mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Gene Heskett <gene.heskett@verizon.net>
To: Denis Vlasenko <vda@port.imtp.ilyichevsk.odessa.ua>
Cc: linux-kernel@vger.kernel.org
Subject: 2.6.7 and 2.6.8-rc2 have similar Oops
Date: Sun, 1 Aug 2004 13:56:43 -0400	[thread overview]
Message-ID: <200408011356.43317.gene.heskett@verizon.net> (raw)
In-Reply-To: <200408011858.13610.vda@port.imtp.ilyichevsk.odessa.ua>

This is a resend with a new subject, the first one apparently didn't 
get there.

On Sunday 01 August 2004 11:58, Denis Vlasenko wrote:
>On Sunday 01 August 2004 15:09, Gene Heskett wrote:
>> >But I digress. OOPS happened in dentry_iput(), which is
>> > (surprise!) an inline function:

This looks as if its as far as we can get Denis, so I'm
just going to post the whole thing to lkml, after snipping
the last Opps out of what is now last weeks log (its
sunday, and so far this log is clean) and inserting it also.

Many thanks for all your help, Denis.  I've learned a bit
about troubleshooting things like this.

To the lkml:  Currently running 2.6.7 because for 2.6.7, its not 
generally fatal.  With 2.8.6-rc2 its nearly 100% fatal.
This data was developed from the 2.6.7 srcs.

I can supply the dcache.txt objdumps, and the
dcache-w-nops.s files to whomever requests them so that they
can draw their own conclusions.  My uptimes here are generally
less than 24 hours because of this Oops.

>> >You need to do more of asm()ing.

>> Ok, I now have them attached.  I did add a pair of curlies to
>> contain another example of the H syndrome.
[...]
>asm("nop #O");
>                if (dentry->d_op && dentry->d_op->d_iput)
>                    ^^^^^^^^^^^^    ^^^^^^^^^^^^^^^^^^^^
>{
>asm("nop #P");
>                        dentry->d_op->d_iput(dentry, inode);
>                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>}
>                else
>{
>asm("nop #Q");
>                        iput(inode);
>asm("nop #R");
>
[... to the objdump]
>
>        nop #O
>        movl    68(%ebx), %eax
>        testl   %eax, %eax <--- this is dentry->d_op. %eax ==
> 00000e00 here. this is bad. pointer can't have such a value
>        je      .L532
>        movl    20(%eax), %edx   <=== it tries to fetch
>				dentry->d_op->d_iput
>				at 00000e20, which is bogus address. OOPS
>        testl   %edx, %edx
>        jne     .L594      <--*
>.L532:
>        nop #Q
>        movl    %esi, (%esp)
>        call    iput
>        nop #R
>.L565:
>        nop #E
>        movl    12(%ebx), %esi
>        nop #F
>        movl    %ebx, (%esp)
>        call    d_free
>        nop #G
>        cmpl    %ebx, %esi
>        je      .L566
>        movl    %esi, (%esp)
>        call    dput
>.L566:
>        nop #I
>        incl    20(%edi)
>        nop #J
>        jmp     .L445
>.L594:                     <--*
>        nop #P
>        movl    %ebx, (%esp)
>        movl    %esi, 4(%esp)
>        call    *20(%eax)   <---- calls dentry->d_op->d_iput(dentry,
> inode) (just checking that I indeed properly matched C to asm) jmp 
>    .L565
>.L593:
>
>Where that dentry came from?
>
>void prune_dcache(int count)
>{
>        for (; count ; count--) {
>                struct dentry *dentry;
>                struct list_head *tmp;
>                tmp = dentry_unused.prev;
>                if (tmp == &dentry_unused) break;
>                list_del_init(tmp);
>                dentry_stat.nr_unused--;
>                dentry = list_entry(tmp, struct dentry, d_lru);
>                /*
>                 * We found an inuse dentry which was not removed
> from * dentry_unused because of laziness during lookup.  Do not
> free * it - just keep it off the dentry_unused list. */
>                if (atomic_read(&dentry->d_count)) continue;
>                }
>                /* If the dentry was recently referenced, don't free
> it. */ if (dentry->d_flags & DCACHE_REFERENCED) {
>			...
>                        continue;
>                }
>                prune_one_dentry(dentry);
>		...
>
>prune_one_dentry(dentry) {
>	...
>        dentry_iput(dentry)
>	...
>}
>
>Heh. I don't know where to go now. At least we can post complete
>bug report on lkml.
>
>You can trace other oopses in a similar way. Maybe there is some
>patterm in them.

This 'pattern' has been pretty consistent, and even attacks kernels
as late as 2.6.8-rc2, with the difference being that from 2.6.7-mm1
on, the Oops is 99% fatal to the machine, as in a total lockup,
reset or power switch to recover.  With 2.6.7, the worst its done
(generally, there have been exceptions) is to screw up the reboot,
eventually needing the reset switch about 5% of the time to get
back to post & get rebooting under way.

Without further prattle, The Oops, from yesterday afternoon:
-------------
Jul 31 13:55:12 coyote kernel: Unable to handle kernel NULL pointer 
dereference at virtual address 00000e14
Jul 31 13:55:12 coyote kernel:  printing eip:
Jul 31 13:55:12 coyote kernel: c0162887
Jul 31 13:55:12 coyote kernel: *pde = 00000000
Jul 31 13:55:12 coyote kernel: Oops: 0000 [#1]
Jul 31 13:55:12 coyote kernel: PREEMPT
Jul 31 13:55:12 coyote kernel: Modules linked in: eeprom snd_seq_oss 
snd_seq_midi_event snd_seq snd_pcm_oss snd_mixer_oss snd
_bt87x snd_intel8x0 snd_ac97_codec snd_pcm snd_timer snd_page_alloc 
snd_mpu401_uart snd_rawmidi snd_seq_device snd forcedeth
sg
Jul 31 13:55:12 coyote kernel: CPU:    0
Jul 31 13:55:12 coyote kernel: EIP:    0060:[<c0162887>]    Not 
tainted
Jul 31 13:55:12 coyote kernel: EFLAGS: 00010206   (2.6.7-nf2)
Jul 31 13:55:12 coyote kernel: EIP is at prune_dcache+0x147/0x1c0
Jul 31 13:55:12 coyote kernel: eax: 00000e00   ebx: d1bde050   ecx: 
f1b3c050   edx: f1b3ac50
Jul 31 13:55:12 coyote kernel: esi: f1b3ac40   edi: c1973000   ebp: 
00000036   esp: c1973ef8
Jul 31 13:55:12 coyote kernel: ds: 007b   es: 007b   ss: 0068
Jul 31 13:55:12 coyote kernel: Process kswapd0 (pid: 65, 
threadinfo=c1973000 task=c1986050)
Jul 31 13:55:12 coyote kernel: Stack: d7721178 c1973ef8 0000007a 
00000000 c1973000 f7ffea48 c0162d1f 0000007a
Jul 31 13:55:12 coyote kernel:        c0139a2b 0000007a 000000d0 
00025528 049dbb00 00000000 000001fa 00000000
Jul 31 13:55:12 coyote kernel:        c0364564 00000001 0000000a 
c0364440 c013add1 00000080 000000d0 00000000
Jul 31 13:55:12 coyote kernel: Call Trace:
Jul 31 13:55:12 coyote kernel:  [<c0162d1f>] 
shrink_dcache_memory+0x1f/0x30
Jul 31 13:55:12 coyote kernel:  [<c0139a2b>] shrink_slab+0x14b/0x190
Jul 31 13:55:12 coyote kernel:  [<c013add1>] balance_pgdat+0x1b1/0x200
Jul 31 13:55:12 coyote kernel:  [<c013aee7>] kswapd+0xc7/0xe0
Jul 31 13:55:12 coyote kernel:  [<c0114270>] 
autoremove_wake_function+0x0/0x60
Jul 31 13:55:12 coyote kernel:  [<c0103e9e>] ret_from_fork+0x6/0x14
Jul 31 13:55:12 coyote kernel:  [<c0114270>] 
autoremove_wake_function+0x0/0x60
Jul 31 13:55:12 coyote kernel:  [<c013ae20>] kswapd+0x0/0xe0
Jul 31 13:55:12 coyote kernel:  [<c01021d1>] 
kernel_thread_helper+0x5/0x14
Jul 31 13:55:12 coyote kernel:
Jul 31 13:55:12 coyote kernel: Code: 8b 50 14 85 d2 75 27 89 34 24 e8 
4a 2b 00 00 8b 73 0c 89 1c
------------------------

-- 
Cheers, Gene
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
99.24% setiathome rank, not too shabby for a WV hillbilly
Yahoo.com attorneys please note, additions to this message
by Gene Heskett are:
Copyright 2004 by Maurice Eugene Heskett, all rights reserved.


      parent reply	other threads:[~2004-08-01 17:56 UTC|newest]

Thread overview: 21+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2004-07-25  1:56 2.6.8-rc2 crash(s)? Gene Heskett
     [not found] ` <200407250116.29651.vda@port.imtp.ilyichevsk.odessa.ua>
2004-07-25  4:12   ` Gene Heskett
2004-07-25  4:37     ` Gene Heskett
2004-07-29  0:12       ` William Lee Irwin III
2004-07-29  0:59         ` Gene Heskett
2004-07-29  1:03           ` William Lee Irwin III
2004-07-29  1:59             ` Gene Heskett
     [not found]     ` <200407250909.00227.vda@port.imtp.ilyichevsk.odessa.ua>
2004-07-25 13:43       ` Gene Heskett
2004-07-25 16:38         ` Kyle Moffett
2004-07-25 20:32           ` Gene Heskett
2004-07-29 22:22       ` Gene Heskett
2004-07-29 22:14         ` Randy.Dunlap
2004-07-30  1:47           ` Gene Heskett
2004-07-30  3:36             ` Randy.Dunlap
2004-07-30  4:50               ` Gene Heskett
2004-07-30  5:03                 ` Randy.Dunlap
2004-07-30  6:49                   ` Denis Vlasenko
2004-07-30 22:13                   ` Randy.Dunlap
2004-07-30  6:45                 ` Denis Vlasenko
     [not found] ` <200408010809.28311.gene.heskett@verizon.net>
     [not found]   ` <200408011858.13610.vda@port.imtp.ilyichevsk.odessa.ua>
2004-08-01 16:55     ` Gene Heskett
2004-08-01 17:56     ` Gene Heskett [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=200408011356.43317.gene.heskett@verizon.net \
    --to=gene.heskett@verizon.net \
    --cc=linux-kernel@vger.kernel.org \
    --cc=vda@port.imtp.ilyichevsk.odessa.ua \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®