| Message ID | 1375198724-22857-1-git-send-email-jlayton@redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Added to for-next branch of cifs-2.6.git On Tue, Jul 30, 2013 at 10:38 AM, Jeff Layton <jlayton@redhat.com> wrote: > Currently, the s_root dentry doesn't get its d_op pointer set to > anything. This breaks lookups in the root of case-insensitive mounts > since that relies on having d_hash and d_compare routines that know to > treat the filename as case-insensitive. > > cifs.ko has been broken this way for a long time, but commit 1c929cfe6 > ("switch cifs"), added a cryptic comment which is removed in the patch > below, which makes me wonder if this was done deliberately for some > reason. It's not clear to me why we'd want the s_root not to have d_op > set properly. > > It may have something to do with d_automount or d_revalidate on the > root, but my suspicion in looking over the code is that Al was just > trying to preserve the existing behavior when changing this code over to > use s_d_op. > > This patch changes it so that we set s_d_op before calling d_make_root > and removes the comment. I tested mounting, accessing and unmounting > several types of shares (including DFS referrals) and everything still > seemed to work OK afterward. I could be missing something however, so > please do let me know if I am. > > Reported-by: Jan-Marek Glogowski <glogow@fbihome.de> > Cc: Al Viro <viro@ZenIV.linux.org.uk> > Cc: Ian Kent <raven@themaw.net> > Signed-off-by: Jeff Layton <jlayton@redhat.com> > --- > fs/cifs/cifsfs.c | 11 +++++------ > 1 file changed, 5 insertions(+), 6 deletions(-) > > diff --git a/fs/cifs/cifsfs.c b/fs/cifs/cifsfs.c > index 4bdd547..85ea98d 100644 > --- a/fs/cifs/cifsfs.c > +++ b/fs/cifs/cifsfs.c > @@ -147,18 +147,17 @@ cifs_read_super(struct super_block *sb) > goto out_no_root; > } > > + if (cifs_sb_master_tcon(cifs_sb)->nocase) > + sb->s_d_op = &cifs_ci_dentry_ops; > + else > + sb->s_d_op = &cifs_dentry_ops; > + > sb->s_root = d_make_root(inode); > if (!sb->s_root) { > rc = -ENOMEM; > goto out_no_root; > } > > - /* do that *after* d_make_root() - we want NULL ->d_op for root here */ > - if (cifs_sb_master_tcon(cifs_sb)->nocase) > - sb->s_d_op = &cifs_ci_dentry_ops; > - else > - sb->s_d_op = &cifs_dentry_ops; > - > #ifdef CONFIG_CIFS_NFSD_EXPORT > if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_SERVER_INUM) { > cifs_dbg(FYI, "export ops supported\n"); > -- > 1.8.3.1 >
diff --git a/fs/cifs/cifsfs.c b/fs/cifs/cifsfs.c index 4bdd547..85ea98d 100644 --- a/fs/cifs/cifsfs.c +++ b/fs/cifs/cifsfs.c @@ -147,18 +147,17 @@ cifs_read_super(struct super_block *sb) goto out_no_root; } + if (cifs_sb_master_tcon(cifs_sb)->nocase) + sb->s_d_op = &cifs_ci_dentry_ops; + else + sb->s_d_op = &cifs_dentry_ops; + sb->s_root = d_make_root(inode); if (!sb->s_root) { rc = -ENOMEM; goto out_no_root; } - /* do that *after* d_make_root() - we want NULL ->d_op for root here */ - if (cifs_sb_master_tcon(cifs_sb)->nocase) - sb->s_d_op = &cifs_ci_dentry_ops; - else - sb->s_d_op = &cifs_dentry_ops; - #ifdef CONFIG_CIFS_NFSD_EXPORT if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_SERVER_INUM) { cifs_dbg(FYI, "export ops supported\n");
Currently, the s_root dentry doesn't get its d_op pointer set to anything. This breaks lookups in the root of case-insensitive mounts since that relies on having d_hash and d_compare routines that know to treat the filename as case-insensitive. cifs.ko has been broken this way for a long time, but commit 1c929cfe6 ("switch cifs"), added a cryptic comment which is removed in the patch below, which makes me wonder if this was done deliberately for some reason. It's not clear to me why we'd want the s_root not to have d_op set properly. It may have something to do with d_automount or d_revalidate on the root, but my suspicion in looking over the code is that Al was just trying to preserve the existing behavior when changing this code over to use s_d_op. This patch changes it so that we set s_d_op before calling d_make_root and removes the comment. I tested mounting, accessing and unmounting several types of shares (including DFS referrals) and everything still seemed to work OK afterward. I could be missing something however, so please do let me know if I am. Reported-by: Jan-Marek Glogowski <glogow@fbihome.de> Cc: Al Viro <viro@ZenIV.linux.org.uk> Cc: Ian Kent <raven@themaw.net> Signed-off-by: Jeff Layton <jlayton@redhat.com> --- fs/cifs/cifsfs.c | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-)