* Oops in rpc_depopulate with 2.5.62
@ 2003-02-22 8:09 Martin J. Bligh
[not found] ` <20030222004930.0240738b.akpm@digeo.com>
0 siblings, 1 reply; 6+ messages in thread
From: Martin J. Bligh @ 2003-02-22 8:09 UTC (permalink / raw)
To: linux-kernel; +Cc: zwane
Getting rpc/nfs bugs on at least two different machines I've seen
and reports of a third from Zwane - all look similar. Anyone got
any bright ideas?
M.
Unable to handle kernel NULL pointer dereference at virtual address 0000006c
printing eip:
c03415e2
*pde = 358d2001
*pte = 00000000
Oops: 0002
CPU: 1
EIP: 0060:[<c03415e2>] Not tainted
EFLAGS: 00010206
EIP is at rpc_depopulate+0x22/0xf0
eax: 00000000 ebx: f5133680 ecx: 0000006c edx: f50afc1c
esi: f50afbf4 edi: f50afbf4 ebp: f50afc08 esp: f50afbec
ds: 007b es: 007b ss: 0068
Process mount (pid: 1166, threadinfo=f50ae000 task=f5766d80)
Stack: c01562f0 00000000 f50afbf4 f50afbf4 f5133680 f5133680 f51c1e00 f50afc40
c0341af5 f5133680 f5117880 f7ff9800 f50afcec 00000000 f50afc34 00000010
00000001 00000000 f5982680 f50afcec 00000000 f50afc50 c0333066 f5982714
Call Trace:
[<c01562f0>] lookup_hash+0x70/0xa0
[<c0341af5>] rpc_rmdir+0x55/0x90
[<c0333066>] rpc_destroy_client+0x46/0x70
[<c03330db>] rpc_release_client+0x4b/0x60
[<c0337bc7>] rpc_release_task+0x1a7/0x1d0
[<c033754b>] __rpc_execute+0x35b/0x370
[<c011a980>] default_wake_function+0x0/0x20
[<c0333274>] rpc_call_sync+0x64/0xa0
[<c0333287>] rpc_call_sync+0x77/0xa0
[<c03366b0>] rpc_run_timer+0x0/0xa0
[<c033e5fb>] rpc_register+0xcb/0x100
[<c0110000>] mask_and_ack_8259A+0x10/0xf0
[<c0339d24>] svc_register+0x94/0x100
[<c0339944>] svc_create+0xd4/0xe0
[<c01c1fb8>] lockd_up+0x58/0x110
[<c01a63f4>] nfs_fill_super+0x374/0x3a0
[<c01a7e90>] nfs_get_sb+0x1f0/0x230
[<c0150052>] do_kern_mount+0x42/0xa0
[<c01631e6>] do_add_mount+0x76/0x150
[<c0133596>] __alloc_pages+0x76/0x2c0
[<c01634d7>] do_mount+0x147/0x160
[<c0163908>] sys_mount+0xa8/0x110
[<c010ae7b>] syscall_call+0x7/0xb
Code: f0 ff 48 6c 0f 88 70 09 00 00 f0 fe 0d 00 d2 44 c0 0f 88 6d
Seems to be crashing on the deference of:
down(&dir->i_sem);
in rpc_depopulate. But there's a big comment just above saying:
-----------------------
/*
* FIXME: This probably has races.
*/
static void
rpc_depopulate(struct dentry *parent)
{
struct inode *dir = parent->d_inode;
LIST_HEAD(head);
struct list_head *pos, *next;
struct dentry *dentry;
down(&dir->i_sem);
------------------
Looks like dir is NULL here ... might be dcache_rcu (a recent dcache_rcu
patch fixed the same file) or a race brought out by the changes.
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Oops in rpc_depopulate with 2.5.62
[not found] ` <20030222004930.0240738b.akpm@digeo.com>
@ 2003-02-24 12:14 ` Maneesh Soni
2003-02-26 16:36 ` Martin J. Bligh
0 siblings, 1 reply; 6+ messages in thread
From: Maneesh Soni @ 2003-02-24 12:14 UTC (permalink / raw)
To: Andrew Morton; +Cc: Martin J. Bligh, Dipankar Sarma, trond.myklebust, LKML
On Sat, Feb 22, 2003 at 12:49:30AM -0800, Andrew Morton wrote:
>
> Maneesh and Dipankar may be able to plug this one. I've looked,
> and apart from a seemingly-unneeded test for !d_unhashed I can't
> see the problem.
>
> Perhaps the problem lies with a different part of the code.
>
> I cannot reproduce the crash.
Same with me also. Surprisingly I am getting same oops on 2.5.61.
Can send .config and ksymoops output if required.
rpc_rmdir does lookup_hash(), which may return a negative dentry.
If a negative dentry is possible at this point then testing for d_inode
in rpc_rmdir() fixes this problem. Like we have in rpc_unlink().
Probably Trond knows better.
Following patch fixes the oops I am seeing in 2.5.61 and should fix
the oops seen on 2.5.62 also.
diff -urN linux-2.5.62-bk6/net/sunrpc/rpc_pipe.c linux-2.5.62-bk6-rpc_depopulate/net/sunrpc/rpc_pipe.c
--- linux-2.5.62-bk6/net/sunrpc/rpc_pipe.c 2003-02-24 16:06:46.000000000 +0530
+++ linux-2.5.62-bk6-rpc_depopulate/net/sunrpc/rpc_pipe.c 2003-02-24 16:57:14.000000000 +0530
@@ -665,8 +665,10 @@
error = PTR_ERR(dentry);
goto out_release;
}
- rpc_depopulate(dentry);
- error = __rpc_rmdir(dir, dentry);
+ if (dentry->d_inode) {
+ rpc_depopulate(dentry);
+ error = __rpc_rmdir(dir, dentry);
+ }
dput(dentry);
out_release:
up(&dir->i_sem);
Regards
Maneesh
> "Martin J. Bligh" <mbligh@aracnet.com> wrote:
> >
> > Getting rpc/nfs bugs on at least two different machines I've seen
> > and reports of a third from Zwane - all look similar. Anyone got
> > any bright ideas?
> >
> > M.
> >
> > Unable to handle kernel NULL pointer dereference at virtual address 0000006c
> > printing eip:
> > c03415e2
> > *pde = 358d2001
> > *pte = 00000000
> > Oops: 0002
> > CPU: 1
> > EIP: 0060:[<c03415e2>] Not tainted
> > EFLAGS: 00010206
> > EIP is at rpc_depopulate+0x22/0xf0
> > eax: 00000000 ebx: f5133680 ecx: 0000006c edx: f50afc1c
> > esi: f50afbf4 edi: f50afbf4 ebp: f50afc08 esp: f50afbec
> > ds: 007b es: 007b ss: 0068
> > Process mount (pid: 1166, threadinfo=f50ae000 task=f5766d80)
> > Stack: c01562f0 00000000 f50afbf4 f50afbf4 f5133680 f5133680 f51c1e00 f50afc40
> > c0341af5 f5133680 f5117880 f7ff9800 f50afcec 00000000 f50afc34 00000010
> > 00000001 00000000 f5982680 f50afcec 00000000 f50afc50 c0333066 f5982714
> > Call Trace:
> > [<c01562f0>] lookup_hash+0x70/0xa0
> > [<c0341af5>] rpc_rmdir+0x55/0x90
> > [<c0333066>] rpc_destroy_client+0x46/0x70
> > [<c03330db>] rpc_release_client+0x4b/0x60
> > [<c0337bc7>] rpc_release_task+0x1a7/0x1d0
> > [<c033754b>] __rpc_execute+0x35b/0x370
> > [<c011a980>] default_wake_function+0x0/0x20
> > [<c0333274>] rpc_call_sync+0x64/0xa0
> > [<c0333287>] rpc_call_sync+0x77/0xa0
> > [<c03366b0>] rpc_run_timer+0x0/0xa0
> > [<c033e5fb>] rpc_register+0xcb/0x100
> > [<c0110000>] mask_and_ack_8259A+0x10/0xf0
> > [<c0339d24>] svc_register+0x94/0x100
> > [<c0339944>] svc_create+0xd4/0xe0
> > [<c01c1fb8>] lockd_up+0x58/0x110
> > [<c01a63f4>] nfs_fill_super+0x374/0x3a0
> > [<c01a7e90>] nfs_get_sb+0x1f0/0x230
> > [<c0150052>] do_kern_mount+0x42/0xa0
> > [<c01631e6>] do_add_mount+0x76/0x150
> > [<c0133596>] __alloc_pages+0x76/0x2c0
> > [<c01634d7>] do_mount+0x147/0x160
> > [<c0163908>] sys_mount+0xa8/0x110
> > [<c010ae7b>] syscall_call+0x7/0xb
> >
> > Code: f0 ff 48 6c 0f 88 70 09 00 00 f0 fe 0d 00 d2 44 c0 0f 88 6d
> >
> > Seems to be crashing on the deference of:
> >
> > down(&dir->i_sem);
> >
> > in rpc_depopulate. But there's a big comment just above saying:
> >
> > -----------------------
> > /*
> > * FIXME: This probably has races.
> > */
> > static void
> > rpc_depopulate(struct dentry *parent)
> > {
> > struct inode *dir = parent->d_inode;
> > LIST_HEAD(head);
> > struct list_head *pos, *next;
> > struct dentry *dentry;
> >
> > down(&dir->i_sem);
> >
> > ------------------
> >
> > Looks like dir is NULL here ... might be dcache_rcu (a recent dcache_rcu
> > patch fixed the same file) or a race brought out by the changes.
> >
> >
> > -
> > To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> > the body of a message to majordomo@vger.kernel.org
> > More majordomo info at http://vger.kernel.org/majordomo-info.html
> > Please read the FAQ at http://www.tux.org/lkml/
--
Maneesh Soni
IBM Linux Technology Center,
IBM India Software Lab, Bangalore.
Phone: +91-80-5044999 email: maneesh@in.ibm.com
http://lse.sourceforge.net/
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Oops in rpc_depopulate with 2.5.62
2003-02-24 12:14 ` Maneesh Soni
@ 2003-02-26 16:36 ` Martin J. Bligh
2003-02-26 22:54 ` Bill Davidsen
2003-02-28 5:07 ` Maneesh Soni
0 siblings, 2 replies; 6+ messages in thread
From: Martin J. Bligh @ 2003-02-26 16:36 UTC (permalink / raw)
To: maneesh, Andrew Morton; +Cc: Dipankar Sarma, trond.myklebust, LKML
This seems to fix the problem. No idea what it's doing, but it works ;-)
M.
> On Sat, Feb 22, 2003 at 12:49:30AM -0800, Andrew Morton wrote:
>>
>> Maneesh and Dipankar may be able to plug this one. I've looked,
>> and apart from a seemingly-unneeded test for !d_unhashed I can't
>> see the problem.
>>
>> Perhaps the problem lies with a different part of the code.
>>
>> I cannot reproduce the crash.
>
> Same with me also. Surprisingly I am getting same oops on 2.5.61.
> Can send .config and ksymoops output if required.
>
> rpc_rmdir does lookup_hash(), which may return a negative dentry.
> If a negative dentry is possible at this point then testing for d_inode
> in rpc_rmdir() fixes this problem. Like we have in rpc_unlink().
> Probably Trond knows better.
>
> Following patch fixes the oops I am seeing in 2.5.61 and should fix
> the oops seen on 2.5.62 also.
>
>
> diff -urN linux-2.5.62-bk6/net/sunrpc/rpc_pipe.c
> linux-2.5.62-bk6-rpc_depopulate/net/sunrpc/rpc_pipe.c ---
> linux-2.5.62-bk6/net/sunrpc/rpc_pipe.c 2003-02-24 16:06:46.000000000 +0530
> +++ linux-2.5.62-bk6-rpc_depopulate/net/sunrpc/rpc_pipe.c 2003-02-24
> 16:57:14.000000000 +0530 @@ -665,8 +665,10 @@
> error = PTR_ERR(dentry);
> goto out_release;
> }
> - rpc_depopulate(dentry);
> - error = __rpc_rmdir(dir, dentry);
> + if (dentry->d_inode) {
> + rpc_depopulate(dentry);
> + error = __rpc_rmdir(dir, dentry);
> + }
> dput(dentry);
> out_release:
> up(&dir->i_sem);
>
>
>
> Regards
> Maneesh
>
>> "Martin J. Bligh" <mbligh@aracnet.com> wrote:
>> >
>> > Getting rpc/nfs bugs on at least two different machines I've seen
>> > and reports of a third from Zwane - all look similar. Anyone got
>> > any bright ideas?
>> >
>> > M.
>> >
>> > Unable to handle kernel NULL pointer dereference at virtual address
>> > 0000006c printing eip:
>> > c03415e2
>> > *pde = 358d2001
>> > *pte = 00000000
>> > Oops: 0002
>> > CPU: 1
>> > EIP: 0060:[<c03415e2>] Not tainted
>> > EFLAGS: 00010206
>> > EIP is at rpc_depopulate+0x22/0xf0
>> > eax: 00000000 ebx: f5133680 ecx: 0000006c edx: f50afc1c
>> > esi: f50afbf4 edi: f50afbf4 ebp: f50afc08 esp: f50afbec
>> > ds: 007b es: 007b ss: 0068
>> > Process mount (pid: 1166, threadinfo=f50ae000 task=f5766d80)
>> > Stack: c01562f0 00000000 f50afbf4 f50afbf4 f5133680 f5133680 f51c1e00
>> > f50afc40 c0341af5 f5133680 f5117880 f7ff9800 f50afcec 00000000
>> > f50afc34 00000010 00000001 00000000 f5982680 f50afcec 00000000
>> > f50afc50 c0333066 f5982714 Call Trace:
>> > [<c01562f0>] lookup_hash+0x70/0xa0
>> > [<c0341af5>] rpc_rmdir+0x55/0x90
>> > [<c0333066>] rpc_destroy_client+0x46/0x70
>> > [<c03330db>] rpc_release_client+0x4b/0x60
>> > [<c0337bc7>] rpc_release_task+0x1a7/0x1d0
>> > [<c033754b>] __rpc_execute+0x35b/0x370
>> > [<c011a980>] default_wake_function+0x0/0x20
>> > [<c0333274>] rpc_call_sync+0x64/0xa0
>> > [<c0333287>] rpc_call_sync+0x77/0xa0
>> > [<c03366b0>] rpc_run_timer+0x0/0xa0
>> > [<c033e5fb>] rpc_register+0xcb/0x100
>> > [<c0110000>] mask_and_ack_8259A+0x10/0xf0
>> > [<c0339d24>] svc_register+0x94/0x100
>> > [<c0339944>] svc_create+0xd4/0xe0
>> > [<c01c1fb8>] lockd_up+0x58/0x110
>> > [<c01a63f4>] nfs_fill_super+0x374/0x3a0
>> > [<c01a7e90>] nfs_get_sb+0x1f0/0x230
>> > [<c0150052>] do_kern_mount+0x42/0xa0
>> > [<c01631e6>] do_add_mount+0x76/0x150
>> > [<c0133596>] __alloc_pages+0x76/0x2c0
>> > [<c01634d7>] do_mount+0x147/0x160
>> > [<c0163908>] sys_mount+0xa8/0x110
>> > [<c010ae7b>] syscall_call+0x7/0xb
>> >
>> > Code: f0 ff 48 6c 0f 88 70 09 00 00 f0 fe 0d 00 d2 44 c0 0f 88 6d
>> >
>> > Seems to be crashing on the deference of:
>> >
>> > down(&dir->i_sem);
>> >
>> > in rpc_depopulate. But there's a big comment just above saying:
>> >
>> > -----------------------
>> > /*
>> > * FIXME: This probably has races.
>> > */
>> > static void
>> > rpc_depopulate(struct dentry *parent)
>> > {
>> > struct inode *dir = parent->d_inode;
>> > LIST_HEAD(head);
>> > struct list_head *pos, *next;
>> > struct dentry *dentry;
>> >
>> > down(&dir->i_sem);
>> >
>> > ------------------
>> >
>> > Looks like dir is NULL here ... might be dcache_rcu (a recent
>> > dcache_rcu patch fixed the same file) or a race brought out by the
>> > changes.
>> >
>> >
>> > -
>> > To unsubscribe from this list: send the line "unsubscribe
>> > linux-kernel" in the body of a message to majordomo@vger.kernel.org
>> > More majordomo info at http://vger.kernel.org/majordomo-info.html
>> > Please read the FAQ at http://www.tux.org/lkml/
>
> --
> Maneesh Soni
> IBM Linux Technology Center,
> IBM India Software Lab, Bangalore.
> Phone: +91-80-5044999 email: maneesh@in.ibm.com
> http://lse.sourceforge.net/
>
>
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Oops in rpc_depopulate with 2.5.62
2003-02-26 16:36 ` Martin J. Bligh
@ 2003-02-26 22:54 ` Bill Davidsen
2003-02-28 5:07 ` Maneesh Soni
1 sibling, 0 replies; 6+ messages in thread
From: Bill Davidsen @ 2003-02-26 22:54 UTC (permalink / raw)
To: Martin J. Bligh; +Cc: LKML
On Wed, 26 Feb 2003, Martin J. Bligh wrote:
>
> This seems to fix the problem. No idea what it's doing, but it works ;-)
In this case the important thing seems to be what it's *not* doing.
--
bill davidsen <davidsen@tmr.com>
CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Oops in rpc_depopulate with 2.5.62
2003-02-26 16:36 ` Martin J. Bligh
2003-02-26 22:54 ` Bill Davidsen
@ 2003-02-28 5:07 ` Maneesh Soni
2003-02-28 5:33 ` Martin J. Bligh
1 sibling, 1 reply; 6+ messages in thread
From: Maneesh Soni @ 2003-02-28 5:07 UTC (permalink / raw)
To: Martin J. Bligh; +Cc: Andrew Morton, Dipankar Sarma, trond.myklebust, LKML
On Wed, Feb 26, 2003 at 08:36:29AM -0800, Martin J. Bligh wrote:
>
> This seems to fix the problem. No idea what it's doing, but it works ;-)
>
> M.
Hi Martin,
Trond has the correct fix for this. The patch is as follows:
--- linux-2.5.61-up/net/sunrpc/clnt.c.orig 2003-02-15 21:05:02.000000000
++0100
+++ linux-2.5.61-up/net/sunrpc/clnt.c 2003-02-17 19:39:20.000000000 +0100
@@ -208,7 +208,8 @@
rpcauth_destroy(clnt->cl_auth);
clnt->cl_auth = NULL;
}
- rpc_rmdir(clnt->cl_pathname);
+ if (clnt->cl_pathname[0])
+ rpc_rmdir(clnt->cl_pathname);
if (clnt->cl_xprt) {
xprt_destroy(clnt->cl_xprt);
clnt->cl_xprt = NULL;
Regards,
Maneesh
--
Maneesh Soni
IBM Linux Technology Center,
IBM India Software Lab, Bangalore.
Phone: +91-80-5044999 email: maneesh@in.ibm.com
http://lse.sourceforge.net/
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Oops in rpc_depopulate with 2.5.62
2003-02-28 5:07 ` Maneesh Soni
@ 2003-02-28 5:33 ` Martin J. Bligh
0 siblings, 0 replies; 6+ messages in thread
From: Martin J. Bligh @ 2003-02-28 5:33 UTC (permalink / raw)
To: maneesh; +Cc: Andrew Morton, Dipankar Sarma, trond.myklebust, LKML
> Hi Martin,
>
> Trond has the correct fix for this. The patch is as follows:
>
> --- linux-2.5.61-up/net/sunrpc/clnt.c.orig 2003-02-15
> 21:05:02.000000000 ++0100
> +++ linux-2.5.61-up/net/sunrpc/clnt.c 2003-02-17 19:39:20.000000000
> +0100 @@ -208,7 +208,8 @@
> rpcauth_destroy(clnt->cl_auth);
> clnt->cl_auth = NULL;
> }
> - rpc_rmdir(clnt->cl_pathname);
> + if (clnt->cl_pathname[0])
> + rpc_rmdir(clnt->cl_pathname);
> if (clnt->cl_xprt) {
> xprt_destroy(clnt->cl_xprt);
> clnt->cl_xprt = NULL;
Cool - I picked this up from LKML, it's in 63-mjb1 ...
Thanks to everyone for the help with this one.
M.
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2003-02-28 5:23 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2003-02-22 8:09 Oops in rpc_depopulate with 2.5.62 Martin J. Bligh
[not found] ` <20030222004930.0240738b.akpm@digeo.com>
2003-02-24 12:14 ` Maneesh Soni
2003-02-26 16:36 ` Martin J. Bligh
2003-02-26 22:54 ` Bill Davidsen
2003-02-28 5:07 ` Maneesh Soni
2003-02-28 5:33 ` Martin J. Bligh
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.