Set module parameter gpu_recovery=1 for amdgpu

install.sh

@@ -264,6 +264,8 @@ for f in libexec_startup/*; do
 done
 install_data misc_etc_files/request-key.d/id_resolver.conf \
                                                           "$DESTDIR$sysconfdir/request-key.d/id_resolver.conf"
+install_data misc_etc_files/modprobe.d/amdgpu_gpu_recovery.conf \
+                                                          "$DESTDIR$sysconfdir/modprobe.d/amdgpu_gpu_recovery.conf"
 
 postinstall
 exit

misc_etc_files/modprobe.d/amdgpu_gpu_recovery.conf

@@ -0,0 +1,18 @@
+# Sometimes we have this on "fenchurch":
+#
+#    kernel: [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx timeout, signaled seq=16634335, emitted seq=16634337
+#    kernel: [drm:amdgpu_job_timedout [amdgpu]] *ERROR* Process information: process Xorg pid 718 thread Xorg:cs0 pid 719
+#    kernel: amdgpu 0000:01:00.0: amdgpu: GPU recovery disabled.
+#
+# From drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c:
+#
+#    **
+#    * DOC: gpu_recovery (int)
+#    * Set to enable GPU recovery mechanism (1 = enable, 0 = disable). The default is -1 (auto, disabled except SRIOV).
+#    */
+#    MODULE_PARM_DESC(gpu_recovery, "Enable GPU recovery mechanism, (2 = advanced tdr mode, 1 = enable, 0 = disable, -1 = auto)");
+#    module_param_named(gpu_recovery, amdgpu_gpu_recovery, int, 0444);
+#
+# So give it a try
+
+options amdgpu gpu_recovery=1