[PATCH] Fix do_path_lookup() to add the check for error in link_path_walk()

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 

Hi,
I encountered an oops with 2.6.16-rc1-git3 kernel ( SLES 10 B2 kernel ), while running racer tests. The problem was hit in audit_inode() with the following stack trace :
Jan 31 19:15:27 x236 klogd:Unable to handle kernel paging request at virtual address 6b6b6b8b 
Jan 31 19:15:27 x236 klogd: CPU:    3
Jan 31 19:15:27 x236 klogd: EIP: 0060:[<c013ffbd>] Tainted: G U VLI 
Jan 31 19:15:27 x236 klogd: EFLAGS: 00010282   (2.6.16-rc1-git3-4-smp)
Jan 31 19:15:27 x236 klogd: EIP is at audit_inode+0x78/0xa9
Jan 31 19:15:27 x236 klogd: eax: d29da000   ebx: ccc23638   ecx: 0000001
edx: ccc23638
Jan 31 19:15:27 x236 klogd: esi: 6b6b6b6b   edi: d29da000   ebp: 0000001
esp: ce4d7ecc
Jan 31 19:15:27 x236 klogd: ds: 007b   es: 007b   ss: 0068
Jan 31 19:15:27 x236 klogd: Process ln (pid: 12674, threadinfo=ce4d6000 task=e49df550) 
Jan 31 19:15:27 x236 klogd: Call Trace:
Jan 31 19:15:27 x236 klogd:  [<c016a843>] do_path_lookup+0x225/0x22f
Jan 31 19:15:27 x236 klogd:  [<c016af42>] __user_walk_fd+0x29/0x3a
Jan 31 19:15:27 x236 klogd:  [<c0164e7e>] vfs_stat_fd+0x15/0x3c
Jan 31 19:15:27 x236 klogd:  [<c014ca1c>] __handle_mm_fault+0x439/0x7a0
Jan 31 19:15:27 x236 klogd:  [<c0164f32>] sys_stat64+0xf/0x23
Jan 31 19:15:27 x236 klogd:  [<c0106d26>] do_syscall_trace+0x123/0x169
Jan 31 19:15:27 x236 klogd:  [<c0103c09>] syscall_call+0x7/0xb
I found the root cause of the problem to be the lack of error-check in do_path_lookup() for the link_path_walk(). 

in do_path_lookup:


                fput_light(file, fput_needed);
        }
        read_unlock(&current->fs->lock);
        current->total_link_count = 0;
        retval = link_path_walk(name, nd); <----- No check for retval !
out:
        if (unlikely(current->audit_context
                     && nd && nd->dentry && nd->dentry->d_inode))
                audit_inode(name, nd->dentry->d_inode, flags);
out_fail:
        return retval;
}
If link_path_walk returns error, the inode may not be reliable. This causes the oops in audit_inode.
The bug is there in 2.6.16-rc2 also. I believe the problem in Bugme #5897 also has the same root cause, though it has different call path.
The patch attached below fixes the issue. I have tested it on 2.6.16-rc1-git3 with racer tests and it works fine. 

Thanks,

Suzuki K P
Linux Technology Centre
IBM Software Labs,


-------------------------------------------------------------------------------------------
Fixes do_path_lookup() to avoid accessing invalid dentry or inode when the link_path_walk() has failed. This should fix Bugme #5897. 

Signed Off by: Suzuki K P <[email protected]>

--- fs/namei.c  2006-02-06 06:10:53.000000000 -0800
+++ fs/namei.c~fix-do-path-lookup       2006-02-06 11:33:59.000000000 -0800
@@ -1122,6 +1122,8 @@ static int fastcall do_path_lookup(int d
        read_unlock(&current->fs->lock);
        current->total_link_count = 0;
        retval = link_path_walk(name, nd);
+       if(retval)
+               goto out_fail;
 out:
        if (unlikely(current->audit_context
                     && nd && nd->dentry && nd->dentry->d_inode))



-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
[Index of Archives]     [Kernel Newbies]     [Netfilter]     [Bugtraq]     [Photo]     [Stuff]     [Gimp]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Video 4 Linux]     [Linux for the blind]     [Linux Resources]
  Powered by Linux