Re: [Intel-gfx] [PATCH v5] drm/i915/mtl: enable local stolen memory
On Thu, Sep 29, 2022 at 05:16:58PM +0530, Aravind Iddamsetty wrote:
> As an integrated GPU, MTL does not have local memory and HAS_LMEM()
> returns false. However the platform's stolen memory is presented via
> BAR2 (i.e., the BAR we traditionally consider to be the GMADR on IGFX)
> and should be managed by the driver the same way that local memory is
> on dgpu platforms (which includes setting the "lmem" bit on page table
> entries). We use the term "local stolen memory" to refer to this
> model.
>
> The major difference from the traditional BAR2 (GMADR) is that
> the stolen area is mapped via the BAR2 while in the former BAR2 is an
> aperture into the GTT VA through which access are made into stolen area.
>
> BSPEC: 53098, 63830
>
> v2:
> 1. dropped is_dsm_invalid, updated valid_stolen_size check from Lucas
> (Jani, Lucas)
> 2. drop lmembar_is_igpu_stolen
> 3. revert to referring GFXMEM_BAR as GEN12_LMEM_BAR (Lucas)
>
> v3:(Jani)
> 1. rename get_mtl_gms_size to mtl_get_gms_size
> 2. define register for MMIO address
>
> v4:(Matt)
> 1. Use REG_FIELD_GET to read GMS value
> 2. replace the calculations with SZ_256M/SZ_8M
>
> v5: Include more details to commit message on how it is different from
> earlier platforms (Anshuman)
>
> Cc: Matt Roper
> Cc: Lucas De Marchi
> Cc: Jani Nikula
>
> Signed-off-by: CQ Tang
> Signed-off-by: Aravind Iddamsetty
> Original-author: CQ Tang
Reviewed-by: Matt Roper
> ---
> drivers/gpu/drm/i915/gem/i915_gem_stolen.c | 83 ++
> drivers/gpu/drm/i915/gt/intel_ggtt.c | 2 +-
> drivers/gpu/drm/i915/i915_drv.h| 3 +
> drivers/gpu/drm/i915/i915_reg.h| 4 ++
> 4 files changed, 76 insertions(+), 16 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
> b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
> index c5a4035c99cd..910086974454 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
> @@ -77,9 +77,9 @@ void i915_gem_stolen_remove_node(struct drm_i915_private
> *i915,
> mutex_unlock(&i915->mm.stolen_lock);
> }
>
> -static bool valid_stolen_size(struct resource *dsm)
> +static bool valid_stolen_size(struct drm_i915_private *i915, struct resource
> *dsm)
> {
> - return dsm->start != 0 && dsm->end > dsm->start;
> + return (dsm->start != 0 || HAS_BAR2_SMEM_STOLEN(i915)) && dsm->end >
> dsm->start;
> }
>
> static int adjust_stolen(struct drm_i915_private *i915,
> @@ -88,7 +88,7 @@ static int adjust_stolen(struct drm_i915_private *i915,
> struct i915_ggtt *ggtt = to_gt(i915)->ggtt;
> struct intel_uncore *uncore = ggtt->vm.gt->uncore;
>
> - if (!valid_stolen_size(dsm))
> + if (!valid_stolen_size(i915, dsm))
> return -EINVAL;
>
> /*
> @@ -135,7 +135,7 @@ static int adjust_stolen(struct drm_i915_private *i915,
> }
> }
>
> - if (!valid_stolen_size(dsm))
> + if (!valid_stolen_size(i915, dsm))
> return -EINVAL;
>
> return 0;
> @@ -149,8 +149,11 @@ static int request_smem_stolen(struct drm_i915_private
> *i915,
> /*
>* With stolen lmem, we don't need to request system memory for the
>* address range since it's local to the gpu.
> + *
> + * Starting MTL, in IGFX devices the stolen memory is exposed via
> + * BAR2 and shall be considered similar to stolen lmem.
>*/
> - if (HAS_LMEM(i915))
> + if (HAS_LMEM(i915) || HAS_BAR2_SMEM_STOLEN(i915))
> return 0;
>
> /*
> @@ -385,8 +388,6 @@ static void icl_get_stolen_reserved(struct
> drm_i915_private *i915,
>
> drm_dbg(&i915->drm, "GEN6_STOLEN_RESERVED = 0x%016llx\n", reg_val);
>
> - *base = reg_val & GEN11_STOLEN_RESERVED_ADDR_MASK;
> -
> switch (reg_val & GEN8_STOLEN_RESERVED_SIZE_MASK) {
> case GEN8_STOLEN_RESERVED_1M:
> *size = 1024 * 1024;
> @@ -404,6 +405,12 @@ static void icl_get_stolen_reserved(struct
> drm_i915_private *i915,
> *size = 8 * 1024 * 1024;
> MISSING_CASE(reg_val & GEN8_STOLEN_RESERVED_SIZE_MASK);
> }
> +
> + if (HAS_BAR2_SMEM_STOLEN(i915))
> + /* the base is initialized to stolen top so subtract size to
> get base */
> + *base -= *size;
> + else
> + *base = reg_val & GEN11_STOLEN_RESERVED_ADDR_MASK;
> }
>
> /*
> @@ -833,6 +840,29 @@ static const struct intel_memory_region_ops
> i915_region_stolen_lmem_ops = {
> .init_object = _i915_gem_object_stolen_init,
> };
>
> +static int mtl_get_gms_size(struct intel_uncore *uncore)
> +{
> + u16 ggc, gms;
> +
> + ggc = intel_uncore_read16(uncore, GGC);
> +
> + /* check GGMS, should be fixed 0x3 (8MB) */
> + if ((ggc & GGMS_MASK) != GGMS_MASK)
> + return -EIO;
> +
> + /* return valid GMS value, -EIO if invalid */
> + gms = REG_FIELD_GET(GMS_MASK, ggc);
> + switch (gms) {
> + case 0x0 ...
[Intel-gfx] [PATCH v5] drm/i915/mtl: enable local stolen memory
As an integrated GPU, MTL does not have local memory and HAS_LMEM()
returns false. However the platform's stolen memory is presented via
BAR2 (i.e., the BAR we traditionally consider to be the GMADR on IGFX)
and should be managed by the driver the same way that local memory is
on dgpu platforms (which includes setting the "lmem" bit on page table
entries). We use the term "local stolen memory" to refer to this
model.
The major difference from the traditional BAR2 (GMADR) is that
the stolen area is mapped via the BAR2 while in the former BAR2 is an
aperture into the GTT VA through which access are made into stolen area.
BSPEC: 53098, 63830
v2:
1. dropped is_dsm_invalid, updated valid_stolen_size check from Lucas
(Jani, Lucas)
2. drop lmembar_is_igpu_stolen
3. revert to referring GFXMEM_BAR as GEN12_LMEM_BAR (Lucas)
v3:(Jani)
1. rename get_mtl_gms_size to mtl_get_gms_size
2. define register for MMIO address
v4:(Matt)
1. Use REG_FIELD_GET to read GMS value
2. replace the calculations with SZ_256M/SZ_8M
v5: Include more details to commit message on how it is different from
earlier platforms (Anshuman)
Cc: Matt Roper
Cc: Lucas De Marchi
Cc: Jani Nikula
Signed-off-by: CQ Tang
Signed-off-by: Aravind Iddamsetty
Original-author: CQ Tang
---
drivers/gpu/drm/i915/gem/i915_gem_stolen.c | 83 ++
drivers/gpu/drm/i915/gt/intel_ggtt.c | 2 +-
drivers/gpu/drm/i915/i915_drv.h| 3 +
drivers/gpu/drm/i915/i915_reg.h| 4 ++
4 files changed, 76 insertions(+), 16 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
index c5a4035c99cd..910086974454 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
@@ -77,9 +77,9 @@ void i915_gem_stolen_remove_node(struct drm_i915_private
*i915,
mutex_unlock(&i915->mm.stolen_lock);
}
-static bool valid_stolen_size(struct resource *dsm)
+static bool valid_stolen_size(struct drm_i915_private *i915, struct resource
*dsm)
{
- return dsm->start != 0 && dsm->end > dsm->start;
+ return (dsm->start != 0 || HAS_BAR2_SMEM_STOLEN(i915)) && dsm->end >
dsm->start;
}
static int adjust_stolen(struct drm_i915_private *i915,
@@ -88,7 +88,7 @@ static int adjust_stolen(struct drm_i915_private *i915,
struct i915_ggtt *ggtt = to_gt(i915)->ggtt;
struct intel_uncore *uncore = ggtt->vm.gt->uncore;
- if (!valid_stolen_size(dsm))
+ if (!valid_stolen_size(i915, dsm))
return -EINVAL;
/*
@@ -135,7 +135,7 @@ static int adjust_stolen(struct drm_i915_private *i915,
}
}
- if (!valid_stolen_size(dsm))
+ if (!valid_stolen_size(i915, dsm))
return -EINVAL;
return 0;
@@ -149,8 +149,11 @@ static int request_smem_stolen(struct drm_i915_private
*i915,
/*
* With stolen lmem, we don't need to request system memory for the
* address range since it's local to the gpu.
+*
+* Starting MTL, in IGFX devices the stolen memory is exposed via
+* BAR2 and shall be considered similar to stolen lmem.
*/
- if (HAS_LMEM(i915))
+ if (HAS_LMEM(i915) || HAS_BAR2_SMEM_STOLEN(i915))
return 0;
/*
@@ -385,8 +388,6 @@ static void icl_get_stolen_reserved(struct drm_i915_private
*i915,
drm_dbg(&i915->drm, "GEN6_STOLEN_RESERVED = 0x%016llx\n", reg_val);
- *base = reg_val & GEN11_STOLEN_RESERVED_ADDR_MASK;
-
switch (reg_val & GEN8_STOLEN_RESERVED_SIZE_MASK) {
case GEN8_STOLEN_RESERVED_1M:
*size = 1024 * 1024;
@@ -404,6 +405,12 @@ static void icl_get_stolen_reserved(struct
drm_i915_private *i915,
*size = 8 * 1024 * 1024;
MISSING_CASE(reg_val & GEN8_STOLEN_RESERVED_SIZE_MASK);
}
+
+ if (HAS_BAR2_SMEM_STOLEN(i915))
+ /* the base is initialized to stolen top so subtract size to
get base */
+ *base -= *size;
+ else
+ *base = reg_val & GEN11_STOLEN_RESERVED_ADDR_MASK;
}
/*
@@ -833,6 +840,29 @@ static const struct intel_memory_region_ops
i915_region_stolen_lmem_ops = {
.init_object = _i915_gem_object_stolen_init,
};
+static int mtl_get_gms_size(struct intel_uncore *uncore)
+{
+ u16 ggc, gms;
+
+ ggc = intel_uncore_read16(uncore, GGC);
+
+ /* check GGMS, should be fixed 0x3 (8MB) */
+ if ((ggc & GGMS_MASK) != GGMS_MASK)
+ return -EIO;
+
+ /* return valid GMS value, -EIO if invalid */
+ gms = REG_FIELD_GET(GMS_MASK, ggc);
+ switch (gms) {
+ case 0x0 ... 0x04:
+ return gms * 32;
+ case 0xf0 ... 0xfe:
+ return (gms - 0xf0 + 1) * 4;
+ default:
+ MISSING_CASE(gms);
+ return -EIO;
+ }
+}
+
struct intel_memory_region *
i915_gem_stolen_lmem_setup(struct drm_i
