[Pvfs2-users] Segmentation fault on insmod pvfs2.ko (pvfs2.6.3-patched, kernel 2.6.22-rc4)

[Sam Lang]

On Jun 27, 2007, at 8:18 AM, Pete Wyckoff wrote:

> murali.vilayannur at gmail.com wrote on Wed, 27 Jun 2007 00:50 -0700:
>> Something really bad is going on with these new kernels. This is the
>> 2nd (or perhaps 3rd) time that someone reported this exact same  
>> error.
>> Weird. I cannot think of any reason based on a cursory code glance
>> that could explain this error.
>> Am I missing something, Pete, Sam? Can you guys reproduce this error?
>>
>> kmem_cache_create gives this error message on
>> - a NULL name - which is not the case
>> - if called from an interrupt context - I doubt if this is  
>> possible at all.
>> - size < 4 bytes: not possible. pvfs2_inode_t is 392 bytes
>> - size > 128K: again not possible
>> - no constructor && a destructor specified - not possible from a code
>> review.
>
> Actually latest kernels don't allow a dtor at all.  Rev
> c59def9f222d44bb7e2f0a559f2906191a0862d7 for those playing
> along at home.

Interesting.  The error message they have could be nicer.  Its  
strange that they keep the parameter in the function and allow it to  
fail at runtime, instead of removing it and making it fail at compile  
time.  Changing function prototypes in the kernel from version to  
version hasn't seemed like a big deal to the kernel developers, which  
can be a pain, but they should at least be consistent.
-sam

>
> We use dtors for the inode cache and devreq cache.  But it's not
> obivous to me how to move that functionality elsewhere, at a quick
> glance.  Maybe you'll know?
>
> 		-- Pete
> _______________________________________________
> Pvfs2-users mailing list
> Pvfs2-users at beowulf-underground.org
> http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users
>