drm/i915/lmem: support kernel mapping

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@linux.intel.com>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Steve Hampson <steven.t.hampson@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20191025153728.23689-3-chris@chris-wilson.co.uk

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_lmem.c b/drivers/gpu/drm/i915/gem/i915_gem_lmem.c
index 6a38d80..926f6c9 100644 (file)
--- a/drivers/gpu/drm/i915/gem/i915_gem_lmem.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_lmem.c
@@ -9,11 +9,50 @@
 #include "i915_drv.h"
 
 const struct drm_i915_gem_object_ops i915_gem_lmem_obj_ops = {
+       .flags = I915_GEM_OBJECT_HAS_IOMEM,
+
        .get_pages = i915_gem_object_get_pages_buddy,
        .put_pages = i915_gem_object_put_pages_buddy,
        .release = i915_gem_object_release_memory_region,
 };
 
+/* 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_wc(&obj->mm.region->iomap, offset, PAGE_SIZE);
+}
+
+void __iomem *
+i915_gem_object_lmem_io_map_page_atomic(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->mm.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(!i915_gem_object_is_contiguous(obj));
+
+       offset = i915_gem_object_get_dma_address(obj, n);
+
+       return io_mapping_map_wc(&obj->mm.region->iomap, offset, size);
+}
+
 bool i915_gem_object_is_lmem(struct drm_i915_gem_object *obj)
 {
        return obj->ops == &i915_gem_lmem_obj_ops;

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_lmem.h b/drivers/gpu/drm/i915/gem/i915_gem_lmem.h
index fc3f155..7c176b8 100644 (file)
--- a/drivers/gpu/drm/i915/gem/i915_gem_lmem.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_lmem.h
@@ -14,6 +14,14 @@ struct intel_memory_region;
 
 extern const struct drm_i915_gem_object_ops i915_gem_lmem_obj_ops;
 
+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);
+void __iomem *
+i915_gem_object_lmem_io_map_page_atomic(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/gem/i915_gem_object_types.h b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
index a387e3e..9600837 100644 (file)
--- a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
@@ -31,10 +31,11 @@ struct i915_lut_handle {
 struct drm_i915_gem_object_ops {
        unsigned int flags;
 #define I915_GEM_OBJECT_HAS_STRUCT_PAGE        BIT(0)
-#define I915_GEM_OBJECT_IS_SHRINKABLE  BIT(1)
-#define I915_GEM_OBJECT_IS_PROXY       BIT(2)
-#define I915_GEM_OBJECT_NO_GGTT                BIT(3)
-#define I915_GEM_OBJECT_ASYNC_CANCEL   BIT(4)
+#define I915_GEM_OBJECT_HAS_IOMEM      BIT(1)
+#define I915_GEM_OBJECT_IS_SHRINKABLE  BIT(2)
+#define I915_GEM_OBJECT_IS_PROXY       BIT(3)
+#define I915_GEM_OBJECT_NO_GGTT                BIT(4)
+#define I915_GEM_OBJECT_ASYNC_CANCEL   BIT(5)
 
        /* Interface between the GEM object and its backing storage.
         * get_pages() is called once prior to the use of the associated set

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pages.c b/drivers/gpu/drm/i915/gem/i915_gem_pages.c
index b0ec095..29f4c28 100644 (file)
--- a/drivers/gpu/drm/i915/gem/i915_gem_pages.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_pages.c
@@ -7,6 +7,7 @@
 #include "i915_drv.h"
 #include "i915_gem_object.h"
 #include "i915_scatterlist.h"
+#include "i915_gem_lmem.h"
 
 void __i915_gem_object_set_pages(struct drm_i915_gem_object *obj,
                                 struct sg_table *pages,
@@ -154,6 +155,16 @@ static void __i915_gem_object_reset_page_iter(struct drm_i915_gem_object *obj)
        rcu_read_unlock();
 }
 
+static void unmap_object(struct drm_i915_gem_object *obj, void *ptr)
+{
+       if (i915_gem_object_is_lmem(obj))
+               io_mapping_unmap((void __force __iomem *)ptr);
+       else if (is_vmalloc_addr(ptr))
+               vunmap(ptr);
+       else
+               kunmap(kmap_to_page(ptr));
+}
+
 struct sg_table *
 __i915_gem_object_unset_pages(struct drm_i915_gem_object *obj)
 {
@@ -169,14 +180,7 @@ __i915_gem_object_unset_pages(struct drm_i915_gem_object *obj)
        i915_gem_object_make_unshrinkable(obj);
 
        if (obj->mm.mapping) {
-               void *ptr;
-
-               ptr = page_mask_bits(obj->mm.mapping);
-               if (is_vmalloc_addr(ptr))
-                       vunmap(ptr);
-               else
-                       kunmap(kmap_to_page(ptr));
-
+               unmap_object(obj, page_mask_bits(obj->mm.mapping));
                obj->mm.mapping = NULL;
        }
 
@@ -231,7 +235,7 @@ unlock:
 }
 
 /* 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;
@@ -244,6 +248,16 @@ 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)) {
+               void __iomem *io;
+
+               if (type != I915_MAP_WC)
+                       return NULL;
+
+               io = i915_gem_object_lmem_io_map(obj, 0, obj->base.size);
+               return (void __force *)io;
+       }
+
        /* A single page can always be kmapped */
        if (n_pages == 1 && type == I915_MAP_WB)
                return kmap(sg_page(sgt->sgl));
@@ -285,11 +299,13 @@ void *i915_gem_object_pin_map(struct drm_i915_gem_object *obj,
                              enum i915_map_type type)
 {
        enum i915_map_type has_type;
+       unsigned int flags;
        bool pinned;
        void *ptr;
        int err;
 
-       if (unlikely(!i915_gem_object_has_struct_page(obj)))
+       flags = I915_GEM_OBJECT_HAS_STRUCT_PAGE | I915_GEM_OBJECT_HAS_IOMEM;
+       if (!i915_gem_object_type_has(obj, flags))
                return ERR_PTR(-ENXIO);
 
        err = mutex_lock_interruptible(&obj->mm.lock);
@@ -321,10 +337,7 @@ void *i915_gem_object_pin_map(struct drm_i915_gem_object *obj,
                        goto err_unpin;
                }
 
-               if (is_vmalloc_addr(ptr))
-                       vunmap(ptr);
-               else
-                       kunmap(kmap_to_page(ptr));
+               unmap_object(obj, ptr);
 
                ptr = obj->mm.mapping = NULL;
        }

diff --git a/drivers/gpu/drm/i915/selftests/intel_memory_region.c b/drivers/gpu/drm/i915/selftests/intel_memory_region.c
index 617a35c..2e0f91f 100644 (file)
--- a/drivers/gpu/drm/i915/selftests/intel_memory_region.c
+++ b/drivers/gpu/drm/i915/selftests/intel_memory_region.c
@@ -13,8 +13,10 @@
 
 #include "gem/i915_gem_lmem.h"
 #include "gem/i915_gem_region.h"
+#include "gem/i915_gem_object_blt.h"
 #include "gem/selftests/mock_context.h"
 #include "gt/intel_gt.h"
+#include "selftests/igt_flush_test.h"
 #include "selftests/i915_random.h"
 
 static void close_objects(struct intel_memory_region *mem,
@@ -275,6 +277,116 @@ out_put:
        return err;
 }
 
+static int igt_lmem_write_cpu(void *arg)
+{
+       struct drm_i915_private *i915 = arg;
+       struct drm_i915_gem_object *obj;
+       I915_RND_STATE(prng);
+       IGT_TIMEOUT(end_time);
+       u32 bytes[] = {
+               0, /* rng placeholder */
+               sizeof(u32),
+               sizeof(u64),
+               64, /* cl */
+               PAGE_SIZE,
+               PAGE_SIZE - sizeof(u32),
+               PAGE_SIZE - sizeof(u64),
+               PAGE_SIZE - 64,
+       };
+       u32 *vaddr;
+       u32 sz;
+       u32 i;
+       int *order;
+       int count;
+       int err;
+
+       if (!HAS_ENGINE(i915, BCS0))
+               return 0;
+
+       sz = round_up(prandom_u32_state(&prng) % SZ_32M, PAGE_SIZE);
+       sz = max_t(u32, 2 * PAGE_SIZE, sz);
+
+       obj = i915_gem_object_create_lmem(i915, sz, I915_BO_ALLOC_CONTIGUOUS);
+       if (IS_ERR(obj))
+               return PTR_ERR(obj);
+
+       vaddr = i915_gem_object_pin_map(obj, I915_MAP_WC);
+       if (IS_ERR(vaddr)) {
+               err = PTR_ERR(vaddr);
+               goto out_put;
+       }
+
+       /* Put the pages into a known state -- from the gpu for added fun */
+       err = i915_gem_object_fill_blt(obj, i915->engine[BCS0]->kernel_context,
+                                      0xdeadbeaf);
+       if (err)
+               goto out_unpin;
+
+       i915_gem_object_lock(obj);
+       err = i915_gem_object_set_to_wc_domain(obj, true);
+       i915_gem_object_unlock(obj);
+       if (err)
+               goto out_unpin;
+
+       count = ARRAY_SIZE(bytes);
+       order = i915_random_order(count * count, &prng);
+       if (!order) {
+               err = -ENOMEM;
+               goto out_unpin;
+       }
+
+       /* We want to throw in a random width/align */
+       bytes[0] = igt_random_offset(&prng, 0, PAGE_SIZE, sizeof(u32),
+                                    sizeof(u32));
+
+       i = 0;
+       do {
+               u32 offset;
+               u32 align;
+               u32 dword;
+               u32 size;
+               u32 val;
+
+               size = bytes[order[i] % count];
+               i = (i + 1) % (count * count);
+
+               align = bytes[order[i] % count];
+               i = (i + 1) % (count * count);
+
+               align = max_t(u32, sizeof(u32), rounddown_pow_of_two(align));
+
+               offset = igt_random_offset(&prng, 0, obj->base.size,
+                                          size, align);
+
+               val = prandom_u32_state(&prng);
+               memset32(vaddr + offset / sizeof(u32), val ^ 0xdeadbeaf,
+                        size / sizeof(u32));
+
+               /*
+                * Sample random dw -- don't waste precious time reading every
+                * single dw.
+                */
+               dword = igt_random_offset(&prng, offset,
+                                         offset + size,
+                                         sizeof(u32), sizeof(u32));
+               dword /= sizeof(u32);
+               if (vaddr[dword] != (val ^ 0xdeadbeaf)) {
+                       pr_err("%s vaddr[%u]=%u, val=%u, size=%u, align=%u, offset=%u\n",
+                              __func__, dword, vaddr[dword], val ^ 0xdeadbeaf,
+                              size, align, offset);
+                       err = -EINVAL;
+                       break;
+               }
+       } while (!__igt_timeout(end_time, NULL));
+
+out_unpin:
+       i915_gem_object_unpin_map(obj);
+out_put:
+       i915_gem_object_put(obj);
+
+       return err;
+}
+
 int intel_memory_region_mock_selftests(void)
 {
        static const struct i915_subtest tests[] = {
@@ -308,6 +420,7 @@ int intel_memory_region_live_selftests(struct drm_i915_private *i915)
 {
        static const struct i915_subtest tests[] = {
                SUBTEST(igt_lmem_create),
+               SUBTEST(igt_lmem_write_cpu),
        };
 
        if (!HAS_LMEM(i915)) {