hi,
the following commit introduced a possible NULL pointer dereference
http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=ae5c79476f36512d1100e162606bb5691f2cce5a
we set cb->cb_client to NULL and pass it to rpc_shutdown_client() which dereferences it.
The easy fix below.
Signed-off-by: Eric Sesterhenn <[email protected]>
--- linux-2.6.18-git5/fs/nfsd/nfs4callback.c.orig 2006-09-26 12:24:23.000000000 +0200
+++ linux-2.6.18-git5/fs/nfsd/nfs4callback.c 2006-09-26 12:24:40.000000000 +0200
@@ -450,7 +450,8 @@ out_rpciod:
rpciod_down();
cb->cb_client = NULL;
out_clnt:
- rpc_shutdown_client(cb->cb_client);
+ if (cb->cb_client)
+ rpc_shutdown_client(cb->cb_client);
out_err:
dprintk("NFSD: warning: no callback path to client %.*s\n",
(int)clp->cl_name.len, clp->cl_name.data);
Eric Sesterhenn wrote:
> hi,
>
> the following commit introduced a possible NULL pointer dereference
> http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=ae5c79476f36512d1100e162606bb5691f2cce5a
>
> we set cb->cb_client to NULL and pass it to rpc_shutdown_client() which dereferences it.
What was I thinking/smoking? ;-)
> The easy fix below.
>
> Signed-off-by: Eric Sesterhenn <[email protected]>
>
> --- linux-2.6.18-git5/fs/nfsd/nfs4callback.c.orig 2006-09-26 12:24:23.000000000 +0200
> +++ linux-2.6.18-git5/fs/nfsd/nfs4callback.c 2006-09-26 12:24:40.000000000 +0200
> @@ -450,7 +450,8 @@ out_rpciod:
> rpciod_down();
> cb->cb_client = NULL;
> out_clnt:
> - rpc_shutdown_client(cb->cb_client);
> + if (cb->cb_client)
> + rpc_shutdown_client(cb->cb_client);
> out_err:
> dprintk("NFSD: warning: no callback path to client %.*s\n",
> (int)clp->cl_name.len, clp->cl_name.data);
Actually I see another problem here.
rpc_create() doesn't return NULL, it returns an ERR_PTR. So the return
code check in the code just before this is also bogus.
I'll send a complete fix through Trond.
On Tue, 26 Sep 2006 12:30:59 +0200, Eric Sesterhenn said:
> we set cb->cb_client to NULL and pass it to rpc_shutdown_client() which dereferences it.
> The easy fix below.
> out_clnt:
> - rpc_shutdown_client(cb->cb_client);
> + if (cb->cb_client)
> + rpc_shutdown_client(cb->cb_client);
OK, I admit not knowing the NFS code well, but this one looks suspiciously
like "easy paper-over" rather than "easy fix". Is there other cod elsewhere
that guards this case from ever actually happening? If it *does* happen,
is it indicative of major borkage and we really should do something like:
if (cb->cb_client)
rpm_shutdown_client(cb->cb_client)
else
printk(KERN_ERR "Yowza - trashed NFS control structures...");
or even maybe a more drastic action (oops/panic)?