On Fri, Sep 16, 2022 at 05:50:33PM +0530, Iddamsetty, Aravind wrote:
On 16-09-2022 02:09, Lucas De Marchi wrote:
Add some helpers: adjust_stolen(), request_smem_stolen_() and
init_reserved_stolen() that are now called by i915_gem_init_stolen() to
initialize each part of the Data Stolen Memory region. Main goal is to
split the reserved part, also known as WOPCM, as its calculation changes
often per platform.
This also fixes a bug in graphics version < 5 (in theory, not tested,
due to no machine available): it would bail out on stolen creation due
to "Stolen reserved area outside stolen memory". Other than that, no
change in behavior.
Signed-off-by: Lucas De Marchi <lucas.demarchi@xxxxxxxxx>
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_stolen.c b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
index c34065fe2ecc..0e57a6d81534 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
@@ -77,22 +77,26 @@ void i915_gem_stolen_remove_node(struct drm_i915_private *i915,
mutex_unlock(&i915->mm.stolen_lock);
}
-static int i915_adjust_stolen(struct drm_i915_private *i915,
- struct resource *dsm)
+static bool valid_stolen_size(struct resource *dsm)
+{
+ return dsm->start != 0 && dsm->end > dsm->start;
+}
+
+static int adjust_stolen(struct drm_i915_private *i915,
+ struct resource *dsm)
{
struct i915_ggtt *ggtt = to_gt(i915)->ggtt;
struct intel_uncore *uncore = ggtt->vm.gt->uncore;
- struct resource *r;
- if (dsm->start == 0 || dsm->end <= dsm->start)
+ if (!valid_stolen_size(dsm))
return -EINVAL;
/*
+ * Make sure we don't clobber the GTT if it's within stolen memory
+ *
* TODO: We have yet too encounter the case where the GTT wasn't at the
* end of stolen. With that assumption we could simplify this.
*/
-
- /* Make sure we don't clobber the GTT if it's within stolen memory */
if (GRAPHICS_VER(i915) <= 4 &&
!IS_G33(i915) && !IS_PINEVIEW(i915) && !IS_G4X(i915)) {
struct resource stolen[2] = {*dsm, *dsm};
@@ -131,10 +135,20 @@ static int i915_adjust_stolen(struct drm_i915_private *i915,
}
}
+ if (!valid_stolen_size(dsm))
+ return -EINVAL;
+
+ return 0;
+}
+
+static int request_smem_stolen(struct drm_i915_private *i915,
+ struct resource *dsm)
+{
+ struct resource *r;
+
/*
- * With stolen lmem, we don't need to check if the address range
- * overlaps with the non-stolen system memory range, since lmem is local
- * to the gpu.
+ * With stolen lmem, we don't need to request if the address range
replace /if/for
+ * since lmem is local to the gpu.
humn.. it seems I skip some words here.
With stolen lmem, we don't need to request system memory since the
stolen region is local to the gpu.
*/
if (HAS_LMEM(i915))
return 0;
@@ -392,39 +406,22 @@ static void icl_get_stolen_reserved(struct drm_i915_private *i915,
}
}
-static int i915_gem_init_stolen(struct intel_memory_region *mem)
+/*
+ * Initialize i915->dsm_reserved to contain the reserved space within the Data
+ * Stolen Memory. This is a range on the top of DSM that is reserved, not to
+ * be used by driver, so must be excluded from the region passed to the
+ * allocator later. In the spec this is also called as WOPCM.
+ *
+ * Our expectation is that the reserved space is at the top of the stolen
+ * region, as it has been the case for every platform, and *never* at the
+ * bottom, so the calculation here can be simplified.
+ */
+static int init_reserved_stolen(struct drm_i915_private *i915)
{
- struct drm_i915_private *i915 = mem->i915;
struct intel_uncore *uncore = &i915->uncore;
resource_size_t reserved_base, stolen_top;
- resource_size_t reserved_total, reserved_size;
-
- mutex_init(&i915->mm.stolen_lock);
-
- if (intel_vgpu_active(i915)) {
- drm_notice(&i915->drm,
- "%s, disabling use of stolen memory\n",
- "iGVT-g active");
- return 0;
- }
-
- if (i915_vtd_active(i915) && GRAPHICS_VER(i915) < 8) {
- drm_notice(&i915->drm,
- "%s, disabling use of stolen memory\n",
- "DMAR active");
- return 0;
- }
-
- if (resource_size(&mem->region) == 0)
- return 0;
-
- if (i915_adjust_stolen(i915, &mem->region))
- return 0;
-
- GEM_BUG_ON(i915->dsm.start == 0);
- GEM_BUG_ON(i915->dsm.end <= i915->dsm.start);
-
- i915->dsm = mem->region;
+ resource_size_t reserved_size;
+ int ret = 0;
stolen_top = i915->dsm.end + 1;
reserved_base = stolen_top;
@@ -453,19 +450,17 @@ static int i915_gem_init_stolen(struct intel_memory_region *mem)
} else if (GRAPHICS_VER(i915) >= 5 || IS_G4X(i915)) {
g4x_get_stolen_reserved(i915, uncore,
&reserved_base, &reserved_size);
+ } else {
+ /* No reserved region */
+ goto bail_out;
better to have a WARN_ON here about STOLEN region wrongly passed on the
region list.
see the follow up I sent on this patch. The bail out here should be outside
this else and is a normal condition. This is about the reserved space
within stolen (aka wopcm), and having it 0 is normal even for the
platforms above that have a function to read the stolen reserved region.
a WARN_ON based on the regions in device_info would be wrong as having
stolen doesn't mean there is a wopcm. Btw I think a great source of
confusion is using "reserved stolen" for this part because we then think
it's about reserving the stolen memory when in fact this is about a
region inside the stolen.
}
- /*
- * Our expectation is that the reserved space is at the top of the
- * stolen region and *never* at the bottom. If we see !reserved_base,
- * it likely means we failed to read the registers correctly.
- */
- if (!reserved_base) {
+ if (!reserved_base || reserved_base == stolen_top) {
drm_err(&i915->drm,
"inconsistent reservation %pa + %pa; ignoring\n",
&reserved_base, &reserved_size);
- reserved_base = stolen_top;
- reserved_size = 0;
+ ret = -EINVAL;
+ goto bail_out;
}
i915->dsm_reserved =
@@ -475,19 +470,55 @@ static int i915_gem_init_stolen(struct intel_memory_region *mem)
drm_err(&i915->drm,
"Stolen reserved area %pR outside stolen memory %pR\n",
&i915->dsm_reserved, &i915->dsm);
+ ret = -EINVAL;
+ goto bail_out;
+ }
+
+ return 0;
+
+bail_out:
+ i915->dsm_reserved =
+ (struct resource)DEFINE_RES_MEM(reserved_base, 0);
+
+ return ret;
+}
+
+static int i915_gem_init_stolen(struct intel_memory_region *mem)
+{
+ struct drm_i915_private *i915 = mem->i915;
+
+ mutex_init(&i915->mm.stolen_lock);
+
+ if (intel_vgpu_active(i915)) {
+ drm_notice(&i915->drm,
+ "%s, disabling use of stolen memory\n",
+ "iGVT-g active");
+ return 0;
+ }
+
+ if (i915_vtd_active(i915) && GRAPHICS_VER(i915) < 8) {
+ drm_notice(&i915->drm,
+ "%s, disabling use of stolen memory\n",
+ "DMAR active");
return 0;
}
+ if (adjust_stolen(i915, &mem->region))
+ return 0;
+
+ if (request_smem_stolen(i915, &mem->region))
+ return 0;
why do we want to subside the errors returned in adjust_stolen and
request_smem_stolen?
This was already there: i915_gem_init_stolen() was only returning 0.
This patch only makes it clearer since it's now a smaller function.
Basically this was the function doing the "log when
appropriate, but consider it non-fatal". On the next patch I move this
decision to the caller.
Lucas De Marchi
+
+ i915->dsm = mem->region;
+
+ if (init_reserved_stolen(i915))
similarly here.
Thanks,
Aravind.
+ return 0;
+
/* Exclude the reserved region from driver use */
- mem->region.end = reserved_base - 1;
+ mem->region.end = i915->dsm_reserved.start - 1;
mem->io_size = min(mem->io_size, resource_size(&mem->region));
- /* It is possible for the reserved area to end before the end of stolen
- * memory, so just consider the start. */
- reserved_total = stolen_top - reserved_base;
-
- i915->stolen_usable_size =
- resource_size(&i915->dsm) - reserved_total;
+ i915->stolen_usable_size = resource_size(&mem->region);
drm_dbg(&i915->drm,
"Memory reserved for graphics device: %lluK, usable: %lluK\n",
@@ -759,11 +790,6 @@ static int init_stolen_lmem(struct intel_memory_region *mem)
if (GEM_WARN_ON(resource_size(&mem->region) == 0))
return -ENODEV;
- /*
- * TODO: For stolen lmem we mostly just care about populating the dsm
- * related bits and setting up the drm_mm allocator for the range.
- * Perhaps split up i915_gem_init_stolen() for this.
- */
err = i915_gem_init_stolen(mem);
if (err)
return err;
