[linux-cifs] Re: [PATCH v2] CIFS: Fix mkdir/rmdir bug for the non-POSIX case


On Sat, 25 Feb 2012 13:11:26 +0400
Pavel Shilovsky  wrote:

> 2012/2/21 Jeff Layton :
> > On Fri, 17 Feb 2012 16:13:30 +0300
> > Pavel Shilovsky  wrote:
> >
> >> Currently we do inc/drop_nlink for a parent directory for every
> >> mkdir/rmdir calls. That's wrong when Unix extensions are disabled
> >> because in this case a server doesn't follow the same semantic and
> >> returns the old value on the next QueryInfo request. As the result,
> >> we update our value with the server one and then decrement it on
> >> every rmdir call - go to negative nlink values.
> >>
> >> Fix this by removing inc/drop_nlink for the parent directory from
> >> mkdir/rmdir, setting it for a revalidation and ignoring NumberOfLinks
> >> for directories when Unix extensions are disabled.
> >>
> >> Signed-off-by: Pavel Shilovsky 
> >> ---
> >>  fs/cifs/inode.c |   28 +++++++++++++++++++---------
> >>  1 files changed, 19 insertions(+), 9 deletions(-)
> >>
> >> diff --git a/fs/cifs/inode.c b/fs/cifs/inode.c
> >> index a5f54b7..745da3d 100644
> >> --- a/fs/cifs/inode.c
> >> +++ b/fs/cifs/inode.c
> >> @@ -534,6 +534,11 @@ cifs_all_info_to_fattr(struct cifs_fattr *fattr, FILE_ALL_INFO *info,
> >>       if (fattr->cf_cifsattrs & ATTR_DIRECTORY) {
> >>               fattr->cf_mode = S_IFDIR | cifs_sb->mnt_dir_mode;
> >>               fattr->cf_dtype = DT_DIR;
> >> +             /*
> >> +              * Server can return wrong NumberOfLinks value for directories
> >> +              * when Unix extensions are disabled - fake it.
> >> +              */
> >> +             fattr->cf_nlink = 2;
> >>       } else {
> >>               fattr->cf_mode = S_IFREG | cifs_sb->mnt_file_mode;
> >>               fattr->cf_dtype = DT_REG;
> >> @@ -541,9 +546,9 @@ cifs_all_info_to_fattr(struct cifs_fattr *fattr, FILE_ALL_INFO *info,
> >>               /* clear write bits if ATTR_READONLY is set */
> >>               if (fattr->cf_cifsattrs & ATTR_READONLY)
> >>                       fattr->cf_mode &= ~(S_IWUGO);
> >> -     }
> >>
> >> -     fattr->cf_nlink = le32_to_cpu(info->NumberOfLinks);
> >> +             fattr->cf_nlink = le32_to_cpu(info->NumberOfLinks);
> >> +     }
> >>
> >>       fattr->cf_uid = cifs_sb->mnt_uid;
> >>       fattr->cf_gid = cifs_sb->mnt_gid;
> >> @@ -1322,7 +1327,6 @@ int cifs_mkdir(struct inode *inode, struct dentry *direntry, umode_t mode)
> >>                       }
> >>  /*BB check (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_SET_UID ) to see if need
> >>       to set uid/gid */
> >> -                     inc_nlink(inode);
> >>
> >>                       cifs_unix_basic_to_fattr(&fattr, pInfo, cifs_sb);
> >>                       cifs_fill_uniqueid(inode->i_sb, &fattr);
> >> @@ -1355,7 +1359,6 @@ mkdir_retry_old:
> >>               d_drop(direntry);
> >>       } else {
> >>  mkdir_get_info:
> >> -             inc_nlink(inode);
> >>               if (pTcon->unix_ext)
> >>                       rc = cifs_get_inode_info_unix(&newinode, full_path,
> >>                                                     inode->i_sb, xid);
> >> @@ -1436,6 +1439,11 @@ mkdir_get_info:
> >>               }
> >>       }
> >>  mkdir_out:
> >> +     /*
> >> +      * Force revalidate to get parent dir info when needed since cached
> >> +      * attributes are invalid now.
> >> +      */
> >> +     CIFS_I(inode)->time = 0;
> >>       kfree(full_path);
> >>       FreeXid(xid);
> >>       cifs_put_tlink(tlink);
> >> @@ -1475,7 +1483,6 @@ int cifs_rmdir(struct inode *inode, struct dentry *direntry)
> >>       cifs_put_tlink(tlink);
> >>
> >>       if (!rc) {
> >> -             drop_nlink(inode);
> >>               spin_lock(&direntry->d_inode->i_lock);
> >>               i_size_write(direntry->d_inode, 0);
> >>               clear_nlink(direntry->d_inode);
> >> @@ -1483,12 +1490,15 @@ int cifs_rmdir(struct inode *inode, struct dentry *direntry)
> >>       }
> >>
> >>       cifsInode = CIFS_I(direntry->d_inode);
> >> -     cifsInode->time = 0;    /* force revalidate to go get info when
> >> -                                needed */
> >> +     /* force revalidate to go get info when needed */
> >> +     cifsInode->time = 0;
> >>
> >>       cifsInode = CIFS_I(inode);
> >> -     cifsInode->time = 0;    /* force revalidate to get parent dir info
> >> -                                since cached search results now invalid */
> >> +     /*
> >> +      * Force revalidate to get parent dir info when needed since cached
> >> +      * attributes are invalid now.
> >> +      */
> >> +     cifsInode->time = 0;
> >>
> >>       direntry->d_inode->i_ctime = inode->i_ctime = inode->i_mtime =
> >>               current_fs_time(inode->i_sb);
> >
> > Looks good.
> >
> > Reviewed-by: Jeff Layton 
> 
> Steve, what's about this patch? I think it should go to stable as well.
> 

I think we probably ought to wait on stable. There is some (small)
potential for regression from this change. AFAIK, this is just fixing a
WARN_ON, so I think that there's no real urgency to stick it into
stable just yet.

We can always reconsider that later too if it turns out that there's
need to do so.

-- 
Jeff Layton 

This message from: http://www.mailbrowse.com/linux-cifs/5535.html
Previous message: Re: [linux-cifs-client] [PATCH] cifs: hard mount option behaviour implementation
Next message:Re: [PATCH v2] CIFS: Fix mkdir/rmdir bug for the non-POSIX case