[Intel-gfx] [RFC PATCH 17/42] drm/i915/lmem: support kernel mapping
Matthew Auld
matthew.auld at intel.com
Thu Feb 14 14:57:15 UTC 2019
From: Abdiel Janulgue <abdiel.janulgue at linux.intel.com>
We can create LMEM objects, but we also need to support mapping them
into kernel space for internal use.
Signed-off-by: Abdiel Janulgue <abdiel.janulgue at linux.intel.com>
Signed-off-by: Matthew Auld <matthew.auld at intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen at linux.intel.com>
---
drivers/gpu/drm/i915/i915_gem.c | 18 ++-
drivers/gpu/drm/i915/intel_memory_region.h | 3 +
drivers/gpu/drm/i915/intel_region_lmem.c | 24 ++++
drivers/gpu/drm/i915/intel_region_lmem.h | 5 +
drivers/gpu/drm/i915/selftests/huge_pages.c | 3 +-
.../drm/i915/selftests/intel_memory_region.c | 107 +++++++++++++++++-
6 files changed, 148 insertions(+), 12 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 33699b3dc395..3f319524beed 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2257,7 +2257,9 @@ __i915_gem_object_unset_pages(struct drm_i915_gem_object *obj)
void *ptr;
ptr = page_mask_bits(obj->mm.mapping);
- if (is_vmalloc_addr(ptr))
+ if (i915_gem_object_is_lmem(obj))
+ io_mapping_unmap(ptr);
+ else if (is_vmalloc_addr(ptr))
vunmap(ptr);
else
kunmap(kmap_to_page(ptr));
@@ -2625,7 +2627,7 @@ int __i915_gem_object_get_pages(struct drm_i915_gem_object *obj)
}
/* The 'mapping' part of i915_gem_object_pin_map() below */
-static void *i915_gem_object_map(const struct drm_i915_gem_object *obj,
+static void *i915_gem_object_map(struct drm_i915_gem_object *obj,
enum i915_map_type type)
{
unsigned long n_pages = obj->base.size >> PAGE_SHIFT;
@@ -2638,6 +2640,11 @@ static void *i915_gem_object_map(const struct drm_i915_gem_object *obj,
pgprot_t pgprot;
void *addr;
+ if (i915_gem_object_is_lmem(obj)) {
+ /* XXX: we are ignoring the type here -- this is simply wc */
+ return i915_gem_object_lmem_io_map(obj, 0, obj->base.size);
+ }
+
/* A single page can always be kmapped */
if (n_pages == 1 && type == I915_MAP_WB)
return kmap(sg_page(sgt->sgl));
@@ -2683,7 +2690,8 @@ void *i915_gem_object_pin_map(struct drm_i915_gem_object *obj,
void *ptr;
int ret;
- if (unlikely(!i915_gem_object_has_struct_page(obj)))
+ if (unlikely(!i915_gem_object_has_struct_page(obj) &&
+ !i915_gem_object_is_lmem(obj)))
return ERR_PTR(-ENXIO);
ret = mutex_lock_interruptible(&obj->mm.lock);
@@ -2715,7 +2723,9 @@ void *i915_gem_object_pin_map(struct drm_i915_gem_object *obj,
goto err_unpin;
}
- if (is_vmalloc_addr(ptr))
+ if (i915_gem_object_is_lmem(obj))
+ io_mapping_unmap(ptr);
+ else if (is_vmalloc_addr(ptr))
vunmap(ptr);
else
kunmap(kmap_to_page(ptr));
diff --git a/drivers/gpu/drm/i915/intel_memory_region.h b/drivers/gpu/drm/i915/intel_memory_region.h
index b1546afb0b6e..7754989ad165 100644
--- a/drivers/gpu/drm/i915/intel_memory_region.h
+++ b/drivers/gpu/drm/i915/intel_memory_region.h
@@ -115,6 +115,9 @@ void i915_memory_region_put_pages_buddy(struct drm_i915_gem_object *obj,
void i915_gem_object_release_memory_region(struct drm_i915_gem_object *obj);
+unsigned long i915_gem_object_lmem_io_pfn(struct drm_i915_gem_object *obj,
+ unsigned long n);
+
struct intel_memory_region *
intel_memory_region_create(struct drm_i915_private *i915,
resource_size_t start,
diff --git a/drivers/gpu/drm/i915/intel_region_lmem.c b/drivers/gpu/drm/i915/intel_region_lmem.c
index b398becb2733..c8bee0c18c88 100644
--- a/drivers/gpu/drm/i915/intel_region_lmem.c
+++ b/drivers/gpu/drm/i915/intel_region_lmem.c
@@ -65,6 +65,30 @@ static const struct intel_memory_region_ops region_lmem_ops = {
.object_create = region_lmem_object_create,
};
+/* XXX: Time to vfunc your life up? */
+void __iomem *i915_gem_object_lmem_io_map_page(struct drm_i915_gem_object *obj,
+ unsigned long n)
+{
+ resource_size_t offset;
+
+ offset = i915_gem_object_get_dma_address(obj, n);
+
+ return io_mapping_map_atomic_wc(&obj->memory_region->iomap, offset);
+}
+
+void __iomem *i915_gem_object_lmem_io_map(struct drm_i915_gem_object *obj,
+ unsigned long n,
+ unsigned long size)
+{
+ resource_size_t offset;
+
+ GEM_BUG_ON(!(obj->flags & I915_BO_ALLOC_CONTIGUOUS));
+
+ offset = i915_gem_object_get_dma_address(obj, n);
+
+ return io_mapping_map_wc(&obj->memory_region->iomap, offset, size);
+}
+
unsigned long i915_gem_object_lmem_io_pfn(struct drm_i915_gem_object *obj,
unsigned long n)
{
diff --git a/drivers/gpu/drm/i915/intel_region_lmem.h b/drivers/gpu/drm/i915/intel_region_lmem.h
index 1aa24bd28690..a76b77b69e71 100644
--- a/drivers/gpu/drm/i915/intel_region_lmem.h
+++ b/drivers/gpu/drm/i915/intel_region_lmem.h
@@ -28,6 +28,11 @@
unsigned long i915_gem_object_lmem_io_pfn(struct drm_i915_gem_object *obj,
unsigned long n);
+void __iomem *i915_gem_object_lmem_io_map(struct drm_i915_gem_object *obj,
+ unsigned long n, unsigned long size);
+void __iomem *i915_gem_object_lmem_io_map_page(struct drm_i915_gem_object *obj,
+ unsigned long n);
+
bool i915_gem_object_is_lmem(struct drm_i915_gem_object *obj);
struct drm_i915_gem_object *
diff --git a/drivers/gpu/drm/i915/selftests/huge_pages.c b/drivers/gpu/drm/i915/selftests/huge_pages.c
index f4283d41c13f..103362f8c553 100644
--- a/drivers/gpu/drm/i915/selftests/huge_pages.c
+++ b/drivers/gpu/drm/i915/selftests/huge_pages.c
@@ -1144,8 +1144,7 @@ static int __cpu_check_lmem(struct drm_i915_gem_object *obj, u32 dword, u32 val)
u32 __iomem *base;
u32 read_val;
- base = (void __force *)io_mapping_map_atomic_wc(&obj->memory_region->iomap,
- i915_gem_object_get_dma_address(obj, n));
+ base = i915_gem_object_lmem_io_map_page(obj, n);
read_val = ioread32(base + dword);
io_mapping_unmap_atomic(base);
diff --git a/drivers/gpu/drm/i915/selftests/intel_memory_region.c b/drivers/gpu/drm/i915/selftests/intel_memory_region.c
index 2c07d8d219e5..ad8d2eeacf53 100644
--- a/drivers/gpu/drm/i915/selftests/intel_memory_region.c
+++ b/drivers/gpu/drm/i915/selftests/intel_memory_region.c
@@ -28,6 +28,9 @@
#include "mock_context.h"
#include "mock_drm.h"
+typedef int (*cpu_check_fn_t)(struct drm_i915_gem_object *obj,
+ u32 __iomem *base, u32 dword, u32 val);
+
static void close_objects(struct list_head *objects)
{
struct drm_i915_gem_object *obj, *on;
@@ -487,7 +490,9 @@ static int igt_gpu_write(struct i915_vma *vma,
return err;
}
-static int igt_cpu_check(struct drm_i915_gem_object *obj, u32 dword, u32 val)
+static int igt_cpu_check(struct drm_i915_gem_object *obj,
+ u32 __iomem *base,
+ u32 dword, u32 val)
{
unsigned long n;
int err;
@@ -504,8 +509,7 @@ static int igt_cpu_check(struct drm_i915_gem_object *obj, u32 dword, u32 val)
u32 __iomem *base;
u32 read_val;
- base = (void __force *) io_mapping_map_atomic_wc(&obj->memory_region->iomap,
- i915_gem_object_get_dma_address(obj, n));
+ base = i915_gem_object_lmem_io_map_page(obj, n);
read_val = ioread32(base + dword);
io_mapping_unmap_atomic(base);
@@ -523,11 +527,13 @@ static int igt_cpu_check(struct drm_i915_gem_object *obj, u32 dword, u32 val)
static int igt_gpu_fill(struct i915_gem_context *ctx,
struct drm_i915_gem_object *obj,
- u32 val)
+ cpu_check_fn_t cpu_check,
+ u32 __iomem *base)
{
struct drm_i915_private *i915 = ctx->i915;
struct i915_address_space *vm = ctx->ppgtt ? &ctx->ppgtt->vm : &i915->ggtt.vm;
struct i915_vma *vma;
+ struct rnd_state prng;
u32 dword;
int err;
@@ -541,12 +547,14 @@ static int igt_gpu_fill(struct i915_gem_context *ctx,
return err;
}
+ prandom_seed_state(&prng, i915_selftest.random_seed);
for (dword = 0; dword < PAGE_SIZE / sizeof(u32); ++dword) {
+ u32 val = prandom_u32_state(&prng);
err = igt_gpu_write(vma, ctx, i915->engine[RCS], dword, val);
if (err)
break;
- err = igt_cpu_check(obj, dword, val);
+ err = cpu_check(obj, base, dword, val);
if (err)
break;
}
@@ -645,7 +653,7 @@ static int igt_lmem_write_gpu(void *arg)
if (err)
goto out_put;
- err = igt_gpu_fill(ctx, obj, 0xdeadbeaf);
+ err = igt_gpu_fill(ctx, obj, igt_cpu_check, NULL);
if (err) {
pr_err("igt_gpu_fill failed(%d)\n", err);
goto out_unpin;
@@ -659,6 +667,92 @@ static int igt_lmem_write_gpu(void *arg)
return err;
}
+static int igt_lmem_cpu_check(struct drm_i915_gem_object *obj,
+ u32 __iomem *base, u32 dword, u32 val)
+{
+ u32 read_val;
+ int err;
+
+ err = i915_gem_object_wait(obj,
+ I915_WAIT_INTERRUPTIBLE |
+ I915_WAIT_LOCKED,
+ MAX_SCHEDULE_TIMEOUT);
+ if (err)
+ return err;
+
+ err = i915_gem_object_pin_pages(obj);
+ if (err)
+ return err;
+
+ read_val = ioread32(base + dword);
+ if (read_val != val) {
+ pr_err("base[%u]=0x%x, val=0x%x\n",
+ dword, read_val, val);
+ return -EINVAL;
+ }
+
+ i915_gem_object_unpin_pages(obj);
+ return 0;
+}
+
+static int igt_lmem_write_cpu(void *arg)
+{
+ struct i915_gem_context *ctx = arg;
+ struct drm_i915_private *i915 = ctx->i915;
+ struct drm_i915_gem_object *obj;
+ struct rnd_state prng;
+ u32 __iomem *vaddr;
+ u32 dword;
+ int ret = 0;
+
+ obj = i915_gem_object_create_lmem(i915, PAGE_SIZE, I915_BO_ALLOC_CONTIGUOUS);
+ if (IS_ERR(obj))
+ return PTR_ERR(obj);
+
+ ret = i915_gem_object_pin_pages(obj);
+ if (ret)
+ goto out_put;
+
+ vaddr = i915_gem_object_pin_map(obj, I915_MAP_WC);
+ if (IS_ERR(vaddr)) {
+ pr_err("Failed to iomap lmembar; err=%d\n", (int)PTR_ERR(vaddr));
+ ret = PTR_ERR(vaddr);
+ goto out_unpin;
+ }
+
+ /* gpu write/cpu read */
+ ret = igt_gpu_fill(ctx, obj, igt_lmem_cpu_check, vaddr);
+ if (ret) {
+ pr_err("igt_gpu_fill failed(%d)\n", ret);
+ goto out_unpin;
+ }
+
+ /* cpu write/cpu read */
+ prandom_seed_state(&prng, i915_selftest.random_seed);
+ for (dword = 0; dword < PAGE_SIZE / sizeof(u32); ++dword) {
+ u32 read_val;
+ u32 val = prandom_u32_state(&prng);
+
+ iowrite32(val, vaddr + dword);
+ wmb();
+
+ read_val = ioread32(vaddr + dword);
+ if (read_val != val) {
+ pr_err("base[%u]=%u, val=%u\n", dword, read_val, val);
+ ret = -EINVAL;
+ break;
+ }
+ }
+
+ i915_gem_object_unpin_map(obj);
+
+out_unpin:
+ i915_gem_object_unpin_pages(obj);
+out_put:
+ i915_gem_object_put(obj);
+ return ret;
+}
+
static int igt_lmem_pages_migrate(void *arg)
{
struct i915_gem_context *ctx = arg;
@@ -776,6 +870,7 @@ int intel_memory_region_live_selftests(struct drm_i915_private *i915)
SUBTEST(igt_smem_create_migrate),
SUBTEST(igt_lmem_create_migrate),
SUBTEST(igt_lmem_write_gpu),
+ SUBTEST(igt_lmem_write_cpu),
SUBTEST(igt_lmem_pages_migrate),
};
struct i915_gem_context *ctx;
--
2.20.1
More information about the Intel-gfx
mailing list