| Summary: | [osmesa] Crash on postprocessing if z buffer is NULL | ||
|---|---|---|---|
| Product: | Mesa | Reporter: | Park, Jeongmin <pjm0616> |
| Component: | Other | Assignee: | mesa-dev |
| Status: | RESOLVED FIXED | QA Contact: | mesa-dev |
| Severity: | normal | ||
| Priority: | medium | ||
| Version: | git | ||
| Hardware: | All | ||
| OS: | All | ||
| Whiteboard: | |||
| i915 platform: | i915 features: | ||
| Attachments: |
patch #1
patch #2 |
||
Created attachment 113147 [details] [review] patch #1 Created attachment 113148 [details] [review] patch #2 Please send the patches to the mesa-dev mailing list if you haven't already. The first patch looks good. I posted something similar to mesa-dev yesterday. But the second patch doesn't seem right. There could be post-process stages that don't care about the depth/stencil buffer but we'd still want them to run. So instead, I think the pp_mlaa.c code should check for a missing/null depth/stencil buffer. Sorry about the first patch. I didn't see that. In dri_sw.c, it checks for z/s buffer: > if (ctx->pp && drawable->textures[ST_ATTACHMENT_DEPTH_STENCIL]) > pp_run(ctx->pp, ptex, ptex, drawable->textures[ST_ATTACHMENT_DEPTH_STENCIL]); and in dri_drawable.c: > if (ctx->pp && src && zsbuf) > pp_run(ctx->pp, src, src, zsbuf); Should I remove these checks and add the checks in pp_mlaa.cc instead? (In reply to Park, Jeongmin from comment #5) > Sorry about the first patch. I didn't see that. > > In dri_sw.c, it checks for z/s buffer: > > if (ctx->pp && drawable->textures[ST_ATTACHMENT_DEPTH_STENCIL]) > > pp_run(ctx->pp, ptex, ptex, drawable->textures[ST_ATTACHMENT_DEPTH_STENCIL]); > and in dri_drawable.c: > > if (ctx->pp && src && zsbuf) > > pp_run(ctx->pp, src, src, zsbuf); > > Should I remove these checks and add the checks in pp_mlaa.cc instead? Yes, I think that would be better. Fixed by 2e6ba6af and 0467a52d. |
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.
It seems that z buffer is not optional to pp_run looking at other codes but it is not checked in osmesa_st_framebuffer_flush_front(). It crashes if zsbuf is NULL with pp_jimenezmlaa. With pp_nored it doesn't crash, but if pipe_sampler_view.reference is not the first member in struct pipe_sampler_view, it would crash in pipe_sampler_view_reference at > if (pipe_reference_described(&(*ptr)->reference, &view->reference, > (debug_reference_descriptor)debug_describe_sampler_view))