You may want to backport patch 1167/3 to 2.4.  The sa1111-pcibuf.c in
rmk's 2.4.x and 2.5.x source tree has several serious bugs.

I dobut that's enough to get you around the "not in interrupt" assertion
...

-ch


> -----Original Message-----
> From: [EMAIL PROTECTED] 
> [mailto:[EMAIL PROTECTED]] On 
> Behalf Of Pering, Trevor
> Sent: Tuesday, July 09, 2002 5:05 PM
> To: [EMAIL PROTECTED]; 
> [EMAIL PROTECTED]
> Subject: PCI/USB consustent_alloc assertion violation.
> 
> 
> Hello -- this is a re-post of an earlier problem from last 
> week I'm still working on.
> 
> I'm using an ARM-based 2.4.18-rmk4 system that is using a USB 
> Bluetooth device to do file transfer. When I transfer a large 
> file, the system hits a "not in interrupt" assertion in 
> consistent_alloc (arch/arm/mm/consistent.c). I've included 
> the stack trace below, and it looks like something in either 
> the USB or PCI subsystem is trying to allocate some memory 
> when it shouldn't be. There is a comment in 
> consistent_alloc() to this effect... how far up the stack 
> should this assertion apply?
> 
> The reason it only happens for large files is that the 
> pci_pool system maintains some buffers to use, which are 
> sufficient for small files. (As a workaround, I made the pci 
> code allocate enough buffers when it initializes, so this 
> problem is not triggered.) 
> 
> I don't know enough about the USB or PCI or SA1111 subsystems 
> to really know/understand what the real problem is here... 
> Max (Bluetooth subsystem) says that the BlueZ stack doesn't 
> deal with the PCI layers, so it probably isn't a problem 
> there (it is valid to call usb_start_wait_urb from an 
> interrupt context)?
> 
>       Trevor
> 
> 
> > Trace; c021af48 <consistent_alloc+0/150>
> > Trace; c021b0ac <pci_alloc_consistent+14/1c>
> > Trace; c021b098 <pci_alloc_consistent+0/1c>
> > Trace; c021e9e0 <pci_pool_alloc+0/1b0>
> > Trace; c021e990 <pci_pool_destroy+80/d0>
> > Trace; c021ec6c <pci_pool_free+80/f0>
> > Trace; c021eb8c <pci_pool_alloc+1ac/1b0>
> > Trace; c021e5f8 <sa1111_map_single+74/a8>
> > Trace; c021e5b4 <sa1111_map_single+30/a8>
> > Trace; c021e784 <pci_pool_create+a0/100>
> > Trace; c021e730 <pci_pool_create+4c/100>
> > Trace; c02c5e68 <td_submit_urb+1bc/3a8>
> > Trace; c02c5d88 <td_submit_urb+dc/3a8>
> > Trace; c02c49a0 <sohci_unlink_urb+d0/244>
> > Trace; c02c44b0 <sohci_submit_urb+dc/4fc>
> > Trace; c02bb580 <usb_start_wait_urb+60/174>
> > Trace; c02bb540 <usb_start_wait_urb+20/174>
> > Trace; c2879820 <END_OF_CODE+2502ff8/????>
> > Trace; c02271f4 <tasklet_init+1c/28>
> > Trace; c0227004 <__tasklet_hi_schedule+38/7c>
> > Trace; c0226f84 <__tasklet_schedule+34/7c>
> > Trace; c0215688 <__do_softirq+8/20>
> > Trace; c02c9250 <sock_recvmsg+28/e4>
> > Trace; c02c9530 <sock_readv_writev+58/80>
> > Trace; c02c9490 <sock_sendpage+38/80>
> > Trace; c0242ddc <do_readv_writev+64/2ec>
> > Trace; c0242d00 <sys_write+dc/154>
> > Trace; c02156a0 <ret_fast_syscall+0/38>
> 
> 
> -------------------------------------------------------------------
> Subscription options: 
> http://lists.arm.linux.org.uk/mailman/listinfo/linux-arm-kerne
l
FAQ/Etiquette:
http://www.arm.linux.org.uk/armlinux/mailinglists.php



-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Stuff, things, and much much more.
http://thinkgeek.com/sf
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel

Reply via email to