:50, Chao Yu wrote: >> Quoted from >> commit 3da40c7b0898 ("ext4: only call ext4_truncate when size <= isize") >> >> " At LSF we decided that if we truncate up from isize we shouldn't trim >> fallocated blocks that were fallocated with KEEP_SIZE and are past the >> new i_size. This patch fixes ext4 to do this. " >> >> And generic/092 of fstest have covered this case for long time, however >> is_quota_modification() didn't adjust based on that rule, so that in >> below condition, we will lose to quota block change: >> - fallocate blocks beyond EOF >> - remount >> - truncate(file_path, file_size) >> >> Fix it. >> >> Signed-off-by: Chao Yu <yuchao0@huawei.com> >> --- >> include/linux/quotaops.h | 2 +- >> 1 file changed, 1 insertion(+), 1 deletion(-) >> >> diff --git a/include/linux/quotaops.h b/include/linux/quotaops.h >> index dc905a4ff8d7..bd30acad3a7f 100644 >> --- a/include/linux/quotaops.h >> +++ b/include/linux/quotaops.h >> @@ -22,7 +22,7 @@ static inline struct quota_info *sb_dqopt(struct super_block *sb) >> /* i_mutex must being held */ >> static inline bool is_quota_modification(struct inode *inode, struct iattr *ia) >> { >> - return (ia->ia_valid & ATTR_SIZE && ia->ia_size != inode->i_size) || >> + return (ia->ia_valid & ATTR_SIZE && ia->ia_size <= inode->i_size) || >> (ia->ia_valid & ATTR_UID && !uid_eq(ia->ia_uid, inode->i_uid)) || >> (ia->ia_valid & ATTR_GID && !gid_eq(ia->ia_gid, inode->i_gid)); >> } > > OK, but your change makes i_size extension not to be quota modification I just try to adapt below rules covered with generic/092, which restrict to not trim preallocate blocks beyond i_size, in that case, filesystem won't change i_blocks. 1) truncate(i_size) will trim all blocks past i_size. 2) truncate(x) where x > i_size will not trim all blocks past i_size. However, I'm okay with your change, because there could be filesystems won't follow above rule. Thanks, > which is IMO wrong. So I think the condition should just be: > > return (ia->ia_valid & ATTR_SIZE) || ... > > I'll fix the patch up and pull it into my tree. > > Honza >
On Mon 16-09-19 10:53:08, Chao Yu wrote: > On 2019/9/12 18:06, Jan Kara wrote: > > On Wed 11-09-19 17:36:50, Chao Yu wrote: > >> diff --git a/include/linux/quotaops.h b/include/linux/quotaops.h > >> index dc905a4ff8d7..bd30acad3a7f 100644 > >> --- a/include/linux/quotaops.h > >> +++ b/include/linux/quotaops.h > >> @@ -22,7 +22,7 @@ static inline struct quota_info *sb_dqopt(struct super_block *sb) > >> /* i_mutex must being held */ > >> static inline bool is_quota_modification(struct inode *inode, struct iattr *ia) > >> { > >> - return (ia->ia_valid & ATTR_SIZE && ia->ia_size != inode->i_size) || > >> + return (ia->ia_valid & ATTR_SIZE && ia->ia_size <= inode->i_size) || > >> (ia->ia_valid & ATTR_UID && !uid_eq(ia->ia_uid, inode->i_uid)) || > >> (ia->ia_valid & ATTR_GID && !gid_eq(ia->ia_gid, inode->i_gid)); > >> } > > > > OK, but your change makes i_size extension not to be quota modification > > I just try to adapt below rules covered with generic/092, which restrict > to not trim preallocate blocks beyond i_size, in that case, filesystem > won't change i_blocks. > > 1) truncate(i_size) will trim all blocks past i_size. > 2) truncate(x) where x > i_size will not trim all blocks past i_size. Ah, OK. > However, I'm okay with your change, because there could be filesystems won't > follow above rule. Yes, I'm concerned that some filesystem may change i_blocks in some corner case when growing inode size (e.g. when it decides to convert inode from inline format to a normal block based format or something like that). So I don't think the optimization is really worth the chance for breakage. Honza
On 2019/9/16 16:23, Jan Kara wrote: > On Mon 16-09-19 10:53:08, Chao Yu wrote: >> On 2019/9/12 18:06, Jan Kara wrote: >>> On Wed 11-09-19 17:36:50, Chao Yu wrote: >>>> diff --git a/include/linux/quotaops.h b/include/linux/quotaops.h >>>> index dc905a4ff8d7..bd30acad3a7f 100644 >>>> --- a/include/linux/quotaops.h >>>> +++ b/include/linux/quotaops.h >>>> @@ -22,7 +22,7 @@ static inline struct quota_info *sb_dqopt(struct super_block *sb) >>>> /* i_mutex must being held */ >>>> static inline bool is_quota_modification(struct inode *inode, struct iattr *ia) >>>> { >>>> - return (ia->ia_valid & ATTR_SIZE && ia->ia_size != inode->i_size) || >>>> + return (ia->ia_valid & ATTR_SIZE && ia->ia_size <= inode->i_size) || >>>> (ia->ia_valid & ATTR_UID && !uid_eq(ia->ia_uid, inode->i_uid)) || >>>> (ia->ia_valid & ATTR_GID && !gid_eq(ia->ia_gid, inode->i_gid)); >>>> } >>> >>> OK, but your change makes i_size extension not to be quota modification >> >> I just try to adapt below rules covered with generic/092, which restrict >> to not trim preallocate blocks beyond i_size, in that case, filesystem >> won't change i_blocks. >> >> 1) truncate(i_size) will trim all blocks past i_size. >> 2) truncate(x) where x > i_size will not trim all blocks past i_size. > > Ah, OK. > >> However, I'm okay with your change, because there could be filesystems won't >> follow above rule. > > Yes, I'm concerned that some filesystem may change i_blocks in some corner > case when growing inode size (e.g. when it decides to convert inode from > inline format to a normal block based format or something like that). So I > don't think the optimization is really worth the chance for breakage. Agreed, :) Thanks, > > Honza >
diff --git a/include/linux/quotaops.h b/include/linux/quotaops.h index dc905a4ff8d7..bd30acad3a7f 100644 --- a/include/linux/quotaops.h +++ b/include/linux/quotaops.h @@ -22,7 +22,7 @@ static inline struct quota_info *sb_dqopt(struct super_block *sb) /* i_mutex must being held */ static inline bool is_quota_modification(struct inode *inode, struct iattr *ia) { - return (ia->ia_valid & ATTR_SIZE && ia->ia_size != inode->i_size) || + return (ia->ia_valid & ATTR_SIZE && ia->ia_size <= inode->i_size) || (ia->ia_valid & ATTR_UID && !uid_eq(ia->ia_uid, inode->i_uid)) || (ia->ia_valid & ATTR_GID && !gid_eq(ia->ia_gid, inode->i_gid)); }
Quoted from commit 3da40c7b0898 ("ext4: only call ext4_truncate when size <= isize") " At LSF we decided that if we truncate up from isize we shouldn't trim fallocated blocks that were fallocated with KEEP_SIZE and are past the new i_size. This patch fixes ext4 to do this. " And generic/092 of fstest have covered this case for long time, however is_quota_modification() didn't adjust based on that rule, so that in below condition, we will lose to quota block change: - fallocate blocks beyond EOF - remount - truncate(file_path, file_size) Fix it. Signed-off-by: Chao Yu <yuchao0@huawei.com> --- include/linux/quotaops.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)