<div dir="ltr"><div>I just pushed this patch: <a href="https://patchwork.freedesktop.org/patch/158551/">https://patchwork.freedesktop.org/patch/158551/</a> </div>Which is the "correct" fix. I'll send out an ISL patch shortly that adds a bunch more asserts and should make this sort of thing easier to catch in the future. </div><div class="gmail_extra"> <div class="gmail_quote">On Mon, Jun 5, 2017 at 2:55 AM, Eero Tamminen <<a href="mailto:eero.t.tamminen@intel.com" target="_blank">eero.t.tamminen@intel.com</a>> wrote: <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Hi, On 04.06.2017 22:26, Kenneth Graunke wrote: <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"> Since commit e9f5004d5ea38a21019e0affcd63d353b98d499a, KDE's Plasma Shell started crashing when intel_disable_rb_aux_buffer tried to dereference irb->mt, and there was no miptree. This happened during intel_texture_view_requires_resolve, coming from a glClear(). </blockquote> That's not the only thing dying because of ISL_FORMAT_UNSUPPORTED. Most of the benchmarks are currently crashing on GEN9. GpuTest 0.7, GLBenchmark 2.7, Lightsmark2008, Unigine Heaven & Valley, SynMark and several GfxBench v4 test-cases (T-Rex, driver, driver2, blending, fill) crash because of ISL_FORMAT_UNSUPPORTED: ------------- Program received signal SIGSEGV, Segmentation fault. 0x00007ffff24c35a1 in isl_format_supports_ccs_e (devinfo=0x7ce998, format=ISL_FORMAT_UNSUPPORTED) at ../../../src/intel/isl/isl_format.c:536 536 if (!format_info[format].exists) ------------- If latest Mesa is installed to the system instead of one running just the test-case it with, one gets for some of those BadWindow errors instead, and Ubuntu compiz also crashes. - Eero<div class="HOEnZb"><div class="h5"> <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"> Apparently intel_tex->_Format is MESA_FORMAT_NONE. We used to translate MESA_FORMAT_NONE to 0, which happens to be ISL_FORMAT_R32G32B32A32_FLOAT, which happens to support CCS_E, so we assumed no resolves were needed. That patch began translating MESA_FORMAT_NONE as ISL_FORMAT_UNSUPPORTED, which claims not to support CCS_E, so we thought we needed resolves, even though the buffer was in a bad state. For now, detect intel_tex->_Format == MESA_FORMAT_NONE and bail, like we used to. This is an ugly band-aid fix, but that's okay, given that Jason has patches on the mailing list to rewrite the resolve code. This makes KDE usable again. --- src/mesa/drivers/dri/i965/brw_context.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/src/mesa/drivers/dri/i965/brw_context.c b/src/mesa/drivers/dri/i965/brw_context.c index 9878cae9d4b..d8219cb57e0 100644 --- a/src/mesa/drivers/dri/i965/brw_context.c +++ b/src/mesa/drivers/dri/i965/brw_context.c @@ -205,6 +205,9 @@ intel_texture_view_requires_resolve(struct brw_context *brw, !intel_miptree_is_lossless_compressed(brw, intel_tex->mt)) return false; + if (intel_tex->_Format == MESA_FORMAT_NONE) + return false; + const enum isl_format isl_format = brw_isl_format_for_mesa_format(intel_tex->_Format); </blockquote> </div></div><div class="HOEnZb"><div class="h5"> _______________________________________________ mesa-dev mailing list <a href="mailto:mesa-dev@lists.freedesktop.org" target="_blank">mesa-dev@lists.freedesktop.org</a> <a href="https://lists.freedesktop.org/mailman/listinfo/mesa-dev" rel="noreferrer" target="_blank">https://lists.freedesktop.org/mailman/listinfo/mesa-dev</a> </div></div></blockquote></div> </div>