CPU underload
Adrian Chadd
adrian.chadd at gmail.com
Sat Oct 31 12:52:18 UTC 2015
hi,
there's a mips callchain kernel option. Sorry, I'm on my mac atm so I
can't tell you. look in sys/conf/options for callchain, callgraph,
etc. It's a build option.
Just rebuild your kernel with that and check.
I haven't tried with fastforward. What else are you doing? pf? ipfw?
-adrian
On 31 October 2015 at 04:56, Eugene Grosbein <eugen at grosbein.net> wrote:
> On 31.10.2015 15:48, Eugene Grosbein wrote:
>> Hi!
>>
>> I'm trying to get familiar with PMC and uncover reason of MIPS32
>> CPU being underloaded - about 50% - while performing traffic forwarding.
>>
>> It seems, kernel code abuses DELAY function as "pmcstat -S instructions -T" shows:
>
> [skip]
>
>> And why so heavy usage of _libmd_SHA256_Transf ?
>
> Well, it seems to be kern.random.harvest is guity for SHA256 abuse.
> I've set sysctl kern.random.harvest.mask=0 and now it' gone:
>
> PMC: [INSTR_EXECUTED] Samples: 7490 (100.0%) , 0 unresolved
>
> %SAMP IMAGE FUNCTION CALLERS
> 4.6 kernel ip_fastforward
> 4.5 kernel memcpy
> 4.4 kernel DELAY
> 3.9 kernel arge_start_locked
> 3.8 kernel _bus_dmamap_load_buf
> 3.2 kernel ether_nh_input
> 3.1 kernel if_inc_counter
> 3.0 kernel arge_intr
> 3.0 kernel ether_output
> 2.9 kernel _bus_dmamap_sync
> 2.6 kernel bzero
> 2.6 kernel critical_enter
> 2.4 kernel critical_exit
> 2.3 kernel bus_dmamap_sync_buf
> 2.2 kernel _rw_runlock_cookie
> 2.1 kernel rn_match
> 1.9 kernel __rw_rlock
> 1.7 kernel netisr_dispatch_src
> 1.5 kernel spinlock_exit
> 1.5 kernel uma_zalloc_arg
> 1.5 kernel generic_bs_barrier
> 1.4 kernel _bus_dmamap_load_mbu
> 1.4 kernel arge_newbuf
> 1.4 kernel _rm_rlock
> 1.4 kernel m_adj
> 1.4 libpmc.so. pmclog_read
> 1.4 kernel mipsNN_pdcache_wb_ra
> 1.4 kernel arpresolve
> 1.3 kernel bcmp
> 1.3 kernel ether_demux
> 1.3 kernel mipsNN_pdcache_inv_r
> 1.3 kernel _bus_dmamap_unload
> 1.2 kernel uma_zfree_arg
> 1.2 kernel if_transmit
> 1.2 kernel bus_dmamap_load_mbuf
> 1.1 kernel _rm_runlock
> 1.0 kernel rtfree
> CONVERSION STATISTICS:oc1_fib
> #exec/elf 1
> #samples/total 102761
> #samples/unknown-function 2939
> #callchain/dubious-frames 23
>
> Still too many of DELAYs.
>
> _______________________________________________
> freebsd-mips at freebsd.org mailing list
> https://lists.freebsd.org/mailman/listinfo/freebsd-mips
> To unsubscribe, send any mail to "freebsd-mips-unsubscribe at freebsd.org"
More information about the freebsd-mips
mailing list