Re: Reverting warning for vmalloc and kmemcheck faults in NMI

[Andrew Morton]

On Thu, 30 Jan 2014 19:06:58 -0800 Michel Lespinasse <walken@xxxxxxxxxx> wrote:

> Hi,
> 
> Way back in 2010, Frederic added commit
> ebc8827f75954fe315492883eee5cb3f355d547d to warn us about cases where
> faults were incorrectly firing during NMI handling on x86, as the IRET
> from such faults would possibly trigger nested NMIs.
> 
> Later (2012), Salman added commit
> 28696f434fef0efa97534b59986ad33b9c4df7f8 to enable nested NMI
> handling. See http://lwn.net/Articles/484932/
> 
> So, I believe such faults nesting under NMI are not an issue anymore,
> and we could revert ebc8827f75954fe315492883eee5cb3f355d547d ?

Seems logical.

> Background: I'm asking this because at Google we like to dump memory
> regions pointed by registers in our arch_trigger_all_cpu_backtrace
> handler, and this occasionally causes the vmalloc_fault in_nmi()
> warning to fire.
> 

It seems odd that x86 arch_trigger_all_cpu_backtrace() uses NMI - why
didn't it use the regular old IPI?  If a CPU is stuck with interrupts
disable, the NMI watchdog will provide the diagnostic?

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/