[openib-general] Re: [ib_at] oops in ib_at when running kdapltest

[Hal Rosenstock]

On Sun, 2005-06-05 at 08:36, Itamar wrote:
> Hi All,
> I was running today some kdapltest and after about 10 tests i got this oops.
> the test was :
> ./kdapltest -T T -s 11.1.4.43 -D mthca0a -d -t 2 -w 8 -i 10 client SR server SR
> (the same test pass before it got the oops)
> my system is :
> dual x86 xeon 2G with 1GB memory + suse9.1 + k2.6.11 + svn rev 2546 + kdapl
> my kernel is with CONFIG_DEBUG_PAGEALLOC
> 
> Jun  5 13:45:32 mtvs42 kernel: Unable to handle kernel paging request at virtual address ec298ff4
> Jun  5 13:45:32 mtvs42 kernel:  printing eip:
> Jun  5 13:45:32 mtvs42 kernel: f8ad3c15
> Jun  5 13:45:32 mtvs42 kernel: *pde = 0047c067
> Jun  5 13:45:32 mtvs42 kernel: Oops: 0002 [#1]
> Jun  5 13:45:32 mtvs42 kernel: SMP DEBUG_PAGEALLOC
> Jun  5 13:45:32 mtvs42 kernel: Modules linked in: kdapltest ib_dat_provider ib_cm ib_at dat ib_umad ib_ipoib ib_sa edd joydev st sd_mod sr_mod scsi_mod ide_cd cdrom nvram usbserial parport_pc lp parport autofs thermal processor fan button ipv6 battery ac e100 mii ib_mthca ib_mad ib_core i2c_i801 i2c_core uhci_hcd usbcore hw_random evdev dm_mod ext3 jbd
> Jun  5 13:45:32 mtvs42 kernel: CPU:    0
> Jun  5 13:45:32 mtvs42 kernel: EIP:    0060:[pg0+946297877/1069155328]    Not tainted VLI
> Jun  5 13:45:32 mtvs42 kernel: EIP:    0060:[<f8ad3c15>]    Not tainted VLI
> Jun  5 13:45:32 mtvs42 kernel: EFLAGS: 00010246   (2.6.11-gen2) 
> Jun  5 13:45:32 mtvs42 kernel: EIP is at resolve_path+0xa5/0x100 [ib_at]
> Jun  5 13:45:32 mtvs42 kernel: eax: 00000000   ebx: ec298f3c   ecx: f21b7e80   edx: 00000000
> Jun  5 13:45:32 mtvs42 kernel: esi: ec79dae8   edi: 00000064   ebp: ec79dae8   esp: ec79daa8
> Jun  5 13:45:32 mtvs42 kernel: ds: 007b   es: 007b   ss: 0068
> Jun  5 13:45:32 mtvs42 kernel: Process DT_Mdep_Thread_ (pid: 9737, threadinfo=ec79c000 task=ebd18ad0)
> Jun  5 13:45:32 mtvs42 kernel: Stack: 000080fe 00000000 00c90200 51d00001 000080fe 00000000 01c90200 71f79d07 
> Jun  5 13:45:32 mtvs42 kernel:        00000000 00000000 00000000 00000000 00000000 ffff0001 00000000 00000000 
> Jun  5 13:45:32 mtvs42 kernel:        000080fe 00000000 00c90200 51d00001 000080fe 00000000 01c90200 71f79d07 
> Jun  5 13:45:32 mtvs42 kernel: Call Trace:
> Jun  5 13:45:32 mtvs42 kernel:  [pg0+946298889/1069155328] ib_at_paths_by_route+0xa9/0xd0 [ib_at]
> Jun  5 13:45:32 mtvs42 kernel:  [<f8ad4009>] ib_at_paths_by_route+0xa9/0xd0 [ib_at]
> Jun  5 13:45:32 mtvs42 kernel:  [pg0+963317968/1069155328] dapl_rt_comp_handler+0x135/0x190 [ib_dat_provider]
> Jun  5 13:45:32 mtvs42 kernel:  [<f9b0f0d0>] dapl_rt_comp_handler+0x135/0x190 [ib_dat_provider]
> Jun  5 13:45:32 mtvs42 kernel:  [pg0+963319747/1069155328] dapl_ib_connect+0x698/0x6ae [ib_dat_provider]
> Jun  5 13:45:32 mtvs42 kernel:  [<f9b0f7c3>] dapl_ib_connect+0x698/0x6ae [ib_dat_provider]
> Jun  5 13:45:32 mtvs42 kernel:  [pg0+963336373/1069155328] dapl_ep_connect+0x73f/0x7b3 [ib_dat_provider]

Can you turn on the DEBUG printing in at.c by changing the following
lines:
#define DEBUG(fmt, ...) while (0) {}
//#define DEBUG(fmt, arg ...)   printk("ib_at: %s: " fmt "\n", __FUNCTION__,  ## arg);

to:
//#define DEBUG(fmt, ...) while (0) {}
#define DEBUG(fmt, arg ...)   printk("ib_at: %s: " fmt "\n", __FUNCTION__,  ## arg);

rebuild ib_at, and send the output of /var/log/messages. Thanks.

-- Hal