Fix "smp_targeted_tlb_shootdown: interrupts disabled" panics

[jtl]

Summary

We have seen several cases recently where we appear to get a double-fault: we have an original panic. Then, instead of writing the core to the dump device, the kernel has a second panic: "smp_targeted_tlb_shootdown: interrupts disabled".

This is an attempt to fix that second panic.

This should be safe since we can't notify other CPUs of the TLB shootdown when they're stopped anyway.

[kib]

I think that it might be unsafe to return from ddb into normal operations with this patch, if ddb activity caused any change in the mappings. Some ddb commands might allocate memory, e.g. coredumping.

You need to flush TLB somewhere in cpustop_handler_post(), most likely, to fix this.

[jtl]

Invalidate TLB when restarting other CPUs.

[jtl]

In D9786#201995, @kib wrote:

I think that it might be unsafe to return from ddb into normal operations with this patch, if ddb activity caused any change in the mappings. Some ddb commands might allocate memory, e.g. coredumping.

You need to flush TLB somewhere in cpustop_handler_post(), most likely, to fix this.

Ugh. You're right.

I should have known better. :-)

I think it should be fixed now...