Hi Guanyou,

On Sat, Nov 2, 2024 at 1:35 AM Guanyou Chen <chenguanyou9338@gmail.com> wrote:
>
> Hi Lianbo, Tao
>
> Remove offline status check, We can query the registers of
> each CPU at any time and obtain their stack.
>
> CPU 0: [OFFLINE]
>     X0: 0000000000000000   X1: 0000000000000000   X2: 0000000000000000
>     X3: 000000000003fcbc   X4: 0000000000000001   X5: 0000000000000000
>     X6: 0000000000000000   X7: 0000000000000000   X8: 00000000ffffffff
>     X9: ffffffc009e6ae48  X10: ffffffc009e6ae20  X11: 0000000000000000
>    X12: 0000000000000002  X13: 0000000000000004  X14: 0000000000000000
>    X15: 0000000000004000  X16: 00000000f90f05f6  X17: 00000000f90f05f6
>    X18: 0000000000000000  X19: 0000000000000002  X20: ffffffc009e3b008
>    X21: ffffffc00a01d020  X22: ffffffc009f798f0  X23: 0000000060001000
>    X24: 0000000000000000  X25: 0000000000000000  X26: 0000000000000000
>    X27: 0000000000000000  X28: ffffff8111eecb00  X29: ffffffc008003f50
>     LR: ffffffc00802df88   SP: ffffffc008003f40   PC: ffffffc00802df94
>    PSTATE: 024003c5   FPVALID: 00000000
>
> crash> bt -c 0
> PID: 1842     TASK: ffffff8111eecb00  CPU: 0    COMMAND: "android.bg"
>  00 [ffffffc008003f50] ipi_handler at ffffffc00802df90
>  01 [ffffffc008003f90] handle_percpu_devid_irq at ffffffc008146f50
>  02 [ffffffc008003fd0] generic_handle_domain_irq at ffffffc00813f484
>  03 [ffffffc008003fe0] gic_handle_irq at ffffffc008010140
> --- <IRQ stack> ---
>  04 [ffffffc019c3be20] call_on_irq_stack at ffffffc008016ed4
>  05 [ffffffc019c3be40] do_interrupt_handler at ffffffc008019cb4
>  06 [ffffffc019c3be60] el0_interrupt at ffffffc008f7b848
>  07 [ffffffc019c3be90] __el0_irq_handler_common at ffffffc008f7b368
>  08 [ffffffc019c3bea0] el0t_64_irq_handler at ffffffc008f7b344
>  09 [ffffffc019c3bfe0] el0t_64_irq at ffffffc008011720
>      PC: 0000000072415108   LR: 00000000724150d0   SP: 0000007691d2bfa0
>     X29: 00000000734f60e0  X28: 000000001a2fa678  X27: 0000000000000063
>     X26: 000000001a2fa678  X25: 000000001a2fa678  X24: 000000001a7bb718
>     X23: 000000001a7ba198  X22: 000000001a7ba190  X21: b4000076f9a828c8
>     X20: 0000000000000000  X19: b4000076f9a82800  X18: 000000768d68a000
>     X17: 00000000708f89f8  X16: 00000000000000f0  X15: 0000000000000000
>     X14: 0000007691d2bca0  X13: 0000000080100000  X12: 0000000000000000
>     X11: 0000000000000000  X10: 0000000000000000   X9: 9636716211228cd4
>      X8: 9636716211228cd4   X7: 0000000000000010   X6: 000000001a7bb728
>      X5: 0000000070845200   X4: 0000000018a40d38   X3: 00000000707e8f98
>      X2: 000000001a2fa678   X1: 000000001a7ba198   X0: 0000000070847aa8
>     ORIG_X0: 00000000ffffff9c  SYSCALLNO: ffffffff  PSTATE: 60001000
>
> Signed-off-by: Guanyou.Chen <chenguanyou@xiaomi.com>
> ---
>  netdump.c | 15 +++++----------
>  1 file changed, 5 insertions(+), 10 deletions(-)
>
> diff --git a/netdump.c b/netdump.c
> index 435793b..455f90e 100644
> --- a/netdump.c
> +++ b/netdump.c
> @@ -101,7 +101,7 @@ map_cpus_to_prstatus(void)
>     nrcpus = (kt->kernel_NR_CPUS ? kt->kernel_NR_CPUS : NR_CPUS);
>
>     for (i = 0; i < nrcpus; i++) {
> -       if (in_cpu_map(ONLINE_MAP, i) && machdep->is_cpu_prstatus_valid(i)) {
> +       if (machdep->is_cpu_prstatus_valid(i)) {
>             nd->nt_prstatus_percpu[i] = nt_ptr[i];

This patch has dependency on your previous "bugfix map cpus register"
patch. I'm not sure about the relations of the 2 patches, but they
don't seem to be independent. So please send them within one patchset
is preferred.

However, for this patch, it will cause regressions after removing
in_cpu_map(ONLINE_MAP, i) check before
machdep->is_cpu_prstatus_valid(i), see the following stacktrace:

...
WARNING: cpu 2027: invalid NT_PRSTATUS note (n_type != NT_PRSTATUS)
WARNING: cpu 2028: invalid NT_PRSTATUS note (n_type != NT_PRSTATUS)
  malloc_bp[1999]: 585a3c0
      smallest: 32
       largest: 65536
      embedded: 2032
  max_embedded: 2032
       mallocs: 2000
         frees: 0
    reqs/total: 2063/837500
  average size: 406

crash: cannot allocate any more memory!
...
(gdb) bt
#0  getbuf (reqsize=368) at tools.c:6130
#1  0x000000000065be0b in have_crash_notes (cpu=2029) at diskdump.c:123
#2  0x000000000065bf57 in diskdump_is_cpu_prstatus_valid (cpu=2029) at
diskdump.c:155
#3  0x000000000064b055 in map_cpus_to_prstatus () at netdump.c:104
...

The reason is, kt->kernel_NR_CPUS might be large(5120 in this case),
without the filter of in_cpu_map(), it will exhaust the memory
buffer.

Thanks,
Tao Liu

>             nd->num_prstatus_notes =
>                 MAX(nd->num_prstatus_notes, i+1);
> @@ -2998,15 +2998,10 @@ dump_registers_for_elf_dumpfiles(void)
>         return;
>     }
>
> -        for (c = 0; c < kt->cpus; c++) {
> -       if (check_offline_cpu(c)) {
> -           fprintf(fp, "%sCPU %d: [OFFLINE]\n", c ? "\n" : "", c);
> -           continue;
> -       }
> -
> -                fprintf(fp, "%sCPU %d:\n", c ? "\n" : "", c);
> -                display_regs_from_elf_notes(c, fp);
> -        }
> +   for (c = 0; c < kt->cpus; c++) {
> +       fprintf(fp, "%sCPU %d: %s\n", c ? "\n" : "", c, check_offline_cpu(c) ? "[OFFLINE]" : "[ONLINE]");
> +       display_regs_from_elf_notes(c, fp);
> +   }
>  }
>
>  struct x86_64_user_regs_struct {
> --
> 2.34.1
>
> Guanyou.
> Thanks.