20:22 <mntirc> i have an apitrace .trace that generates MMU faults on playback with eglretrace on gc7000. interesting?

21:22 <mntirc> austriancoder: does it make sense that etnaviv would cause mmu faults when glVertexAttribPointer passes an unusual pointer?

21:43 <mntirc> hmm, i've boiled down the mmu fault trace to 5 operations

21:44 <mntirc> glGenBuffers(1, [1]); glBindBuffer(GL_ARRAY_BUFFER, 1); glEnableVertexAttribArray(0); glVertexAttribPointer(0,2,GL_FLOAT,GL_FALSE,0,0xaaab00819370); glDrawArrays(GL_TRIANGLE_FAN,0,4);

21:59 <mntirc> lol

22:00 <mntirc> i compiled a simple example program from the web to make a test case and it also mmu faults, even if i didn't do anything interesting yet

22:01 <mntirc> is it possible that glVertexAttribPointer() is broken?

22:17 <mntirc> ah. if the first argument to glVertexAttribPointer() is invalid, it will cause mmu faults or memory scribbling

22:17 <mntirc> i.e. it can trash textures of other applications

22:21 <mntirc> wow

22:22 <mntirc> i can cause weird kernel errors by doing sth like this: glVertexAttribPointer(0, 3, GL_FLOAT, GL_FALSE, 5 * sizeof(float), (void *)0xffff);

22:22 <mntirc> this causes > [ 5781.229602] Huh VM_FAULT_OOM leaked out to the #PF handler. Retrying PF

22:22 <mntirc> cc lynxeye austriancoder

22:51 <mntirc> issue filed https://gitlab.freedesktop.org/mesa/mesa/-/issues/8180

23:27 <mntirc> hmm, alyssa says this is kernel responsibility