drm/i915/gt: Fix reservation address in ggtt_reserve_guc_top
authorJavier Pello <devel@otheo.eu>
Sat, 2 Sep 2023 15:10:39 +0000 (17:10 +0200)
committerRodrigo Vivi <rodrigo.vivi@intel.com>
Tue, 26 Sep 2023 17:58:12 +0000 (13:58 -0400)
There is an assertion in ggtt_reserve_guc_top that the global GTT
is of size at least GUC_GGTT_TOP, which is not the case on a 32-bit
platform; see commit 562d55d991b39ce376c492df2f7890fd6a541ffc
("drm/i915/bdw: Only use 2g GGTT for 32b platforms"). If GEM_BUG_ON
is enabled, this triggers a BUG(); if GEM_BUG_ON is disabled, the
subsequent reservation fails and the driver fails to initialise
the device:

i915 0000:00:02.0: [drm:i915_init_ggtt [i915]] Failed to reserve top of GGTT for GuC
i915 0000:00:02.0: Device initialization failed (-28)
i915 0000:00:02.0: Please file a bug on drm/i915; see https://gitlab.freedesktop.org/drm/intel/-/wikis/How-to-file-i915-bugs for details.
i915: probe of 0000:00:02.0 failed with error -28

Make the reservation at the top of the available space, whatever
that is, instead of assuming that the top will be GUC_GGTT_TOP.

Fixes: 911800765ef6 ("drm/i915/uc: Reserve upper range of GGTT")
Link: https://gitlab.freedesktop.org/drm/intel/-/issues/9080
Signed-off-by: Javier Pello <devel@otheo.eu>
Reviewed-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Fernando Pacheco <fernando.pacheco@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: intel-gfx@lists.freedesktop.org
Cc: stable@vger.kernel.org # v5.3+
Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230902171039.2229126186d697dbcf62d6d8@otheo.eu
(cherry picked from commit 0f3fa942d91165c2702577e9274d2ee1c7212afc)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
drivers/gpu/drm/i915/gt/intel_ggtt.c

index dd0ed94..da21f27 100644 (file)
@@ -511,20 +511,31 @@ void intel_ggtt_unbind_vma(struct i915_address_space *vm,
        vm->clear_range(vm, vma_res->start, vma_res->vma_size);
 }
 
+/*
+ * Reserve the top of the GuC address space for firmware images. Addresses
+ * beyond GUC_GGTT_TOP in the GuC address space are inaccessible by GuC,
+ * which makes for a suitable range to hold GuC/HuC firmware images if the
+ * size of the GGTT is 4G. However, on a 32-bit platform the size of the GGTT
+ * is limited to 2G, which is less than GUC_GGTT_TOP, but we reserve a chunk
+ * of the same size anyway, which is far more than needed, to keep the logic
+ * in uc_fw_ggtt_offset() simple.
+ */
+#define GUC_TOP_RESERVE_SIZE (SZ_4G - GUC_GGTT_TOP)
+
 static int ggtt_reserve_guc_top(struct i915_ggtt *ggtt)
 {
-       u64 size;
+       u64 offset;
        int ret;
 
        if (!intel_uc_uses_guc(&ggtt->vm.gt->uc))
                return 0;
 
-       GEM_BUG_ON(ggtt->vm.total <= GUC_GGTT_TOP);
-       size = ggtt->vm.total - GUC_GGTT_TOP;
+       GEM_BUG_ON(ggtt->vm.total <= GUC_TOP_RESERVE_SIZE);
+       offset = ggtt->vm.total - GUC_TOP_RESERVE_SIZE;
 
-       ret = i915_gem_gtt_reserve(&ggtt->vm, NULL, &ggtt->uc_fw, size,
-                                  GUC_GGTT_TOP, I915_COLOR_UNEVICTABLE,
-                                  PIN_NOEVICT);
+       ret = i915_gem_gtt_reserve(&ggtt->vm, NULL, &ggtt->uc_fw,
+                                  GUC_TOP_RESERVE_SIZE, offset,
+                                  I915_COLOR_UNEVICTABLE, PIN_NOEVICT);
        if (ret)
                drm_dbg(&ggtt->vm.i915->drm,
                        "Failed to reserve top of GGTT for GuC\n");