Hi Lucas,

Thanks for the fix.

On Fri, Apr 24, 2026 at 9:17 AM Lucas Oakley <soakley@redhat.com> wrote:
>
>     In RHEL 9.4 ppc64le, the stack size was adjusted from 16k to
>     32k. As a result, ppc64_back_trace() can bail prematurely when
>     checking if the stack pointer exists in the range of the
>     range of the irq stacks, since SIZE(irq_ctx), used in
>     ppc64_in_irqstack(), is set to 16k. This patch ensures that
>     irq_ctx is updated to 32k if a 32k stack size is used.
>
>     Tested against:
>     el6    x86_64
>     el7    ppc64le s390x x86_64
>     el8    aarch64 ppc64le s390x x86_64
>     el9    aarch64 ppc64le s390x x86_64
>     el10   aarch64 ppc64le s390x x86_64
>
>     Without the commit:
>
>     crash> bt -c 3
>     PID: 17524    TASK: c0000000b2c0e400  CPU: 3    COMMAND: "xyz"
>     cannot find the stack info.
>
>     With the commit:
>
>     crash> bt -c 3
>     PID: 17524    TASK: c0000000b2c0e400  CPU: 3    COMMAND: "xyz"
>      #0 [c000001dff7d7c10] smp_call_function_single_async at c00000000028dd38
>      #1 [c000001dff7d7d30] _raw_spin_lock_irqsave at c000000001023f1c
>      #2 [c000001dff7d7d60] ibmvscsi_handle_crq at c0080000044635ec [ibmvscsi]
>      #3 [c000001dff7d7de0] ibmvscsi_task at c008000004463804 [ibmvscsi]
>      #4 [c000001dff7d7e30] tasklet_action_common.constprop.0 at c0000000001624cc
>      #5 [c000001dff7d7e90] __do_softirq at c0000000010244cc
>      #6 [c000001dff7d7f90] do_softirq_own_stack at c000000000016480
>      #7 [c000000140d67700] __irq_exit_rcu at c0000000001613b8
>      #8 [c000000140d67730] irq_exit at c000000000162170
>      #9 [c000000140d67750] do_IRQ at c000000000015fa4
>      #10 [c000000140d67780] hardware_interrupt_common_virt at c000000000009080
>      Hardware Interrupt [500] exception frame:
>      R0:  c000000001023de0    R1:  c000000140d67a90    R2:  c000000002c02500
>      R3:  c00800000b30269c    R4:  0000000000000001    R5:  0000000000000001
>      R6:  ffffffffffffffff    R7:  0000000000000000    R8:  0000000000000000
>      R9:  fffffffffffe0000    R10: 0000000000000002    R11: 0000000048422824
>      R12: c000000001023d70    R13: c000001dffffd480    R14: 00007ffd6ef6d238
>      R15: 0000000000000028    R16: c000001dfc1e2280    R17: c000001dfc1e2280
>      R18: 00000000ab97fa48    R19: 0000000000000000    R20: 0000000000000001
>      R21: 0000001df9ff0000    R22: 0000000000000028    R23: c0000000021f2280
>      R24: 0000000000000000    R25: c0000000021f2280    R26: c0000000021f2380
>      R27: 0000000000000000    R28: c00800000b30269c    R29: 0000000000000000
>      R30: c000000002c47190    R31: 000000000020000b
>      NIP: c0000000000aea14    MSR: 800000000280b033    OR3: c0000000000ae944
>      CTR: c000000001023d70    LR:  c000000001023de0    XER: 0000000020040001
>      CCR: 0000000088422824    MQ:  0000000000000000    DAR: 0000000000000001
>      DSISR: c0080000073b1e94     Syscall Result: 0000000000000000
>      [NIP  : queued_spin_lock_slowpath+1204]
>      [LR   : _raw_spin_lock+112]
>      #11 [c000000140d67a90] queued_spin_lock_slowpath at c0000000000aea14
>      #12 [c000000140d67bb0] _raw_spin_lock at c000000001023de0  (unreliable)
>      #13 [c000000140d67bd0] dm_blk_close at c00800000b2c6850 [dm_mod]
>      #14 [c000000140d67c00] blkdev_put_whole at c0000000007d2738
>      #15 [c000000140d67c30] bdev_release at c0000000007d3a38
>      #16 [c000000140d67c90] blkdev_release at c0000000007d4224
>      #17 [c000000140d67cb0] __fput at c0000000005d2e98
>      #18 [c000000140d67d00] task_work_run at c00000000018fb14
>      #19 [c000000140d67d50] do_notify_resume at c000000000020bd4
>      #20 [c000000140d67d80] interrupt_exit_user_prepare_main at c00000000002ed98
>      #21 [c000000140d67de0] syscall_exit_prepare at c00000000002f240
>      #22 [c000000140d67e10] system_call_vectored_common at c00000000000bff4
>
> Signed-off-by: Lucas Oakley <soakley@redhat.com>
> ---
>  task.c | 12 ++++++++++++
>  1 file changed, 12 insertions(+)
>
> diff --git a/task.c b/task.c
> index ec04b55..a398a8d 100644
> --- a/task.c
> +++ b/task.c
> @@ -739,6 +739,18 @@ irqstacks_init(void)
>         if (!(tt->softirq_tasks = (ulong *)calloc(NR_CPUS, sizeof(ulong))))
>                 error(FATAL, "cannot malloc softirq_tasks space.");
>
> +       /*
> +        *  RHEL 9.4.z adjusted the stack size from 16k to 32k for
> +        *  ppc64le only. We need to ensure that SIZE(irq_ctx) is
> +        *  correctly set so the unwinder doesn't prematurely bail
> +        *  when switching between the kernel stack and irq stacks.
> +        *  The stack size is updated in task_init(), which calls
> +        *  this routine, irqstacks_init() after checking for the
> +        *  existence of irq_ctx.
> +        */
> +       if ((SIZE(irq_ctx) != -1) && (STACKSIZE() > SIZE(irq_ctx)))
> +               ASSIGN_SIZE(irq_ctx) = STACKSIZE();
> +
I think the (SIZE(irq_ctx) != -1) check is not needed. Because when
entering the function, VALID_STRUCT(irq_ctx) already ensure
size_table.irq_ctx >= 0, so (SIZE(irq_ctx) != -1) is always true. I
have no other concerns for the patch.

Thanks,
Tao Liu

>         thread_info_buf = GETBUF(SIZE(irq_ctx));
>
>         if ((hard_sp = per_cpu_symbol_search("per_cpu__hardirq_ctx")) ||
> --
> 2.52.0
> --
> Crash-utility mailing list -- devel@lists.crash-utility.osci.io
> To unsubscribe send an email to devel-leave@lists.crash-utility.osci.io
> https://${domain_name}/admin/lists/devel.lists.crash-utility.osci.io/
> Contribution Guidelines: https://github.com/crash-utility/crash/wiki
>