On Thursday, May 15, 2008 10:10 am Arnaldo Carvalho de Melo wrote: > Em Thu, May 15, 2008 at 11:04:07AM -0600, Matthew Wilcox escreveu: > > On Thu, May 15, 2008 at 01:04:26PM -0300, Arnaldo Carvalho de Melo wrote: > > > So I implemented pci_find_capability_cached and made MSI use it > > > for good measure, please consider applying. > > > > As I told you on IRC, this is just the MSI code being complete crap. > > It should be caching the offset itself. We shouldn't have this extra > > array in the struct pci_dev just because MSI is broken. > > Well, we can certainly do that, its just that I did this first and > thought that perhaps there could be some other users, but I see that 44 > extra bytes per pci_dev can be a pain if the only one to reap benefits > is MSI, can't you think of any other users? I couldn't detect any so far > in my admitedly limited testing. There are a few other common cap checks, but I don't think they compare to MSI in terms of latency sensitivity (though I didn't audit all the CAP_ID_EXP checks, there are quite a few of those). Since we know MSI is a problem, let's just go with fixing that for now. If we find that other caps are also causing problems we can revisit caching all of them; the patch is simple enough. Thanks, Jesse -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
