Quoting Daniele Ceraolo Spurio (2019-08-16 19:43:47) > > > On 8/16/19 10:58 AM, Chris Wilson wrote: > > Add a redzone to our context image and check the HW does not write into > > after a context save, to verify that we have the correct context size. > > (This does vary with feature bits, so test with a live setup that should > > match how we run userspace.) > > > > On newer gens the data saved during the ctx switch is variable, based on > the type of switch and the current state of the HW, e.g. some state is > only saved during a preemption, and the ctx layout is compressed > accordingly. We'd need the test to generate the maximum possible size > (and I have no idea which usage case produces that), but I don't think > that'd scale well from gen to gen. I'd take this as a starting point, and we can definitely generate preemption events easily etc -- it's only when it start depending on state set by userspace do we run into logistical problems (I can only dread encountering such a bug in the wild). So v2, add a redzone everywhere and check on context unpinning. -Chris _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
