From: Kaixu Xia <kaixuxia@tencent.com> Here we use the READ_ONCE to fix race conditions in ->d_compare() and ->d_hash() when they are called in RCU-walk mode, seems we can use the normal helper d_inode_rcu() to get the actual inode. Signed-off-by: Kaixu Xia <kaixuxia@tencent.com> --- fs/ext4/dir.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/fs/ext4/dir.c b/fs/ext4/dir.c index 1d82336b1cd4..3bf6cb8e55f6 100644 --- a/fs/ext4/dir.c +++ b/fs/ext4/dir.c @@ -674,7 +674,7 @@ static int ext4_d_compare(const struct dentry *dentry, unsigned int len, { struct qstr qstr = {.name = str, .len = len }; const struct dentry *parent = READ_ONCE(dentry->d_parent); - const struct inode *inode = READ_ONCE(parent->d_inode); + const struct inode *inode = d_inode_rcu(parent); char strbuf[DNAME_INLINE_LEN]; if (!inode || !IS_CASEFOLDED(inode) || @@ -706,7 +706,7 @@ static int ext4_d_hash(const struct dentry *dentry, struct qstr *str) { const struct ext4_sb_info *sbi = EXT4_SB(dentry->d_sb); const struct unicode_map *um = sbi->s_encoding; - const struct inode *inode = READ_ONCE(dentry->d_inode); + const struct inode *inode = d_inode_rcu(dentry); unsigned char *norm; int len, ret = 0; -- 2.20.0
On Sat, Oct 10, 2020 at 04:10:16PM +0800, xiakaixu1987@gmail.com wrote: > From: Kaixu Xia <kaixuxia@tencent.com> > > Here we use the READ_ONCE to fix race conditions in ->d_compare() and > ->d_hash() when they are called in RCU-walk mode, seems we can use > the normal helper d_inode_rcu() to get the actual inode. Looks good to me. Thanks! Reviewed-by: Lukas Czerner <lczerner@redhat.com> > > Signed-off-by: Kaixu Xia <kaixuxia@tencent.com> > --- > fs/ext4/dir.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/fs/ext4/dir.c b/fs/ext4/dir.c > index 1d82336b1cd4..3bf6cb8e55f6 100644 > --- a/fs/ext4/dir.c > +++ b/fs/ext4/dir.c > @@ -674,7 +674,7 @@ static int ext4_d_compare(const struct dentry *dentry, unsigned int len, > { > struct qstr qstr = {.name = str, .len = len }; > const struct dentry *parent = READ_ONCE(dentry->d_parent); > - const struct inode *inode = READ_ONCE(parent->d_inode); > + const struct inode *inode = d_inode_rcu(parent); > char strbuf[DNAME_INLINE_LEN]; > > if (!inode || !IS_CASEFOLDED(inode) || > @@ -706,7 +706,7 @@ static int ext4_d_hash(const struct dentry *dentry, struct qstr *str) > { > const struct ext4_sb_info *sbi = EXT4_SB(dentry->d_sb); > const struct unicode_map *um = sbi->s_encoding; > - const struct inode *inode = READ_ONCE(dentry->d_inode); > + const struct inode *inode = d_inode_rcu(dentry); > unsigned char *norm; > int len, ret = 0; > > -- > 2.20.0 >
On Mon, Oct 12, 2020 at 11:33:04AM +0200, Lukas Czerner wrote:
> On Sat, Oct 10, 2020 at 04:10:16PM +0800, xiakaixu1987@gmail.com wrote:
> > From: Kaixu Xia <kaixuxia@tencent.com>
> >
> > Here we use the READ_ONCE to fix race conditions in ->d_compare() and
> > ->d_hash() when they are called in RCU-walk mode, seems we can use
> > the normal helper d_inode_rcu() to get the actual inode.
>
> Looks good to me.
> Thanks!
>
> Reviewed-by: Lukas Czerner <lczerner@redhat.com>
Thanks, applied.
- Ted