[PATCH] Weird stack trace for HostAP in 2.6.14-rc2

[Jouni Malinen]

On Tue, Sep 20, 2005 at 03:59:25PM -0700, Jean Tourrilhes wrote:

> 	2 patches for HostAP in the kernel...

Thanks!

> 	This is the trace I got :
> ----------------------------------------------------
> [  224.918985] Debug: sleeping function called from invalid context at mm/slab.c:2459
> [  224.919046] in_atomic():1, irqs_disabled():0
> [  224.919090]  [<c01198a7>] __might_sleep+0xa7/0xb0
> [  224.919159]  [<d08b815d>] hfa384x_setup_bap+0xcd/0x150 [hostap_cs]
> [  224.919238]  [<c0146cfc>] kmem_cache_alloc+0x7c/0x80
> [  224.919298]  [<d092c4c6>] __prism2_translate_scan+0x386/0x6f0 [hostap]
> [  224.919408]  [<d092c9ab>] prism2_ioctl_giwscan+0x17b/0x320 [hostap]

> 	Weird thing is that I never got that with the driver compiled
> externally.

Odd.. I can see the same behavior, but cannot really explain why that
happens and it is just the kmalloc() in __prism2_translate_scan() that
triggers this. Anyway, your patch removes this, so I added it to my
queue for kernel updates.
 
> 	Also, now that we are compiling with -fno-strict-aliasing
> (this is the kernel default), we can drop the following kludge :

Agreed, added to the queue.

-- 
Jouni Malinen                                            PGP id EFC895FA