Date: Fri, 1 Dec 2023 18:19:49 -0500 From: George Mitchell <george+freebsd@m5p.com> To: FreeBSD Hackers <freebsd-hackers@FreeBSD.org> Subject: Radeon DRM kmod problem after LLVM upgrade!? Message-ID: <ffbfb81c-7545-4d6c-bb72-e63b78a09b7b@m5p.com>
next in thread | raw e-mail | index | archive | help
This afternoon after upgrading llvm15 from 15.0.7_5 to 15.0.7_7, Xorg no longer runs properly -- if I use drm-510-kmod instead of VESA. My display card is Raven Ridge [Radeon Vega Series / Radeon Vega Mobile Series] (USB id vendor=0x1002 device=0x15dd). When I boot in single-user mode, load the amdgpu.ko module from graphics/drm-510-kmod, and attempt to start Xorg, I get a pretty pattern of dots all over the screen, and Xorg reports (on the console): ac_rtld error: !data || data->d_size != shdr->sh_size LLVM failed to upload shader ac_rtld error: !data || data->d_size != shdr->sh_size LLVM failed to upload shader EE ../src/gallium/drivers/radeonsi/si_state_shaders.cpp:2505 si_build_shader_variant - Failed to build shader variant (type=0) ac_rtld error: !data || data->d_size != shdr->sh_size LLVM failed to upload shader EE ../src/gallium/drivers/radeonsi/si_state_shaders.cpp:2505 si_build_shader_variant - Failed to build shader variant (type=0) ac_rtld error: !data || data->d_size != shdr->sh_size LLVM failed to upload shader EE ../src/gallium/drivers/radeonsi/si_state_shaders.cpp:2505 si_build_shader_variant - Failed to build shader variant (type=0) ac_rtld error: !data || data->d_size != shdr->sh_size LLVM failed to upload shader EE ../src/gallium/drivers/radeonsi/si_state_shaders.cpp:2505 si_build_shader_variant - Failed to build shader variant (type=0) ac_rtld error: !data || data->d_size != shdr->sh_size LLVM failed to upload shader EE ../src/gallium/drivers/radeonsi/si_state_shaders.cpp:2505 si_build_shader_variant - Failed to build shader variant (type=0) Everything is fine if I use VESA mode, thankfully. My system has never been completely trouble-free with amd-510-kmod, but it has so far worked wonderfully well when it doesn't crash. But how that driver could interact with LLVM is beyond me. Should I try simply recompiling the module with the latest LLVM? -- George
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?ffbfb81c-7545-4d6c-bb72-e63b78a09b7b>