lkml.org 
[lkml]   [2005]   [Sep]   [25]   [last100]   RSS Feed
Views: [more markup]   [less markup]   [headers]   [forward]  
 
Messages in this thread
/
DateSun, 25 Sep 2005 21:22:47 +1000
FromNick Piggin <>
SubjectRe: [PATCH linux-2.6 01/04] brsem: implement big reader semaphore
Tejun Heo wrote:

>  I did a busy-loop microbenchmark, and I think it's informative enough. :-)
> 

Great!

>  The following command is run on three versions - vanilla version, one 
> with read_down/up(->s_umount) added to vfs_write(), and one with 
> brsem_read_down/up(->s_umount) added to vfs_write().
> 
> # time -p dd if=/dev/zero of=out bs=32 count=10M
> 
>  The test is run three times and the results are averaged.
> 
> a. vanilla
> 
> real 58.63
> user 5.61
> sys 52.37
> 
> b. rwsem
> 
> real 59.24
> user 6.06
> sys 52.29
> 
> c. brsem
> 
> real 61.74
> user 5.78
> sys 55.04
> 
>  I don't think brsem has any chance of being faster than rwsem if it's 
> slower in this micro benchmark.  One weird thing is that the result of 
> rwsem consistently showed higher user time and lower system time than 
> vanilla (no synchronization) case, maybe oprofiling will tell something.
> 

Yep, probably just some timing or cache anomaly, generally just
sum the user and system time in the case where you are running
identical userspace code... I wouldn't worry too much about it.

>  Anyways, you were absolutely right.  My brsem was a pretty stupid idea 
> after all.  Let's hope at least I learned something from it.  :-(
> 

I wouldn't say stupid. Implementation was too complex, but some
clever ideas were required to solve the problems you identified.

There are definitely a couple of possible places where a brsem
may be useful, and I think cpucontrol semaphore is one of those.

Al can probably comment on its use in the superblock. It seems
fair enough, though I think we'll want to slim down my naive
rwsem array and maybe think about using alloc_percpu.

Is the remount case much rarer than mount/unmount? Do we need to
be careful about bloating the size of the superblock on large
machines? In that case maybe a global remount brsem will be
enough to handle this race?

-- 
SUSE Labs, Novell Inc.

Send instant messages to your online friends http://au.messenger.yahoo.com 

-
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/

\
 
 \ /
  Last update: 2005-09-25 11:30    [from the cache]
©2003-2008