[PATCH 2/3] drm/i915: Implement GGTT update method with blitter
Nirmoy Das
nirmoy.das at intel.com
Thu Aug 10 09:20:28 UTC 2023
Implement GGTT update method with blitter command, MI_UPDATE_GTT.
Signed-off-by: Nirmoy Das <nirmoy.das at intel.com>
---
drivers/gpu/drm/i915/gt/intel_ggtt.c | 176 +++++++++++++++++++++++++++
1 file changed, 176 insertions(+)
diff --git a/drivers/gpu/drm/i915/gt/intel_ggtt.c b/drivers/gpu/drm/i915/gt/intel_ggtt.c
index dd0ed941441a..970b375d1456 100644
--- a/drivers/gpu/drm/i915/gt/intel_ggtt.c
+++ b/drivers/gpu/drm/i915/gt/intel_ggtt.c
@@ -15,12 +15,16 @@
#include "display/intel_display.h"
#include "gem/i915_gem_lmem.h"
+#include "intel_context.h"
#include "intel_ggtt_gmch.h"
+#include "intel_gpu_commands.h"
#include "intel_gt.h"
#include "intel_gt_regs.h"
#include "intel_pci_config.h"
+#include "intel_ring.h"
#include "i915_drv.h"
#include "i915_pci.h"
+#include "i915_request.h"
#include "i915_scatterlist.h"
#include "i915_utils.h"
#include "i915_vgpu.h"
@@ -252,6 +256,84 @@ u64 gen8_ggtt_pte_encode(dma_addr_t addr,
return pte;
}
+static bool update_ggtt_with_blit(struct i915_ggtt *ggtt)
+{
+ struct intel_gt *gt = ggtt->vm.gt;
+
+ return intel_engine_blitter_context_ready(gt);
+}
+
+static bool gen8_ggtt_blit_fixed_pte(struct i915_ggtt *ggtt, u32 offset,
+ u32 num_entries, const gen8_pte_t pte)
+{
+ struct intel_gt *gt = ggtt->vm.gt;
+ struct i915_request *rq;
+ struct intel_context *ce;
+ struct i915_sched_attr attr = {};
+ u32 *cs;
+
+ ce = gt->engine[BCS0]->blitter_context;
+ if (!ce) {
+ drm_dbg(&ggtt->vm.i915->drm, "Failed to get blitter context\n");
+ goto err_print;
+ }
+
+ while (num_entries) {
+ /* MI_UPDATE_GTT can update 1022 dwords in a single command */
+ u32 n_ptes = min_t(u32, 511, num_entries);
+
+ mutex_lock(&ce->timeline->mutex);
+
+ intel_context_enter(ce);
+ rq = __i915_request_create(ce, GFP_NOWAIT | GFP_ATOMIC);
+ intel_context_exit(ce);
+ if (!rq) {
+ drm_warn(&ggtt->vm.i915->drm,
+ "Failed to get blitter request\n");
+ goto err_unlock;
+ }
+
+ cs = intel_ring_begin(rq, 2 * n_ptes + 2);
+ if (IS_ERR(cs)) {
+ drm_warn(&ggtt->vm.i915->drm,
+ "Failed to begin ring for GGTT blitter\n");
+ goto err_rq;
+ }
+
+ *cs++ = MI_UPDATE_GTT | (2 * n_ptes);
+ *cs++ = offset << 12;
+ memset64((u64 *)cs, pte, n_ptes);
+ cs += n_ptes * 2;
+ intel_ring_advance(rq, cs);
+
+ i915_request_get(rq);
+ __i915_request_commit(rq);
+ __i915_request_queue(rq, &attr);
+
+ mutex_unlock(&ce->timeline->mutex);
+ if (i915_request_wait(rq, 0, HZ) < 0) {
+ drm_warn(&ggtt->vm.i915->drm, "GGTT update timed out\n");
+ goto err_rq;
+ }
+
+ i915_request_put(rq);
+ num_entries -= n_ptes;
+ }
+
+ ggtt->invalidate(ggtt);
+ return true;
+
+err_rq:
+ i915_request_put(rq);
+err_unlock:
+ mutex_unlock(&ce->timeline->mutex);
+err_print:
+ drm_warn(&ggtt->vm.i915->drm,
+ "GGTT update failed with blitter, falling back to pci bar based update\n");
+ intel_engine_blitter_context_set_ready(gt, false);
+ return false;
+}
+
static void gen8_set_pte(void __iomem *addr, gen8_pte_t pte)
{
writeq(pte, addr);
@@ -272,6 +354,22 @@ static void gen8_ggtt_insert_page(struct i915_address_space *vm,
ggtt->invalidate(ggtt);
}
+static void gen8_ggtt_insert_page_blit(struct i915_address_space *vm,
+ dma_addr_t addr, u64 offset,
+ unsigned int pat_index, u32 flags)
+{
+ struct i915_ggtt *ggtt = i915_vm_to_ggtt(vm);
+ gen8_pte_t pte;
+
+ pte = ggtt->vm.pte_encode(addr, pat_index, flags);
+ if (update_ggtt_with_blit(i915_vm_to_ggtt(vm)))
+ gen8_ggtt_blit_fixed_pte(ggtt, offset, 1, pte);
+ else
+ gen8_ggtt_insert_page(vm, addr, offset, pat_index, flags);
+
+ ggtt->invalidate(ggtt);
+}
+
static void gen8_ggtt_insert_entries(struct i915_address_space *vm,
struct i915_vma_resource *vma_res,
unsigned int pat_index,
@@ -311,6 +409,53 @@ static void gen8_ggtt_insert_entries(struct i915_address_space *vm,
ggtt->invalidate(ggtt);
}
+static void __gen8_ggtt_insert_entries_blit(struct i915_address_space *vm,
+ struct i915_vma_resource *vma_res,
+ unsigned int pat_index, u32 flags)
+{
+ gen8_pte_t pte_encode;
+ struct i915_ggtt *ggtt = i915_vm_to_ggtt(vm);
+ struct sgt_iter iter;
+ dma_addr_t addr;
+ u64 start, end;
+
+ pte_encode = ggtt->vm.pte_encode(0, pat_index, flags);
+ start = (vma_res->start - vma_res->guard) / I915_GTT_PAGE_SIZE;
+ end = start + vma_res->guard / I915_GTT_PAGE_SIZE;
+ if (gen8_ggtt_blit_fixed_pte(ggtt, start, end - start, vm->scratch[0]->encode))
+ goto err;
+ start = end;
+
+ end += (vma_res->node_size + vma_res->guard) / I915_GTT_PAGE_SIZE;
+ /* TODO: MI_UPDATE_GTT can update 511 entries in a single command. */
+ for_each_sgt_daddr(addr, iter, vma_res->bi.pages) {
+ if (gen8_ggtt_blit_fixed_pte(ggtt, start++, 1, pte_encode | addr))
+ goto err;
+ }
+
+ if (gen8_ggtt_blit_fixed_pte(ggtt, start, end - start, vm->scratch[0]->encode))
+ goto err;
+
+ ggtt->invalidate(ggtt);
+ return;
+
+err:
+ drm_dbg(&ggtt->vm.i915->drm, "falling back to gen8_ggtt_insert_entries\n");
+ gen8_ggtt_insert_entries(vm, vma_res, pat_index, flags);
+}
+
+static void gen8_ggtt_insert_entries_blit(struct i915_address_space *vm,
+ struct i915_vma_resource *vma_res,
+ unsigned int pat_index, u32 flags)
+{
+ if (!update_ggtt_with_blit(i915_vm_to_ggtt(vm))) {
+ gen8_ggtt_insert_entries(vm, vma_res, pat_index, flags);
+ return;
+ }
+
+ __gen8_ggtt_insert_entries_blit(vm, vma_res, pat_index, flags);
+}
+
static void gen8_ggtt_clear_range(struct i915_address_space *vm,
u64 start, u64 length)
{
@@ -332,6 +477,31 @@ static void gen8_ggtt_clear_range(struct i915_address_space *vm,
gen8_set_pte(>t_base[i], scratch_pte);
}
+static void gen8_ggtt_clear_range_blit(struct i915_address_space *vm,
+ u64 start, u64 length)
+{
+ struct i915_ggtt *ggtt = i915_vm_to_ggtt(vm);
+ unsigned int first_entry = start / I915_GTT_PAGE_SIZE;
+ unsigned int num_entries = length / I915_GTT_PAGE_SIZE;
+ const gen8_pte_t scratch_pte = vm->scratch[0]->encode;
+ gen8_pte_t __iomem *gtt_base =
+ (gen8_pte_t __iomem *)ggtt->gsm + first_entry;
+ const int max_entries = ggtt_total_entries(ggtt) - first_entry;
+ int i;
+
+ if (WARN(num_entries > max_entries,
+ "First entry = %d; Num entries = %d (max=%d)\n",
+ first_entry, num_entries, max_entries))
+ num_entries = max_entries;
+
+ if (update_ggtt_with_blit(ggtt)) {
+ gen8_ggtt_blit_fixed_pte(ggtt, first_entry, num_entries, scratch_pte);
+ } else {
+ for (i = 0; i < num_entries; i++)
+ gen8_set_pte(>t_base[i], scratch_pte);
+ }
+}
+
static void gen6_ggtt_insert_page(struct i915_address_space *vm,
dma_addr_t addr,
u64 offset,
@@ -997,6 +1167,12 @@ static int gen8_gmch_probe(struct i915_ggtt *ggtt)
I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND;
}
+ if (i915_ggtt_require_blitter(i915)) {
+ ggtt->vm.scratch_range = gen8_ggtt_clear_range_blit;
+ ggtt->vm.raw_insert_page = gen8_ggtt_insert_page_blit;
+ ggtt->vm.insert_entries = gen8_ggtt_insert_entries_blit;
+ }
+
if (intel_uc_wants_guc(&ggtt->vm.gt->uc))
ggtt->invalidate = guc_ggtt_invalidate;
else
--
2.39.0
More information about the Intel-gfx-trybot
mailing list