From mboxrd@z Thu Jan 1 00:00:00 1970 From: Robin Holt Date: Mon, 03 Nov 2008 11:46:39 +0000 Subject: Re: [PATCH V2 0/2] enable irqs when waiting for rwlocks Message-Id: <20081103114638.GD8483@sgi.com> List-Id: References: <1224777451.11530.52.camel@elijah.suse.cz> <1225443142.6825.4.camel@elijah.suse.cz> <57C9024A16AD2D4C97DC78E552063EA35A7ED9D3@orsmsx505.amr.corp.intel.com> In-Reply-To: <57C9024A16AD2D4C97DC78E552063EA35A7ED9D3@orsmsx505.amr.corp.intel.com> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: "Luck, Tony" Cc: Petr Tesarik , "linux-kernel@vger.kernel.org" , "linux-ia64@vger.kernel.org" On Fri, Oct 31, 2008 at 01:22:56PM -0700, Luck, Tony wrote: > > Any comments on my second patch series? Not even an Acked-by? Dislike of > > the concept? Should I post it again? > > Better response to interrupts is good, but it comes at the > cost of longer latency acquiring the lock (in the case where > an interrupt happens while we are waiting for the lock, and > the lock is freed while we are off executing the interrupt > handler). > > Any suggestions on how to measure the trade-off here? Possibly > it doesn't matter because this may only be significant when > the lock is heavily contended and you are probably aleady > hosed in this case. Just a few years of experiencal evidence. The equivalent of this patch has been in the SuSE SLES10 kernel for years and not been detected as being negative. Sorry I don't have more detail. The person at SGI who first detected this problem has long since left, and actually passed away from an aneurysm a couple years ago. The first version of the patch was in our one-off kernel based on Redhat's 2.4 kernel. The patch was not pushed to SuSE and the community for SLES9. It was in SLES10, but I can not find our internal tracking tool's record of it (searching is failing me this morning). Without that, I have not been able to find why it was not pushed to the community. Thanks, Robin