arm64: fix infinite stacktrace
authorMikulas Patocka <mpatocka@redhat.com>
Thu, 14 Jun 2018 18:58:21 +0000 (14:58 -0400)
committerWill Deacon <will.deacon@arm.com>
Wed, 4 Jul 2018 17:34:23 +0000 (18:34 +0100)
I've got this infinite stacktrace when debugging another problem:
[  908.795225] INFO: rcu_preempt detected stalls on CPUs/tasks:
[  908.796176]  1-...!: (1 GPs behind) idle=952/1/4611686018427387904 softirq=1462/1462 fqs=355
[  908.797692]  2-...!: (1 GPs behind) idle=f42/1/4611686018427387904 softirq=1550/1551 fqs=355
[  908.799189]  (detected by 0, t=2109 jiffies, g=130, c=129, q=235)
[  908.800284] Task dump for CPU 1:
[  908.800871] kworker/1:1     R  running task        0    32      2 0x00000022
[  908.802127] Workqueue: writecache-writeabck writecache_writeback [dm_writecache]
[  908.820285] Call trace:
[  908.824785]  __switch_to+0x68/0x90
[  908.837661]  0xfffffe00603afd90
[  908.844119]  0xfffffe00603afd90
[  908.850091]  0xfffffe00603afd90
[  908.854285]  0xfffffe00603afd90
[  908.863538]  0xfffffe00603afd90
[  908.865523]  0xfffffe00603afd90

The machine just locked up and kept on printing the same line over and
over again. This patch fixes it.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
arch/arm64/kernel/stacktrace.c

index d5718a060672e1696618904c8844447daa042007..e160ca123da377aec69449d183ae48b1552c9fdc 100644 (file)
@@ -56,6 +56,9 @@ int notrace unwind_frame(struct task_struct *tsk, struct stackframe *frame)
        frame->fp = READ_ONCE_NOCHECK(*(unsigned long *)(fp));
        frame->pc = READ_ONCE_NOCHECK(*(unsigned long *)(fp + 8));
 
+       if (frame->fp <= fp)
+               return -EINVAL;
+
 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
        if (tsk->ret_stack &&
                        (frame->pc == (unsigned long)return_to_handler)) {