From: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
To: intel-xe@lists.freedesktop.org
Cc: "Himal Prasad Ghimiray" <himal.prasad.ghimiray@intel.com>,
"Matthew Brost" <matthew.brost@intel.com>,
"Thomas Hellström" <thomas.hellstrom@linux.intel.com>
Subject: [PATCH v4 2/5] drm/xe: Merge adjacent default-attribute VMAs during garbage collection
Date: Mon, 3 Nov 2025 11:39:54 +0530 [thread overview]
Message-ID: <20251103060957.957760-3-himal.prasad.ghimiray@intel.com> (raw)
In-Reply-To: <20251103060957.957760-1-himal.prasad.ghimiray@intel.com>
While restoring default memory attributes for VMAs during garbage
collection, extend the target range by checking neighboring VMAs. If
adjacent VMAs are CPU-address-mirrored and have default attributes,
include them in the mergeable range to reduce fragmentation and improve
VMA reuse.
v2
-Rebase
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Signed-off-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
---
drivers/gpu/drm/xe/xe_svm.c | 62 +++++++++++++++++++++----------------
1 file changed, 36 insertions(+), 26 deletions(-)
diff --git a/drivers/gpu/drm/xe/xe_svm.c b/drivers/gpu/drm/xe/xe_svm.c
index 13af589715a7..585e3eccc876 100644
--- a/drivers/gpu/drm/xe/xe_svm.c
+++ b/drivers/gpu/drm/xe/xe_svm.c
@@ -286,19 +286,21 @@ static int __xe_svm_garbage_collector(struct xe_vm *vm,
return 0;
}
-static int xe_svm_range_set_default_attr(struct xe_vm *vm, u64 range_start, u64 range_end)
+static void xe_vma_set_default_attributes(struct xe_vma *vma)
+{
+ vma->attr.preferred_loc.devmem_fd = DRM_XE_PREFERRED_LOC_DEFAULT_DEVICE;
+ vma->attr.preferred_loc.migration_policy = DRM_XE_MIGRATE_ALL_PAGES;
+ vma->attr.pat_index = vma->attr.default_pat_index;
+ vma->attr.atomic_access = DRM_XE_ATOMIC_UNDEFINED;
+}
+
+static int xe_svm_range_set_default_attr(struct xe_vm *vm, u64 start, u64 end)
{
struct xe_vma *vma;
- struct xe_vma_mem_attr default_attr = {
- .preferred_loc = {
- .devmem_fd = DRM_XE_PREFERRED_LOC_DEFAULT_DEVICE,
- .migration_policy = DRM_XE_MIGRATE_ALL_PAGES,
- },
- .atomic_access = DRM_XE_ATOMIC_UNDEFINED,
- };
- int err = 0;
+ bool has_default_attr;
+ int err;
- vma = xe_vm_find_vma_by_addr(vm, range_start);
+ vma = xe_vm_find_vma_by_addr(vm, start);
if (!vma)
return -EINVAL;
@@ -307,25 +309,33 @@ static int xe_svm_range_set_default_attr(struct xe_vm *vm, u64 range_start, u64
return 0;
}
- if (xe_vma_has_default_mem_attrs(vma))
- return 0;
-
vm_dbg(&vm->xe->drm, "Existing VMA start=0x%016llx, vma_end=0x%016llx",
xe_vma_start(vma), xe_vma_end(vma));
- if (xe_vma_start(vma) == range_start && xe_vma_end(vma) == range_end) {
- default_attr.pat_index = vma->attr.default_pat_index;
- default_attr.default_pat_index = vma->attr.default_pat_index;
- vma->attr = default_attr;
- } else {
- vm_dbg(&vm->xe->drm, "Split VMA start=0x%016llx, vma_end=0x%016llx",
- range_start, range_end);
- err = xe_vm_alloc_cpu_addr_mirror_vma(vm, range_start, range_end - range_start);
- if (err) {
- drm_warn(&vm->xe->drm, "VMA SPLIT failed: %pe\n", ERR_PTR(err));
- xe_vm_kill(vm, true);
- return err;
- }
+ has_default_attr = xe_vma_has_default_mem_attrs(vma);
+
+ if (has_default_attr) {
+ if (xe_svm_has_mapping(vm, xe_vma_start(vma), xe_vma_end(vma)))
+ return 0;
+
+ start = xe_vma_start(vma);
+ end = xe_vma_end(vma);
+ } else if (xe_vma_start(vma) == start && xe_vma_end(vma) == end) {
+ xe_vma_set_default_attributes(vma);
+ }
+
+ xe_vm_find_cpu_addr_mirror_vma_range(vm, &start, &end);
+
+ if (xe_vma_start(vma) == start && xe_vma_end(vma) == end && has_default_attr)
+ return 0;
+
+ vm_dbg(&vm->xe->drm, "New VMA start=0x%016llx, vma_end=0x%016llx", start, end);
+
+ err = xe_vm_alloc_cpu_addr_mirror_vma(vm, start, end - start);
+ if (err) {
+ drm_warn(&vm->xe->drm, "New VMA MAP failed: %pe\n", ERR_PTR(err));
+ xe_vm_kill(vm, true);
+ return err;
}
/*
--
2.34.1
next prev parent reply other threads:[~2025-11-03 5:41 UTC|newest]
Thread overview: 15+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-11-03 6:09 [PATCH v4 0/5] Improve VMA merging for CPU-address-mirrored mappings Himal Prasad Ghimiray
2025-11-03 5:53 ` ✓ CI.KUnit: success for Improve VMA merging for CPU-address-mirrored mappings (rev4) Patchwork
2025-11-03 6:09 ` [PATCH v4 1/5] drm/xe: Add helper to extend CPU-mirrored VMA range for merge Himal Prasad Ghimiray
2025-11-22 21:48 ` Matthew Brost
2025-11-03 6:09 ` Himal Prasad Ghimiray [this message]
2025-11-22 21:55 ` [PATCH v4 2/5] drm/xe: Merge adjacent default-attribute VMAs during garbage collection Matthew Brost
2025-11-03 6:09 ` [PATCH v4 3/5] drm/xe/svm: Extend MAP range to reduce vma fragmentation Himal Prasad Ghimiray
2025-11-03 6:09 ` [PATCH v4 4/5] drm/xe/svm: Enable UNMAP for VMA merging operations Himal Prasad Ghimiray
2025-11-22 22:01 ` Matthew Brost
2025-11-24 6:01 ` Ghimiray, Himal Prasad
2025-11-24 16:58 ` Matthew Brost
2025-11-03 6:09 ` [PATCH v4 5/5] drm/xe/vm: Skip ufence association for CPU address mirror VMA during MAP Himal Prasad Ghimiray
2025-11-22 22:02 ` Matthew Brost
2025-11-03 7:04 ` ✓ Xe.CI.BAT: success for Improve VMA merging for CPU-address-mirrored mappings (rev4) Patchwork
2025-11-03 8:20 ` ✗ Xe.CI.Full: failure " 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=20251103060957.957760-3-himal.prasad.ghimiray@intel.com \
--to=himal.prasad.ghimiray@intel.com \
--cc=intel-xe@lists.freedesktop.org \
--cc=matthew.brost@intel.com \
--cc=thomas.hellstrom@linux.intel.com \
/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