Re: 2.6.23 spinlock hang in kswapd under heavy disk write loads

[Nick Piggin <nickpiggin@yahoo.com.au>]

On Thursday 11 October 2007 01:33, Berkley Shands wrote:
> 2.6.23 with CONFIG_DEBUG_SPINLOCK on does not hang under very high write
> loads to either an LSI8888ELP (write rate 1.1GB/Sec) or to a highpoint
> RR2340 (write rate 1.0GB/Sec). With CONFIG_DEBUG_SPINLOCK off however, the
> system hangs with kswapd getting 100% of the cpu and most if not all other
> processes are locked out. Sometimes even the keyboard is locked out.
>
> This is seen on a supermicro H8DM3-2 and H8DMi-2 motherboards, 16 GB RAM,
> 2222 SE or 2216 processors. Not seen on a Uniwide 3546ES, or on a
> Supermicro H8DM8 with 8222 CPUS.
>
> kswapd sits at (according to KDB)
> __spinlock_irqsave + 0x15
> prepare_to_wait + 0x15
> kswapd + 0xe3
> kthread + 0x47


This is the pgdat->kswapd_wait waitqueue lock, by the looks?
It should be basically impossible to get this lock wrong, so
perhaps what happens is some memory corruption which is padded
(or some other random change) by spinlock debugging?

Is this very reproducable? Does your dmesg have any funny
messages after performing this heavy write load with
CONFIG_DEBUG_SPINLOCK set?

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