Hello, On Tue, Oct 08, 2013 at 02:22:16PM +0200, Alexander Gordeev wrote: > If we talk about pSeries quota, then the current pSeries pci_enable_msix() > implementation is racy internally and could fail if the quota went down > *while* pci_enable_msix() is executing. In this case the loop will have to > exit rather than retry with a lower number (what number?).
Ah, okay, so that one is already broken. > In this regard the new scheme does not bring anything new and relies on > the fact this race does not hit and therefore does not worry. > > If we talk about quota as it has to be, then yes - the loop scheme seems > more preferable. > > Overall, looks like we just need to fix the pSeries implementation, > if the guys want it, he-he :) If we can't figure out a better interface for the retry case, I think what can really help is having a simple interface for the simpler cases. > > The problem case is where multiple msi(x) allocation fails completely > > because the global limit went down before inquiry and allocation. In > > the loop based interface, it'd retry with the lower number. > > I am probably missing something here. If the global limit went down before > inquiry then the inquiry will get what is available and try to allocate with > than number. Oh, I should have written between inquiry and allocation. Sorry. Thanks. -- tejun ------------------------------------------------------------------------------ October Webinars: Code for Performance Free Intel webinars can help you accelerate application performance. Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from the latest Intel processors and coprocessors. See abstracts and register > http://pubads.g.doubleclick.net/gampad/clk?id=60134071&iu=/4140/ostg.clktrk _______________________________________________ E1000-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/e1000-devel To learn more about Intel® Ethernet, visit http://communities.intel.com/community/wired
