From: Nareshkumar Gollakoti <naresh.kumar.g@intel.com>
To: intel-xe@lists.freedesktop.org
Cc: himal.prasad.ghimiray@intel.com, naresh.kumar.g@intel.com
Subject: [PATCH v10 5/6] drm/xe/pt: allow selecting the bind leaf PTE level
Date: Wed, 29 Jul 2026 14:10:45 +0530 [thread overview]
Message-ID: <20260729084046.1050783-6-naresh.kumar.g@intel.com> (raw)
In-Reply-To: <20260729084046.1050783-1-naresh.kumar.g@intel.com>
Add a target_leaf_level field to the page-table bind walk and use it to
control the level at which leaf entries are emitted.
By default, the bind walk emits level-0 leaf PTEs and relies on
xe_pt_hugepte_possible() to select huge mappings when possible. Add an
explicit target leaf level so the walk can stop earlier when the VMA
requests a larger mapping size.
Use level 1 for 2M PDE mappings and level 2 for 1G PDP mappings, while
keeping level 0 for normal mappings. The existing huge-page heuristic
is preserved for the default level-0 path.
This allows the bind path to emit 2M and 1G leaf entries when requested
by the VMA, while still validating alignment and size requirements.
v2
- avoid using max_level to control walk depth
- use target_leaf_level to preserve the normal walk behavior
- keep the default huge-page heuristic only for the level-0 path
- refine commit message
v3
- reword commit message
v4
- allow fallback to smaller huge-page levels for non-zero
target_leaf_level
- avoid constraining clear_pt walks by target_leaf_level
v5(Himal)
- Restrict only intended level in debug page size policy mode
- Allow the normal path to proceed smoothly when
no debug page-size mode is selected.
v8 (Himal)
- Drop
https://patchwork.freedesktop.org/patch/740059/?series=168905&rev=5
patch and populate target_leaf_level from bo flags
- populate target_leaf_level if it is in debug page size mode
otherwise fill with 0 which is having no effect on the normal
flow
v10 (Himal)
- use xe_bo_is_vram() instead of raw VRAM flag checks
so huge-page selection is based on BO VRAM placement.
Signed-off-by: Nareshkumar Gollakoti <naresh.kumar.g@intel.com>
Reviewed-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
---
drivers/gpu/drm/xe/xe_pt.c | 76 +++++++++++++++++++++++++++++++++++++-
1 file changed, 74 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/xe/xe_pt.c b/drivers/gpu/drm/xe/xe_pt.c
index 598c6b2571e7..9517968ed414 100644
--- a/drivers/gpu/drm/xe/xe_pt.c
+++ b/drivers/gpu/drm/xe/xe_pt.c
@@ -302,6 +302,14 @@ struct xe_pt_stage_bind_walk {
bool needs_64K;
/** @clear_pt: clear page table entries during the bind walk */
bool clear_pt;
+ /**
+ * @target_leaf_level: Page-table level at which to emit leaf PTEs
+ * 0 for normal 4K/64K mappings, 1 for 2M huge pages, and 2 for 1G huge
+ * pages. The walk still traverses from the root down; this field tells
+ * xe_pt_stage_bind_entry() to treat the selected level as a leaf instead
+ * of descending further.
+ */
+ u32 target_leaf_level;
/**
* @vma: VMA being mapped
*/
@@ -514,6 +522,39 @@ xe_pt_is_pte_ps64K(u64 addr, u64 next, struct xe_pt_stage_bind_walk *xe_walk)
return xe_walk->found_64K;
}
+static bool xe_pt_huge_leaf_allowed(u64 addr, u64 next, unsigned int level,
+ struct xe_pt_stage_bind_walk *xe_walk)
+{
+ if (xe_walk->clear_pt)
+ return xe_pt_hugepte_possible(addr, next, level, xe_walk);
+
+ if (!xe_debug_page_size_supported(xe_walk->vm->xe))
+ return xe_pt_hugepte_possible(addr, next, level, xe_walk);
+
+ if (!xe_walk->target_leaf_level)
+ return xe_pt_hugepte_possible(addr, next, level, xe_walk);
+
+ if (level == xe_walk->target_leaf_level)
+ return xe_pt_hugepte_possible(addr, next, level, xe_walk);
+
+ return false;
+}
+
+static bool xe_pt_exact_leaf_required_but_invalid(u64 addr, u64 next,
+ unsigned int level,
+ struct xe_pt_stage_bind_walk *xe_walk)
+{
+ struct xe_device *xe = xe_walk->vm->xe;
+
+ if (!xe_debug_page_size_mode_not_none(xe))
+ return false;
+
+ return !xe_walk->clear_pt &&
+ xe_walk->target_leaf_level &&
+ level == xe_walk->target_leaf_level &&
+ !xe_pt_hugepte_possible(addr, next, level, xe_walk);
+}
+
static int
xe_pt_stage_bind_entry(struct xe_ptw *parent, pgoff_t offset,
unsigned int level, u64 addr, u64 next,
@@ -531,8 +572,18 @@ xe_pt_stage_bind_entry(struct xe_ptw *parent, pgoff_t offset,
int ret = 0;
u64 pte;
- /* Is this a leaf entry ?*/
- if (level == 0 || xe_pt_hugepte_possible(addr, next, level, xe_walk)) {
+ if (xe_pt_exact_leaf_required_but_invalid(addr, next, level, xe_walk))
+ return -EINVAL;
+
+ /*
+ * Is this a leaf entry?
+ * Always create a 4K leaf at level 0. For huge pages (level > 0),
+ * validate alignment and size with xe_pt_hugepte_possible().
+ * When target_leaf_level is non-zero, only that huge-page level is
+ * accepted for normal bind walks. Clear walks remain unconstrained so
+ * existing huge leaves can be cleared without descending further.
+ */
+ if (level == 0 || xe_pt_huge_leaf_allowed(addr, next, level, xe_walk)) {
struct xe_res_cursor *curs = xe_walk->curs;
struct xe_bo *bo = xe_vma_bo(xe_walk->vma);
bool is_null_or_purged = xe_vma_is_null(xe_walk->vma) ||
@@ -682,6 +733,26 @@ static bool xe_atomic_for_system(struct xe_vm *vm, struct xe_vma *vma)
(bo && xe_bo_has_single_placement(bo))));
}
+static u32 xe_pt_target_leaf_level_from_bo(struct xe_device *xe,
+ struct xe_vma *vma)
+{
+ struct xe_bo *bo = xe_vma_bo(vma);
+
+ if (!xe_debug_page_size_mode_not_none(xe))
+ return 0;
+
+ if (!bo || !xe_bo_is_vram(bo) || !(bo->flags & XE_BO_FLAG_USER))
+ return 0;
+
+ if (bo->flags & XE_BO_FLAG_NEEDS_1G)
+ return 2;
+
+ if (bo->flags & XE_BO_FLAG_NEEDS_2M)
+ return 1;
+
+ return 0;
+}
+
/**
* xe_pt_stage_bind() - Build a disconnected page-table tree for a given address
* range.
@@ -774,6 +845,7 @@ xe_pt_stage_bind(struct xe_tile *tile, struct xe_vma *vma,
xe_svm_notifier_unlock(vm);
}
+ xe_walk.target_leaf_level = xe_pt_target_leaf_level_from_bo(xe, vma);
xe_walk.needs_64K = (vm->flags & XE_VM_FLAG_64K);
if (clear_pt)
goto walk_pt;
--
2.43.0
next prev parent reply other threads:[~2026-07-29 8:40 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-07-29 8:40 [PATCH v10 0/6] drm/xe: add page size allocation mode control and coverage Nareshkumar Gollakoti
2026-07-29 8:40 ` [PATCH v10 1/6] drm/xe: add page size allocation control state to xe_device Nareshkumar Gollakoti
2026-07-29 8:40 ` [PATCH v10 2/6] drm/xe/debugfs: add page-size allocation mode knob Nareshkumar Gollakoti
2026-07-29 8:40 ` [PATCH v10 3/6] drm/xe: add XE_BO_FLAG_NEEDS_1G for minimum page-size sizing Nareshkumar Gollakoti
2026-07-29 8:40 ` [PATCH v10 4/6] drm/xe: apply debug page-size allocation policy to user BOs Nareshkumar Gollakoti
2026-07-29 8:40 ` Nareshkumar Gollakoti [this message]
2026-07-29 8:40 ` [PATCH v10 6/6] drm/xe/tests: add live KUnit coverage for BO page-size allocation modes Nareshkumar Gollakoti
-- strict thread matches above, loose matches on Subject: below --
2026-07-29 12:18 [PATCH v10 0/6] drm/xe: add page size allocation mode control and coverage Nareshkumar Gollakoti
2026-07-29 12:18 ` [PATCH v10 5/6] drm/xe/pt: allow selecting the bind leaf PTE level Nareshkumar Gollakoti
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=20260729084046.1050783-6-naresh.kumar.g@intel.com \
--to=naresh.kumar.g@intel.com \
--cc=himal.prasad.ghimiray@intel.com \
--cc=intel-xe@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