Re: 2.6.23-rc1-mm1 -- INFO: possible recursive locking detected -- (&hashbin->hb_spinlock){....}, at: [<c03d95d3>] irias_seq_show+0xba/0x1a8

["Paul Menage" <menage@google.com>]

On 7/28/07, Andrew Morton <akpm@linux-foundation.org> wrote:
> >  [<c0154ad2>] proc_cpuset_show+0x5e/0xb9
> >  [<c01976b1>] seq_read+0xef/0x266
> >  [<c0181398>] vfs_read+0x8e/0x117
> >  [<c01817c9>] sys_read+0x3d/0x61
> >  [<c010401e>] sysenter_past_esp+0x5f/0x99
> >  =======================
> > INFO: lockdep is turned off.
> > Code: 00 89 d8 83 c4 0c 5b 5e 5f 5d c3 55 89 e5 57 56 53 83 ec 08 89
> > 45 f0 89 55 ec 89 d3 01 cb 8d 43 ff c6 43 ff 00 8b 55 f0 8b 72 1c <8b>
> > 56 3c 29 d0 3b 45 ec 72 45 89 d1 c1 e9 02 8b 76 40 89 c7 f3
> > EIP: [<c01536a7>] container_path+0x20/0x7b SS:ESP 0068:c5ccff00
>

Oops, this is a case of trying to read /proc/self/cpuset when cpusets
are completely unmounted. It works fine when cpusets is mounted as
part of any hierarchy, or part of an active but unmounted hierarchy.
Attached patch fixes it, and should be rolled in to the first patch of
the series. (basic framework patch).

Paul

---
 kernel/container.c |    9 +++++++++
 1 file changed, 9 insertions(+)

Index: container-2.6.22-rc6-mm1/kernel/container.c
===================================================================
--- container-2.6.22-rc6-mm1.orig/kernel/container.c
+++ container-2.6.22-rc6-mm1/kernel/container.c
@@ -684,6 +684,15 @@ int container_path(const struct containe
 {
 	char *start;
 
+	if (cont == dummytop) {
+		/*
+		 * Inactive subsystems have no dentry for their root
+		 * container
+		 */
+		strcpy(buf, "/");
+		return 0;
+	}
+
 	start = buf + buflen;
 
 	*--start = '\0';