Fwd: [RFD]:Is there any method to snapshot all information from both tasks on cpus and cpus when kernel crash occurs

Fwd: [RFD]:Is there any method to snapshot all information from both tasks on cpus and cpus when kernel crash occurs

[Liu hua]

Hi all,

For arm32 platform. I want to know whether or not a mechanism in kernel can snapshot tasks and cpus
information as the follow. If not, such a mechanism can be accepted by kernel.



...

</module list>

<task info>

TASK: [pid 1], [tgid 1], [linuxrc] [state S] [policy 0] [cpu 3]

 EIP address: [<401945a4>] 0x401945a4

       r0 = ffffffff;   r1 = 00000000

       r2 = 00000000;      r3 = 00000000

       r4 = 00000000;      r5 = 00000000

       r6 = ffffffff;   r7 = 00000072

       r8 = 00000050;      r9 = 00000000

       r10 = beec1e34;     fp = 00000000

       ip = 00095414;      sp = beec1bc8

       lr = 00069f2c;pc = 401945a4

       cpsr = 60000210;

[<c0cd9338>] (schedule+0x4fc/0x61c)

[<c0a56418>] (do_wait+0x1c0/0x22c)

[<c0a57670>] (sys_wait4+0xa0/0xc0)

[<c0a295c0>] (ret_fast_syscall+0x0/0x3c)

TASK: [pid 2], [tgid 2], [kthreadd] [state S] [policy 0] [cpu 1]

 EIP address: [<c0a6ea94>] kthreadd+0x8c/0x14c

[<c0cd9338>] (schedule+0x4fc/0x61c)

[<c0a6ea94>] (kthreadd+0x8c/0x14c)

[<c0a2a7d8>] (kernel_thread_exit+0x0/0x8)

...

</task info>

<cpu info cpu_id = 0>

TASK: [pid 29666], [tgid 29666], [kstop/1] [state R] [policy 1] [cpu 1]

[<c0a31310>] (ksnapshot_unwind_backtrace.part.2+0x50/0x154)

[<bf068144>] (us_dump_stack+0x38/0x44 [rtos_snapshot])

[<bf068f34>] (snapshot_cpu_info+0x60/0x4a8 [rtos_snapshot])

[<bf068b84>] (ksnapshot_taskinfo_buffer+0x320/0x4a8 [rtos_snapshot])

[<bf071018>] (fiq_callcack_handler+0x18/0x30 [test_ks])

[<c0a2eddc>] (fiq_real_handle+0x24/0x30)

[<c0a2836c>] (do_IPI+0x64/0x1c0)

[<c0cdb8e4>] (__irq_svc+0x44/0xe0)

[<c0a90088>] (stop_cpu+0x104/0x124)

[<c0a6a670>] (worker_thread+0x1a4/0x248)

[<c0a6e7a4>] (kthread+0x78/0x84)

[<c0a2a7d8>] (kernel_thread_exit+0x0/0x8)

</cpu info>

...

<cpu info cpu_id = n>

...

</cpu info>

Thanks
Yan