Skip to contentSkip to conversation

paging request BUG in 2.6.19-rc5 on resume - X60s

DISCUSSION Martin Lorenz 5 participants 9 messages Nov 13, 2006 – Nov 15, 2006

lore
Share & tools
lore
#0 lore raw

Hallo again,

here is another one:

I reported a black screen on resume with my latest kernel build earlyer. But this was not reproducible. Only occured once.

BUT I suspended with the ipw3945 module loaded once again now and got a BUG report in the log instead of a black screen.

I only see this when ipw3945 is loaded.

[226156.057000] BUG: unable to handle kernel paging request at virtual
address 756e6567
[226156.057000]  printing eip:
[226156.057000] c016ffb7
[226156.057000] *pde = 00000000
[226156.057000] Oops: 0000 [#1]
[226156.057000] SMP
[226156.057000] Modules linked in: tun ipw3945 ieee80211 ieee80211_crypt

nls_iso8859_1 nls_cp437 vfat fat usb_storage snd_hda_intel snd_hda_codec snd_pcm_oss snd_mixer_oss snd_pcm snd_timer snd soundcore snd_page_alloc vmnet(P) vmmon(P) i915 binfmt_misc nfs nfsd exportfs lockd nfs_acl sunrpc cpufreq_ondemand container video thermal i2c_ec fan dock button battery ac mmc_block speedstep_centrino freq_table processor ibm_acpi sbp2 nvram eth1394 irtty_sir sir_dev pcmcia ehci_hcd uhci_hcd firmware_class nsc_ircc generic usbcore psmouse irda ohci1394 ieee1394 sdhci ide_core yenta_socket rsrc_nonstatic pcmcia_core serio_raw crc_ccitt pcspkr mmc_core evdev

[226156.058000] CPU:    1
[226156.058000] EIP:    0060:[<c016ffb7>]    Tainted: P      VLI
[226156.058000] EFLAGS: 00010282
(2.6.19-rc5+ieee80211+e1000-45.3+1909-g6a4abeae-dirty #1)
[226156.058000] EIP is at iput+0xd/0x66
[226156.058000] eax: 756e6547   ebx: c0416e10   ecx: c016ee14   edx:
c55c7114
[226156.058000] esi: c046f1c0   edi: c046f21c   ebp: f7feb800   esp:
dcfbfde4
[226156.058000] ds: 007b   es: 007b   ss: 0068
[226156.058000] Process mount (pid: 22076, ti=dcfbe000 task=f7df1550
task.ti=dcfbe000)
[226156.058000] Stack: c046f21c c016ef85 c046f244 c046f1c0 c016f2e0 fffffff3
00000000 f7feb800
[226156.058000]        c8b73000 c01619bb 00000000 f7feb83c 00000000 f7feb800
00000000 c0172f49
[226156.058000]        00000000 c8b73000 00000000 e613a000 dcfb0000 00000444
00000020 0cf68720
[226156.058000] Call Trace:
[226156.058000]  [<c016ef85>] prune_one_dentry+0x53/0x74
[226156.058000]  [<c016f2e0>] shrink_dcache_sb+0x8f/0xb3
[226156.058000]  [<c01619bb>] do_remount_sb+0x40/0x120
[226156.058000]  [<c0172f49>] do_mount+0x1b0/0x66c
[226156.058000]  [<c017347c>] sys_mount+0x77/0xb3
[226156.058000]  [<c0102dc7>] syscall_call+0x7/0xb
[226156.058000] DWARF2 unwinder stuck at syscall_call+0x7/0xb
[226156.058000]
[226156.058000] Leftover inexact backtrace:
[226156.058000]
[226156.058000]  =======================
[226156.058000] Code: ba 03 00 00 00 e9 ee fc fb ff 83 a0 2c 01 00 00 b7 e9
e0 ff ff ff e8 d1 3e 17 00 31 c0 c3 53 89 c3 85 c0 74 5d 8b 80 98 00 00 00
<8b> 40 20 83 bb 2c 01 00 00 20 75 08 0f 0b 5d 04 dc 61 30 c0 85
[226156.058000] EIP: [<c016ffb7>] iput+0xd/0x66 SS:ESP 0068:dcfbfde4
[226156.058000]  <7>bridge-eth2: disabling the bridge
[226206.083000] bridge-eth2: down
[226206.190000] ACPI: PCI interrupt for device 0000:03:00.0 disabled
[226206.258000] ieee80211_crypt: unregistered algorithm 'NULL'

dmesg output and log is at http://www.lorenz.eu.org/~mlo/kernel/

http://www.lorenz.eu.org/~mlo/kernel/dmesg-2.6.19-rc5+ieee80211+e1000-45.3+1909-g6a4abeae-dirty-resume.out

http://www.lorenz.eu.org/~mlo/kernel/messages-2.6.19-rc5+ieee80211+e1000-45.3+1909-g6a4abeae-dirty-resume
this one includes a SysRq-t output

gruss
mlo
--
Dipl.-Ing. Martin Lorenz

They that can give up essential liberty to obtain a little temporary safety deserve neither liberty nor safety.
Benjamin Franklin

please encrypt your mail to me
GnuPG key-ID: F1AAD37D
get it here:
http://blackhole.pca.dfn.de:11371/pks/lookup?op=get&search=0xF1AAD37D

ICQ UIN: 33588107

#1 lore raw
Quoted 11 lines · Martin Lorenz

Hallo again,

here is another one:

I reported a black screen on resume with my latest kernel build earlyer. But this was not reproducible. Only occured once.

BUT I suspended with the ipw3945 module loaded once again now and got a BUG report in the log instead of a black screen.

I only see this when ipw3945 is loaded.

Interesting oops... another one trying to dereference "Genu".

Repeatable? Repeatable without vmware modules ever having been loaded?

Quoted 47 lines
[226156.057000] BUG: unable to handle kernel paging request at virtual
address 756e6567
[226156.057000]  printing eip:
[226156.057000] c016ffb7
[226156.057000] *pde = 00000000
[226156.057000] Oops: 0000 [#1]
[226156.057000] SMP
[226156.057000] Modules linked in: tun ipw3945 ieee80211 ieee80211_crypt

nls_iso8859_1 nls_cp437 vfat fat usb_storage snd_hda_intel snd_hda_codec snd_pcm_oss snd_mixer_oss snd_pcm snd_timer snd soundcore snd_page_alloc vmnet(P) vmmon(P) i915 binfmt_misc nfs nfsd exportfs lockd nfs_acl sunrpc cpufreq_ondemand container video thermal i2c_ec fan dock button battery ac mmc_block speedstep_centrino freq_table processor ibm_acpi sbp2 nvram eth1394 irtty_sir sir_dev pcmcia ehci_hcd uhci_hcd firmware_class nsc_ircc generic usbcore psmouse irda ohci1394 ieee1394 sdhci ide_core yenta_socket rsrc_nonstatic pcmcia_core serio_raw crc_ccitt pcspkr mmc_core evdev

[226156.058000] CPU:    1
[226156.058000] EIP:    0060:[<c016ffb7>]    Tainted: P      VLI
[226156.058000] EFLAGS: 00010282
(2.6.19-rc5+ieee80211+e1000-45.3+1909-g6a4abeae-dirty #1)
[226156.058000] EIP is at iput+0xd/0x66
[226156.058000] eax: 756e6547   ebx: c0416e10   ecx: c016ee14   edx:
c55c7114
[226156.058000] esi: c046f1c0   edi: c046f21c   ebp: f7feb800   esp:
dcfbfde4
[226156.058000] ds: 007b   es: 007b   ss: 0068
[226156.058000] Process mount (pid: 22076, ti=dcfbe000 task=f7df1550
task.ti=dcfbe000)
[226156.058000] Stack: c046f21c c016ef85 c046f244 c046f1c0 c016f2e0 fffffff3
00000000 f7feb800
[226156.058000]        c8b73000 c01619bb 00000000 f7feb83c 00000000 f7feb800
00000000 c0172f49
[226156.058000]        00000000 c8b73000 00000000 e613a000 dcfb0000 00000444
00000020 0cf68720
[226156.058000] Call Trace:
[226156.058000]  [<c016ef85>] prune_one_dentry+0x53/0x74
[226156.058000]  [<c016f2e0>] shrink_dcache_sb+0x8f/0xb3
[226156.058000]  [<c01619bb>] do_remount_sb+0x40/0x120
[226156.058000]  [<c0172f49>] do_mount+0x1b0/0x66c
[226156.058000]  [<c017347c>] sys_mount+0x77/0xb3
[226156.058000]  [<c0102dc7>] syscall_call+0x7/0xb
[226156.058000] DWARF2 unwinder stuck at syscall_call+0x7/0xb
[226156.058000]
[226156.058000] Leftover inexact backtrace:
[226156.058000]
[226156.058000]  =======================
[226156.058000] Code: ba 03 00 00 00 e9 ee fc fb ff 83 a0 2c 01 00 00 b7 e9
Quoted 2 lines

e0 ff ff ff e8 d1 3e 17 00 31 c0 c3 53 89 c3 85 c0 74 5d 8b 80 98 00 00 00 <8b> 40 20 83 bb 2c 01 00 00 20 75 08 0f 0b 5d 04 dc 61 30 c0 85

Per ksymoops, that code is:

   0:   ba 03 00 00 00            mov    $0x3,%edx
   5:   e9 ee fc fb ff            jmp    fffbfcf8 <_EIP+0xfffbfcf8>
   a:   83 a0 2c 01 00 00 b7      andl   $0xffffffb7,0x12c(%eax)
  11:   e9 00 00 00 00            jmp    16 <_EIP+0x16>

There is no such andl with an offset of 0x12c and that mask (I_LOCK| I_NEW?) anywhere in my kernel or modules. How about yours?

-Mike

#2 lore raw
Quoted 9 lines · Mike Galbraith

Per ksymoops, that code is:

   0:   ba 03 00 00 00            mov    $0x3,%edx
   5:   e9 ee fc fb ff            jmp    fffbfcf8 <_EIP+0xfffbfcf8>
   a:   83 a0 2c 01 00 00 b7      andl   $0xffffffb7,0x12c(%eax)
  11:   e9 00 00 00 00            jmp    16 <_EIP+0x16>

There is no such andl with an offset of 0x12c and that mask (I_LOCK| I_NEW?) anywhere in my kernel or modules. How about yours?

$ objdump -D vmlinux | grep -5 'andl   $0xffffffb7,0x12c'
c016ff87:       05 2c 01 00 00          add    $0x12c,%eax
c016ff8c:       ba 03 00 00 00          mov    $0x3,%edx
c016ff91:       e9 ee fc fb ff          jmp    c012fc84 <wake_up_bit>

c016ff96 <unlock_new_inode>:

c016ff96:       83 a0 2c 01 00 00 b7    andl   $0xffffffb7,0x12c(%eax)
c016ff9d:       e9 e0 ff ff ff          jmp    c016ff82 <wake_up_inode>

c016ffa2 <inode_wait>:

c016ffa2:       e8 d1 3e 17 00          call   c02e3e78 <schedule>
c016ffa7:       31 c0                   xor    %eax,%eax

gruss
mlo
--
Dipl.-Ing. Martin Lorenz

They that can give up essential liberty to obtain a little temporary safety deserve neither liberty nor safety.
Benjamin Franklin

please encrypt your mail to me
GnuPG key-ID: F1AAD37D
get it here:
http://blackhole.pca.dfn.de:11371/pks/lookup?op=get&search=0xF1AAD37D

ICQ UIN: 33588107

#3 lore raw
Quoted 3 lines · Martin Lorenz

c016ff96 <unlock_new_inode>:

c016ff96:       83 a0 2c 01 00 00 b7    andl   $0xffffffb7,0x12c(%eax)
c016ff9d:       e9 e0 ff ff ff          jmp    c016ff82 <wake_up_inode>

Ok, that's what I figured it had to be with that mask (though I can't convince either of my compilers to produce that offset), so now we just have to figure out how the heck it can get there and find a corrupted pointer.

Can you enable frame-pointers, and capture another explosion? A more complete trace might help. It would definitely help to reproduce without the proprietary modules having ever been loaded.

-Mike

#4 lore raw
In reply to Martin Lorenz
Quoted 7 lines · Martin Lorenz

here is another one:

I reported a black screen on resume with my latest kernel build earlyer. But this was not reproducible. Only occured once.

BUT I suspended with the ipw3945 module loaded once again now and got a BUG report in the log instead of a black screen.

I get nearly the same oopses on my thinkpad t60, too.
Always only after resuming (never after a clean reboot), and after the (otherwise successfull) resume it can take hours until the oops shows.

Did not report this problem anywhere yet, because I am using a heavily modified 2.6.17 based on the ubuntu edgy tree plus lots of addon patches (suspend2, linux-vserver, loop-aes) and most of the time with the evil fglrx loaded, too (to get any x11 outputs at all).

Quoted 1 line

I only see this when ipw3945 is loaded.

Will try to shutdown wireless and unload the module before the next suspend, and see if it helps.

c'ya
sven

Signature · 2 lines
The Internet treats censorship as a routing problem, and routes around it.
(John Gilmore on http://www.cygnus.com/~gnu/)
#5 lore raw
In reply to Martin Lorenz
Quoted 6 lines · Martin Lorenz

Hallo again,

here is another one:

I reported a black screen on resume with my latest kernel build earlyer. But this was not reproducible. Only occured once.

Is this a resume from disk? If so, which kernel are you using?

Rafael

Signature · 2 lines
You never change things by fighting the existing reality.
		R. Buckminster Fuller
#6 lore raw
Quoted 10 lines · Rafael J. Wysocki
Quoted 6 lines · Martin Lorenz

Hallo again,

here is another one:

I reported a black screen on resume with my latest kernel build earlyer. But this was not reproducible. Only occured once.

Is this a resume from disk? If so, which kernel are you using?

no from suspend to ram

Linux gimli 2.6.19-rc5+ieee80211+e1000-45.3+1909-g6a4abeae-dirty #1 SMP Wed Nov 8 20:14:31 CET 2006 i686 GNU/Linux

gruss
mlo
--
Dipl.-Ing. Martin Lorenz

They that can give up essential liberty to obtain a little temporary safety deserve neither liberty nor safety.
Benjamin Franklin

please encrypt your mail to me
GnuPG key-ID: F1AAD37D
get it here:
http://blackhole.pca.dfn.de:11371/pks/lookup?op=get&search=0xF1AAD37D

ICQ UIN: 33588107

#7 lore raw
Quoted 6 lines · Martin Lorenz
Quoted 5 lines · Rafael J. Wysocki
Quoted 4 lines · Martin Lorenz

Hallo again,

here is another one:

Quoted 7 lines
Quoted 5 lines
Quoted 2 lines

I reported a black screen on resume with my latest kernel build earlyer. But this was not reproducible. Only occured once.

Is this a resume from disk? If so, which kernel are you using?

no from suspend to ram

Interesting. See http://lkml.org/lkml/2006/10/3/19

-Mike

#8 lore raw
In reply to Martin Lorenz
Quoted 5 lines · Martin Lorenz

I only see this when ipw3945 is loaded.

[226156.057000] BUG: unable to handle kernel paging request at virtual

address 756e6567

OK, very bizarre. Another instance of this pattern:

1. Recent Core Duo Thinkpad (X60, T60, X60s)
2. tainting wireless driver loaded (ipw3945, madwifi)
3. fault at "Genu" somewhere in filesystem code
4. not long after a resume from ram (?)

Not exactly the same backtrace as before
(https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=208488
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=207658), but pretty close.

The only things I can think of are:

1. ipw3945 and madwifi are sharing some 802.11 code, which splats this pattern into memory for some reason
2. some firmware/smm bug which end up corrupting a register (?)
3. erm? anyone?

J