Re: [PATCH 7/8] drm/ttm: Introduce a huge page aligning TTM range manager.

Thomas Hellström (VMware) <thomas_os@xxxxxxxxxxxx> · Wed, 4 Dec 2019 12:45:01 +0100

On 12/4/19 12:13 PM, Christian König wrote:
Am 03.12.19 um 14:22 schrieb Thomas Hellström (VMware):
From: Thomas Hellstrom <thellstrom@xxxxxxxxxx>

Using huge page-table entries require that the start of a buffer object
is huge page size aligned. So introduce a ttm_bo_man_get_node_huge()
function that attempts to accomplish this for allocations that are 
larger
than the huge page size, and provide a new range-manager instance that
uses that function.

I still don't think that this is a good idea.

Again, can you elaborate with some specific concerns?


The driver/userspace should just use a proper alignment if it wants to 
use huge pages.

There are drawbacks with this approach. The TTM alignment is a hard 
constraint. Assume that you want to fit a 1GB buffer object into limited 
VRAM space, and _if possible_ use PUD size huge pages. Let's say there 
is 1GB available, but not 1GB aligned. The proper alignment approach 
would fail and possibly start to evict stuff from VRAM just to be able 
to accomodate the PUD alignment. That's bad. The approach I suggest 
would instead fall back to PMD alignment and use 2MB page table entries 
if possible, and as a last resort use 4K page table entries.

Or do I misunderstand how you envision enforcing the alignment?

Thanks,

Thomas







Regards,
Christian.


Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
Cc: Michal Hocko <mhocko@xxxxxxxx>
Cc: "Matthew Wilcox (Oracle)" <willy@xxxxxxxxxxxxx>
Cc: "Kirill A. Shutemov" <kirill.shutemov@xxxxxxxxxxxxxxx>
Cc: Ralph Campbell <rcampbell@xxxxxxxxxx>
Cc: "Jérôme Glisse" <jglisse@xxxxxxxxxx>
Cc: "Christian König" <christian.koenig@xxxxxxx>
Signed-off-by: Thomas Hellstrom <thellstrom@xxxxxxxxxx>
---
  drivers/gpu/drm/ttm/ttm_bo_manager.c | 92 ++++++++++++++++++++++++++++
  include/drm/ttm/ttm_bo_driver.h      |  1 +
  2 files changed, 93 insertions(+)

diff --git a/drivers/gpu/drm/ttm/ttm_bo_manager.c 
b/drivers/gpu/drm/ttm/ttm_bo_manager.c
index 18d3debcc949..26aa1a2ae7f1 100644
--- a/drivers/gpu/drm/ttm/ttm_bo_manager.c
+++ b/drivers/gpu/drm/ttm/ttm_bo_manager.c
@@ -89,6 +89,89 @@ static int ttm_bo_man_get_node(struct 
ttm_mem_type_manager *man,
      return 0;
  }
  +#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+static int ttm_bo_insert_aligned(struct drm_mm *mm, struct 
drm_mm_node *node,
+                 unsigned long align_pages,
+                 const struct ttm_place *place,
+                 struct ttm_mem_reg *mem,
+                 unsigned long lpfn,
+                 enum drm_mm_insert_mode mode)
+{
+    if (align_pages >= mem->page_alignment &&
+        (!mem->page_alignment || align_pages % mem->page_alignment 
== 0)) {
+        return drm_mm_insert_node_in_range(mm, node,
+                           mem->num_pages,
+                           align_pages, 0,
+                           place->fpfn, lpfn, mode);
+    }
+
+    return -ENOSPC;
+}
+
+static int ttm_bo_man_get_node_huge(struct ttm_mem_type_manager *man,
+                    struct ttm_buffer_object *bo,
+                    const struct ttm_place *place,
+                    struct ttm_mem_reg *mem)
+{
+    struct ttm_range_manager *rman = (struct ttm_range_manager *) 
man->priv;
+    struct drm_mm *mm = &rman->mm;
+    struct drm_mm_node *node;
+    unsigned long align_pages;
+    unsigned long lpfn;
+    enum drm_mm_insert_mode mode = DRM_MM_INSERT_BEST;
+    int ret;
+
+    node = kzalloc(sizeof(*node), GFP_KERNEL);
+    if (!node)
+        return -ENOMEM;
+
+    lpfn = place->lpfn;
+    if (!lpfn)
+        lpfn = man->size;
+
+    mode = DRM_MM_INSERT_BEST;
+    if (place->flags & TTM_PL_FLAG_TOPDOWN)
+        mode = DRM_MM_INSERT_HIGH;
+
+    spin_lock(&rman->lock);
+    if (IS_ENABLED(CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD)) {
+        align_pages = (HPAGE_PUD_SIZE >> PAGE_SHIFT);
+        if (mem->num_pages >= align_pages) {
+            ret = ttm_bo_insert_aligned(mm, node, align_pages,
+                            place, mem, lpfn, mode);
+            if (!ret)
+                goto found_unlock;
+        }
+    }
+
+    align_pages = (HPAGE_PMD_SIZE >> PAGE_SHIFT);
+    if (mem->num_pages >= align_pages) {
+        ret = ttm_bo_insert_aligned(mm, node, align_pages, place, mem,
+                        lpfn, mode);
+        if (!ret)
+            goto found_unlock;
+    }
+
+    ret = drm_mm_insert_node_in_range(mm, node, mem->num_pages,
+                      mem->page_alignment, 0,
+                      place->fpfn, lpfn, mode);
+found_unlock:
+    spin_unlock(&rman->lock);
+
+    if (unlikely(ret)) {
+        kfree(node);
+    } else {
+        mem->mm_node = node;
+        mem->start = node->start;
+    }
+
+    return 0;
+}
+#else
+#define ttm_bo_man_get_node_huge ttm_bo_man_get_node
+#endif
+
+
  static void ttm_bo_man_put_node(struct ttm_mem_type_manager *man,
                  struct ttm_mem_reg *mem)
  {
@@ -154,3 +237,12 @@ const struct ttm_mem_type_manager_func 
ttm_bo_manager_func = {
      .debug = ttm_bo_man_debug
  };
  EXPORT_SYMBOL(ttm_bo_manager_func);
+
+const struct ttm_mem_type_manager_func ttm_bo_manager_huge_func = {
+    .init = ttm_bo_man_init,
+    .takedown = ttm_bo_man_takedown,
+    .get_node = ttm_bo_man_get_node_huge,
+    .put_node = ttm_bo_man_put_node,
+    .debug = ttm_bo_man_debug
+};
+EXPORT_SYMBOL(ttm_bo_manager_huge_func);
diff --git a/include/drm/ttm/ttm_bo_driver.h 
b/include/drm/ttm/ttm_bo_driver.h
index cac7a8a0825a..868bd0d4be6a 100644
--- a/include/drm/ttm/ttm_bo_driver.h
+++ b/include/drm/ttm/ttm_bo_driver.h
@@ -888,5 +888,6 @@ int ttm_bo_pipeline_gutting(struct 
ttm_buffer_object *bo);
  pgprot_t ttm_io_prot(uint32_t caching_flags, pgprot_t tmp);
    extern const struct ttm_mem_type_manager_func ttm_bo_manager_func;
+extern const struct ttm_mem_type_manager_func ttm_bo_manager_huge_func;
    #endif










