BUG: BUS error while returning from read() in /dev/oprofile/buffer

BUG: BUS error while returning from read() in /dev/oprofile/buffer

[Anirban Sinha]

[-- Attachment #1: Type: text/plain, Size: 754 bytes --]

Hi:

 

I have been trying to hunt down this bug for several days now. What
mainly happens is that when oprofiled wakes up from read() in
/dev/oprofile/buffer on receiving a signal USR1 (i.e, when someone does
opcontrol -start after doing opcontrol-start-daemon), it somehow gets
SIGBUS within glibc read().  We are using a mips machine with Sybyte SB1
processor. On intel, this error does not show up. Interestingly, when I
tried running a small test program that simply reads
/dev/oprofile/buffer, the error can't be reproduced! 

 

Ralf and others, any insights, suggestions or useful comments from
experience will be really really appreciated. I am spending a lot of
time trying to fix this bug.

 

Cheers,

 

Ani

 


[-- Attachment #2: Type: text/html, Size: 2715 bytes --]

BUG: BUS error while returning from read() in /dev/oprofile/buffer

[Anirban Sinha]

[-- Attachment #1: Type: text/plain, Size: 754 bytes --]

Hi:

 

I have been trying to hunt down this bug for several days now. What
mainly happens is that when oprofiled wakes up from read() in
/dev/oprofile/buffer on receiving a signal USR1 (i.e, when someone does
opcontrol -start after doing opcontrol-start-daemon), it somehow gets
SIGBUS within glibc read().  We are using a mips machine with Sybyte SB1
processor. On intel, this error does not show up. Interestingly, when I
tried running a small test program that simply reads
/dev/oprofile/buffer, the error can't be reproduced! 

 

Ralf and others, any insights, suggestions or useful comments from
experience will be really really appreciated. I am spending a lot of
time trying to fix this bug.

 

Cheers,

 

Ani

 


[-- Attachment #2: Type: text/html, Size: 2715 bytes --]

Re: BUG: BUS error while returning from read() in /dev/oprofile/buffer

[David Daney]

Anirban Sinha wrote:
> Hi:
> 
>  
> 
> I have been trying to hunt down this bug for several days now. What 
> mainly happens is that when oprofiled wakes up from read() in 
> /dev/oprofile/buffer on receiving a signal USR1 (i.e, when someone does 
> opcontrol –start after doing opcontrol—start-daemon), it somehow gets 
> SIGBUS within glibc read().  We are using a mips machine with Sybyte SB1 
> processor. On intel, this error does not show up. Interestingly, when I 
> tried running a small test program that simply reads 
> /dev/oprofile/buffer, the error can’t be reproduced!
> 
>  
> 
> Ralf and others, any insights, suggestions or useful comments from 
> experience will be really really appreciated. I am spending a lot of 
> time trying to fix this bug.
> 

Likely you did a cross build of bash and your signal numbers are incorrect.

On the system where you built bash, SIGUSR1 is 10.

On the mips-linux target signal 10 is SIGBUS so when the bash kill 
builtin thinks it is sending SIGUSR1 it is really sending SIGBUS.

David Daney

RE: BUS error while returning from read() in /dev/oprofile/buffer

[Anirban Sinha]

[-- Attachment #1: Type: text/plain, Size: 1554 bytes --]

First a correction. My test program does terminate with a bus error on
receiving signal USR1. Just in case anyone is interested - It turned out
that the BUS error results from the fact that when coreutils was built,
it used the signal numbers from the devel machine (x86). These numbers
are different from our mips kernel on which we want to use oprofile
(mips has slightly different signal numbers from x86). Many thanks to
David Daney for drawing my attention to this.

 

Cheers,

 

Ani

 

 

From: linux-mips-bounce@linux-mips.org
[mailto:linux-mips-bounce@linux-mips.org] On Behalf Of Anirban Sinha
Sent: Thursday, January 03, 2008 5:18 PM
To: linux-mips@linux-mips.org; Ralf Baechle
Subject: BUG: BUS error while returning from read() in
/dev/oprofile/buffer

 

Hi:

 

I have been trying to hunt down this bug for several days now. What
mainly happens is that when oprofiled wakes up from read() in
/dev/oprofile/buffer on receiving a signal USR1 (i.e, when someone does
opcontrol -start after doing opcontrol-start-daemon), it somehow gets
SIGBUS within glibc read().  We are using a mips machine with Sybyte SB1
processor. On intel, this error does not show up. Interestingly, when I
tried running a small test program that simply reads
/dev/oprofile/buffer, the error can't be reproduced! 

 

Ralf and others, any insights, suggestions or useful comments from
experience will be really really appreciated. I am spending a lot of
time trying to fix this bug.

 

Cheers,

 

Ani

 


[-- Attachment #2: Type: text/html, Size: 4683 bytes --]

RE: BUS error while returning from read() in /dev/oprofile/buffer

[Anirban Sinha]

[-- Attachment #1: Type: text/plain, Size: 1554 bytes --]

First a correction. My test program does terminate with a bus error on
receiving signal USR1. Just in case anyone is interested - It turned out
that the BUS error results from the fact that when coreutils was built,
it used the signal numbers from the devel machine (x86). These numbers
are different from our mips kernel on which we want to use oprofile
(mips has slightly different signal numbers from x86). Many thanks to
David Daney for drawing my attention to this.

 

Cheers,

 

Ani

 

 

From: linux-mips-bounce@linux-mips.org
[mailto:linux-mips-bounce@linux-mips.org] On Behalf Of Anirban Sinha
Sent: Thursday, January 03, 2008 5:18 PM
To: linux-mips@linux-mips.org; Ralf Baechle
Subject: BUG: BUS error while returning from read() in
/dev/oprofile/buffer

 

Hi:

 

I have been trying to hunt down this bug for several days now. What
mainly happens is that when oprofiled wakes up from read() in
/dev/oprofile/buffer on receiving a signal USR1 (i.e, when someone does
opcontrol -start after doing opcontrol-start-daemon), it somehow gets
SIGBUS within glibc read().  We are using a mips machine with Sybyte SB1
processor. On intel, this error does not show up. Interestingly, when I
tried running a small test program that simply reads
/dev/oprofile/buffer, the error can't be reproduced! 

 

Ralf and others, any insights, suggestions or useful comments from
experience will be really really appreciated. I am spending a lot of
time trying to fix this bug.

 

Cheers,

 

Ani

 


[-- Attachment #2: Type: text/html, Size: 4683 bytes --]