Skip to contentSkip to conversation

Oops on removing via-rhine [2.4.10-ac11]

DISCUSSION Andrew Over 3 participants 5 messages Oct 11, 2001 – Oct 15, 2001 Network drivers

lore
Share & tools
lore
#0 lore raw

Received the following oops when trying to remove via-rhine.

The kernel is 'tainted' by the ALSA modules (no NVdriver!), which don't yet have the MODULE_LICENSE stuff done.

This occurs with kernel 2.4.10-ac11 (and also 2.4.10-ac8). System in question is an Athlon 1400 running on a KT266 mobo.

After the oops, via-rhine is reported by lsmod as "(deleted)".

This oops is reasonably easy to reproduce (reinsert and remove via-rhine enough times and it happens), and any further information is available upon request. Next time I reboot, I'll try reproducing the problem without the ALSA modules.

ksymoops output:

Warning (compare_ksyms_lsmod): module via-rhine is in lsmod but not in ksyms, pr obably no symbols exported

invalid operand: 0000
CPU:    0
EIP:    0010:[<c0129841>]    Tainted: P 

Using defaults from ksymoops -t elf32-i386 -a i386
EFLAGS: 00010282
eax: c18542d0 ebx: 00000000 ecx: c18542d0 edx: 00000000
esi: c1896c00 edi: c18542d0 ebp: bfffe8fc esp: dcb6df28
ds: 0018 es: 0018 ss: 0018
Process rmmod (pid: 6729, stackpage=dcb6d000)

Stack: dfbfa400 c1896c00 dfbfa540 bfffe8fc ed0800ff 0000001c c1896c00 c01acd3f 
       c0129f1a c0129f44 c010b6bd e08fdbab c1896c00 00000200 df5b0000 1f5b0000 
       c1896c00 dfbfa400 c1896c00 e08fe860 e08fc000 c01ad01f c1896c00 e08fc000 
Call Trace: [<c01acd3f>] [<c0129f1a>] [<c0129f44>] [<c010b6bd>] [<e08fdbab>] 
   [<e08fe860>] [<c01ad01f>] [<e08fdc0a>] [<e08fe860>] [<c01150c7>] [<c0114457>]

[<c0106c4b>]

Code: 0f 0b 8b 57 08 85 d2 74 02 0f 0b 89 f8 2b 05 2c 04 27 c0 69 
Quoted 1 line
Quoted 1 line

EIP; c0129840 <__free_pages_ok+10/1c0> <=====

Trace; c01acd3e <pci_release_regions+6e/80>
Trace; c0129f1a <__free_pages+1a/20>
Trace; c0129f44 <free_pages+24/30>
Trace; c010b6bc <pci_free_consistent+1c/20>
Trace; e08fdbaa <_end+2066e636/20670a8c>
Trace; e08fe860 <_end+2066f2ec/20670a8c>
Trace; c01ad01e <pci_unregister_driver+3e/60>
Trace; e08fdc0a <_end+2066e696/20670a8c>
Trace; e08fe860 <_end+2066f2ec/20670a8c>
Trace; c01150c6 <free_module+16/a0>
Trace; c0114456 <sys_delete_module+f6/1c0>
Trace; c0106c4a <system_call+32/38>
Code; c0129840 <__free_pages_ok+10/1c0>
00000000 <_EIP>:
Code; c0129840 <__free_pages_ok+10/1c0> <=====
0: 0f 0b ud2a <=====
Code; c0129842 <__free_pages_ok+12/1c0>
2: 8b 57 08 mov 0x8(%edi),%edx
Code; c0129844 <__free_pages_ok+14/1c0>
5: 85 d2 test %edx,%edx
Code; c0129846 <__free_pages_ok+16/1c0>
7: 74 02 je b <_EIP+0xb> c012984a <__free_pages_ok+1a/1c0>
Code; c0129848 <__free_pages_ok+18/1c0>
9: 0f 0b ud2a
Code; c012984a <__free_pages_ok+1a/1c0>
b: 89 f8 mov %edi,%eax
Code; c012984c <__free_pages_ok+1c/1c0>
d: 2b 05 2c 04 27 c0 sub 0xc027042c,%eax
Code; c0129852 <__free_pages_ok+22/1c0>
13: 69 00 00 00 00 00 imul $0x0,(%eax),%eax

ver_linux output:

Linux cartman 2.4.10-ac11 #1 Wed Oct 10 19:40:56 PDT 2001 i686 unknown

Gnu C                  2.95.4
Gnu make               3.79.1
util-linux             2.11l
mount                  2.11l
modutils               2.4.10
e2fsprogs              1.25
Linux C Library        2.2.4
Dynamic linker (ldd)   2.2.4
Procps                 2.0.7
Net-tools              1.60
Console-tools          0.2.3
Sh-utils               2.0.11
Modules Loaded         3c59x isofs snd-pcm-oss snd-mixer-oss serial snd-card-cmipci snd-pcm snd-opl3 snd-timer snd-hwdep snd-mpu401-uart snd-rawmidi snd-seq-device snd soundcore via-rhine
#1 lore raw
Quoted 12 lines · Andrew Over

This oops is reasonably easy to reproduce (reinsert and remove via-rhine enough times and it happens)

Quoted 1 line
Quoted 1 line

EIP; c0129840 <__free_pages_ok+10/1c0> <=====

Trace; c01acd3e <pci_release_regions+6e/80>
Trace; c0129f1a <__free_pages+1a/20>
Trace; c0129f44 <free_pages+24/30>
Trace; c010b6bc <pci_free_consistent+1c/20>
Trace; e08fdbaa <_end+2066e636/20670a8c>
Trace; e08fe860 <_end+2066f2ec/20670a8c>
Trace; c01ad01e <pci_unregister_driver+3e/60>
Trace; e08fdc0a <_end+2066e696/20670a8c>
Trace; e08fe860 <_end+2066f2ec/20670a8c>

Oops in module delete may not get the symbol tables for the deleted module if they have already been removed from /proc/ksyms. I suggest you create /var/log/ksymoops, man insmod. Then insmod and rmmod will save the symbol tables after each module load or unload, you can point ksymoops at the saved symbols from before the failing rmmod. That will give a better ksymoops decode.

#2 lore raw
In reply to Andrew Over
Quoted 6 lines · Keith Owens

Oops in module delete may not get the symbol tables for the deleted module if they have already been removed from /proc/ksyms. I suggest you create /var/log/ksymoops, man insmod. Then insmod and rmmod will save the symbol tables after each module load or unload, you can point ksymoops at the saved symbols from before the failing rmmod. That will give a better ksymoops decode.

Thanks for that. Already setup courtesy of Debian.

Revised oops as follows:

invalid operand: 0000
CPU:    0
EIP:    0010:[<c0129841>]    Tainted: P 

Using defaults from ksymoops -t elf32-i386 -a i386
EFLAGS: 00010282
eax: c18542d0 ebx: 00000000 ecx: c18542d0 edx: 00000000
esi: c1896c00 edi: c18542d0 ebp: bfffe8fc esp: dcb6df28
ds: 0018 es: 0018 ss: 0018
Process rmmod (pid: 6729, stackpage=dcb6d000)

Stack: dfbfa400 c1896c00 dfbfa540 bfffe8fc ed0800ff 0000001c c1896c00 c01acd3f 
       c0129f1a c0129f44 c010b6bd e08fdbab c1896c00 00000200 df5b0000 1f5b0000 
       c1896c00 dfbfa400 c1896c00 e08fe860 e08fc000 c01ad01f c1896c00 e08fc000 
Call Trace: [<c01acd3f>] [<c0129f1a>] [<c0129f44>] [<c010b6bd>] [<e08fdbab>] 
   [<e08fe860>] [<c01ad01f>] [<e08fdc0a>] [<e08fe860>] [<c01150c7>] [<c0114457>] 
   [<c0106c4b>] 
Code: 0f 0b 8b 57 08 85 d2 74 02 0f 0b 89 f8 2b 05 2c 04 27 c0 69 
Quoted 1 line
Quoted 1 line

EIP; c0129840 <__free_pages_ok+10/1c0> <=====

Trace; c01acd3e <pci_release_regions+6e/80> Trace; c0129f1a <__free_pages+1a/20>
Trace; c0129f44 <free_pages+24/30>
Trace; c010b6bc <pci_free_consistent+1c/20> Trace; e08fdbaa <[via-rhine]via_rhine_remove_one+2a/40> Trace; e08fe860 <[via-rhine]via_rhine_driver+0/3e> Trace; c01ad01e <pci_unregister_driver+3e/60> Trace; e08fdc0a <[via-rhine]via_rhine_cleanup+a/10> Trace; e08fe860 <[via-rhine]via_rhine_driver+0/3e> Trace; c01150c6 <free_module+16/a0>
Trace; c0114456 <sys_delete_module+f6/1c0> Trace; c0106c4a <system_call+32/38>
Code; c0129840 <__free_pages_ok+10/1c0> 00000000 <_EIP>:
Code; c0129840 <__free_pages_ok+10/1c0> <=====
0: 0f 0b ud2a <=====
Code; c0129842 <__free_pages_ok+12/1c0>
2: 8b 57 08 mov 0x8(%edi),%edx
Code; c0129844 <__free_pages_ok+14/1c0>
5: 85 d2 test %edx,%edx
Code; c0129846 <__free_pages_ok+16/1c0>
7: 74 02 je b <_EIP+0xb> c012984a <__free_pages_ok+1a/1c0>
Code; c0129848 <__free_pages_ok+18/1c0>
9: 0f 0b ud2a
Code; c012984a <__free_pages_ok+1a/1c0>
b: 89 f8 mov %edi,%eax
Code; c012984c <__free_pages_ok+1c/1c0>
d: 2b 05 2c 04 27 c0 sub 0xc027042c,%eax
Code; c0129852 <__free_pages_ok+22/1c0>
13: 69 00 00 00 00 00 imul $0x0,(%eax),%eax

--Andrew

#3 lore raw
Quoted 2 lines · Andrew Over

Trace; c010b6bc <pci_free_consistent+1c/20> Trace; e08fdbaa <[via-rhine]via_rhine_remove_one+2a/40>

remove_one shouldn't call pci_free_consistent as that is already done in via_rhine_close (and I am assuming that if you open you must always close before remove ...).

I could repeat some kind of crash, but could you please test if the patch below fixes the crash you see. patch vs 2.4.13-pre2, but should apply vs others as well.

/Urban

drivers/net/via-rhine.c +1 -5
diff -urN -X exclude linux-2.4.13-pre2-orig/drivers/net/via-rhine.c linux/drivers/net/via-rhine.c
@@ -764,6 +764,7 @@
 			    RX_RING_SIZE * sizeof(struct rx_desc) +
 			    TX_RING_SIZE * sizeof(struct tx_desc),
 			    np->rx_ring, np->rx_ring_dma);
+	np->tx_ring = NULL;
 
 	if (np->tx_bufs)
 		pci_free_consistent(np->pdev, PKT_BUF_SZ * TX_RING_SIZE,
@@ -1595,11 +1596,6 @@
 #ifndef USE_IO
 	iounmap((char *)(dev->base_addr));
 #endif
-
-	pci_free_consistent(pdev, 
-			    RX_RING_SIZE * sizeof(struct rx_desc) +
-			    TX_RING_SIZE * sizeof(struct tx_desc),
-			    np->rx_ring, np->rx_ring_dma);
 
 	kfree(dev);
 
#4 lore raw
In reply to Andrew Over
Quoted 3 lines · Urban Widmark

I could repeat some kind of crash, but could you please test if the patch below fixes the crash you see. patch vs 2.4.13-pre2, but should apply vs others as well.

Can no longer reproduce the problem on removal.

Thanks!
--Andrew