Crash when loading a module (without executing any code of the module!)

Crash when loading a module (without executing any code of the module!)

[Michael Tasche]

Hi,
currently I'm developping a small kernel module for a hardware (which was developped by a friend of mine), which is 
supposed to load a firmware onto a PCI-card.
 
The development is done together with the friend, who's developping the firmware.
We tried the following:
He compiled a kernel-independent object (containing the firmware) on his system using kbuild
(Dual-AthlonMP, SuSE 8.2 with kernel.org-kernel 2.6.3, module-init-tools 0.9.14-pre2, gcc 3.3.1,
ld 2.14.90.0.5 20030722).
Afterwards I tried to link it to my kernel-module (using the same kbuild makefile with
firmware.o_shipped) on my machine (Fedora2, 2.6.5-3.1smp, module-init-tools 3.0-pre10, gcc 3.3.3, ld 2.15.90.0.3 
20040415).
 
This is what happened:
 
Unable to handle kernel paging request at virtual address 82d90700
 printing eip:
02135657
*pde = 00000000
Oops: 0002 [#1]
PREEMPT SMP
CPU:    1
EIP:    0060:[<02135657>]    Not tainted
EFLAGS: 00010246   (2.6.5-3.1smp)
EIP is at module_unload_init+0xa/0x4d
eax: 82d90700   ebx: 82c2387c   ecx: 82d8f600   edx: 00000000
esi: 82c38f33   edi: 82c40027   ebp: 000005f0   esp: 763c3f38
ds: 007b   es: 007b   ss: 0068
Process insmod (pid: 1812, threadinfo=763c2000 task=7f3c60b0)
Stack: 02136dc0 7864cc40 8282a000 00000000 82d8f600 00000000 00000000 00000000
       00000000 00000000 0000000b 00000000 00000010 00000000 00000000 00000009
       00000025 00000024 00000026 82c3829c 82c23727 82a94000 0856a008 763c3fc4
Call Trace:
 [<02136dc0>] load_module+0x53e/0x7fa
 [<021370da>] sys_init_module+0x5e/0x293
 
Code: 89 81 00 11 00 00 89 81 04 11 00 00 89 c8 c7 80 00 01 00 00
 
 
What puzzles me, is that I don't see any of my code in the calltrace. I had a look into the
kernel-code and it seems to crash, before it even jumps into my code. What am I missing?
By the way, everything works fine, if I compile the entire module on my machine.
Some more testing showed, that we do also expierence a crash, if we do everything vice-versa.
 
Regards,
Michael
 
P.S: This was also posted by the driver developer to comp.os.linux.development.system.
____________________________________________________
Aufnehmen, abschicken, nah sein - So einfach ist 
WEB.DE Video-Mail: http://freemail.web.de/?mc=021200

Re: Crash when loading a module (without executing any code of the module!)

[Randy.Dunlap]

Hi-

On Wed, 30 Jun 2004 15:41:27 +0200 Michael Tasche wrote:

| Hi,
| currently I'm developping a small kernel module for a hardware (which was developped by a friend of mine), which is 
| supposed to load a firmware onto a PCI-card.
|  
| The development is done together with the friend, who's developping the firmware.
| We tried the following:
| He compiled a kernel-independent object (containing the firmware) on his system using kbuild
| (Dual-AthlonMP, SuSE 8.2 with kernel.org-kernel 2.6.3, module-init-tools 0.9.14-pre2, gcc 3.3.1,
| ld 2.14.90.0.5 20030722).
| Afterwards I tried to link it to my kernel-module (using the same kbuild makefile with
| firmware.o_shipped) on my machine (Fedora2, 2.6.5-3.1smp, module-init-tools 3.0-pre10, gcc 3.3.3, ld 2.15.90.0.3 
| 20040415).
|  
| This is what happened:
|  
| Unable to handle kernel paging request at virtual address 82d90700
|  printing eip:
| 02135657
| *pde = 00000000
| Oops: 0002 [#1]
| PREEMPT SMP
| CPU:    1
| EIP:    0060:[<02135657>]    Not tainted
| EFLAGS: 00010246   (2.6.5-3.1smp)
| EIP is at module_unload_init+0xa/0x4d
| eax: 82d90700   ebx: 82c2387c   ecx: 82d8f600   edx: 00000000
| esi: 82c38f33   edi: 82c40027   ebp: 000005f0   esp: 763c3f38
| ds: 007b   es: 007b   ss: 0068
| Process insmod (pid: 1812, threadinfo=763c2000 task=7f3c60b0)
| Stack: 02136dc0 7864cc40 8282a000 00000000 82d8f600 00000000 00000000 00000000
|        00000000 00000000 0000000b 00000000 00000010 00000000 00000000 00000009
|        00000025 00000024 00000026 82c3829c 82c23727 82a94000 0856a008 763c3fc4
| Call Trace:
|  [<02136dc0>] load_module+0x53e/0x7fa
|  [<021370da>] sys_init_module+0x5e/0x293
|  
| Code: 89 81 00 11 00 00 89 81 04 11 00 00 89 c8 c7 80 00 01 00 00

Some of those stack addresses look odd to me.
Please send me your kernel .config file.

| What puzzles me, is that I don't see any of my code in the calltrace. I had a look into the
| kernel-code and it seems to crash, before it even jumps into my code. What am I missing?
| By the way, everything works fine, if I compile the entire module on my machine.
| Some more testing showed, that we do also expierence a crash, if we do everything vice-versa.
|  
| Regards,
| Michael
|  
| P.S: This was also posted by the driver developer to comp.os.linux.development.system.

Can you post the module source code, or a subset of it that causes
the problem?

--
~Randy

Re: Crash when loading a module (without executing any code of the module!)

[Michael Tasche]

Hi,
 
attached you will find the .configs of the two kernel-trees.
 
-Michael
 
On Wed, 30 Jun 2004 14:29:48 -0700, Randy.Dunlap <rddunlap@osdl.org> wrote:
>
>Hi-
> 
>On Wed, 30 Jun 2004 15:41:27 +0200 Michael Tasche wrote:
>| Hi,
>| currently I'm developping a small kernel module for a hardware (which was developped by a friend of mine), which is
>| supposed to load a firmware onto a PCI-card.
>|
>| The development is done together with the friend, who's developping the firmware.
>| We tried the following:
>| He compiled a kernel-independent object (containing the firmware) on his system using kbuild
>| (Dual-AthlonMP, SuSE 8.2 with kernel.org-kernel 2.6.3, module-init-tools 0.9.14-pre2, gcc 3.3.1,
>| ld 2.14.90.0.5 20030722).
>| Afterwards I tried to link it to my kernel-module (using the same kbuild makefile with
>| firmware.o_shipped) on my machine (Fedora2, 2.6.5-3.1smp, module-init-tools 3.0-pre10, gcc 3.3.3, ld 2.15.90.0.3
>| 20040415).
>|
>| This is what happened:
>|
>| Unable to handle kernel paging request at virtual address 82d90700
>|  printing eip:
>| 02135657
>| *pde = 00000000
>| Oops: 0002 [#1]
>| PREEMPT SMP
>| CPU:    1
>| EIP:    0060:[<02135657>]    Not tainted
>| EFLAGS: 00010246   (2.6.5-3.1smp)
>| EIP is at module_unload_init+0xa/0x4d
>| eax: 82d90700   ebx: 82c2387c   ecx: 82d8f600   edx: 00000000
>| esi: 82c38f33   edi: 82c40027   ebp: 000005f0   esp: 763c3f38
>| ds: 007b   es: 007b   ss: 0068
>| Process insmod (pid: 1812, threadinfo=763c2000 task=7f3c60b0)
>| Stack: 02136dc0 7864cc40 8282a000 00000000 82d8f600 00000000 00000000 00000000
>|        00000000 00000000 0000000b 00000000 00000010 00000000 00000000 00000009
>|        00000025 00000024 00000026 82c3829c 82c23727 82a94000 0856a008 763c3fc4
>| Call Trace:
>|  [<02136dc0>] load_module+0x53e/0x7fa
>|  [<021370da>] sys_init_module+0x5e/0x293
>|
>| Code: 89 81 00 11 00 00 89 81 04 11 00 00 89 c8 c7 80 00 01 00 00
> 
>Some of those stack addresses look odd to me.
>Please send me your kernel .config file.
> 
>| What puzzles me, is that I don't see any of my code in the calltrace. I had a look into the
>| kernel-code and it seems to crash, before it even jumps into my code. What am I missing?
>| By the way, everything works fine, if I compile the entire module on my machine.
>| Some more testing showed, that we do also expierence a crash, if we do everything vice-versa.
>|
>| Regards,
>| Michael
>|
>| P.S: This was also posted by the driver developer to comp.os.linux.development.system.
> 
>Can you post the module source code, or a subset of it that causes
>the problem?
> 
>--
>~Randy
>-
>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/

________________________________________________________________
Verschicken Sie romantische, coole und witzige Bilder per SMS!
Jetzt neu bei WEB.DE FreeMail: http://freemail.web.de/?mc=021193

Re: Crash when loading a module (without executing any code of the module!)

[Randy.Dunlap]

On Fri, 2 Jul 2004 17:15:29 +0200 Michael Tasche wrote:

| Hi,
|  
| attached you will find the .configs of the two kernel-trees.

Not.  and there was a part 2 of the request also:

| >Can you post the module source code, or a subset of it that causes
| >the problem?

My modules load OK, so without some broken source code, it's
gonna be tough.


| -Michael
|  
| On Wed, 30 Jun 2004 14:29:48 -0700, Randy.Dunlap <rddunlap@osdl.org> wrote:
| >
| >Hi-
| > 
| >On Wed, 30 Jun 2004 15:41:27 +0200 Michael Tasche wrote:
| >| Hi,
| >| currently I'm developping a small kernel module for a hardware (which was developped by a friend of mine), which is
| >| supposed to load a firmware onto a PCI-card.
| >|
| >| The development is done together with the friend, who's developping the firmware.
| >| We tried the following:
| >| He compiled a kernel-independent object (containing the firmware) on his system using kbuild
| >| (Dual-AthlonMP, SuSE 8.2 with kernel.org-kernel 2.6.3, module-init-tools 0.9.14-pre2, gcc 3.3.1,
| >| ld 2.14.90.0.5 20030722).
| >| Afterwards I tried to link it to my kernel-module (using the same kbuild makefile with
| >| firmware.o_shipped) on my machine (Fedora2, 2.6.5-3.1smp, module-init-tools 3.0-pre10, gcc 3.3.3, ld 2.15.90.0.3
| >| 20040415).
| >|
| >| This is what happened:
| >|
| >| Unable to handle kernel paging request at virtual address 82d90700
| >|  printing eip:
| >| 02135657
| >| *pde = 00000000
| >| Oops: 0002 [#1]
| >| PREEMPT SMP
| >| CPU:    1
| >| EIP:    0060:[<02135657>]    Not tainted
| >| EFLAGS: 00010246   (2.6.5-3.1smp)
| >| EIP is at module_unload_init+0xa/0x4d
| >| eax: 82d90700   ebx: 82c2387c   ecx: 82d8f600   edx: 00000000
| >| esi: 82c38f33   edi: 82c40027   ebp: 000005f0   esp: 763c3f38
| >| ds: 007b   es: 007b   ss: 0068
| >| Process insmod (pid: 1812, threadinfo=763c2000 task=7f3c60b0)
| >| Stack: 02136dc0 7864cc40 8282a000 00000000 82d8f600 00000000 00000000 00000000
| >|        00000000 00000000 0000000b 00000000 00000010 00000000 00000000 00000009
| >|        00000025 00000024 00000026 82c3829c 82c23727 82a94000 0856a008 763c3fc4
| >| Call Trace:
| >|  [<02136dc0>] load_module+0x53e/0x7fa
| >|  [<021370da>] sys_init_module+0x5e/0x293
| >|
| >| Code: 89 81 00 11 00 00 89 81 04 11 00 00 89 c8 c7 80 00 01 00 00
| > 
| >Some of those stack addresses look odd to me.
| >Please send me your kernel .config file.
| > 
| >| What puzzles me, is that I don't see any of my code in the calltrace. I had a look into the
| >| kernel-code and it seems to crash, before it even jumps into my code. What am I missing?
| >| By the way, everything works fine, if I compile the entire module on my machine.
| >| Some more testing showed, that we do also expierence a crash, if we do everything vice-versa.
| >|
| >| Regards,
| >| Michael
| >|
| >| P.S: This was also posted by the driver developer to comp.os.linux.development.system.
| > 
| >Can you post the module source code, or a subset of it that causes
| >the problem?
| > 
| >--


--
~Randy

Re: Crash when loading a module (without executing any code of the module!)

[Michael Tasche]

[-- Attachment #1: Type: text/plain, Size: 4196 bytes --]

Hi, 

On Fri, 2 Jul 2004 08:49:15 -0700, Randy.Dunlap <rddunlap@osdl.org> wrote:

> On Fri, 2 Jul 2004 17:15:29 +0200 Michael Tasche wrote:
>
> | Hi,
> |
> | attached you will find the .configs of the two kernel-trees.
>
> Not.  and there was a part 2 of the request also:
>
> | >Can you post the module source code, or a subset of it that causes
> | >the problem?
>
> My modules load OK, so without some broken source code, it's
> gonna be tough.

we are just preparing a down stripped driver source for you.

Well, we do not think to solve the problem with the source, because the the crash occurs, before any of our source-code is executed.

Again the problem:  If our linux indepedant firmware-loader object is generated with "kbuild" on the other kernel-tree, we get the crash, if we build both objects against the same kernel-tree, everything works fine.

I have attached the .configs again, sorry my fault. 

- Michael

>
>
> | -Michael
> |
> | On Wed, 30 Jun 2004 14:29:48 -0700, Randy.Dunlap <rddunlap@osdl.org>  
> wrote:
> | >
> | >Hi-
> | >
> | >On Wed, 30 Jun 2004 15:41:27 +0200 Michael Tasche wrote:
> | >| Hi,
> | >| currently I'm developping a small kernel module for a hardware  
> (which was developped by a friend of mine), which is
> | >| supposed to load a firmware onto a PCI-card.
> | >|
> | >| The development is done together with the friend, who's developping  
> the firmware.
> | >| We tried the following:
> | >| He compiled a kernel-independent object (containing the firmware)  
> on his system using kbuild
> | >| (Dual-AthlonMP, SuSE 8.2 with kernel.org-kernel 2.6.3,  
> module-init-tools 0.9.14-pre2, gcc 3.3.1,
> | >| ld 2.14.90.0.5 20030722).
> | >| Afterwards I tried to link it to my kernel-module (using the same  
> kbuild makefile with
> | >| firmware.o_shipped) on my machine (Fedora2, 2.6.5-3.1smp,  
> module-init-tools 3.0-pre10, gcc 3.3.3, ld 2.15.90.0.3
> | >| 20040415).
> | >|
> | >| This is what happened:
> | >|
> | >| Unable to handle kernel paging request at virtual address 82d90700
> | >|  printing eip:
> | >| 02135657
> | >| *pde = 00000000
> | >| Oops: 0002 [#1]
> | >| PREEMPT SMP
> | >| CPU:    1
> | >| EIP:    0060:[<02135657>]    Not tainted
> | >| EFLAGS: 00010246   (2.6.5-3.1smp)
> | >| EIP is at module_unload_init+0xa/0x4d
> | >| eax: 82d90700   ebx: 82c2387c   ecx: 82d8f600   edx: 00000000
> | >| esi: 82c38f33   edi: 82c40027   ebp: 000005f0   esp: 763c3f38
> | >| ds: 007b   es: 007b   ss: 0068
> | >| Process insmod (pid: 1812, threadinfo=763c2000 task=7f3c60b0)
> | >| Stack: 02136dc0 7864cc40 8282a000 00000000 82d8f600 00000000  
> 00000000 00000000
> | >|        00000000 00000000 0000000b 00000000 00000010 00000000  
> 00000000 00000009
> | >|        00000025 00000024 00000026 82c3829c 82c23727 82a94000  
> 0856a008 763c3fc4
> | >| Call Trace:
> | >|  [<02136dc0>] load_module+0x53e/0x7fa
> | >|  [<021370da>] sys_init_module+0x5e/0x293
> | >|
> | >| Code: 89 81 00 11 00 00 89 81 04 11 00 00 89 c8 c7 80 00 01 00 00
> | >
> | >Some of those stack addresses look odd to me.
> | >Please send me your kernel .config file.
> | >
> | >| What puzzles me, is that I don't see any of my code in the  
> calltrace. I had a look into the
> | >| kernel-code and it seems to crash, before it even jumps into my  
> code. What am I missing?
> | >| By the way, everything works fine, if I compile the entire module  
> on my machine.
> | >| Some more testing showed, that we do also expierence a crash, if we  
> do everything vice-versa.
> | >|
> | >| Regards,
> | >| Michael
> | >|
> | >| P.S: This was also posted by the driver developer to  
> comp.os.linux.development.system.
> | >
> | >Can you post the module source code, or a subset of it that causes
> | >the problem?
> | >
> | >--
>
>
> --
> ~Randy
> -
> 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/
>

____________________________________________________
Aufnehmen, abschicken, nah sein - So einfach ist 
WEB.DE Video-Mail: http://freemail.web.de/?mc=021200

[-- Attachment #2: 26-configs.tgz --]
[-- Type: application/x-gzip, Size: 24292 bytes --]