RE: Regression in reading /proc/stat in the newer kernels withlarge SMP and NUMA configurations

[Eric Dumazet]

Le vendredi 14 octobre 2011 Ã 17:27 +0100, Oberman, Laurence (HAS GSE) a
Ãcrit :
> Eric,
> 
> Got the system ready and ran the tests.
> 
> This is a DL980 128 CPCUS, 256GB memory.
> 
> [root@linuxbeast ~]# cat /proc/cpuinfo | grep processor | wc -l
> 128
> 
> [root@linuxbeast ~]# uname -a
> Linux linuxbeast.gse.mvlabs.corp.hp.com 2.6.32-131.6.1.el6.x86_64 #1 SMP Mon Jun 20 14:15:38 EDT 2011 x86_64 x86_64 x86_64 GNU/Linux
> 
> In this example we take >4ms per read. We are spending most of the time in 
> 
>     35.09%        t  [kernel.kallsyms]  [k] kstat_irqs_cpu
>     32.13%        t  [kernel.kallsyms]  [k] find_next_bit
>     24.32%        t  [kernel.kallsyms]  [k] show_stat
>      1.98%        t  [kernel.kallsyms]  [k] vsnprintf
>      1.63%        t  [kernel.kallsyms]  [k] format_decode
>      1.23%        t  [kernel.kallsyms]  [k] number

Hmm, have you compiled your kernel with NR_CPUS=4096 ???

--
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/