Re: [RFC] eventpoll: Move a kmem_cache_alloc and kmem_cache_free

2013-09-24 Thread Nathan Zimmer
On Mon, Sep 23, 2013 at 11:47:41AM -0500, Nathan Zimmer wrote: > On Mon, Sep 23, 2013 at 11:17:39AM -0400, Jason Baron wrote: > > On 09/19/2013 12:37 PM, Nathan Zimmer wrote: > > > On 09/18/2013 02:09 PM, Jason Baron wrote: > > >> On 09/13/2013 11:54 AM, Nathan Zimmer wrote: > > >>> We noticed some

Re: [RFC] eventpoll: Move a kmem_cache_alloc and kmem_cache_free

2013-09-23 Thread Nathan Zimmer
On Mon, Sep 23, 2013 at 11:17:39AM -0400, Jason Baron wrote: > On 09/19/2013 12:37 PM, Nathan Zimmer wrote: > > On 09/18/2013 02:09 PM, Jason Baron wrote: > >> On 09/13/2013 11:54 AM, Nathan Zimmer wrote: > >>> We noticed some scaling issue in the SPECjbb benchmark. Running perf > >>> we found tha

Re: [RFC] eventpoll: Move a kmem_cache_alloc and kmem_cache_free

2013-09-23 Thread Jason Baron
On 09/22/2013 04:41 PM, Eric Wong wrote: > Jason Baron wrote: >> epoll: reduce usage of global 'epmutex' lock >> >> Epoll file descriptors that are 1 link from a wakeup source and >> are not nested within other epoll descriptors, or pointing to >> other epoll descriptors, don't nee

Re: [RFC] eventpoll: Move a kmem_cache_alloc and kmem_cache_free

2013-09-23 Thread Jason Baron
On 09/19/2013 12:37 PM, Nathan Zimmer wrote: > On 09/18/2013 02:09 PM, Jason Baron wrote: >> On 09/13/2013 11:54 AM, Nathan Zimmer wrote: >>> We noticed some scaling issue in the SPECjbb benchmark. Running perf >>> we found that the it was spending lots of time in SYS_epoll_ctl. >>> In particular

Re: [RFC] eventpoll: Move a kmem_cache_alloc and kmem_cache_free

2013-09-22 Thread Eric Wong
Jason Baron wrote: > epoll: reduce usage of global 'epmutex' lock > > Epoll file descriptors that are 1 link from a wakeup source and > are not nested within other epoll descriptors, or pointing to > other epoll descriptors, don't need to check for loop creation or > the creati

Re: [RFC] eventpoll: Move a kmem_cache_alloc and kmem_cache_free

2013-09-19 Thread Nathan Zimmer
On 09/18/2013 02:09 PM, Jason Baron wrote: On 09/13/2013 11:54 AM, Nathan Zimmer wrote: We noticed some scaling issue in the SPECjbb benchmark. Running perf we found that the it was spending lots of time in SYS_epoll_ctl. In particular it is holding the epmutex. This patch helps by moving out t

Re: [RFC] eventpoll: Move a kmem_cache_alloc and kmem_cache_free

2013-09-18 Thread Jason Baron
On 09/13/2013 11:54 AM, Nathan Zimmer wrote: > We noticed some scaling issue in the SPECjbb benchmark. Running perf > we found that the it was spending lots of time in SYS_epoll_ctl. > In particular it is holding the epmutex. > This patch helps by moving out the kmem_cache_alloc and kmem_cache_fre

[RFC] eventpoll: Move a kmem_cache_alloc and kmem_cache_free

2013-09-13 Thread Nathan Zimmer
We noticed some scaling issue in the SPECjbb benchmark. Running perf we found that the it was spending lots of time in SYS_epoll_ctl. In particular it is holding the epmutex. This patch helps by moving out the kmem_cache_alloc and kmem_cache_free out from under the lock. It improves throughput by