From: Matthew Auld <matthew.auld@intel.com>
To: intel-gfx@lists.freedesktop.org
Subject: [PATCH 10/15] drm/i915: support huge gtt pages for the 48b PPGTT
Date: Wed, 31 May 2017 19:52:05 +0100 [thread overview]
Message-ID: <20170531185210.29189-11-matthew.auld@intel.com> (raw)
In-Reply-To: <20170531185210.29189-1-matthew.auld@intel.com>
Support inserting huge gtt pages into the 48b PPGTT, including
mixed-mode where we allow a mixture of gtt page sizes.
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel@ffwll.ch>
---
drivers/gpu/drm/i915/i915_gem_gtt.c | 79 ++++++++++++++++++++++++++++++++++---
drivers/gpu/drm/i915/i915_gem_gtt.h | 5 +++
2 files changed, 79 insertions(+), 5 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 935656802f09..924aec4adf6d 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -934,17 +934,86 @@ static void gen8_ppgtt_insert_4lvl(struct i915_address_space *vm,
u32 unused)
{
struct i915_hw_ppgtt *ppgtt = i915_vm_to_ppgtt(vm);
+ struct gen8_insert_pte idx = gen8_insert_pte(start);
+ struct i915_page_directory_pointer **pdps = ppgtt->pml4.pdps;
+ struct i915_page_directory_pointer *pdp = pdps[idx.pml4e];
+ struct i915_page_directory *pd = pdp->page_directory[idx.pdpe];
+ struct i915_page_table *pt = pd->page_table[idx.pde];
+ gen8_pte_t *pdp_vaddr = kmap_atomic_px(pdp);
+ gen8_pte_t *pd_vaddr = kmap_atomic_px(pd);
+ gen8_pte_t *pt_vaddr = kmap_atomic_px(pt);
+ const gen8_pte_t pte_encode = gen8_pte_encode(0, cache_level);
struct sgt_dma iter = {
.sg = pages->sgl,
.dma = sg_dma_address(iter.sg),
.max = iter.dma + iter.sg->length,
};
- struct i915_page_directory_pointer **pdps = ppgtt->pml4.pdps;
- struct gen8_insert_pte idx = gen8_insert_pte(start);
- while (gen8_ppgtt_insert_pte_entries(ppgtt, pdps[idx.pml4e++], &iter,
- &idx, cache_level))
- GEM_BUG_ON(idx.pml4e >= GEN8_PML4ES_PER_PML4);
+ do {
+ unsigned int page_size;
+
+ pt_vaddr[idx.pte] = pte_encode | iter.dma;
+ page_size = I915_GTT_PAGE_SIZE;
+
+ if (!idx.pte && page_sizes->sg > I915_GTT_PAGE_SIZE) {
+ dma_addr_t remaining = iter.max - iter.dma;
+
+ if (unlikely(page_sizes->sg & I915_GTT_PAGE_SIZE_1G) &&
+ remaining >= I915_GTT_PAGE_SIZE_1G && !idx.pde) {
+ pdp_vaddr[idx.pdpe] = pte_encode | GEN8_PDPE_PS_1G | iter.dma;
+ page_size = I915_GTT_PAGE_SIZE_1G;
+ } else if (page_sizes->sg & I915_GTT_PAGE_SIZE_2M &&
+ remaining >= I915_GTT_PAGE_SIZE_2M) {
+ pd_vaddr[idx.pde] = pte_encode | GEN8_PDE_PS_2M | iter.dma;
+ page_size = I915_GTT_PAGE_SIZE_2M;
+ /* We don't support 64K in mixed mode for now */
+ } else if (page_sizes->sg == I915_GTT_PAGE_SIZE_64K) {
+ pd_vaddr[idx.pde] |= GEN8_PDE_IPS_64K;
+ }
+ }
+
+ start += page_size;
+ iter.dma += page_size;
+ if (iter.dma >= iter.max) {
+ iter.sg = __sg_next(iter.sg);
+ if (!iter.sg)
+ break;
+
+ iter.dma = sg_dma_address(iter.sg);
+ iter.max = iter.dma + iter.sg->length;
+ }
+
+ idx.pte = gen8_pte_index(start);
+ if (!idx.pte) {
+ idx.pde = gen8_pde_index(start);
+
+ if (!idx.pde) {
+ idx.pdpe = gen8_pdpe_index(start);
+
+ if (!idx.pdpe) {
+ GEM_BUG_ON(idx.pml4e >= GEN8_PML4ES_PER_PML4);
+ pdp = pdps[idx.pml4e++];
+
+ kunmap_atomic(pdp_vaddr);
+ pdp_vaddr = kmap_atomic_px(pdp);
+ }
+
+ pd = pdp->page_directory[idx.pdpe];
+
+ kunmap_atomic(pd_vaddr);
+ pd_vaddr = kmap_atomic_px(pd);
+ }
+
+ pt = pd->page_table[idx.pde];
+
+ kunmap_atomic(pt_vaddr);
+ pt_vaddr = kmap_atomic_px(pt);
+ }
+ } while (1);
+
+ kunmap_atomic(pt_vaddr);
+ kunmap_atomic(pd_vaddr);
+ kunmap_atomic(pdp_vaddr);
}
static void gen8_free_page_tables(struct i915_address_space *vm,
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
index d45729b9da0c..0811859b3a55 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
@@ -149,6 +149,11 @@ typedef u64 gen8_ppgtt_pml4e_t;
#define GEN8_PPAT_ELLC_OVERRIDE (0<<2)
#define GEN8_PPAT(i, x) ((u64)(x) << ((i) * 8))
+#define GEN8_PDE_IPS_64K BIT(11)
+#define GEN8_PDE_PS_2M BIT(7)
+
+#define GEN8_PDPE_PS_1G BIT(7)
+
struct sg_table;
struct intel_rotation_info {
--
2.9.4
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
next prev parent reply other threads:[~2017-05-31 18:52 UTC|newest]
Thread overview: 30+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-05-31 18:51 [PATCH 00/15] huge gtt pages Matthew Auld
2017-05-31 18:51 ` [PATCH 01/15] drm/i915: really simple gemfs Matthew Auld
2017-05-31 19:16 ` Chris Wilson
2017-06-01 10:49 ` Joonas Lahtinen
2017-06-01 12:33 ` Matthew Auld
2017-05-31 18:51 ` [PATCH 02/15] drm/i915: enable THP for gemfs Matthew Auld
2017-06-01 10:27 ` Joonas Lahtinen
2017-05-31 18:51 ` [PATCH 03/15] drm/i915: introduce page_size_mask to dev_info Matthew Auld
2017-05-31 19:19 ` Chris Wilson
2017-06-01 10:57 ` Joonas Lahtinen
2017-06-01 12:10 ` Jani Nikula
2017-05-31 18:51 ` [PATCH 04/15] drm/i915: introduce gem object page_sizes Matthew Auld
2017-05-31 19:25 ` Chris Wilson
2017-06-01 9:49 ` Chris Wilson
2017-05-31 18:52 ` [PATCH 05/15] drm/i915: align the vma start to the largest gtt page size Matthew Auld
2017-05-31 19:31 ` Chris Wilson
2017-05-31 18:52 ` [PATCH 06/15] drm/i915: align 64K objects to 2M Matthew Auld
2017-06-01 9:55 ` Chris Wilson
2017-05-31 18:52 ` [PATCH 07/15] drm/i915: pass mm.gtt_page_sizes to ppgtt insert_entries Matthew Auld
2017-05-31 19:10 ` Chris Wilson
2017-05-31 18:52 ` [PATCH 08/15] drm/i915: enable IPS bit for 64K pages Matthew Auld
2017-05-31 18:52 ` [PATCH 09/15] drm/i915: disable GTT cache for 2M/1G pages Matthew Auld
2017-06-01 9:56 ` Chris Wilson
2017-05-31 18:52 ` Matthew Auld [this message]
2017-05-31 18:52 ` [PATCH 11/15] drm/i915: accurate page size tracking for the ppgtt Matthew Auld
2017-05-31 18:52 ` [PATCH 12/15] drm/i915/debugfs: include some gtt page size metrics Matthew Auld
2017-05-31 18:52 ` [PATCH 13/15] drm/i915: enable platform support for 64K pages Matthew Auld
2017-05-31 18:52 ` [PATCH 14/15] drm/i915: enable platform support for 2M pages Matthew Auld
2017-05-31 18:52 ` [PATCH 15/15] drm/i915: enable platform support for 1G pages Matthew Auld
2017-05-31 19:11 ` ✓ Fi.CI.BAT: success for huge gtt pages Patchwork
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20170531185210.29189-11-matthew.auld@intel.com \
--to=matthew.auld@intel.com \
--cc=intel-gfx@lists.freedesktop.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).