TeamForge : Post

Forum Topic - devb-eide core dump:

View: as

Mario Charest

01/29/2008 9:28 AM

post4644

I have attached a core dump of devb-eide.

This started to happened after I updated my runtime image to 6.3.2.  I doubled checked and I am confident I didn't miss 
a file during the process.

I can get it to crash consistenly, just have to copy a directory with lots (100) of files/directory in it.  The files 
are copied over the network, but I get the feeling that's not related.

Attachment:

devb-eide.core.7z 2.16 MB

Mario Charest

01/30/2008 1:58 PM

post4662

Re: devb-eide core dump

> 
> I have attached a core dump of devb-eide.
> 
> This started to happened after I updated my runtime image to 6.3.2.  I doubled
>  checked and I am confident I didn't miss a file during the process.
> 
> I can get it to crash consistenly, just have to copy a directory with lots (
> 100) of files/directory in it.  The files are copied over the network, but I 
> get the feeling that's not related.
> 


Somebody else is having the exact same problem.  It crashes at the same address.

http://www.openqnx.com/index.php?name=PNphpBB2&file=viewtopic&p=41381#41381

Mario Charest

01/30/2008 7:49 PM

post4667

Re: devb-eide core dump

Is this the right place to post for this?  

Any holiday I wasn't aware of?

> > 
> > I have attached a core dump of devb-eide.
> > 
> > This started to happened after I updated my runtime image to 6.3.2.  I 
> doubled
> >  checked and I am confident I didn't miss a file during the process.
> > 
> > I can get it to crash consistenly, just have to copy a directory with lots (
> 
> > 100) of files/directory in it.  The files are copied over the network, but I
>  
> > get the feeling that's not related.
> > 
> 
> 
> Somebody else is having the exact same problem.  It crashes at the same 
> address.
> 
> http://www.openqnx.com/index.php?name=PNphpBB2&file=viewtopic&p=41381#41381 
>

Jeevan Mathew

01/31/2008 8:34 AM

post4675

Re: devb-eide core dump

Hello Mario,

> Is this the right place to post for this?  

IMHO, BSP and Drivers would have been better. 

> 
> Any holiday I wasn't aware of?
> 

No.

> > > 
> > > I have attached a core dump of devb-eide.
> > > 
> > > This started to happened after I updated my runtime image to 6.3.2.  I 
> > doubled
> > >  checked and I am confident I didn't miss a file during the process.

Waht was Your  exact update procedure ?

As a starting point I would suggest that You issue a 
pidin -pdevb-eide mem

and post all cksums of the involved libs/*.so's  and tell the place where they are accessed from. 

Just want to exclude a stupid library version missmatch.


> > > 
> > > I can get it to crash consistenly, just have to copy a directory with lots
>  (
> > 
> > > 100) of files/directory in it.  The files are copied over the network, but
>  I
> >  
> > > get the feeling that's not related.
> > > 
> > 
> > 
> > Somebody else is having the exact same problem.  It crashes at the same 
> > address.
> > 
> > http://www.openqnx.com/index.php?name=PNphpBB2&file=viewtopic&p=41381#41381 
> 
> > 
> 
> 

Regards,
Jeevan

Peter Mitsis

02/04/2008 11:56 AM

post4775

Re: devb-eide core dump

We believe that we have located the problem.  It appears to lie within the io-blk.so code.

When the blk ramdisk=xxx option is used, a path has been found where a ramdisk's buffer can be cleaned when invalidated.
  When using the ramdisk (not to be confused with devb-ram), the buffers should not be cleaned when invalidated.  In the
 case of this forum's thread, this eventually leads to crashing the devb-eide driver.

Until fixed, the suggested workaround is to use devb-ram instead of using the blk ramdisk=xxx option with devb-eide or 
the other devb-xxx drivers.

Peter Mitsis

Mario Charest

02/05/2008 3:29 PM

post4816

Re: devb-eide core dump

> We believe that we have located the problem.  It appears to lie within the io-
> blk.so code.
> 
> When the blk ramdisk=xxx option is used, a path has been found where a 
> ramdisk's buffer can be cleaned when invalidated.  When using the ramdisk (not
>  to be confused with devb-ram), the buffers should not be cleaned when 
> invalidated.  In the case of this forum's thread, this eventually leads to 
> crashing the devb-eide driver.
> 
> Until fixed, the suggested workaround is to use devb-ram instead of using the 
> blk ramdisk=xxx option with devb-eide or the other devb-xxx drivers.
> 
> Peter Mitsis

Just want to confirm that not using the ramdisk (devb-eide) solved the crashing problem for us.

Thanks.

Return