Re: [PATCH] Revert: audit: Fix check of return value of strnlen_user()

[Jan Kara <jack@suse.cz>]

On Wed 08-07-15 11:26:07, Steven Rostedt wrote:
> When testing 4.2-rc1 I hit this WARNING:
> 
> ------------[ cut here ]------------
> WARNING: CPU: 1 PID: 1639 at /work/autotest/nobackup/linux-test.git/kernel/auditsc.c:1025 audit_log_exit+0x734/0xb0d()
> Modules linked in:
> CPU: 1 PID: 1639 Comm: fstab-decode Not tainted 4.2.0-rc1-test+ #2
> Hardware name: MSI MS-7823/CSM-H87M-G43 (MS-7823), BIOS V1.6 02/22/2014
>  00000000 00000000 f0817eb4 c0cd118d c10224d4 f0817ee4 c0440fbe c1011b30
>  00000001 00000667 c10224d4 00000401 c04ba081 c04ba081 f18e3c00 f1cd5080^M
>  00000000 f0817ef4 c0440ff7 00000009 00000000 f0817f84 c04ba081 f0817f60
> Call Trace:
>  [<c0cd118d>] dump_stack+0x41/0x52
>  [<c0440fbe>] warn_slowpath_common+0x9d/0xb4
>  [<c04ba081>] ? audit_log_exit+0x734/0xb0d
>  [<c04ba081>] ? audit_log_exit+0x734/0xb0d
>  [<c0440ff7>] warn_slowpath_null+0x22/0x24
>  [<c04ba081>] audit_log_exit+0x734/0xb0d
>  [<c04bb7f4>] __audit_syscall_exit+0x43/0xf6
>  [<c040ce27>] syscall_trace_leave+0x30/0xe5
>  [<c0cdbced>] syscall_exit_work+0x19/0x1e
> ---[ end trace 156b2a7afa592deb ]---
> 
> Debugging it, I found that it was triggered by this commit:
> 
> commit 0b08c5e5944 ("audit: Fix check of return value of strnlen_user()")
> 
> Yes, strnlen_user() returns 0 on fault, but if you look at what len is
> set to, than you would notice that on fault len would be -1.

You're right. Thanks for catching this. I don't know what I was thinking
when writing that "fix"...

								Honza

>  len = strnlen_user(p, MAX_ARG_STRLEN) - 1;
> 
> Now the warning triggers on a string of size zero ("\0"), which is a
> legitimate entry.
> 
> Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
> ---
> diff --git a/kernel/auditsc.c b/kernel/auditsc.c
> index 09c65640cad6..ee097948b0a8 100644
> --- a/kernel/auditsc.c
> +++ b/kernel/auditsc.c
> @@ -1021,7 +1021,7 @@ static int audit_log_single_execve_arg(struct audit_context *context,
>  	 * for strings that are too long, we should not have created
>  	 * any.
>  	 */
> -	if (unlikely((len == 0) || len > MAX_ARG_STRLEN - 1)) {
> +	if (unlikely((len == -1) || len > MAX_ARG_STRLEN - 1)) {
>  		WARN_ON(1);
>  		send_sig(SIGKILL, current, 0);
>  		return -1;
-- 
Jan Kara <jack@suse.cz>
SUSE Labs, CR