intel-gfx.lists.freedesktop.org archive mirror
 help / color / mirror / Atom feed
* [Intel-gfx] [PATCH 1/8] drm/i915/gt: Lock intel_engine_apply_whitelist with uncore->lock
@ 2020-11-17 11:01 Chris Wilson
  2020-11-17 11:01 ` [Intel-gfx] [PATCH 2/8] drm/i915/gt: Refactor _wa_add to reuse wa_index and wa_list_grow Chris Wilson
                   ` (7 more replies)
  0 siblings, 8 replies; 18+ messages in thread
From: Chris Wilson @ 2020-11-17 11:01 UTC (permalink / raw)
  To: intel-gfx; +Cc: Chris Wilson

From: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>

Refactor intel_engine_apply_whitelist into locked and unlocked versions
so that a caller who already has the lock can apply whitelist.

v2: Fix sparse warning
v3: (Chris)
- Drop prefix and suffix for static function
- Use longest to shortest line ordering for variable declaration

Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
Reviewed-by: Chris Wilson <chris.p.wilson@intel.com>
---
 drivers/gpu/drm/i915/gt/intel_workarounds.c | 44 +++++++++++++++------
 1 file changed, 31 insertions(+), 13 deletions(-)

diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.c b/drivers/gpu/drm/i915/gt/intel_workarounds.c
index adc9a8ea410a..c49083957074 100644
--- a/drivers/gpu/drm/i915/gt/intel_workarounds.c
+++ b/drivers/gpu/drm/i915/gt/intel_workarounds.c
@@ -1339,7 +1339,8 @@ void intel_gt_init_workarounds(struct drm_i915_private *i915)
 }
 
 static enum forcewake_domains
-wal_get_fw_for_rmw(struct intel_uncore *uncore, const struct i915_wa_list *wal)
+wal_get_fw(struct intel_uncore *uncore, const struct i915_wa_list *wal,
+	   unsigned int op)
 {
 	enum forcewake_domains fw = 0;
 	struct i915_wa *wa;
@@ -1348,8 +1349,7 @@ wal_get_fw_for_rmw(struct intel_uncore *uncore, const struct i915_wa_list *wal)
 	for (i = 0, wa = wal->list; i < wal->count; i++, wa++)
 		fw |= intel_uncore_forcewake_for_reg(uncore,
 						     wa->reg,
-						     FW_REG_READ |
-						     FW_REG_WRITE);
+						     op);
 
 	return fw;
 }
@@ -1379,7 +1379,7 @@ wa_list_apply(struct intel_uncore *uncore, const struct i915_wa_list *wal)
 	if (!wal->count)
 		return;
 
-	fw = wal_get_fw_for_rmw(uncore, wal);
+	fw = wal_get_fw(uncore, wal, FW_REG_READ | FW_REG_WRITE);
 
 	spin_lock_irqsave(&uncore->lock, flags);
 	intel_uncore_forcewake_get__locked(uncore, fw);
@@ -1705,27 +1705,45 @@ void intel_engine_init_whitelist(struct intel_engine_cs *engine)
 	wa_init_finish(w);
 }
 
-void intel_engine_apply_whitelist(struct intel_engine_cs *engine)
+static void __engine_apply_whitelist(struct intel_engine_cs *engine)
 {
 	const struct i915_wa_list *wal = &engine->whitelist;
 	struct intel_uncore *uncore = engine->uncore;
 	const u32 base = engine->mmio_base;
+	enum forcewake_domains fw;
 	struct i915_wa *wa;
 	unsigned int i;
 
-	if (!wal->count)
-		return;
+	lockdep_assert_held(&uncore->lock);
+
+	fw = wal_get_fw(uncore, wal, FW_REG_WRITE);
+	intel_uncore_forcewake_get__locked(uncore, fw);
 
 	for (i = 0, wa = wal->list; i < wal->count; i++, wa++)
-		intel_uncore_write(uncore,
-				   RING_FORCE_TO_NONPRIV(base, i),
-				   i915_mmio_reg_offset(wa->reg));
+		intel_uncore_write_fw(uncore,
+				      RING_FORCE_TO_NONPRIV(base, i),
+				      i915_mmio_reg_offset(wa->reg));
 
 	/* And clear the rest just in case of garbage */
 	for (; i < RING_MAX_NONPRIV_SLOTS; i++)
-		intel_uncore_write(uncore,
-				   RING_FORCE_TO_NONPRIV(base, i),
-				   i915_mmio_reg_offset(RING_NOPID(base)));
+		intel_uncore_write_fw(uncore,
+				      RING_FORCE_TO_NONPRIV(base, i),
+				      i915_mmio_reg_offset(RING_NOPID(base)));
+
+	intel_uncore_forcewake_put__locked(uncore, fw);
+}
+
+void intel_engine_apply_whitelist(struct intel_engine_cs *engine)
+{
+	const struct i915_wa_list *wal = &engine->whitelist;
+	unsigned long flags;
+
+	if (!wal->count)
+		return;
+
+	spin_lock_irqsave(&engine->uncore->lock, flags);
+	__engine_apply_whitelist(engine);
+	spin_unlock_irqrestore(&engine->uncore->lock, flags);
 }
 
 static void
-- 
2.20.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

^ permalink raw reply related	[flat|nested] 18+ messages in thread
* [Intel-gfx] [PATCH 2/8] drm/i915/gt: Refactor _wa_add to reuse wa_index and wa_list_grow
  2020-11-17 11:01 [Intel-gfx] [PATCH 1/8] drm/i915/gt: Lock intel_engine_apply_whitelist with uncore->lock Chris Wilson
@ 2020-11-17 11:01 ` Chris Wilson
  2020-11-17 18:14   ` Umesh Nerlige Ramappa
  2020-11-17 11:01 ` [Intel-gfx] [PATCH 3/8] drm/i915/gt: Check for conflicting RING_NONPRIV Chris Wilson
                   ` (6 subsequent siblings)
  7 siblings, 1 reply; 18+ messages in thread
From: Chris Wilson @ 2020-11-17 11:01 UTC (permalink / raw)
  To: intel-gfx

From: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>

Switch the search and grow code of the _wa_add to use _wa_index and
_wa_list_grow.

Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
---
 drivers/gpu/drm/i915/gt/intel_workarounds.c | 124 +++++++++++---------
 1 file changed, 71 insertions(+), 53 deletions(-)

diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.c b/drivers/gpu/drm/i915/gt/intel_workarounds.c
index c49083957074..e50c72d2b3f1 100644
--- a/drivers/gpu/drm/i915/gt/intel_workarounds.c
+++ b/drivers/gpu/drm/i915/gt/intel_workarounds.c
@@ -91,20 +91,19 @@ static void wa_init_start(struct i915_wa_list *wal, const char *name, const char
 
 #define WA_LIST_CHUNK (1 << 4)
 
-static void wa_init_finish(struct i915_wa_list *wal)
+static void wa_trim(struct i915_wa_list *wal, gfp_t gfp)
 {
+	struct i915_wa *list;
+
 	/* Trim unused entries. */
-	if (!IS_ALIGNED(wal->count, WA_LIST_CHUNK)) {
-		struct i915_wa *list = kmemdup(wal->list,
-					       wal->count * sizeof(*list),
-					       GFP_KERNEL);
-
-		if (list) {
-			kfree(wal->list);
-			wal->list = list;
-		}
-	}
+	list = krealloc(wal->list, wal->count * sizeof(*list), gfp);
+	if (list)
+		wal->list = list;
+}
 
+static void wa_init_finish(struct i915_wa_list *wal)
+{
+	wa_trim(wal, GFP_KERNEL);
 	if (!wal->count)
 		return;
 
@@ -112,57 +111,60 @@ static void wa_init_finish(struct i915_wa_list *wal)
 			 wal->wa_count, wal->name, wal->engine_name);
 }
 
-static void _wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
+static int wa_index(struct i915_wa_list *wal, i915_reg_t reg)
 {
-	unsigned int addr = i915_mmio_reg_offset(wa->reg);
-	unsigned int start = 0, end = wal->count;
-	const unsigned int grow = WA_LIST_CHUNK;
-	struct i915_wa *wa_;
+	unsigned int addr = i915_mmio_reg_offset(reg);
+	int start = 0, end = wal->count;
 
-	GEM_BUG_ON(!is_power_of_2(grow));
+	/* addr and wal->list[].reg, both include the R/W flags */
+	while (start < end) {
+		unsigned int mid = start + (end - start) / 2;
 
-	if (IS_ALIGNED(wal->count, grow)) { /* Either uninitialized or full. */
-		struct i915_wa *list;
+		if (i915_mmio_reg_offset(wal->list[mid].reg) < addr)
+			start = mid + 1;
+		else if (i915_mmio_reg_offset(wal->list[mid].reg) > addr)
+			end = mid;
+		else
+			return mid;
+	}
 
-		list = kmalloc_array(ALIGN(wal->count + 1, grow), sizeof(*wa),
-				     GFP_KERNEL);
-		if (!list) {
-			DRM_ERROR("No space for workaround init!\n");
-			return;
-		}
+	return -ENOENT;
+}
 
-		if (wal->list) {
-			memcpy(list, wal->list, sizeof(*wa) * wal->count);
-			kfree(wal->list);
-		}
+static int wa_list_grow(struct i915_wa_list *wal, size_t count, gfp_t gfp)
+{
+	struct i915_wa *list;
 
-		wal->list = list;
-	}
+	list = krealloc(wal->list, count * sizeof(*list), gfp);
+	if (!list)
+		return -ENOMEM;
 
-	while (start < end) {
-		unsigned int mid = start + (end - start) / 2;
+	wal->list = list;
+	return 0;
+}
 
-		if (i915_mmio_reg_offset(wal->list[mid].reg) < addr) {
-			start = mid + 1;
-		} else if (i915_mmio_reg_offset(wal->list[mid].reg) > addr) {
-			end = mid;
-		} else {
-			wa_ = &wal->list[mid];
-
-			if ((wa->clr | wa_->clr) && !(wa->clr & ~wa_->clr)) {
-				DRM_ERROR("Discarding overwritten w/a for reg %04x (clear: %08x, set: %08x)\n",
-					  i915_mmio_reg_offset(wa_->reg),
-					  wa_->clr, wa_->set);
-
-				wa_->set &= ~wa->clr;
-			}
-
-			wal->wa_count++;
-			wa_->set |= wa->set;
-			wa_->clr |= wa->clr;
-			wa_->read |= wa->read;
-			return;
+static void __wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
+{
+	struct i915_wa *wa_;
+	int index;
+
+	index = wa_index(wal, wa->reg);
+	if (index >= 0) {
+		wa_ = &wal->list[index];
+
+		if ((wa->clr | wa_->clr) && !(wa->clr & ~wa_->clr)) {
+			DRM_ERROR("Discarding overwritten w/a for reg %04x (clear: %08x, set: %08x)\n",
+				  i915_mmio_reg_offset(wa_->reg),
+				  wa_->clr, wa_->set);
+
+			wa_->set &= ~wa->clr;
 		}
+
+		wal->wa_count++;
+		wa_->set |= wa->set;
+		wa_->clr |= wa->clr;
+		wa_->read |= wa->read;
+		return;
 	}
 
 	wal->wa_count++;
@@ -180,6 +182,22 @@ static void _wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
 	}
 }
 
+static void _wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
+{
+	const unsigned int grow = WA_LIST_CHUNK;
+
+	GEM_BUG_ON(!is_power_of_2(grow));
+
+	if (IS_ALIGNED(wal->count, grow) && /* Either uninitialized or full. */
+	    wa_list_grow(wal, ALIGN(wal->count + 1, grow), GFP_KERNEL)) {
+		DRM_ERROR("Unable to store w/a for reg %04x\n",
+			  i915_mmio_reg_offset(wa->reg));
+		return;
+	}
+
+	__wa_add(wal, wa);
+}
+
 static void wa_add(struct i915_wa_list *wal, i915_reg_t reg,
 		   u32 clear, u32 set, u32 read_mask)
 {
-- 
2.20.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

^ permalink raw reply related	[flat|nested] 18+ messages in thread
* [Intel-gfx] [PATCH 3/8] drm/i915/gt: Check for conflicting RING_NONPRIV
  2020-11-17 11:01 [Intel-gfx] [PATCH 1/8] drm/i915/gt: Lock intel_engine_apply_whitelist with uncore->lock Chris Wilson
  2020-11-17 11:01 ` [Intel-gfx] [PATCH 2/8] drm/i915/gt: Refactor _wa_add to reuse wa_index and wa_list_grow Chris Wilson
@ 2020-11-17 11:01 ` Chris Wilson
  2020-11-17 18:08   ` Umesh Nerlige Ramappa
  2020-11-17 11:01 ` [Intel-gfx] [PATCH 4/8] drm/i915/gt: Enable dynamic adjustment of RING_NONPRIV Chris Wilson
                   ` (5 subsequent siblings)
  7 siblings, 1 reply; 18+ messages in thread
From: Chris Wilson @ 2020-11-17 11:01 UTC (permalink / raw)
  To: intel-gfx; +Cc: Chris Wilson

Strip the encoded bits from the register offset so that we only use the
address for looking up the RING_NONPRIV entry.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
 drivers/gpu/drm/i915/gt/intel_workarounds.c | 64 +++++++++++++--------
 1 file changed, 41 insertions(+), 23 deletions(-)

diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.c b/drivers/gpu/drm/i915/gt/intel_workarounds.c
index e50c72d2b3f1..290aa277ab10 100644
--- a/drivers/gpu/drm/i915/gt/intel_workarounds.c
+++ b/drivers/gpu/drm/i915/gt/intel_workarounds.c
@@ -111,18 +111,43 @@ static void wa_init_finish(struct i915_wa_list *wal)
 			 wal->wa_count, wal->name, wal->engine_name);
 }
 
+static u32 reg_offset(i915_reg_t reg)
+{
+	return i915_mmio_reg_offset(reg) & RING_FORCE_TO_NONPRIV_ADDRESS_MASK;
+}
+
+static u32 reg_flags(i915_reg_t reg)
+{
+	return i915_mmio_reg_offset(reg) & ~RING_FORCE_TO_NONPRIV_ADDRESS_MASK;
+}
+
+static inline bool is_nonpriv_flags_valid(u32 flags)
+{
+	/* Check only valid flag bits are set */
+	if (flags & ~RING_FORCE_TO_NONPRIV_MASK_VALID)
+		return false;
+
+	/* NB: Only 3 out of 4 enum values are valid for access field */
+	if ((flags & RING_FORCE_TO_NONPRIV_ACCESS_MASK) ==
+	    RING_FORCE_TO_NONPRIV_ACCESS_INVALID)
+		return false;
+
+	return true;
+}
+
 static int wa_index(struct i915_wa_list *wal, i915_reg_t reg)
 {
-	unsigned int addr = i915_mmio_reg_offset(reg);
 	int start = 0, end = wal->count;
+	u32 addr = reg_offset(reg);
 
 	/* addr and wal->list[].reg, both include the R/W flags */
 	while (start < end) {
 		unsigned int mid = start + (end - start) / 2;
+		u32 pos = reg_offset(wal->list[mid].reg);
 
-		if (i915_mmio_reg_offset(wal->list[mid].reg) < addr)
+		if (pos < addr)
 			start = mid + 1;
-		else if (i915_mmio_reg_offset(wal->list[mid].reg) > addr)
+		else if (pos > addr)
 			end = mid;
 		else
 			return mid;
@@ -148,13 +173,22 @@ static void __wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
 	struct i915_wa *wa_;
 	int index;
 
+	GEM_BUG_ON(!is_nonpriv_flags_valid(reg_flags(wa->reg)));
+
 	index = wa_index(wal, wa->reg);
 	if (index >= 0) {
 		wa_ = &wal->list[index];
 
+		if (i915_mmio_reg_offset(wa->reg) !=
+		    i915_mmio_reg_offset(wa_->reg)) {
+			DRM_ERROR("Discarding incompatible w/a for reg %04x\n",
+				  reg_offset(wa->reg));
+			return;
+		}
+
 		if ((wa->clr | wa_->clr) && !(wa->clr & ~wa_->clr)) {
 			DRM_ERROR("Discarding overwritten w/a for reg %04x (clear: %08x, set: %08x)\n",
-				  i915_mmio_reg_offset(wa_->reg),
+				  reg_offset(wa_->reg),
 				  wa_->clr, wa_->set);
 
 			wa_->set &= ~wa->clr;
@@ -172,10 +206,8 @@ static void __wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
 	*wa_ = *wa;
 
 	while (wa_-- > wal->list) {
-		GEM_BUG_ON(i915_mmio_reg_offset(wa_[0].reg) ==
-			   i915_mmio_reg_offset(wa_[1].reg));
-		if (i915_mmio_reg_offset(wa_[1].reg) >
-		    i915_mmio_reg_offset(wa_[0].reg))
+		GEM_BUG_ON(reg_offset(wa_[0].reg) == reg_offset(wa_[1].reg));
+		if (reg_offset(wa_[1].reg) > reg_offset(wa_[0].reg))
 			break;
 
 		swap(wa_[1], wa_[0]);
@@ -191,7 +223,7 @@ static void _wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
 	if (IS_ALIGNED(wal->count, grow) && /* Either uninitialized or full. */
 	    wa_list_grow(wal, ALIGN(wal->count + 1, grow), GFP_KERNEL)) {
 		DRM_ERROR("Unable to store w/a for reg %04x\n",
-			  i915_mmio_reg_offset(wa->reg));
+			  reg_offset(wa->reg));
 		return;
 	}
 
@@ -1443,20 +1475,6 @@ bool intel_gt_verify_workarounds(struct intel_gt *gt, const char *from)
 	return wa_list_verify(gt->uncore, &gt->i915->gt_wa_list, from);
 }
 
-static inline bool is_nonpriv_flags_valid(u32 flags)
-{
-	/* Check only valid flag bits are set */
-	if (flags & ~RING_FORCE_TO_NONPRIV_MASK_VALID)
-		return false;
-
-	/* NB: Only 3 out of 4 enum values are valid for access field */
-	if ((flags & RING_FORCE_TO_NONPRIV_ACCESS_MASK) ==
-	    RING_FORCE_TO_NONPRIV_ACCESS_INVALID)
-		return false;
-
-	return true;
-}
-
 static void
 whitelist_reg_ext(struct i915_wa_list *wal, i915_reg_t reg, u32 flags)
 {
-- 
2.20.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

^ permalink raw reply related	[flat|nested] 18+ messages in thread
* [Intel-gfx] [PATCH 4/8] drm/i915/gt: Enable dynamic adjustment of RING_NONPRIV
  2020-11-17 11:01 [Intel-gfx] [PATCH 1/8] drm/i915/gt: Lock intel_engine_apply_whitelist with uncore->lock Chris Wilson
  2020-11-17 11:01 ` [Intel-gfx] [PATCH 2/8] drm/i915/gt: Refactor _wa_add to reuse wa_index and wa_list_grow Chris Wilson
  2020-11-17 11:01 ` [Intel-gfx] [PATCH 3/8] drm/i915/gt: Check for conflicting RING_NONPRIV Chris Wilson
@ 2020-11-17 11:01 ` Chris Wilson
  2020-11-17 20:34   ` Umesh Nerlige Ramappa
  2020-11-17 11:01 ` [Intel-gfx] [PATCH 5/8] drm/i915/perf: Ensure observation logic is not clock gated Chris Wilson
                   ` (4 subsequent siblings)
  7 siblings, 1 reply; 18+ messages in thread
From: Chris Wilson @ 2020-11-17 11:01 UTC (permalink / raw)
  To: intel-gfx; +Cc: Chris Wilson

The OA subsystem would like to enable its privileged clients access to
the OA registers from execbuf. This requires temporarily removing the
HW validation from those registers for the duration of the OA client,
for which we need to allow OA to dynamically adjust the set of
RING_NONPRIV.

Care must still be taken since the RING_NONPRIV are global, so any and
all contexts that run at the same time as the OA client, will also be
able to adjust the registers from their execbuf.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
 drivers/gpu/drm/i915/gt/intel_workarounds.c   |  59 +++++
 drivers/gpu/drm/i915/gt/intel_workarounds.h   |   7 +
 .../gpu/drm/i915/gt/selftest_workarounds.c    | 236 ++++++++++++++++++
 3 files changed, 302 insertions(+)

diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.c b/drivers/gpu/drm/i915/gt/intel_workarounds.c
index 290aa277ab10..c918739df6a6 100644
--- a/drivers/gpu/drm/i915/gt/intel_workarounds.c
+++ b/drivers/gpu/drm/i915/gt/intel_workarounds.c
@@ -230,6 +230,18 @@ static void _wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
 	__wa_add(wal, wa);
 }
 
+static void _wa_del(struct i915_wa_list *wal, i915_reg_t reg)
+{
+	struct i915_wa *wa = wal->list;
+	int index;
+
+	index = wa_index(wal, reg);
+	if (GEM_DEBUG_WARN_ON(index < 0))
+		return;
+
+	memmove(wa + index, wa + index + 1, --wal->count - index);
+}
+
 static void wa_add(struct i915_wa_list *wal, i915_reg_t reg,
 		   u32 clear, u32 set, u32 read_mask)
 {
@@ -2116,6 +2128,53 @@ void intel_engine_init_workarounds(struct intel_engine_cs *engine)
 	wa_init_finish(wal);
 }
 
+int intel_engine_allow_user_register_access(struct intel_engine_cs *engine,
+					    const i915_reg_t *reg,
+					    unsigned int count)
+{
+	struct i915_wa_list *wal = &engine->whitelist;
+	unsigned long flags;
+	int err;
+
+	if (GEM_DEBUG_WARN_ON(wal->count + count >= RING_MAX_NONPRIV_SLOTS))
+		return -ENOSPC;
+
+	spin_lock_irqsave(&engine->uncore->lock, flags);
+
+	err = wa_list_grow(wal, wal->count + count, GFP_ATOMIC | __GFP_NOWARN);
+	if (err)
+		goto out;
+
+	while (count--) {
+		struct i915_wa wa = { .reg = *reg++ };
+
+		__wa_add(wal, &wa);
+	}
+
+	__engine_apply_whitelist(engine);
+
+out:
+	spin_unlock_irqrestore(&engine->uncore->lock, flags);
+	return err;
+}
+
+void intel_engine_deny_user_register_access(struct intel_engine_cs *engine,
+					    const i915_reg_t *reg,
+					    unsigned int count)
+{
+	struct i915_wa_list *wal = &engine->whitelist;
+	unsigned long flags;
+
+	spin_lock_irqsave(&engine->uncore->lock, flags);
+
+	while (count--)
+		_wa_del(wal, *reg++);
+
+	__engine_apply_whitelist(engine);
+
+	spin_unlock_irqrestore(&engine->uncore->lock, flags);
+}
+
 void intel_engine_apply_workarounds(struct intel_engine_cs *engine)
 {
 	wa_list_apply(engine->uncore, &engine->wa_list);
diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.h b/drivers/gpu/drm/i915/gt/intel_workarounds.h
index 8c9c769c2204..9c691493ea56 100644
--- a/drivers/gpu/drm/i915/gt/intel_workarounds.h
+++ b/drivers/gpu/drm/i915/gt/intel_workarounds.h
@@ -37,4 +37,11 @@ void intel_engine_apply_workarounds(struct intel_engine_cs *engine);
 int intel_engine_verify_workarounds(struct intel_engine_cs *engine,
 				    const char *from);
 
+int intel_engine_allow_user_register_access(struct intel_engine_cs *engine,
+					    const i915_reg_t *reg,
+					    unsigned int count);
+void intel_engine_deny_user_register_access(struct intel_engine_cs *engine,
+					    const i915_reg_t *reg,
+					    unsigned int count);
+
 #endif
diff --git a/drivers/gpu/drm/i915/gt/selftest_workarounds.c b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
index 61a0532d0f3d..6f526199a11a 100644
--- a/drivers/gpu/drm/i915/gt/selftest_workarounds.c
+++ b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
@@ -1127,6 +1127,241 @@ static int live_isolated_whitelist(void *arg)
 	return err;
 }
 
+static int rmw_reg(struct intel_engine_cs *engine, const i915_reg_t reg)
+{
+	const u32 values[] = {
+		0x00000000,
+		0x01010101,
+		0x10100101,
+		0x03030303,
+		0x30300303,
+		0x05050505,
+		0x50500505,
+		0x0f0f0f0f,
+		0xf00ff00f,
+		0x10101010,
+		0xf0f01010,
+		0x30303030,
+		0xa0a03030,
+		0x50505050,
+		0xc0c05050,
+		0xf0f0f0f0,
+		0x11111111,
+		0x33333333,
+		0x55555555,
+		0x0000ffff,
+		0x00ff00ff,
+		0xff0000ff,
+		0xffff00ff,
+		0xffffffff,
+	};
+	struct i915_vma *vma, *batch;
+	struct intel_context *ce;
+	struct i915_request *rq;
+	u32 srm, lrm, idx;
+	u32 *cs, *results;
+	u64 addr;
+	int err;
+	int v;
+
+	ce = intel_context_create(engine);
+	if (IS_ERR(ce))
+		return PTR_ERR(ce);
+
+	vma = create_scratch(ce->vm, 1);
+	if (IS_ERR(vma)) {
+		err = PTR_ERR(vma);
+		goto out_context;
+	}
+
+	batch = create_batch(ce->vm);
+	if (IS_ERR(batch)) {
+		err = PTR_ERR(batch);
+		goto out_vma;
+	}
+
+	srm = MI_STORE_REGISTER_MEM;
+	lrm = MI_LOAD_REGISTER_MEM;
+	if (INTEL_GEN(ce->vm->i915) >= 8)
+		lrm++, srm++;
+
+	addr = vma->node.start;
+
+	cs = i915_gem_object_pin_map(batch->obj, I915_MAP_WC);
+	if (IS_ERR(cs)) {
+		err = PTR_ERR(cs);
+		goto out_batch;
+	}
+
+	/* SRM original */
+	*cs++ = srm;
+	*cs++ = i915_mmio_reg_offset(reg);
+	*cs++ = lower_32_bits(addr);
+	*cs++ = upper_32_bits(addr);
+
+	idx = 1;
+	for (v = 0; v < ARRAY_SIZE(values); v++) {
+		/* LRI garbage */
+		*cs++ = MI_LOAD_REGISTER_IMM(1);
+		*cs++ = i915_mmio_reg_offset(reg);
+		*cs++ = values[v];
+
+		/* SRM result */
+		*cs++ = srm;
+		*cs++ = i915_mmio_reg_offset(reg);
+		*cs++ = lower_32_bits(addr + sizeof(u32) * idx);
+		*cs++ = upper_32_bits(addr + sizeof(u32) * idx);
+		idx++;
+	}
+	for (v = 0; v < ARRAY_SIZE(values); v++) {
+		/* LRI garbage */
+		*cs++ = MI_LOAD_REGISTER_IMM(1);
+		*cs++ = i915_mmio_reg_offset(reg);
+		*cs++ = ~values[v];
+
+		/* SRM result */
+		*cs++ = srm;
+		*cs++ = i915_mmio_reg_offset(reg);
+		*cs++ = lower_32_bits(addr + sizeof(u32) * idx);
+		*cs++ = upper_32_bits(addr + sizeof(u32) * idx);
+		idx++;
+	}
+
+	/* LRM original -- don't leave garbage in the context! */
+	*cs++ = lrm;
+	*cs++ = i915_mmio_reg_offset(reg);
+	*cs++ = lower_32_bits(addr);
+	*cs++ = upper_32_bits(addr);
+
+	*cs++ = MI_BATCH_BUFFER_END;
+
+	i915_gem_object_flush_map(batch->obj);
+	i915_gem_object_unpin_map(batch->obj);
+	intel_gt_chipset_flush(engine->gt);
+
+	rq = intel_context_create_request(ce);
+	if (IS_ERR(rq)) {
+		err = PTR_ERR(rq);
+		goto out_batch;
+	}
+
+	if (engine->emit_init_breadcrumb) { /* Be nice if we hang */
+		err = engine->emit_init_breadcrumb(rq);
+		if (err)
+			goto err_request;
+	}
+
+	i915_vma_lock(batch);
+	err = i915_request_await_object(rq, batch->obj, false);
+	if (err == 0)
+		err = i915_vma_move_to_active(batch, rq, 0);
+	i915_vma_unlock(batch);
+	if (err)
+		goto err_request;
+
+	i915_vma_lock(vma);
+	err = i915_request_await_object(rq, vma->obj, true);
+	if (err == 0)
+		err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
+	i915_vma_unlock(vma);
+	if (err)
+		goto err_request;
+
+	err = engine->emit_bb_start(rq, batch->node.start, PAGE_SIZE, 0);
+	if (err)
+		goto err_request;
+
+err_request:
+	err = request_add_sync(rq, err);
+	if (err) {
+		pr_err("%s: Futzing %04x timedout; cancelling test\n",
+		       engine->name, i915_mmio_reg_offset(reg));
+		intel_gt_set_wedged(engine->gt);
+		goto out_batch;
+	}
+
+	results = i915_gem_object_pin_map(vma->obj, I915_MAP_WB);
+	if (IS_ERR(results)) {
+		err = PTR_ERR(results);
+		goto out_batch;
+	}
+
+	for (v = 0, idx = 0; v < 2 * ARRAY_SIZE(values); v++) {
+		if (results[++idx] != results[0]) {
+			err = idx;
+			break;
+		}
+	}
+
+out_batch:
+	i915_vma_put(batch);
+out_vma:
+	i915_vma_put(vma);
+out_context:
+	intel_context_put(ce);
+	return err;
+}
+
+static int live_dynamic_whitelist(void *arg)
+{
+	struct intel_gt *gt = arg;
+	struct intel_engine_cs *engine;
+	enum intel_engine_id id;
+	int err = 0;
+
+	if (INTEL_GEN(gt->i915) < 8)
+		return 0;
+
+	for_each_engine(engine, gt, id) {
+		const i915_reg_t reg = RING_MAX_IDLE(engine->mmio_base);
+
+		err = rmw_reg(engine, reg);
+		if (err < 0)
+			break;
+
+		if (err) {
+			pr_err("%s: Able to write to protected reg:%04x!\n",
+			       engine->name, i915_mmio_reg_offset(reg));
+			err = -EINVAL;
+			break;
+		}
+
+		err = intel_engine_allow_user_register_access(engine, &reg, 1);
+		if (err)
+			break;
+
+		err = rmw_reg(engine, reg);
+		intel_engine_deny_user_register_access(engine, &reg, 1);
+		if (err < 0)
+			break;
+
+		if (!err) {
+			pr_err("%s: Unable to write to allowed reg:%04x!\n",
+			       engine->name, i915_mmio_reg_offset(reg));
+			err = -EINVAL;
+			break;
+		}
+
+		err = rmw_reg(engine, reg);
+		if (err < 0)
+			break;
+
+		if (err) {
+			pr_err("%s: Able to write to denied reg:%04x!\n",
+			       engine->name, i915_mmio_reg_offset(reg));
+			err = -EINVAL;
+			break;
+		}
+
+		err = 0;
+	}
+
+	if (igt_flush_test(gt->i915))
+		err = -EIO;
+
+	return err;
+}
+
 static bool
 verify_wa_lists(struct i915_gem_context *ctx, struct wa_lists *lists,
 		const char *str)
@@ -1288,6 +1523,7 @@ int intel_workarounds_live_selftests(struct drm_i915_private *i915)
 		SUBTEST(live_dirty_whitelist),
 		SUBTEST(live_reset_whitelist),
 		SUBTEST(live_isolated_whitelist),
+		SUBTEST(live_dynamic_whitelist),
 		SUBTEST(live_gpu_reset_workarounds),
 		SUBTEST(live_engine_reset_workarounds),
 	};
-- 
2.20.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

^ permalink raw reply related	[flat|nested] 18+ messages in thread
* [Intel-gfx] [PATCH 5/8] drm/i915/perf: Ensure observation logic is not clock gated
  2020-11-17 11:01 [Intel-gfx] [PATCH 1/8] drm/i915/gt: Lock intel_engine_apply_whitelist with uncore->lock Chris Wilson
                   ` (2 preceding siblings ...)
  2020-11-17 11:01 ` [Intel-gfx] [PATCH 4/8] drm/i915/gt: Enable dynamic adjustment of RING_NONPRIV Chris Wilson
@ 2020-11-17 11:01 ` Chris Wilson
  2020-11-17 11:01 ` [Intel-gfx] [PATCH 6/8] drm/i915/perf: Whitelist OA report trigger registers Chris Wilson
                   ` (3 subsequent siblings)
  7 siblings, 0 replies; 18+ messages in thread
From: Chris Wilson @ 2020-11-17 11:01 UTC (permalink / raw)
  To: intel-gfx; +Cc: Piotr Maciejewski

From: Piotr Maciejewski <piotr.maciejewski@intel.com>

A clock gating switch can control if the performance monitoring and
observation logic is enaled or not. Ensure that we enable the clocks.

v2: Separate code from other patches (Lionel)
v3: Reset PMON enable when disabling perf to save power (Lionel)
v4: Use intel_uncore_rmw and REG_BIT (Chris)

Fixes: 00a7f0d7155c ("drm/i915/tgl: Add perf support on TGL")
Signed-off-by: Piotr Maciejewski <piotr.maciejewski@intel.com>
Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
---
 drivers/gpu/drm/i915/i915_perf.c | 9 +++++++++
 drivers/gpu/drm/i915/i915_reg.h  | 2 ++
 2 files changed, 11 insertions(+)

diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
index 3ced82820483..7bef68874d80 100644
--- a/drivers/gpu/drm/i915/i915_perf.c
+++ b/drivers/gpu/drm/i915/i915_perf.c
@@ -2522,6 +2522,12 @@ gen12_enable_metric_set(struct i915_perf_stream *stream,
 			    (period_exponent << GEN12_OAG_OAGLBCTXCTRL_TIMER_PERIOD_SHIFT))
 			    : 0);
 
+	/*
+	 * Initialize Super Queue Internal Cnt Register
+	 * Set PMON Enable in order to collect valid metrics.
+	 */
+	intel_uncore_rmw(uncore, GEN12_SQCNT1, 0, GEN12_SQCNT1_PMON_ENABLE);
+
 	/*
 	 * Update all contexts prior writing the mux configurations as we need
 	 * to make sure all slices/subslices are ON before writing to NOA
@@ -2581,6 +2587,9 @@ static void gen12_disable_metric_set(struct i915_perf_stream *stream)
 
 	/* Make sure we disable noa to save power. */
 	intel_uncore_rmw(uncore, RPM_CONFIG1, GEN10_GT_NOA_ENABLE, 0);
+
+	/* Reset PMON Enable to save power. */
+	intel_uncore_rmw(uncore, GEN12_SQCNT1, GEN12_SQCNT1_PMON_ENABLE, 0);
 }
 
 static void gen7_oa_enable(struct i915_perf_stream *stream)
diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index 7ea70b7ffcc6..4137ca9d3315 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -699,6 +699,8 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg)
 #define OABUFFER_SIZE_16M   (7 << 3)
 
 #define GEN12_OA_TLB_INV_CR _MMIO(0xceec)
+#define GEN12_SQCNT1 _MMIO(0x8718)
+#define  GEN12_SQCNT1_PMON_ENABLE REG_BIT(30)
 
 /* Gen12 OAR unit */
 #define GEN12_OAR_OACONTROL _MMIO(0x2960)
-- 
2.20.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

^ permalink raw reply related	[flat|nested] 18+ messages in thread
* [Intel-gfx] [PATCH 6/8] drm/i915/perf: Whitelist OA report trigger registers
  2020-11-17 11:01 [Intel-gfx] [PATCH 1/8] drm/i915/gt: Lock intel_engine_apply_whitelist with uncore->lock Chris Wilson
                   ` (3 preceding siblings ...)
  2020-11-17 11:01 ` [Intel-gfx] [PATCH 5/8] drm/i915/perf: Ensure observation logic is not clock gated Chris Wilson
@ 2020-11-17 11:01 ` Chris Wilson
  2020-11-17 11:01 ` [Intel-gfx] [PATCH 7/8] drm/i915/perf: Whitelist OA counter and buffer registers Chris Wilson
                   ` (2 subsequent siblings)
  7 siblings, 0 replies; 18+ messages in thread
From: Chris Wilson @ 2020-11-17 11:01 UTC (permalink / raw)
  To: intel-gfx; +Cc: Piotr Maciejewski

From: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>

OA reports can be triggered into the OA buffer by writing into the
OAREPORTTRIG registers. Whitelist the registers to allow non-privileged
user to trigger reports.

Whitelist registers only if perf_stream_paranoid is set to 0. In
i915_perf_open_ioctl, this setting is checked and the whitelist is
enabled accordingly. On closing the perf fd, the whitelist is removed.

This ensures that the access to the whitelist is gated by
perf_stream_paranoid.

v2:
- Move related change to this patch (Lionel)
- Bump up perf revision (Lionel)

v3: Pardon whitelisted registers for selftest (Umesh)
v4: Document supported gens for the feature (Lionel)
v5: Whitelist registers only if perf_stream_paranoid is set to 0 (Jon)
v6: Move oa whitelist array to i915_perf (Chris)
v7: Fix OA writing beyond the wal->list memory (CI)
v8: Protect write to engine whitelist registers

v9: (Umesh)
- Use uncore->lock to protect write to forcepriv regs
- In case wal->count falls to zero on _wa_remove, make sure you still
  clear the registers. Remove wal->count check when applying whitelist.

v10: (Umesh)
- Split patches modifying intel_workarounds
- On some platforms there are no whitelisted regs. intel_engine_resume
  applies whitelist on these platforms too and the wal->count gates such
  platforms. Bring back the wal->count check.
- intel_engine_allow/deny_user_register_access modifies the engine
  whitelist and the wal->count. Use uncore->lock to serialize it with
  intel_engine_apply_whitelist.
- Grow the wal->list when adding whitelist registers after driver load.

v11:
- Fix memory leak in _wa_list_grow (Chris)
- Serialize kfree with engine resume using uncore->lock (Umesh)
- Grow the list only if wal->count is not aligned (Umesh)

Signed-off-by: Piotr Maciejewski <piotr.maciejewski@intel.com>
Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
---
 drivers/gpu/drm/i915/i915_perf.c       | 79 +++++++++++++++++++++++++-
 drivers/gpu/drm/i915/i915_perf_types.h |  8 +++
 drivers/gpu/drm/i915/i915_reg.h        | 12 ++--
 3 files changed, 92 insertions(+), 7 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
index 7bef68874d80..8abd2cfd1c2f 100644
--- a/drivers/gpu/drm/i915/i915_perf.c
+++ b/drivers/gpu/drm/i915/i915_perf.c
@@ -1364,12 +1364,56 @@ free_noa_wait(struct i915_perf_stream *stream)
 	i915_vma_unpin_and_release(&stream->noa_wait, 0);
 }
 
+static const i915_reg_t gen9_oa_wl_regs[] = {
+	{ __OAREPORTTRIG2 | RING_FORCE_TO_NONPRIV_ACCESS_RW },
+	{ __OAREPORTTRIG6 | RING_FORCE_TO_NONPRIV_ACCESS_RW },
+};
+
+static const i915_reg_t gen12_oa_wl_regs[] = {
+	{ __GEN12_OAG_OAREPORTTRIG2 | RING_FORCE_TO_NONPRIV_ACCESS_RW },
+	{ __GEN12_OAG_OAREPORTTRIG6 | RING_FORCE_TO_NONPRIV_ACCESS_RW },
+};
+
+static int intel_engine_apply_oa_whitelist(struct i915_perf_stream *stream)
+{
+	struct intel_engine_cs *engine = stream->engine;
+	int ret;
+
+	if (i915_perf_stream_paranoid ||
+	    !(stream->sample_flags & SAMPLE_OA_REPORT) ||
+	    !stream->perf->oa_wl)
+		return 0;
+
+	ret = intel_engine_allow_user_register_access(engine,
+						      stream->perf->oa_wl,
+						      stream->perf->num_oa_wl);
+	if (ret < 0)
+		return ret;
+
+	stream->oa_whitelisted = true;
+	return 0;
+}
+
+static void intel_engine_remove_oa_whitelist(struct i915_perf_stream *stream)
+{
+	struct intel_engine_cs *engine = stream->engine;
+
+	if (!stream->oa_whitelisted)
+		return;
+
+	intel_engine_deny_user_register_access(engine,
+					       stream->perf->oa_wl,
+					       stream->perf->num_oa_wl);
+}
+
 static void i915_oa_stream_destroy(struct i915_perf_stream *stream)
 {
 	struct i915_perf *perf = stream->perf;
 
 	BUG_ON(stream != perf->exclusive_stream);
 
+	intel_engine_remove_oa_whitelist(stream);
+
 	/*
 	 * Unset exclusive_stream first, it will be checked while disabling
 	 * the metric set on gen8+.
@@ -1465,7 +1509,8 @@ static void gen8_init_oa_buffer(struct i915_perf_stream *stream)
 	 *  bit."
 	 */
 	intel_uncore_write(uncore, GEN8_OABUFFER, gtt_offset |
-		   OABUFFER_SIZE_16M | GEN8_OABUFFER_MEM_SELECT_GGTT);
+			   OABUFFER_SIZE_16M | GEN8_OABUFFER_MEM_SELECT_GGTT |
+			   GEN7_OABUFFER_EDGE_TRIGGER);
 	intel_uncore_write(uncore, GEN8_OATAILPTR, gtt_offset & GEN8_OATAILPTR_MASK);
 
 	/* Mark that we need updated tail pointers to read from... */
@@ -1518,7 +1563,8 @@ static void gen12_init_oa_buffer(struct i915_perf_stream *stream)
 	 *  bit."
 	 */
 	intel_uncore_write(uncore, GEN12_OAG_OABUFFER, gtt_offset |
-			   OABUFFER_SIZE_16M | GEN8_OABUFFER_MEM_SELECT_GGTT);
+			   OABUFFER_SIZE_16M | GEN8_OABUFFER_MEM_SELECT_GGTT |
+			   GEN7_OABUFFER_EDGE_TRIGGER);
 	intel_uncore_write(uncore, GEN12_OAG_OATAILPTR,
 			   gtt_offset & GEN12_OAG_OATAILPTR_MASK);
 
@@ -3499,6 +3545,20 @@ i915_perf_open_ioctl_locked(struct i915_perf *perf,
 	if (!(param->flags & I915_PERF_FLAG_DISABLED))
 		i915_perf_enable_locked(stream);
 
+	/*
+	 * OA whitelist allows non-privileged access to some OA counters for
+	 * triggering reports into the OA buffer. This is only allowed if
+	 * perf_stream_paranoid is set to 0 by the sysadmin.
+	 *
+	 * We want to make sure this is almost the last thing we do before
+	 * returning the stream fd. If we do end up checking for errors in code
+	 * that follows this, we MUST call intel_engine_remove_oa_whitelist in
+	 * the error handling path to remove the whitelisted registers.
+	 */
+	ret = intel_engine_apply_oa_whitelist(stream);
+	if (ret < 0)
+		goto err_flags;
+
 	/* Take a reference on the driver that will be kept with stream_fd
 	 * until its release.
 	 */
@@ -4325,6 +4385,8 @@ void i915_perf_init(struct drm_i915_private *i915)
 				perf->ctx_flexeu0_offset = 0x3de;
 
 				perf->gen8_valid_ctx_bit = BIT(16);
+				perf->oa_wl = gen9_oa_wl_regs;
+				perf->num_oa_wl = ARRAY_SIZE(gen9_oa_wl_regs);
 			}
 		} else if (IS_GEN_RANGE(i915, 10, 11)) {
 			perf->oa_formats = gen8_plus_oa_formats;
@@ -4342,6 +4404,9 @@ void i915_perf_init(struct drm_i915_private *i915)
 			perf->ops.disable_metric_set = gen10_disable_metric_set;
 			perf->ops.oa_hw_tail_read = gen8_oa_hw_tail_read;
 
+			perf->oa_wl = gen9_oa_wl_regs;
+			perf->num_oa_wl = ARRAY_SIZE(gen9_oa_wl_regs);
+
 			if (IS_GEN(i915, 10)) {
 				perf->ctx_oactxctrl_offset = 0x128;
 				perf->ctx_flexeu0_offset = 0x3de;
@@ -4368,6 +4433,9 @@ void i915_perf_init(struct drm_i915_private *i915)
 
 			perf->ctx_flexeu0_offset = 0;
 			perf->ctx_oactxctrl_offset = 0x144;
+
+			perf->oa_wl = gen12_oa_wl_regs;
+			perf->num_oa_wl = ARRAY_SIZE(gen12_oa_wl_regs);
 		}
 	}
 
@@ -4470,8 +4538,13 @@ int i915_perf_ioctl_version(void)
 	 *
 	 * 5: Add DRM_I915_PERF_PROP_POLL_OA_PERIOD parameter that controls the
 	 *    interval for the hrtimer used to check for OA data.
+	 *
+	 * 6: Whitelist OATRIGGER registers to allow user to trigger reports
+	 *    into the OA buffer. This applies only to gen8+. The feature can
+	 *    only be accessed if perf_stream_paranoid is set to 0 by privileged
+	 *    user.
 	 */
-	return 5;
+	return 6;
 }
 
 #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
diff --git a/drivers/gpu/drm/i915/i915_perf_types.h b/drivers/gpu/drm/i915/i915_perf_types.h
index a36a455ae336..f54dabcbfc34 100644
--- a/drivers/gpu/drm/i915/i915_perf_types.h
+++ b/drivers/gpu/drm/i915/i915_perf_types.h
@@ -311,6 +311,11 @@ struct i915_perf_stream {
 	 * buffer should be checked for available data.
 	 */
 	u64 poll_oa_period;
+
+	/**
+	 * @oa_whitelisted: Indicates that the oa registers are whitelisted.
+	 */
+	bool oa_whitelisted;
 };
 
 /**
@@ -431,6 +436,9 @@ struct i915_perf {
 	u32 ctx_oactxctrl_offset;
 	u32 ctx_flexeu0_offset;
 
+	const i915_reg_t *oa_wl;
+	unsigned int num_oa_wl;
+
 	/**
 	 * The RPT_ID/reason field for Gen8+ includes a bit
 	 * to determine if the CTX ID in the report is valid
diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index 4137ca9d3315..7d27c271d607 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -875,7 +875,8 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg)
 #define OAREPORTTRIG1_THRESHOLD_MASK 0xffff
 #define OAREPORTTRIG1_EDGE_LEVEL_TRIGGER_SELECT_MASK 0xffff0000 /* 0=level */
 
-#define OAREPORTTRIG2 _MMIO(0x2744)
+#define __OAREPORTTRIG2 0x2744
+#define OAREPORTTRIG2 _MMIO(__OAREPORTTRIG2)
 #define OAREPORTTRIG2_INVERT_A_0  (1 << 0)
 #define OAREPORTTRIG2_INVERT_A_1  (1 << 1)
 #define OAREPORTTRIG2_INVERT_A_2  (1 << 2)
@@ -928,7 +929,8 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg)
 #define OAREPORTTRIG5_THRESHOLD_MASK 0xffff
 #define OAREPORTTRIG5_EDGE_LEVEL_TRIGGER_SELECT_MASK 0xffff0000 /* 0=level */
 
-#define OAREPORTTRIG6 _MMIO(0x2754)
+#define __OAREPORTTRIG6 0x2754
+#define OAREPORTTRIG6 _MMIO(__OAREPORTTRIG6)
 #define OAREPORTTRIG6_INVERT_A_0  (1 << 0)
 #define OAREPORTTRIG6_INVERT_A_1  (1 << 1)
 #define OAREPORTTRIG6_INVERT_A_2  (1 << 2)
@@ -989,11 +991,13 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg)
 
 /* Same layout as OAREPORTTRIGX */
 #define GEN12_OAG_OAREPORTTRIG1 _MMIO(0xd920)
-#define GEN12_OAG_OAREPORTTRIG2 _MMIO(0xd924)
+#define __GEN12_OAG_OAREPORTTRIG2 0xd924
+#define GEN12_OAG_OAREPORTTRIG2 _MMIO(__GEN12_OAG_OAREPORTTRIG2)
 #define GEN12_OAG_OAREPORTTRIG3 _MMIO(0xd928)
 #define GEN12_OAG_OAREPORTTRIG4 _MMIO(0xd92c)
 #define GEN12_OAG_OAREPORTTRIG5 _MMIO(0xd930)
-#define GEN12_OAG_OAREPORTTRIG6 _MMIO(0xd934)
+#define __GEN12_OAG_OAREPORTTRIG6 0xd934
+#define GEN12_OAG_OAREPORTTRIG6 _MMIO(__GEN12_OAG_OAREPORTTRIG6)
 #define GEN12_OAG_OAREPORTTRIG7 _MMIO(0xd938)
 #define GEN12_OAG_OAREPORTTRIG8 _MMIO(0xd93c)
 
-- 
2.20.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

^ permalink raw reply related	[flat|nested] 18+ messages in thread
* [Intel-gfx] [PATCH 7/8] drm/i915/perf: Whitelist OA counter and buffer registers
  2020-11-17 11:01 [Intel-gfx] [PATCH 1/8] drm/i915/gt: Lock intel_engine_apply_whitelist with uncore->lock Chris Wilson
                   ` (4 preceding siblings ...)
  2020-11-17 11:01 ` [Intel-gfx] [PATCH 6/8] drm/i915/perf: Whitelist OA report trigger registers Chris Wilson
@ 2020-11-17 11:01 ` Chris Wilson
  2020-11-17 11:01 ` [Intel-gfx] [PATCH 8/8] drm/i915/perf: Map OA buffer to user space for gen12 performance query Chris Wilson
  2020-11-17 13:18 ` [Intel-gfx] ✗ Fi.CI.BAT: failure for series starting with [1/8] drm/i915/gt: Lock intel_engine_apply_whitelist with uncore->lock Patchwork
  7 siblings, 0 replies; 18+ messages in thread
From: Chris Wilson @ 2020-11-17 11:01 UTC (permalink / raw)
  To: intel-gfx; +Cc: Piotr Maciejewski

From: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>

It is useful to have markers in the OA reports to identify triggered
reports. Whitelist some OA counters that can be used as markers.

A triggered report can be found faster if we can sample the HW tail and
head registers when the report was triggered. Whitelist OA buffer
specific registers.

v2:
- Bump up the perf revision (Lionel)
- Use indexing for counters (Lionel)
- Fix selftest for oa ticking register (Umesh)

v3: Pardon whitelisted registers for selftest (Umesh)

v4:
- Document whitelisted registers (Lionel)
- Fix live isolated whitelist for OA regs (Umesh)

v5:
- Free up whitelist slots. Remove GPU_TICKS and A20 counter (Piotr)
- Whitelist registers only if perf_stream_paranoid is set to 0 (Jon)

v6: Move oa whitelist array to i915_perf (Chris)

Signed-off-by: Piotr Maciejewski <piotr.maciejewski@intel.com>
Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
---
 drivers/gpu/drm/i915/i915_perf.c | 18 +++++++++++++++++-
 drivers/gpu/drm/i915/i915_reg.h  | 16 ++++++++++++++--
 2 files changed, 31 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
index 8abd2cfd1c2f..c91f2da84189 100644
--- a/drivers/gpu/drm/i915/i915_perf.c
+++ b/drivers/gpu/drm/i915/i915_perf.c
@@ -1367,11 +1367,19 @@ free_noa_wait(struct i915_perf_stream *stream)
 static const i915_reg_t gen9_oa_wl_regs[] = {
 	{ __OAREPORTTRIG2 | RING_FORCE_TO_NONPRIV_ACCESS_RW },
 	{ __OAREPORTTRIG6 | RING_FORCE_TO_NONPRIV_ACCESS_RW },
+	{ __OA_PERF_COUNTER_A(18) | (RING_FORCE_TO_NONPRIV_ACCESS_RW |
+				     RING_FORCE_TO_NONPRIV_RANGE_4) },
+	{ __GEN8_OASTATUS | (RING_FORCE_TO_NONPRIV_ACCESS_RD |
+			     RING_FORCE_TO_NONPRIV_RANGE_4) },
 };
 
 static const i915_reg_t gen12_oa_wl_regs[] = {
 	{ __GEN12_OAG_OAREPORTTRIG2 | RING_FORCE_TO_NONPRIV_ACCESS_RW },
 	{ __GEN12_OAG_OAREPORTTRIG6 | RING_FORCE_TO_NONPRIV_ACCESS_RW },
+	{ __GEN12_OAG_PERF_COUNTER_A(18) | (RING_FORCE_TO_NONPRIV_ACCESS_RW |
+					    RING_FORCE_TO_NONPRIV_RANGE_4) },
+	{ __GEN12_OAG_OASTATUS | (RING_FORCE_TO_NONPRIV_ACCESS_RD |
+				  RING_FORCE_TO_NONPRIV_RANGE_4) },
 };
 
 static int intel_engine_apply_oa_whitelist(struct i915_perf_stream *stream)
@@ -4543,8 +4551,16 @@ int i915_perf_ioctl_version(void)
 	 *    into the OA buffer. This applies only to gen8+. The feature can
 	 *    only be accessed if perf_stream_paranoid is set to 0 by privileged
 	 *    user.
+	 *
+	 * 7: Whitelist below OA registers for user to identify the location of
+	 *    triggered reports in the OA buffer. This applies only to gen8+.
+	 *    The feature can only be accessed if perf_stream_paranoid is set to
+	 *    0 by privileged user.
+	 *
+	 *    - OA buffer head/tail/status/buffer registers for read only
+	 *    - OA counters A18, A19, A20 for read/write
 	 */
-	return 6;
+	return 7;
 }
 
 #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index 7d27c271d607..479ecbdf4d3a 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -678,7 +678,8 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg)
 #define  GEN7_OASTATUS2_HEAD_MASK           0xffffffc0
 #define  GEN7_OASTATUS2_MEM_SELECT_GGTT     (1 << 0) /* 0: PPGTT, 1: GGTT */
 
-#define GEN8_OASTATUS _MMIO(0x2b08)
+#define __GEN8_OASTATUS 0x2b08
+#define GEN8_OASTATUS _MMIO(__GEN8_OASTATUS)
 #define  GEN8_OASTATUS_OVERRUN_STATUS	    (1 << 3)
 #define  GEN8_OASTATUS_COUNTER_OVERFLOW     (1 << 2)
 #define  GEN8_OASTATUS_OABUFFER_OVERFLOW    (1 << 1)
@@ -736,7 +737,8 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg)
 #define  GEN12_OAG_OA_DEBUG_DISABLE_GO_1_0_REPORTS     (1 << 2)
 #define  GEN12_OAG_OA_DEBUG_DISABLE_CTX_SWITCH_REPORTS (1 << 1)
 
-#define GEN12_OAG_OASTATUS _MMIO(0xdafc)
+#define __GEN12_OAG_OASTATUS 0xdafc
+#define GEN12_OAG_OASTATUS _MMIO(__GEN12_OAG_OASTATUS)
 #define  GEN12_OAG_OASTATUS_COUNTER_OVERFLOW (1 << 2)
 #define  GEN12_OAG_OASTATUS_BUFFER_OVERFLOW  (1 << 1)
 #define  GEN12_OAG_OASTATUS_REPORT_LOST      (1 << 0)
@@ -979,6 +981,16 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg)
 #define OAREPORTTRIG8_NOA_SELECT_6_SHIFT    24
 #define OAREPORTTRIG8_NOA_SELECT_7_SHIFT    28
 
+/* Performance counters registers */
+#define __OA_PERF_COUNTER_A(idx)	(0x2800 + 8 * (idx))
+#define OA_PERF_COUNTER_A(idx)		_MMIO(__OA_PERF_COUNTER_A(idx))
+#define OA_PERF_COUNTER_A_UDW(idx)		_MMIO(__OA_PERF_COUNTER_A(idx) + 4)
+
+/* Gen12 Performance counters registers */
+#define __GEN12_OAG_PERF_COUNTER_A(idx)	(0xD980 + 8 * (idx))
+#define GEN12_OAG_PERF_COUNTER_A(idx)	_MMIO(__GEN12_OAG_PERF_COUNTER_A(idx))
+#define GEN12_OAG_PERF_COUNTER_A_UDW(idx) _MMIO(__GEN12_OAG_PERF_COUNTER_A(idx) + 4)
+
 /* Same layout as OASTARTTRIGX */
 #define GEN12_OAG_OASTARTTRIG1 _MMIO(0xd900)
 #define GEN12_OAG_OASTARTTRIG2 _MMIO(0xd904)
-- 
2.20.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

^ permalink raw reply related	[flat|nested] 18+ messages in thread
* [Intel-gfx] [PATCH 8/8] drm/i915/perf: Map OA buffer to user space for gen12 performance query
  2020-11-17 11:01 [Intel-gfx] [PATCH 1/8] drm/i915/gt: Lock intel_engine_apply_whitelist with uncore->lock Chris Wilson
                   ` (5 preceding siblings ...)
  2020-11-17 11:01 ` [Intel-gfx] [PATCH 7/8] drm/i915/perf: Whitelist OA counter and buffer registers Chris Wilson
@ 2020-11-17 11:01 ` Chris Wilson
  2020-11-18 11:41   ` Joonas Lahtinen
  2020-11-17 13:18 ` [Intel-gfx] ✗ Fi.CI.BAT: failure for series starting with [1/8] drm/i915/gt: Lock intel_engine_apply_whitelist with uncore->lock Patchwork
  7 siblings, 1 reply; 18+ messages in thread
From: Chris Wilson @ 2020-11-17 11:01 UTC (permalink / raw)
  To: intel-gfx; +Cc: Piotr Maciejewski, Chris Wilson

From: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>

i915 used to support time based sampling mode which is good for overall
system monitoring, but is not enough for query mode used to measure a
single draw call or dispatch. Gen9-Gen11 are using current i915 perf
implementation for query, but Gen12+ requires a new approach for query
based on triggered reports within oa buffer.

Triggering reports into the OA buffer is achieved by writing into a
a trigger register. Optionally an unused counter/register is set with a
marker value such that a triggered report can be identified in the OA
buffer. Reports are usually triggered at the start and end of work that
is measured.

Since OA buffer is large and queries can be frequent, an efficient way
to look for triggered reports is required. By knowing the current head
and tail offsets into the OA buffer, it is easier to determine the
locality of the reports of interest.

Current perf OA interface does not expose head/tail information to the
user and it filters out invalid reports before sending data to user.
Also considering limited size of user buffer used during a query,
creating a 1:1 copy of the OA buffer at the user space added undesired
complexity.

The solution was to map the OA buffer to user space provided

(1) that it is accessed from a privileged user.
(2) OA report filtering is not used.

These 2 conditions would satisfy the safety criteria that the current
perf interface addresses.

To enable the query:
- Add an ioctl to expose head and tail to the user
- Add an ioctl to return size and offset of the OA buffer
- Map the OA buffer to the user space

v2:
- Improve commit message (Chris)
- Do not mmap based on gem object filp. Instead, use perf_fd and support
  mmap syscall (Chris)
- Pass non-zero offset in mmap to enforce the right object is
  mapped (Chris)
- Do not expose gpu_address (Chris)
- Verify start and length of vma for page alignment (Lionel)
- Move SQNTL config out (Lionel)

v3: (Chris)
- Omit redundant checks
- Return VM_FAULT_SIGBUS is old stream is closed
- Maintain reference counts to stream in vm_open and vm_close
- Use switch to identify object to be mapped

v4: Call kref_put on closing perf fd (Chris)
v5:
- Strip access to OA buffer from unprivileged child of a privileged
  parent. Use VM_DONTCOPY
- Enforce MAP_PRIVATE by checking for VM_MAYSHARE

v6:
(Chris)
- Use len of -1 in unmap_mapping_range
- Don't use stream->oa_buffer.vma->obj in vm_fault_oa
- Use kernel block comment style
- do_mmap gets a reference to the file and puts it in do_munmap, so
  no need to maintain a reference to i915_perf_stream. Hence, remove
  vm_open/vm_close and stream->closed hooks/checks.
(Umesh)
- Do not allow mmap if SAMPLE_OA_REPORT is not set during
  i915_perf_open_ioctl.
- Drop ioctl returning head/tail since this information is already
  whitelisted. Remove hooks to read head register.

v7: (Chris)
- unmap before destroy
- change ioctl argument struct

v8: Documentation and more checks (Chris)

Signed-off-by: Piotr Maciejewski <piotr.maciejewski@intel.com>
Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
---
 drivers/gpu/drm/i915/gem/i915_gem_mman.c |   2 +-
 drivers/gpu/drm/i915/gem/i915_gem_mman.h |   2 +
 drivers/gpu/drm/i915/i915_perf.c         | 126 ++++++++++++++++++++++-
 include/uapi/drm/i915_drm.h              |  33 ++++++
 4 files changed, 161 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_mman.c b/drivers/gpu/drm/i915/gem/i915_gem_mman.c
index 3d69e51f3e4d..2ab08b152b9d 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_mman.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_mman.c
@@ -204,7 +204,7 @@ compute_partial_view(const struct drm_i915_gem_object *obj,
 	return view;
 }
 
-static vm_fault_t i915_error_to_vmf_fault(int err)
+vm_fault_t i915_error_to_vmf_fault(int err)
 {
 	switch (err) {
 	default:
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_mman.h b/drivers/gpu/drm/i915/gem/i915_gem_mman.h
index efee9e0d2508..1190a3a228ea 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_mman.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_mman.h
@@ -29,4 +29,6 @@ void i915_gem_object_release_mmap_gtt(struct drm_i915_gem_object *obj);
 
 void i915_gem_object_release_mmap_offset(struct drm_i915_gem_object *obj);
 
+vm_fault_t i915_error_to_vmf_fault(int err);
+
 #endif
diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
index c91f2da84189..6fd669b520d8 100644
--- a/drivers/gpu/drm/i915/i915_perf.c
+++ b/drivers/gpu/drm/i915/i915_perf.c
@@ -192,10 +192,12 @@
  */
 
 #include <linux/anon_inodes.h>
+#include <linux/mman.h>
 #include <linux/sizes.h>
 #include <linux/uuid.h>
 
 #include "gem/i915_gem_context.h"
+#include "gem/i915_gem_mman.h"
 #include "gt/intel_engine_pm.h"
 #include "gt/intel_engine_user.h"
 #include "gt/intel_gt.h"
@@ -3291,6 +3293,44 @@ static long i915_perf_config_locked(struct i915_perf_stream *stream,
 	return ret;
 }
 
+#define I915_PERF_OA_BUFFER_MMAP_OFFSET 1
+
+/**
+ * i915_perf_oa_buffer_info_locked - size and offset of the OA buffer
+ * @stream: i915 perf stream
+ * @cmd: ioctl command
+ * @arg: pointer to oa buffer info filled by this function.
+ */
+static int i915_perf_oa_buffer_info_locked(struct i915_perf_stream *stream,
+					   unsigned int cmd,
+					   unsigned long arg)
+{
+	struct drm_i915_perf_oa_buffer_info info;
+	void __user *output = (void __user *)arg;
+
+	if (i915_perf_stream_paranoid && !perfmon_capable()) {
+		DRM_DEBUG("Insufficient privileges to access OA buffer info\n");
+		return -EACCES;
+	}
+
+	if (_IOC_SIZE(cmd) != sizeof(info))
+		return -EINVAL;
+
+	if (copy_from_user(&info, output, sizeof(info)))
+		return -EFAULT;
+
+	if (info.type || info.flags || info.rsvd)
+		return -EINVAL;
+
+	info.size = stream->oa_buffer.vma->size;
+	info.offset = I915_PERF_OA_BUFFER_MMAP_OFFSET * PAGE_SIZE;
+
+	if (copy_to_user(output, &info, sizeof(info)))
+		return -EFAULT;
+
+	return 0;
+}
+
 /**
  * i915_perf_ioctl_locked - support ioctl() usage with i915 perf stream FDs
  * @stream: An i915 perf stream
@@ -3316,6 +3356,8 @@ static long i915_perf_ioctl_locked(struct i915_perf_stream *stream,
 		return 0;
 	case I915_PERF_IOCTL_CONFIG:
 		return i915_perf_config_locked(stream, arg);
+	case I915_PERF_IOCTL_GET_OA_BUFFER_INFO:
+		return i915_perf_oa_buffer_info_locked(stream, cmd, arg);
 	}
 
 	return -EINVAL;
@@ -3387,6 +3429,14 @@ static int i915_perf_release(struct inode *inode, struct file *file)
 	struct i915_perf_stream *stream = file->private_data;
 	struct i915_perf *perf = stream->perf;
 
+	/*
+	 * User could have multiple vmas from multiple mmaps. We want to zap
+	 * them all here. Note that a fresh fault cannot occur as the mmap holds
+	 * a reference to the stream via the vma->vm_file, so before user's
+	 * munmap, the stream cannot be destroyed.
+	 */
+	unmap_mapping_range(file->f_mapping, 0, -1, 1);
+
 	mutex_lock(&perf->lock);
 	i915_perf_destroy_locked(stream);
 	mutex_unlock(&perf->lock);
@@ -3397,6 +3447,75 @@ static int i915_perf_release(struct inode *inode, struct file *file)
 	return 0;
 }
 
+static vm_fault_t vm_fault_oa(struct vm_fault *vmf)
+{
+	struct vm_area_struct *vma = vmf->vma;
+	struct i915_perf_stream *stream = vma->vm_private_data;
+	int err;
+
+	err = remap_io_sg(vma,
+			  vma->vm_start, vma->vm_end - vma->vm_start,
+			  stream->oa_buffer.vma->pages->sgl, -1);
+
+	return i915_error_to_vmf_fault(err);
+}
+
+static const struct vm_operations_struct vm_ops_oa = {
+	.fault = vm_fault_oa,
+};
+
+static int i915_perf_mmap(struct file *file, struct vm_area_struct *vma)
+{
+	struct i915_perf_stream *stream = file->private_data;
+
+	/* mmap-ing OA buffer to user space MUST absolutely be privileged */
+	if (i915_perf_stream_paranoid && !perfmon_capable()) {
+		DRM_DEBUG("Insufficient privileges to map OA buffer\n");
+		return -EACCES;
+	}
+
+	switch (vma->vm_pgoff) {
+	/*
+	 * A non-zero offset ensures that we are mapping the right object. Also
+	 * leaves room for future objects added to this implementation.
+	 */
+	case I915_PERF_OA_BUFFER_MMAP_OFFSET:
+		if (!(stream->sample_flags & SAMPLE_OA_REPORT))
+			return -EINVAL;
+
+		if (vma->vm_end - vma->vm_start > OA_BUFFER_SIZE)
+			return -EINVAL;
+
+		/*
+		 * Only support VM_READ. Enforce MAP_PRIVATE by checking for
+		 * VM_MAYSHARE.
+		 */
+		if (vma->vm_flags & (VM_WRITE | VM_EXEC |
+				     VM_SHARED | VM_MAYSHARE))
+			return -EINVAL;
+
+		vma->vm_flags &= ~(VM_MAYWRITE | VM_MAYEXEC);
+
+		/*
+		 * If the privileged parent forks and child drops root
+		 * privilege, we do not want the child to retain access to the
+		 * mapped OA buffer. Explicitly set VM_DONTCOPY to avoid such
+		 * cases.
+		 */
+		vma->vm_flags |= VM_PFNMAP | VM_DONTEXPAND |
+				 VM_DONTDUMP | VM_DONTCOPY;
+		break;
+
+	default:
+		return -EINVAL;
+	}
+
+	vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
+	vma->vm_private_data = stream;
+	vma->vm_ops = &vm_ops_oa;
+
+	return 0;
+}
 
 static const struct file_operations fops = {
 	.owner		= THIS_MODULE,
@@ -3409,6 +3528,7 @@ static const struct file_operations fops = {
 	 * to handle 32bits compatibility.
 	 */
 	.compat_ioctl   = i915_perf_ioctl,
+	.mmap		= i915_perf_mmap,
 };
 
 
@@ -4559,8 +4679,12 @@ int i915_perf_ioctl_version(void)
 	 *
 	 *    - OA buffer head/tail/status/buffer registers for read only
 	 *    - OA counters A18, A19, A20 for read/write
+	 *
+	 * 8: Added an option to map oa buffer at umd driver level and trigger
+	 *    oa reports within oa buffer from command buffer. See
+	 *    I915_PERF_IOCTL_GET_OA_BUFFER_INFO.
 	 */
-	return 7;
+	return 8;
 }
 
 #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
index fa1f3d62f9a6..cc1702ddc859 100644
--- a/include/uapi/drm/i915_drm.h
+++ b/include/uapi/drm/i915_drm.h
@@ -2101,6 +2101,39 @@ struct drm_i915_perf_open_param {
  */
 #define I915_PERF_IOCTL_CONFIG	_IO('i', 0x2)
 
+/**
+ * Returns OA buffer properties to be used with mmap.
+ *
+ * This ioctl is available in perf revision 8.
+ */
+#define I915_PERF_IOCTL_GET_OA_BUFFER_INFO _IOWR('i', 0x3, struct drm_i915_perf_oa_buffer_info)
+
+/**
+ * OA buffer size and offset.
+ *
+ * OA output buffer
+ *   type: 0
+ *   flags: mbz
+ *
+ *   After querying the info, pass (size,offset) to mmap(),
+ *
+ *   mmap(0, info.size, PROT_READ, MAP_PRIVATE, perf_fd, info.offset).
+ *
+ *   Note that only a private (not shared between processes, or across fork())
+ *   read-only mmapping is allowed.
+ *
+ *   Userspace must treat the incoming data as tainted, but it conforms to the OA
+ *   format as specified by user config. The buffer provides reports that have
+ *   OA counters - A, B and C.
+ */
+struct drm_i915_perf_oa_buffer_info {
+	__u32 type;   /* in */
+	__u32 flags;  /* in */
+	__u64 size;   /* out */
+	__u64 offset; /* out */
+	__u64 rsvd;   /* mbz */
+};
+
 /**
  * Common to all i915 perf records
  */
-- 
2.20.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

^ permalink raw reply related	[flat|nested] 18+ messages in thread
* [Intel-gfx] ✗ Fi.CI.BAT: failure for series starting with [1/8] drm/i915/gt: Lock intel_engine_apply_whitelist with uncore->lock
  2020-11-17 11:01 [Intel-gfx] [PATCH 1/8] drm/i915/gt: Lock intel_engine_apply_whitelist with uncore->lock Chris Wilson
                   ` (6 preceding siblings ...)
  2020-11-17 11:01 ` [Intel-gfx] [PATCH 8/8] drm/i915/perf: Map OA buffer to user space for gen12 performance query Chris Wilson
@ 2020-11-17 13:18 ` Patchwork
  7 siblings, 0 replies; 18+ messages in thread
From: Patchwork @ 2020-11-17 13:18 UTC (permalink / raw)
  To: Chris Wilson; +Cc: intel-gfx


[-- Attachment #1.1: Type: text/plain, Size: 8717 bytes --]

== Series Details ==

Series: series starting with [1/8] drm/i915/gt: Lock intel_engine_apply_whitelist with uncore->lock
URL   : https://patchwork.freedesktop.org/series/83949/
State : failure

== Summary ==

CI Bug Log - changes from CI_DRM_9343 -> Patchwork_18919
====================================================

Summary
-------

  **FAILURE**

  Serious unknown changes coming with Patchwork_18919 absolutely need to be
  verified manually.
  
  If you think the reported changes have nothing to do with the changes
  introduced in Patchwork_18919, please notify your bug team to allow them
  to document this new failure mode, which will reduce false positives in CI.

  External URL: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18919/index.html

Possible new issues
-------------------

  Here are the unknown changes that may have been introduced in Patchwork_18919:

### IGT changes ###

#### Possible regressions ####

  * igt@gem_exec_suspend@basic-s0:
    - fi-cml-u2:          [PASS][1] -> [INCOMPLETE][2]
   [1]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_9343/fi-cml-u2/igt@gem_exec_suspend@basic-s0.html
   [2]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18919/fi-cml-u2/igt@gem_exec_suspend@basic-s0.html

  * igt@i915_selftest@live@workarounds:
    - fi-bsw-n3050:       [PASS][3] -> [DMESG-FAIL][4]
   [3]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_9343/fi-bsw-n3050/igt@i915_selftest@live@workarounds.html
   [4]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18919/fi-bsw-n3050/igt@i915_selftest@live@workarounds.html
    - fi-skl-6600u:       [PASS][5] -> [DMESG-FAIL][6]
   [5]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_9343/fi-skl-6600u/igt@i915_selftest@live@workarounds.html
   [6]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18919/fi-skl-6600u/igt@i915_selftest@live@workarounds.html
    - fi-bsw-kefka:       [PASS][7] -> [DMESG-FAIL][8]
   [7]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_9343/fi-bsw-kefka/igt@i915_selftest@live@workarounds.html
   [8]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18919/fi-bsw-kefka/igt@i915_selftest@live@workarounds.html
    - fi-bsw-nick:        [PASS][9] -> [DMESG-FAIL][10]
   [9]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_9343/fi-bsw-nick/igt@i915_selftest@live@workarounds.html
   [10]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18919/fi-bsw-nick/igt@i915_selftest@live@workarounds.html
    - fi-bdw-5557u:       NOTRUN -> [DMESG-FAIL][11]
   [11]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18919/fi-bdw-5557u/igt@i915_selftest@live@workarounds.html

  
#### Warnings ####

  * igt@i915_selftest@live@gt_pm:
    - fi-tgl-y:           [DMESG-FAIL][12] ([i915#1759]) -> [DMESG-FAIL][13]
   [12]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_9343/fi-tgl-y/igt@i915_selftest@live@gt_pm.html
   [13]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18919/fi-tgl-y/igt@i915_selftest@live@gt_pm.html

  
New tests
---------

  New tests have been introduced between CI_DRM_9343 and Patchwork_18919:

### New CI tests (1) ###

  * boot:
    - Statuses : 41 pass(s)
    - Exec time: [0.0] s

  

Known issues
------------

  Here are the changes found in Patchwork_18919 that come from known issues:

### IGT changes ###

#### Issues hit ####

  * igt@gem_tiled_fence_blits@basic:
    - fi-tgl-y:           [PASS][14] -> [DMESG-WARN][15] ([i915#402]) +2 similar issues
   [14]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_9343/fi-tgl-y/igt@gem_tiled_fence_blits@basic.html
   [15]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18919/fi-tgl-y/igt@gem_tiled_fence_blits@basic.html

  * igt@i915_selftest@live@workarounds:
    - fi-bxt-dsi:         [PASS][16] -> [DMESG-FAIL][17] ([i915#1635])
   [16]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_9343/fi-bxt-dsi/igt@i915_selftest@live@workarounds.html
   [17]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18919/fi-bxt-dsi/igt@i915_selftest@live@workarounds.html

  * igt@kms_chamelium@dp-crc-fast:
    - fi-kbl-7500u:       [PASS][18] -> [FAIL][19] ([i915#1161] / [i915#262])
   [18]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_9343/fi-kbl-7500u/igt@kms_chamelium@dp-crc-fast.html
   [19]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18919/fi-kbl-7500u/igt@kms_chamelium@dp-crc-fast.html

  * igt@kms_cursor_legacy@basic-busy-flip-before-cursor-atomic:
    - fi-bsw-kefka:       [PASS][20] -> [DMESG-WARN][21] ([i915#1982])
   [20]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_9343/fi-bsw-kefka/igt@kms_cursor_legacy@basic-busy-flip-before-cursor-atomic.html
   [21]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18919/fi-bsw-kefka/igt@kms_cursor_legacy@basic-busy-flip-before-cursor-atomic.html

  * igt@kms_pipe_crc_basic@compare-crc-sanitycheck-pipe-b:
    - fi-tgl-y:           [PASS][22] -> [DMESG-WARN][23] ([i915#1982]) +1 similar issue
   [22]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_9343/fi-tgl-y/igt@kms_pipe_crc_basic@compare-crc-sanitycheck-pipe-b.html
   [23]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18919/fi-tgl-y/igt@kms_pipe_crc_basic@compare-crc-sanitycheck-pipe-b.html

  
#### Possible fixes ####

  * igt@debugfs_test@read_all_entries:
    - fi-tgl-y:           [DMESG-WARN][24] ([i915#402]) -> [PASS][25] +2 similar issues
   [24]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_9343/fi-tgl-y/igt@debugfs_test@read_all_entries.html
   [25]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18919/fi-tgl-y/igt@debugfs_test@read_all_entries.html

  * igt@i915_module_load@reload:
    - fi-kbl-soraka:      [DMESG-WARN][26] ([i915#1982]) -> [PASS][27]
   [26]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_9343/fi-kbl-soraka/igt@i915_module_load@reload.html
   [27]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18919/fi-kbl-soraka/igt@i915_module_load@reload.html

  * igt@i915_pm_rpm@basic-pci-d3-state:
    - fi-bsw-kefka:       [DMESG-WARN][28] ([i915#1982]) -> [PASS][29]
   [28]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_9343/fi-bsw-kefka/igt@i915_pm_rpm@basic-pci-d3-state.html
   [29]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18919/fi-bsw-kefka/igt@i915_pm_rpm@basic-pci-d3-state.html

  * igt@kms_busy@basic@flip:
    - {fi-kbl-7560u}:     [DMESG-WARN][30] ([i915#1982]) -> [PASS][31]
   [30]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_9343/fi-kbl-7560u/igt@kms_busy@basic@flip.html
   [31]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18919/fi-kbl-7560u/igt@kms_busy@basic@flip.html

  * igt@kms_psr@sprite_plane_onoff:
    - fi-tgl-y:           [DMESG-WARN][32] ([i915#1982]) -> [PASS][33] +1 similar issue
   [32]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_9343/fi-tgl-y/igt@kms_psr@sprite_plane_onoff.html
   [33]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18919/fi-tgl-y/igt@kms_psr@sprite_plane_onoff.html

  
  {name}: This element is suppressed. This means it is ignored when computing
          the status of the difference (SUCCESS, WARNING, or FAILURE).

  [i915#1161]: https://gitlab.freedesktop.org/drm/intel/issues/1161
  [i915#1635]: https://gitlab.freedesktop.org/drm/intel/issues/1635
  [i915#1759]: https://gitlab.freedesktop.org/drm/intel/issues/1759
  [i915#1982]: https://gitlab.freedesktop.org/drm/intel/issues/1982
  [i915#262]: https://gitlab.freedesktop.org/drm/intel/issues/262
  [i915#402]: https://gitlab.freedesktop.org/drm/intel/issues/402


Participating hosts (45 -> 41)
------------------------------

  Additional (1): fi-blb-e6850 
  Missing    (5): fi-ilk-m540 fi-hsw-4200u fi-bsw-cyan fi-ctg-p8600 fi-bdw-samus 


Build changes
-------------

  * Linux: CI_DRM_9343 -> Patchwork_18919

  CI-20190529: 20190529
  CI_DRM_9343: e1757932e1d12120875a6f468de2583f1842eb2e @ git://anongit.freedesktop.org/gfx-ci/linux
  IGT_5855: d9b3c7058efe41e5224dd1e43fac05dc6d049380 @ git://anongit.freedesktop.org/xorg/app/intel-gpu-tools
  Patchwork_18919: 97cc6d5326a953ea4a90e7e424f5a1b86cd19961 @ git://anongit.freedesktop.org/gfx-ci/linux


== Linux commits ==

97cc6d5326a9 drm/i915/perf: Map OA buffer to user space for gen12 performance query
73445f675e90 drm/i915/perf: Whitelist OA counter and buffer registers
57558da9c81e drm/i915/perf: Whitelist OA report trigger registers
5136579becd4 drm/i915/perf: Ensure observation logic is not clock gated
be0360b5350b drm/i915/gt: Enable dynamic adjustment of RING_NONPRIV
64703e8c9ffa drm/i915/gt: Check for conflicting RING_NONPRIV
d80973c9056b drm/i915/gt: Refactor _wa_add to reuse wa_index and wa_list_grow
20c62532edcd drm/i915/gt: Lock intel_engine_apply_whitelist with uncore->lock

== Logs ==

For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18919/index.html

[-- Attachment #1.2: Type: text/html, Size: 10221 bytes --]

[-- Attachment #2: Type: text/plain, Size: 160 bytes --]

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

^ permalink raw reply	[flat|nested] 18+ messages in thread
* Re: [Intel-gfx] [PATCH 3/8] drm/i915/gt: Check for conflicting RING_NONPRIV
  2020-11-17 11:01 ` [Intel-gfx] [PATCH 3/8] drm/i915/gt: Check for conflicting RING_NONPRIV Chris Wilson
@ 2020-11-17 18:08   ` Umesh Nerlige Ramappa
  0 siblings, 0 replies; 18+ messages in thread
From: Umesh Nerlige Ramappa @ 2020-11-17 18:08 UTC (permalink / raw)
  To: Chris Wilson; +Cc: intel-gfx

On Tue, Nov 17, 2020 at 11:01:27AM +0000, Chris Wilson wrote:
>Strip the encoded bits from the register offset so that we only use the
>address for looking up the RING_NONPRIV entry.
>
>Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

Reviewed-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>

Thanks,
Umesh
>---
> drivers/gpu/drm/i915/gt/intel_workarounds.c | 64 +++++++++++++--------
> 1 file changed, 41 insertions(+), 23 deletions(-)
>
>diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.c b/drivers/gpu/drm/i915/gt/intel_workarounds.c
>index e50c72d2b3f1..290aa277ab10 100644
>--- a/drivers/gpu/drm/i915/gt/intel_workarounds.c
>+++ b/drivers/gpu/drm/i915/gt/intel_workarounds.c
>@@ -111,18 +111,43 @@ static void wa_init_finish(struct i915_wa_list *wal)
> 			 wal->wa_count, wal->name, wal->engine_name);
> }
>
>+static u32 reg_offset(i915_reg_t reg)
>+{
>+	return i915_mmio_reg_offset(reg) & RING_FORCE_TO_NONPRIV_ADDRESS_MASK;
>+}
>+
>+static u32 reg_flags(i915_reg_t reg)
>+{
>+	return i915_mmio_reg_offset(reg) & ~RING_FORCE_TO_NONPRIV_ADDRESS_MASK;
>+}
>+
>+static inline bool is_nonpriv_flags_valid(u32 flags)
>+{
>+	/* Check only valid flag bits are set */
>+	if (flags & ~RING_FORCE_TO_NONPRIV_MASK_VALID)
>+		return false;
>+
>+	/* NB: Only 3 out of 4 enum values are valid for access field */
>+	if ((flags & RING_FORCE_TO_NONPRIV_ACCESS_MASK) ==
>+	    RING_FORCE_TO_NONPRIV_ACCESS_INVALID)
>+		return false;
>+
>+	return true;
>+}
>+
> static int wa_index(struct i915_wa_list *wal, i915_reg_t reg)
> {
>-	unsigned int addr = i915_mmio_reg_offset(reg);
> 	int start = 0, end = wal->count;
>+	u32 addr = reg_offset(reg);
>
> 	/* addr and wal->list[].reg, both include the R/W flags */
> 	while (start < end) {
> 		unsigned int mid = start + (end - start) / 2;
>+		u32 pos = reg_offset(wal->list[mid].reg);
>
>-		if (i915_mmio_reg_offset(wal->list[mid].reg) < addr)
>+		if (pos < addr)
> 			start = mid + 1;
>-		else if (i915_mmio_reg_offset(wal->list[mid].reg) > addr)
>+		else if (pos > addr)
> 			end = mid;
> 		else
> 			return mid;
>@@ -148,13 +173,22 @@ static void __wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
> 	struct i915_wa *wa_;
> 	int index;
>
>+	GEM_BUG_ON(!is_nonpriv_flags_valid(reg_flags(wa->reg)));
>+
> 	index = wa_index(wal, wa->reg);
> 	if (index >= 0) {
> 		wa_ = &wal->list[index];
>
>+		if (i915_mmio_reg_offset(wa->reg) !=
>+		    i915_mmio_reg_offset(wa_->reg)) {
>+			DRM_ERROR("Discarding incompatible w/a for reg %04x\n",
>+				  reg_offset(wa->reg));
>+			return;
>+		}
>+
> 		if ((wa->clr | wa_->clr) && !(wa->clr & ~wa_->clr)) {
> 			DRM_ERROR("Discarding overwritten w/a for reg %04x (clear: %08x, set: %08x)\n",
>-				  i915_mmio_reg_offset(wa_->reg),
>+				  reg_offset(wa_->reg),
> 				  wa_->clr, wa_->set);
>
> 			wa_->set &= ~wa->clr;
>@@ -172,10 +206,8 @@ static void __wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
> 	*wa_ = *wa;
>
> 	while (wa_-- > wal->list) {
>-		GEM_BUG_ON(i915_mmio_reg_offset(wa_[0].reg) ==
>-			   i915_mmio_reg_offset(wa_[1].reg));
>-		if (i915_mmio_reg_offset(wa_[1].reg) >
>-		    i915_mmio_reg_offset(wa_[0].reg))
>+		GEM_BUG_ON(reg_offset(wa_[0].reg) == reg_offset(wa_[1].reg));
>+		if (reg_offset(wa_[1].reg) > reg_offset(wa_[0].reg))
> 			break;
>
> 		swap(wa_[1], wa_[0]);
>@@ -191,7 +223,7 @@ static void _wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
> 	if (IS_ALIGNED(wal->count, grow) && /* Either uninitialized or full. */
> 	    wa_list_grow(wal, ALIGN(wal->count + 1, grow), GFP_KERNEL)) {
> 		DRM_ERROR("Unable to store w/a for reg %04x\n",
>-			  i915_mmio_reg_offset(wa->reg));
>+			  reg_offset(wa->reg));
> 		return;
> 	}
>
>@@ -1443,20 +1475,6 @@ bool intel_gt_verify_workarounds(struct intel_gt *gt, const char *from)
> 	return wa_list_verify(gt->uncore, &gt->i915->gt_wa_list, from);
> }
>
>-static inline bool is_nonpriv_flags_valid(u32 flags)
>-{
>-	/* Check only valid flag bits are set */
>-	if (flags & ~RING_FORCE_TO_NONPRIV_MASK_VALID)
>-		return false;
>-
>-	/* NB: Only 3 out of 4 enum values are valid for access field */
>-	if ((flags & RING_FORCE_TO_NONPRIV_ACCESS_MASK) ==
>-	    RING_FORCE_TO_NONPRIV_ACCESS_INVALID)
>-		return false;
>-
>-	return true;
>-}
>-
> static void
> whitelist_reg_ext(struct i915_wa_list *wal, i915_reg_t reg, u32 flags)
> {
>-- 
>2.20.1
>
>_______________________________________________
>Intel-gfx mailing list
>Intel-gfx@lists.freedesktop.org
>https://lists.freedesktop.org/mailman/listinfo/intel-gfx
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

^ permalink raw reply	[flat|nested] 18+ messages in thread
* Re: [Intel-gfx] [PATCH 2/8] drm/i915/gt: Refactor _wa_add to reuse wa_index and wa_list_grow
  2020-11-17 11:01 ` [Intel-gfx] [PATCH 2/8] drm/i915/gt: Refactor _wa_add to reuse wa_index and wa_list_grow Chris Wilson
@ 2020-11-17 18:14   ` Umesh Nerlige Ramappa
  2021-02-27  6:58     ` Lucas De Marchi
  0 siblings, 1 reply; 18+ messages in thread
From: Umesh Nerlige Ramappa @ 2020-11-17 18:14 UTC (permalink / raw)
  To: Chris Wilson; +Cc: intel-gfx

This is quite different and much cleane from my original patch :), so 
you should be the author. With that, this is

Reviewed-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>

Thanks,
Umesh

On Tue, Nov 17, 2020 at 11:01:26AM +0000, Chris Wilson wrote:
>From: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
>
>Switch the search and grow code of the _wa_add to use _wa_index and
>_wa_list_grow.
>
>Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
>---
> drivers/gpu/drm/i915/gt/intel_workarounds.c | 124 +++++++++++---------
> 1 file changed, 71 insertions(+), 53 deletions(-)
>
>diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.c b/drivers/gpu/drm/i915/gt/intel_workarounds.c
>index c49083957074..e50c72d2b3f1 100644
>--- a/drivers/gpu/drm/i915/gt/intel_workarounds.c
>+++ b/drivers/gpu/drm/i915/gt/intel_workarounds.c
>@@ -91,20 +91,19 @@ static void wa_init_start(struct i915_wa_list *wal, const char *name, const char
>
> #define WA_LIST_CHUNK (1 << 4)
>
>-static void wa_init_finish(struct i915_wa_list *wal)
>+static void wa_trim(struct i915_wa_list *wal, gfp_t gfp)
> {
>+	struct i915_wa *list;
>+
> 	/* Trim unused entries. */
>-	if (!IS_ALIGNED(wal->count, WA_LIST_CHUNK)) {
>-		struct i915_wa *list = kmemdup(wal->list,
>-					       wal->count * sizeof(*list),
>-					       GFP_KERNEL);
>-
>-		if (list) {
>-			kfree(wal->list);
>-			wal->list = list;
>-		}
>-	}
>+	list = krealloc(wal->list, wal->count * sizeof(*list), gfp);
>+	if (list)
>+		wal->list = list;
>+}
>
>+static void wa_init_finish(struct i915_wa_list *wal)
>+{
>+	wa_trim(wal, GFP_KERNEL);
> 	if (!wal->count)
> 		return;
>
>@@ -112,57 +111,60 @@ static void wa_init_finish(struct i915_wa_list *wal)
> 			 wal->wa_count, wal->name, wal->engine_name);
> }
>
>-static void _wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
>+static int wa_index(struct i915_wa_list *wal, i915_reg_t reg)
> {
>-	unsigned int addr = i915_mmio_reg_offset(wa->reg);
>-	unsigned int start = 0, end = wal->count;
>-	const unsigned int grow = WA_LIST_CHUNK;
>-	struct i915_wa *wa_;
>+	unsigned int addr = i915_mmio_reg_offset(reg);
>+	int start = 0, end = wal->count;
>
>-	GEM_BUG_ON(!is_power_of_2(grow));
>+	/* addr and wal->list[].reg, both include the R/W flags */
>+	while (start < end) {
>+		unsigned int mid = start + (end - start) / 2;
>
>-	if (IS_ALIGNED(wal->count, grow)) { /* Either uninitialized or full. */
>-		struct i915_wa *list;
>+		if (i915_mmio_reg_offset(wal->list[mid].reg) < addr)
>+			start = mid + 1;
>+		else if (i915_mmio_reg_offset(wal->list[mid].reg) > addr)
>+			end = mid;
>+		else
>+			return mid;
>+	}
>
>-		list = kmalloc_array(ALIGN(wal->count + 1, grow), sizeof(*wa),
>-				     GFP_KERNEL);
>-		if (!list) {
>-			DRM_ERROR("No space for workaround init!\n");
>-			return;
>-		}
>+	return -ENOENT;
>+}
>
>-		if (wal->list) {
>-			memcpy(list, wal->list, sizeof(*wa) * wal->count);
>-			kfree(wal->list);
>-		}
>+static int wa_list_grow(struct i915_wa_list *wal, size_t count, gfp_t gfp)
>+{
>+	struct i915_wa *list;
>
>-		wal->list = list;
>-	}
>+	list = krealloc(wal->list, count * sizeof(*list), gfp);
>+	if (!list)
>+		return -ENOMEM;
>
>-	while (start < end) {
>-		unsigned int mid = start + (end - start) / 2;
>+	wal->list = list;
>+	return 0;
>+}
>
>-		if (i915_mmio_reg_offset(wal->list[mid].reg) < addr) {
>-			start = mid + 1;
>-		} else if (i915_mmio_reg_offset(wal->list[mid].reg) > addr) {
>-			end = mid;
>-		} else {
>-			wa_ = &wal->list[mid];
>-
>-			if ((wa->clr | wa_->clr) && !(wa->clr & ~wa_->clr)) {
>-				DRM_ERROR("Discarding overwritten w/a for reg %04x (clear: %08x, set: %08x)\n",
>-					  i915_mmio_reg_offset(wa_->reg),
>-					  wa_->clr, wa_->set);
>-
>-				wa_->set &= ~wa->clr;
>-			}
>-
>-			wal->wa_count++;
>-			wa_->set |= wa->set;
>-			wa_->clr |= wa->clr;
>-			wa_->read |= wa->read;
>-			return;
>+static void __wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
>+{
>+	struct i915_wa *wa_;
>+	int index;
>+
>+	index = wa_index(wal, wa->reg);
>+	if (index >= 0) {
>+		wa_ = &wal->list[index];
>+
>+		if ((wa->clr | wa_->clr) && !(wa->clr & ~wa_->clr)) {
>+			DRM_ERROR("Discarding overwritten w/a for reg %04x (clear: %08x, set: %08x)\n",
>+				  i915_mmio_reg_offset(wa_->reg),
>+				  wa_->clr, wa_->set);
>+
>+			wa_->set &= ~wa->clr;
> 		}
>+
>+		wal->wa_count++;
>+		wa_->set |= wa->set;
>+		wa_->clr |= wa->clr;
>+		wa_->read |= wa->read;
>+		return;
> 	}
>
> 	wal->wa_count++;
>@@ -180,6 +182,22 @@ static void _wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
> 	}
> }
>
>+static void _wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
>+{
>+	const unsigned int grow = WA_LIST_CHUNK;
>+
>+	GEM_BUG_ON(!is_power_of_2(grow));
>+
>+	if (IS_ALIGNED(wal->count, grow) && /* Either uninitialized or full. */
>+	    wa_list_grow(wal, ALIGN(wal->count + 1, grow), GFP_KERNEL)) {
>+		DRM_ERROR("Unable to store w/a for reg %04x\n",
>+			  i915_mmio_reg_offset(wa->reg));
>+		return;
>+	}
>+
>+	__wa_add(wal, wa);
>+}
>+
> static void wa_add(struct i915_wa_list *wal, i915_reg_t reg,
> 		   u32 clear, u32 set, u32 read_mask)
> {
>-- 
>2.20.1
>
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

^ permalink raw reply	[flat|nested] 18+ messages in thread
* Re: [Intel-gfx] [PATCH 4/8] drm/i915/gt: Enable dynamic adjustment of RING_NONPRIV
  2020-11-17 11:01 ` [Intel-gfx] [PATCH 4/8] drm/i915/gt: Enable dynamic adjustment of RING_NONPRIV Chris Wilson
@ 2020-11-17 20:34   ` Umesh Nerlige Ramappa
  0 siblings, 0 replies; 18+ messages in thread
From: Umesh Nerlige Ramappa @ 2020-11-17 20:34 UTC (permalink / raw)
  To: Chris Wilson; +Cc: intel-gfx

On Tue, Nov 17, 2020 at 11:01:28AM +0000, Chris Wilson wrote:
>The OA subsystem would like to enable its privileged clients access to
>the OA registers from execbuf. This requires temporarily removing the
>HW validation from those registers for the duration of the OA client,
>for which we need to allow OA to dynamically adjust the set of
>RING_NONPRIV.
>
>Care must still be taken since the RING_NONPRIV are global, so any and
>all contexts that run at the same time as the OA client, will also be
>able to adjust the registers from their execbuf.
>
>Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
>---
> drivers/gpu/drm/i915/gt/intel_workarounds.c   |  59 +++++
> drivers/gpu/drm/i915/gt/intel_workarounds.h   |   7 +
> .../gpu/drm/i915/gt/selftest_workarounds.c    | 236 ++++++++++++++++++
> 3 files changed, 302 insertions(+)
>
>diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.c b/drivers/gpu/drm/i915/gt/intel_workarounds.c
>index 290aa277ab10..c918739df6a6 100644
>--- a/drivers/gpu/drm/i915/gt/intel_workarounds.c
>+++ b/drivers/gpu/drm/i915/gt/intel_workarounds.c
>@@ -230,6 +230,18 @@ static void _wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
> 	__wa_add(wal, wa);
> }
>
>+static void _wa_del(struct i915_wa_list *wal, i915_reg_t reg)
>+{
>+	struct i915_wa *wa = wal->list;
>+	int index;
>+
>+	index = wa_index(wal, reg);
>+	if (GEM_DEBUG_WARN_ON(index < 0))
>+		return;
>+
>+	memmove(wa + index, wa + index + 1, --wal->count - index);

size to move = (--wal->count - index) * sizeof(*wa)

with that fixed, this is

Reviewed-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>

Thanks,
Umesh

>+}
>+
> static void wa_add(struct i915_wa_list *wal, i915_reg_t reg,
> 		   u32 clear, u32 set, u32 read_mask)
> {
>@@ -2116,6 +2128,53 @@ void intel_engine_init_workarounds(struct intel_engine_cs *engine)
> 	wa_init_finish(wal);
> }
>
>+int intel_engine_allow_user_register_access(struct intel_engine_cs *engine,
>+					    const i915_reg_t *reg,
>+					    unsigned int count)
>+{
>+	struct i915_wa_list *wal = &engine->whitelist;
>+	unsigned long flags;
>+	int err;
>+
>+	if (GEM_DEBUG_WARN_ON(wal->count + count >= RING_MAX_NONPRIV_SLOTS))
>+		return -ENOSPC;
>+
>+	spin_lock_irqsave(&engine->uncore->lock, flags);
>+
>+	err = wa_list_grow(wal, wal->count + count, GFP_ATOMIC | __GFP_NOWARN);
>+	if (err)
>+		goto out;
>+
>+	while (count--) {
>+		struct i915_wa wa = { .reg = *reg++ };
>+
>+		__wa_add(wal, &wa);
>+	}
>+
>+	__engine_apply_whitelist(engine);
>+
>+out:
>+	spin_unlock_irqrestore(&engine->uncore->lock, flags);
>+	return err;
>+}
>+
>+void intel_engine_deny_user_register_access(struct intel_engine_cs *engine,
>+					    const i915_reg_t *reg,
>+					    unsigned int count)
>+{
>+	struct i915_wa_list *wal = &engine->whitelist;
>+	unsigned long flags;
>+
>+	spin_lock_irqsave(&engine->uncore->lock, flags);
>+
>+	while (count--)
>+		_wa_del(wal, *reg++);
>+
>+	__engine_apply_whitelist(engine);
>+
>+	spin_unlock_irqrestore(&engine->uncore->lock, flags);
>+}
>+
> void intel_engine_apply_workarounds(struct intel_engine_cs *engine)
> {
> 	wa_list_apply(engine->uncore, &engine->wa_list);
>diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.h b/drivers/gpu/drm/i915/gt/intel_workarounds.h
>index 8c9c769c2204..9c691493ea56 100644
>--- a/drivers/gpu/drm/i915/gt/intel_workarounds.h
>+++ b/drivers/gpu/drm/i915/gt/intel_workarounds.h
>@@ -37,4 +37,11 @@ void intel_engine_apply_workarounds(struct intel_engine_cs *engine);
> int intel_engine_verify_workarounds(struct intel_engine_cs *engine,
> 				    const char *from);
>
>+int intel_engine_allow_user_register_access(struct intel_engine_cs *engine,
>+					    const i915_reg_t *reg,
>+					    unsigned int count);
>+void intel_engine_deny_user_register_access(struct intel_engine_cs *engine,
>+					    const i915_reg_t *reg,
>+					    unsigned int count);
>+
> #endif
>diff --git a/drivers/gpu/drm/i915/gt/selftest_workarounds.c b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
>index 61a0532d0f3d..6f526199a11a 100644
>--- a/drivers/gpu/drm/i915/gt/selftest_workarounds.c
>+++ b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
>@@ -1127,6 +1127,241 @@ static int live_isolated_whitelist(void *arg)
> 	return err;
> }
>
>+static int rmw_reg(struct intel_engine_cs *engine, const i915_reg_t reg)
>+{
>+	const u32 values[] = {
>+		0x00000000,
>+		0x01010101,
>+		0x10100101,
>+		0x03030303,
>+		0x30300303,
>+		0x05050505,
>+		0x50500505,
>+		0x0f0f0f0f,
>+		0xf00ff00f,
>+		0x10101010,
>+		0xf0f01010,
>+		0x30303030,
>+		0xa0a03030,
>+		0x50505050,
>+		0xc0c05050,
>+		0xf0f0f0f0,
>+		0x11111111,
>+		0x33333333,
>+		0x55555555,
>+		0x0000ffff,
>+		0x00ff00ff,
>+		0xff0000ff,
>+		0xffff00ff,
>+		0xffffffff,
>+	};
>+	struct i915_vma *vma, *batch;
>+	struct intel_context *ce;
>+	struct i915_request *rq;
>+	u32 srm, lrm, idx;
>+	u32 *cs, *results;
>+	u64 addr;
>+	int err;
>+	int v;
>+
>+	ce = intel_context_create(engine);
>+	if (IS_ERR(ce))
>+		return PTR_ERR(ce);
>+
>+	vma = create_scratch(ce->vm, 1);
>+	if (IS_ERR(vma)) {
>+		err = PTR_ERR(vma);
>+		goto out_context;
>+	}
>+
>+	batch = create_batch(ce->vm);
>+	if (IS_ERR(batch)) {
>+		err = PTR_ERR(batch);
>+		goto out_vma;
>+	}
>+
>+	srm = MI_STORE_REGISTER_MEM;
>+	lrm = MI_LOAD_REGISTER_MEM;
>+	if (INTEL_GEN(ce->vm->i915) >= 8)
>+		lrm++, srm++;
>+
>+	addr = vma->node.start;
>+
>+	cs = i915_gem_object_pin_map(batch->obj, I915_MAP_WC);
>+	if (IS_ERR(cs)) {
>+		err = PTR_ERR(cs);
>+		goto out_batch;
>+	}
>+
>+	/* SRM original */
>+	*cs++ = srm;
>+	*cs++ = i915_mmio_reg_offset(reg);
>+	*cs++ = lower_32_bits(addr);
>+	*cs++ = upper_32_bits(addr);
>+
>+	idx = 1;
>+	for (v = 0; v < ARRAY_SIZE(values); v++) {
>+		/* LRI garbage */
>+		*cs++ = MI_LOAD_REGISTER_IMM(1);
>+		*cs++ = i915_mmio_reg_offset(reg);
>+		*cs++ = values[v];
>+
>+		/* SRM result */
>+		*cs++ = srm;
>+		*cs++ = i915_mmio_reg_offset(reg);
>+		*cs++ = lower_32_bits(addr + sizeof(u32) * idx);
>+		*cs++ = upper_32_bits(addr + sizeof(u32) * idx);
>+		idx++;
>+	}
>+	for (v = 0; v < ARRAY_SIZE(values); v++) {
>+		/* LRI garbage */
>+		*cs++ = MI_LOAD_REGISTER_IMM(1);
>+		*cs++ = i915_mmio_reg_offset(reg);
>+		*cs++ = ~values[v];
>+
>+		/* SRM result */
>+		*cs++ = srm;
>+		*cs++ = i915_mmio_reg_offset(reg);
>+		*cs++ = lower_32_bits(addr + sizeof(u32) * idx);
>+		*cs++ = upper_32_bits(addr + sizeof(u32) * idx);
>+		idx++;
>+	}
>+
>+	/* LRM original -- don't leave garbage in the context! */
>+	*cs++ = lrm;
>+	*cs++ = i915_mmio_reg_offset(reg);
>+	*cs++ = lower_32_bits(addr);
>+	*cs++ = upper_32_bits(addr);
>+
>+	*cs++ = MI_BATCH_BUFFER_END;
>+
>+	i915_gem_object_flush_map(batch->obj);
>+	i915_gem_object_unpin_map(batch->obj);
>+	intel_gt_chipset_flush(engine->gt);
>+
>+	rq = intel_context_create_request(ce);
>+	if (IS_ERR(rq)) {
>+		err = PTR_ERR(rq);
>+		goto out_batch;
>+	}
>+
>+	if (engine->emit_init_breadcrumb) { /* Be nice if we hang */
>+		err = engine->emit_init_breadcrumb(rq);
>+		if (err)
>+			goto err_request;
>+	}
>+
>+	i915_vma_lock(batch);
>+	err = i915_request_await_object(rq, batch->obj, false);
>+	if (err == 0)
>+		err = i915_vma_move_to_active(batch, rq, 0);
>+	i915_vma_unlock(batch);
>+	if (err)
>+		goto err_request;
>+
>+	i915_vma_lock(vma);
>+	err = i915_request_await_object(rq, vma->obj, true);
>+	if (err == 0)
>+		err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
>+	i915_vma_unlock(vma);
>+	if (err)
>+		goto err_request;
>+
>+	err = engine->emit_bb_start(rq, batch->node.start, PAGE_SIZE, 0);
>+	if (err)
>+		goto err_request;
>+
>+err_request:
>+	err = request_add_sync(rq, err);
>+	if (err) {
>+		pr_err("%s: Futzing %04x timedout; cancelling test\n",
>+		       engine->name, i915_mmio_reg_offset(reg));
>+		intel_gt_set_wedged(engine->gt);
>+		goto out_batch;
>+	}
>+
>+	results = i915_gem_object_pin_map(vma->obj, I915_MAP_WB);
>+	if (IS_ERR(results)) {
>+		err = PTR_ERR(results);
>+		goto out_batch;
>+	}
>+
>+	for (v = 0, idx = 0; v < 2 * ARRAY_SIZE(values); v++) {
>+		if (results[++idx] != results[0]) {
>+			err = idx;
>+			break;
>+		}
>+	}
>+
>+out_batch:
>+	i915_vma_put(batch);
>+out_vma:
>+	i915_vma_put(vma);
>+out_context:
>+	intel_context_put(ce);
>+	return err;
>+}
>+
>+static int live_dynamic_whitelist(void *arg)
>+{
>+	struct intel_gt *gt = arg;
>+	struct intel_engine_cs *engine;
>+	enum intel_engine_id id;
>+	int err = 0;
>+
>+	if (INTEL_GEN(gt->i915) < 8)
>+		return 0;
>+
>+	for_each_engine(engine, gt, id) {
>+		const i915_reg_t reg = RING_MAX_IDLE(engine->mmio_base);
>+
>+		err = rmw_reg(engine, reg);
>+		if (err < 0)
>+			break;
>+
>+		if (err) {
>+			pr_err("%s: Able to write to protected reg:%04x!\n",
>+			       engine->name, i915_mmio_reg_offset(reg));
>+			err = -EINVAL;
>+			break;
>+		}
>+
>+		err = intel_engine_allow_user_register_access(engine, &reg, 1);
>+		if (err)
>+			break;
>+
>+		err = rmw_reg(engine, reg);
>+		intel_engine_deny_user_register_access(engine, &reg, 1);
>+		if (err < 0)
>+			break;
>+
>+		if (!err) {
>+			pr_err("%s: Unable to write to allowed reg:%04x!\n",
>+			       engine->name, i915_mmio_reg_offset(reg));
>+			err = -EINVAL;
>+			break;
>+		}
>+
>+		err = rmw_reg(engine, reg);
>+		if (err < 0)
>+			break;
>+
>+		if (err) {
>+			pr_err("%s: Able to write to denied reg:%04x!\n",
>+			       engine->name, i915_mmio_reg_offset(reg));
>+			err = -EINVAL;
>+			break;
>+		}
>+
>+		err = 0;
>+	}
>+
>+	if (igt_flush_test(gt->i915))
>+		err = -EIO;
>+
>+	return err;
>+}
>+
> static bool
> verify_wa_lists(struct i915_gem_context *ctx, struct wa_lists *lists,
> 		const char *str)
>@@ -1288,6 +1523,7 @@ int intel_workarounds_live_selftests(struct drm_i915_private *i915)
> 		SUBTEST(live_dirty_whitelist),
> 		SUBTEST(live_reset_whitelist),
> 		SUBTEST(live_isolated_whitelist),
>+		SUBTEST(live_dynamic_whitelist),
> 		SUBTEST(live_gpu_reset_workarounds),
> 		SUBTEST(live_engine_reset_workarounds),
> 	};
>-- 
>2.20.1
>
>_______________________________________________
>Intel-gfx mailing list
>Intel-gfx@lists.freedesktop.org
>https://lists.freedesktop.org/mailman/listinfo/intel-gfx
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

^ permalink raw reply	[flat|nested] 18+ messages in thread
* Re: [Intel-gfx] [PATCH 8/8] drm/i915/perf: Map OA buffer to user space for gen12 performance query
  2020-11-17 11:01 ` [Intel-gfx] [PATCH 8/8] drm/i915/perf: Map OA buffer to user space for gen12 performance query Chris Wilson
@ 2020-11-18 11:41   ` Joonas Lahtinen
  2020-11-18 11:57     ` Lionel Landwerlin
  0 siblings, 1 reply; 18+ messages in thread
From: Joonas Lahtinen @ 2020-11-18 11:41 UTC (permalink / raw)
  To: Chris Wilson, intel-gfx, Umesh Nerlige Ramappa, Lionel Landwerlin
  Cc: Piotr Maciejewski, Chris Wilson

+ Umesh, Lionel

Do we have a link to the userspace changes and IGT tests? Those are
absolutely needed before we can do a final review and merge.

We should really test and review the kernel and userspace changes
together to make sure that we're coming up with a solid uAPI.

Regards, Joonas

Quoting Chris Wilson (2020-11-17 13:01:32)
> From: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
> 
> i915 used to support time based sampling mode which is good for overall
> system monitoring, but is not enough for query mode used to measure a
> single draw call or dispatch. Gen9-Gen11 are using current i915 perf
> implementation for query, but Gen12+ requires a new approach for query
> based on triggered reports within oa buffer.
> 
> Triggering reports into the OA buffer is achieved by writing into a
> a trigger register. Optionally an unused counter/register is set with a
> marker value such that a triggered report can be identified in the OA
> buffer. Reports are usually triggered at the start and end of work that
> is measured.
> 
> Since OA buffer is large and queries can be frequent, an efficient way
> to look for triggered reports is required. By knowing the current head
> and tail offsets into the OA buffer, it is easier to determine the
> locality of the reports of interest.
> 
> Current perf OA interface does not expose head/tail information to the
> user and it filters out invalid reports before sending data to user.
> Also considering limited size of user buffer used during a query,
> creating a 1:1 copy of the OA buffer at the user space added undesired
> complexity.
> 
> The solution was to map the OA buffer to user space provided
> 
> (1) that it is accessed from a privileged user.
> (2) OA report filtering is not used.
> 
> These 2 conditions would satisfy the safety criteria that the current
> perf interface addresses.
> 
> To enable the query:
> - Add an ioctl to expose head and tail to the user
> - Add an ioctl to return size and offset of the OA buffer
> - Map the OA buffer to the user space
> 
> v2:
> - Improve commit message (Chris)
> - Do not mmap based on gem object filp. Instead, use perf_fd and support
>   mmap syscall (Chris)
> - Pass non-zero offset in mmap to enforce the right object is
>   mapped (Chris)
> - Do not expose gpu_address (Chris)
> - Verify start and length of vma for page alignment (Lionel)
> - Move SQNTL config out (Lionel)
> 
> v3: (Chris)
> - Omit redundant checks
> - Return VM_FAULT_SIGBUS is old stream is closed
> - Maintain reference counts to stream in vm_open and vm_close
> - Use switch to identify object to be mapped
> 
> v4: Call kref_put on closing perf fd (Chris)
> v5:
> - Strip access to OA buffer from unprivileged child of a privileged
>   parent. Use VM_DONTCOPY
> - Enforce MAP_PRIVATE by checking for VM_MAYSHARE
> 
> v6:
> (Chris)
> - Use len of -1 in unmap_mapping_range
> - Don't use stream->oa_buffer.vma->obj in vm_fault_oa
> - Use kernel block comment style
> - do_mmap gets a reference to the file and puts it in do_munmap, so
>   no need to maintain a reference to i915_perf_stream. Hence, remove
>   vm_open/vm_close and stream->closed hooks/checks.
> (Umesh)
> - Do not allow mmap if SAMPLE_OA_REPORT is not set during
>   i915_perf_open_ioctl.
> - Drop ioctl returning head/tail since this information is already
>   whitelisted. Remove hooks to read head register.
> 
> v7: (Chris)
> - unmap before destroy
> - change ioctl argument struct
> 
> v8: Documentation and more checks (Chris)
> 
> Signed-off-by: Piotr Maciejewski <piotr.maciejewski@intel.com>
> Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
> ---
>  drivers/gpu/drm/i915/gem/i915_gem_mman.c |   2 +-
>  drivers/gpu/drm/i915/gem/i915_gem_mman.h |   2 +
>  drivers/gpu/drm/i915/i915_perf.c         | 126 ++++++++++++++++++++++-
>  include/uapi/drm/i915_drm.h              |  33 ++++++
>  4 files changed, 161 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_mman.c b/drivers/gpu/drm/i915/gem/i915_gem_mman.c
> index 3d69e51f3e4d..2ab08b152b9d 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_mman.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_mman.c
> @@ -204,7 +204,7 @@ compute_partial_view(const struct drm_i915_gem_object *obj,
>         return view;
>  }
>  
> -static vm_fault_t i915_error_to_vmf_fault(int err)
> +vm_fault_t i915_error_to_vmf_fault(int err)
>  {
>         switch (err) {
>         default:
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_mman.h b/drivers/gpu/drm/i915/gem/i915_gem_mman.h
> index efee9e0d2508..1190a3a228ea 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_mman.h
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_mman.h
> @@ -29,4 +29,6 @@ void i915_gem_object_release_mmap_gtt(struct drm_i915_gem_object *obj);
>  
>  void i915_gem_object_release_mmap_offset(struct drm_i915_gem_object *obj);
>  
> +vm_fault_t i915_error_to_vmf_fault(int err);
> +
>  #endif
> diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
> index c91f2da84189..6fd669b520d8 100644
> --- a/drivers/gpu/drm/i915/i915_perf.c
> +++ b/drivers/gpu/drm/i915/i915_perf.c
> @@ -192,10 +192,12 @@
>   */
>  
>  #include <linux/anon_inodes.h>
> +#include <linux/mman.h>
>  #include <linux/sizes.h>
>  #include <linux/uuid.h>
>  
>  #include "gem/i915_gem_context.h"
> +#include "gem/i915_gem_mman.h"
>  #include "gt/intel_engine_pm.h"
>  #include "gt/intel_engine_user.h"
>  #include "gt/intel_gt.h"
> @@ -3291,6 +3293,44 @@ static long i915_perf_config_locked(struct i915_perf_stream *stream,
>         return ret;
>  }
>  
> +#define I915_PERF_OA_BUFFER_MMAP_OFFSET 1
> +
> +/**
> + * i915_perf_oa_buffer_info_locked - size and offset of the OA buffer
> + * @stream: i915 perf stream
> + * @cmd: ioctl command
> + * @arg: pointer to oa buffer info filled by this function.
> + */
> +static int i915_perf_oa_buffer_info_locked(struct i915_perf_stream *stream,
> +                                          unsigned int cmd,
> +                                          unsigned long arg)
> +{
> +       struct drm_i915_perf_oa_buffer_info info;
> +       void __user *output = (void __user *)arg;
> +
> +       if (i915_perf_stream_paranoid && !perfmon_capable()) {
> +               DRM_DEBUG("Insufficient privileges to access OA buffer info\n");
> +               return -EACCES;
> +       }
> +
> +       if (_IOC_SIZE(cmd) != sizeof(info))
> +               return -EINVAL;
> +
> +       if (copy_from_user(&info, output, sizeof(info)))
> +               return -EFAULT;
> +
> +       if (info.type || info.flags || info.rsvd)
> +               return -EINVAL;
> +
> +       info.size = stream->oa_buffer.vma->size;
> +       info.offset = I915_PERF_OA_BUFFER_MMAP_OFFSET * PAGE_SIZE;
> +
> +       if (copy_to_user(output, &info, sizeof(info)))
> +               return -EFAULT;
> +
> +       return 0;
> +}
> +
>  /**
>   * i915_perf_ioctl_locked - support ioctl() usage with i915 perf stream FDs
>   * @stream: An i915 perf stream
> @@ -3316,6 +3356,8 @@ static long i915_perf_ioctl_locked(struct i915_perf_stream *stream,
>                 return 0;
>         case I915_PERF_IOCTL_CONFIG:
>                 return i915_perf_config_locked(stream, arg);
> +       case I915_PERF_IOCTL_GET_OA_BUFFER_INFO:
> +               return i915_perf_oa_buffer_info_locked(stream, cmd, arg);
>         }
>  
>         return -EINVAL;
> @@ -3387,6 +3429,14 @@ static int i915_perf_release(struct inode *inode, struct file *file)
>         struct i915_perf_stream *stream = file->private_data;
>         struct i915_perf *perf = stream->perf;
>  
> +       /*
> +        * User could have multiple vmas from multiple mmaps. We want to zap
> +        * them all here. Note that a fresh fault cannot occur as the mmap holds
> +        * a reference to the stream via the vma->vm_file, so before user's
> +        * munmap, the stream cannot be destroyed.
> +        */
> +       unmap_mapping_range(file->f_mapping, 0, -1, 1);
> +
>         mutex_lock(&perf->lock);
>         i915_perf_destroy_locked(stream);
>         mutex_unlock(&perf->lock);
> @@ -3397,6 +3447,75 @@ static int i915_perf_release(struct inode *inode, struct file *file)
>         return 0;
>  }
>  
> +static vm_fault_t vm_fault_oa(struct vm_fault *vmf)
> +{
> +       struct vm_area_struct *vma = vmf->vma;
> +       struct i915_perf_stream *stream = vma->vm_private_data;
> +       int err;
> +
> +       err = remap_io_sg(vma,
> +                         vma->vm_start, vma->vm_end - vma->vm_start,
> +                         stream->oa_buffer.vma->pages->sgl, -1);
> +
> +       return i915_error_to_vmf_fault(err);
> +}
> +
> +static const struct vm_operations_struct vm_ops_oa = {
> +       .fault = vm_fault_oa,
> +};
> +
> +static int i915_perf_mmap(struct file *file, struct vm_area_struct *vma)
> +{
> +       struct i915_perf_stream *stream = file->private_data;
> +
> +       /* mmap-ing OA buffer to user space MUST absolutely be privileged */
> +       if (i915_perf_stream_paranoid && !perfmon_capable()) {
> +               DRM_DEBUG("Insufficient privileges to map OA buffer\n");
> +               return -EACCES;
> +       }
> +
> +       switch (vma->vm_pgoff) {
> +       /*
> +        * A non-zero offset ensures that we are mapping the right object. Also
> +        * leaves room for future objects added to this implementation.
> +        */
> +       case I915_PERF_OA_BUFFER_MMAP_OFFSET:
> +               if (!(stream->sample_flags & SAMPLE_OA_REPORT))
> +                       return -EINVAL;
> +
> +               if (vma->vm_end - vma->vm_start > OA_BUFFER_SIZE)
> +                       return -EINVAL;
> +
> +               /*
> +                * Only support VM_READ. Enforce MAP_PRIVATE by checking for
> +                * VM_MAYSHARE.
> +                */
> +               if (vma->vm_flags & (VM_WRITE | VM_EXEC |
> +                                    VM_SHARED | VM_MAYSHARE))
> +                       return -EINVAL;
> +
> +               vma->vm_flags &= ~(VM_MAYWRITE | VM_MAYEXEC);
> +
> +               /*
> +                * If the privileged parent forks and child drops root
> +                * privilege, we do not want the child to retain access to the
> +                * mapped OA buffer. Explicitly set VM_DONTCOPY to avoid such
> +                * cases.
> +                */
> +               vma->vm_flags |= VM_PFNMAP | VM_DONTEXPAND |
> +                                VM_DONTDUMP | VM_DONTCOPY;
> +               break;
> +
> +       default:
> +               return -EINVAL;
> +       }
> +
> +       vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
> +       vma->vm_private_data = stream;
> +       vma->vm_ops = &vm_ops_oa;
> +
> +       return 0;
> +}
>  
>  static const struct file_operations fops = {
>         .owner          = THIS_MODULE,
> @@ -3409,6 +3528,7 @@ static const struct file_operations fops = {
>          * to handle 32bits compatibility.
>          */
>         .compat_ioctl   = i915_perf_ioctl,
> +       .mmap           = i915_perf_mmap,
>  };
>  
>  
> @@ -4559,8 +4679,12 @@ int i915_perf_ioctl_version(void)
>          *
>          *    - OA buffer head/tail/status/buffer registers for read only
>          *    - OA counters A18, A19, A20 for read/write
> +        *
> +        * 8: Added an option to map oa buffer at umd driver level and trigger
> +        *    oa reports within oa buffer from command buffer. See
> +        *    I915_PERF_IOCTL_GET_OA_BUFFER_INFO.
>          */
> -       return 7;
> +       return 8;
>  }
>  
>  #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
> diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
> index fa1f3d62f9a6..cc1702ddc859 100644
> --- a/include/uapi/drm/i915_drm.h
> +++ b/include/uapi/drm/i915_drm.h
> @@ -2101,6 +2101,39 @@ struct drm_i915_perf_open_param {
>   */
>  #define I915_PERF_IOCTL_CONFIG _IO('i', 0x2)
>  
> +/**
> + * Returns OA buffer properties to be used with mmap.
> + *
> + * This ioctl is available in perf revision 8.
> + */
> +#define I915_PERF_IOCTL_GET_OA_BUFFER_INFO _IOWR('i', 0x3, struct drm_i915_perf_oa_buffer_info)
> +
> +/**
> + * OA buffer size and offset.
> + *
> + * OA output buffer
> + *   type: 0
> + *   flags: mbz
> + *
> + *   After querying the info, pass (size,offset) to mmap(),
> + *
> + *   mmap(0, info.size, PROT_READ, MAP_PRIVATE, perf_fd, info.offset).
> + *
> + *   Note that only a private (not shared between processes, or across fork())
> + *   read-only mmapping is allowed.
> + *
> + *   Userspace must treat the incoming data as tainted, but it conforms to the OA
> + *   format as specified by user config. The buffer provides reports that have
> + *   OA counters - A, B and C.
> + */
> +struct drm_i915_perf_oa_buffer_info {
> +       __u32 type;   /* in */
> +       __u32 flags;  /* in */
> +       __u64 size;   /* out */
> +       __u64 offset; /* out */
> +       __u64 rsvd;   /* mbz */
> +};
> +
>  /**
>   * Common to all i915 perf records
>   */
> -- 
> 2.20.1
> 
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/intel-gfx
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

^ permalink raw reply	[flat|nested] 18+ messages in thread
* Re: [Intel-gfx] [PATCH 8/8] drm/i915/perf: Map OA buffer to user space for gen12 performance query
  2020-11-18 11:41   ` Joonas Lahtinen
@ 2020-11-18 11:57     ` Lionel Landwerlin
  2020-11-19 10:56       ` Maciejewski, Piotr
  0 siblings, 1 reply; 18+ messages in thread
From: Lionel Landwerlin @ 2020-11-18 11:57 UTC (permalink / raw)
  To: Joonas Lahtinen, Chris Wilson, intel-gfx, Umesh Nerlige Ramappa
  Cc: Piotr Maciejewski

Tests are on the IGT ml : 
https://patchwork.freedesktop.org/series/82352/ 
<https://patchwork.freedesktop.org/series/82352/>

This feature was requested by the metrics-discovery team.
Pretty sure this was tested with it, maybe a branch can be shared by Piotr?

Cheers,

-Lionel

On 18/11/2020 13:41, Joonas Lahtinen wrote:
> + Umesh, Lionel
>
> Do we have a link to the userspace changes and IGT tests? Those are
> absolutely needed before we can do a final review and merge.
>
> We should really test and review the kernel and userspace changes
> together to make sure that we're coming up with a solid uAPI.
>
> Regards, Joonas
>
> Quoting Chris Wilson (2020-11-17 13:01:32)
>> From: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
>>
>> i915 used to support time based sampling mode which is good for overall
>> system monitoring, but is not enough for query mode used to measure a
>> single draw call or dispatch. Gen9-Gen11 are using current i915 perf
>> implementation for query, but Gen12+ requires a new approach for query
>> based on triggered reports within oa buffer.
>>
>> Triggering reports into the OA buffer is achieved by writing into a
>> a trigger register. Optionally an unused counter/register is set with a
>> marker value such that a triggered report can be identified in the OA
>> buffer. Reports are usually triggered at the start and end of work that
>> is measured.
>>
>> Since OA buffer is large and queries can be frequent, an efficient way
>> to look for triggered reports is required. By knowing the current head
>> and tail offsets into the OA buffer, it is easier to determine the
>> locality of the reports of interest.
>>
>> Current perf OA interface does not expose head/tail information to the
>> user and it filters out invalid reports before sending data to user.
>> Also considering limited size of user buffer used during a query,
>> creating a 1:1 copy of the OA buffer at the user space added undesired
>> complexity.
>>
>> The solution was to map the OA buffer to user space provided
>>
>> (1) that it is accessed from a privileged user.
>> (2) OA report filtering is not used.
>>
>> These 2 conditions would satisfy the safety criteria that the current
>> perf interface addresses.
>>
>> To enable the query:
>> - Add an ioctl to expose head and tail to the user
>> - Add an ioctl to return size and offset of the OA buffer
>> - Map the OA buffer to the user space
>>
>> v2:
>> - Improve commit message (Chris)
>> - Do not mmap based on gem object filp. Instead, use perf_fd and support
>>    mmap syscall (Chris)
>> - Pass non-zero offset in mmap to enforce the right object is
>>    mapped (Chris)
>> - Do not expose gpu_address (Chris)
>> - Verify start and length of vma for page alignment (Lionel)
>> - Move SQNTL config out (Lionel)
>>
>> v3: (Chris)
>> - Omit redundant checks
>> - Return VM_FAULT_SIGBUS is old stream is closed
>> - Maintain reference counts to stream in vm_open and vm_close
>> - Use switch to identify object to be mapped
>>
>> v4: Call kref_put on closing perf fd (Chris)
>> v5:
>> - Strip access to OA buffer from unprivileged child of a privileged
>>    parent. Use VM_DONTCOPY
>> - Enforce MAP_PRIVATE by checking for VM_MAYSHARE
>>
>> v6:
>> (Chris)
>> - Use len of -1 in unmap_mapping_range
>> - Don't use stream->oa_buffer.vma->obj in vm_fault_oa
>> - Use kernel block comment style
>> - do_mmap gets a reference to the file and puts it in do_munmap, so
>>    no need to maintain a reference to i915_perf_stream. Hence, remove
>>    vm_open/vm_close and stream->closed hooks/checks.
>> (Umesh)
>> - Do not allow mmap if SAMPLE_OA_REPORT is not set during
>>    i915_perf_open_ioctl.
>> - Drop ioctl returning head/tail since this information is already
>>    whitelisted. Remove hooks to read head register.
>>
>> v7: (Chris)
>> - unmap before destroy
>> - change ioctl argument struct
>>
>> v8: Documentation and more checks (Chris)
>>
>> Signed-off-by: Piotr Maciejewski <piotr.maciejewski@intel.com>
>> Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
>> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
>> ---
>>   drivers/gpu/drm/i915/gem/i915_gem_mman.c |   2 +-
>>   drivers/gpu/drm/i915/gem/i915_gem_mman.h |   2 +
>>   drivers/gpu/drm/i915/i915_perf.c         | 126 ++++++++++++++++++++++-
>>   include/uapi/drm/i915_drm.h              |  33 ++++++
>>   4 files changed, 161 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_mman.c b/drivers/gpu/drm/i915/gem/i915_gem_mman.c
>> index 3d69e51f3e4d..2ab08b152b9d 100644
>> --- a/drivers/gpu/drm/i915/gem/i915_gem_mman.c
>> +++ b/drivers/gpu/drm/i915/gem/i915_gem_mman.c
>> @@ -204,7 +204,7 @@ compute_partial_view(const struct drm_i915_gem_object *obj,
>>          return view;
>>   }
>>   
>> -static vm_fault_t i915_error_to_vmf_fault(int err)
>> +vm_fault_t i915_error_to_vmf_fault(int err)
>>   {
>>          switch (err) {
>>          default:
>> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_mman.h b/drivers/gpu/drm/i915/gem/i915_gem_mman.h
>> index efee9e0d2508..1190a3a228ea 100644
>> --- a/drivers/gpu/drm/i915/gem/i915_gem_mman.h
>> +++ b/drivers/gpu/drm/i915/gem/i915_gem_mman.h
>> @@ -29,4 +29,6 @@ void i915_gem_object_release_mmap_gtt(struct drm_i915_gem_object *obj);
>>   
>>   void i915_gem_object_release_mmap_offset(struct drm_i915_gem_object *obj);
>>   
>> +vm_fault_t i915_error_to_vmf_fault(int err);
>> +
>>   #endif
>> diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
>> index c91f2da84189..6fd669b520d8 100644
>> --- a/drivers/gpu/drm/i915/i915_perf.c
>> +++ b/drivers/gpu/drm/i915/i915_perf.c
>> @@ -192,10 +192,12 @@
>>    */
>>   
>>   #include <linux/anon_inodes.h>
>> +#include <linux/mman.h>
>>   #include <linux/sizes.h>
>>   #include <linux/uuid.h>
>>   
>>   #include "gem/i915_gem_context.h"
>> +#include "gem/i915_gem_mman.h"
>>   #include "gt/intel_engine_pm.h"
>>   #include "gt/intel_engine_user.h"
>>   #include "gt/intel_gt.h"
>> @@ -3291,6 +3293,44 @@ static long i915_perf_config_locked(struct i915_perf_stream *stream,
>>          return ret;
>>   }
>>   
>> +#define I915_PERF_OA_BUFFER_MMAP_OFFSET 1
>> +
>> +/**
>> + * i915_perf_oa_buffer_info_locked - size and offset of the OA buffer
>> + * @stream: i915 perf stream
>> + * @cmd: ioctl command
>> + * @arg: pointer to oa buffer info filled by this function.
>> + */
>> +static int i915_perf_oa_buffer_info_locked(struct i915_perf_stream *stream,
>> +                                          unsigned int cmd,
>> +                                          unsigned long arg)
>> +{
>> +       struct drm_i915_perf_oa_buffer_info info;
>> +       void __user *output = (void __user *)arg;
>> +
>> +       if (i915_perf_stream_paranoid && !perfmon_capable()) {
>> +               DRM_DEBUG("Insufficient privileges to access OA buffer info\n");
>> +               return -EACCES;
>> +       }
>> +
>> +       if (_IOC_SIZE(cmd) != sizeof(info))
>> +               return -EINVAL;
>> +
>> +       if (copy_from_user(&info, output, sizeof(info)))
>> +               return -EFAULT;
>> +
>> +       if (info.type || info.flags || info.rsvd)
>> +               return -EINVAL;
>> +
>> +       info.size = stream->oa_buffer.vma->size;
>> +       info.offset = I915_PERF_OA_BUFFER_MMAP_OFFSET * PAGE_SIZE;
>> +
>> +       if (copy_to_user(output, &info, sizeof(info)))
>> +               return -EFAULT;
>> +
>> +       return 0;
>> +}
>> +
>>   /**
>>    * i915_perf_ioctl_locked - support ioctl() usage with i915 perf stream FDs
>>    * @stream: An i915 perf stream
>> @@ -3316,6 +3356,8 @@ static long i915_perf_ioctl_locked(struct i915_perf_stream *stream,
>>                  return 0;
>>          case I915_PERF_IOCTL_CONFIG:
>>                  return i915_perf_config_locked(stream, arg);
>> +       case I915_PERF_IOCTL_GET_OA_BUFFER_INFO:
>> +               return i915_perf_oa_buffer_info_locked(stream, cmd, arg);
>>          }
>>   
>>          return -EINVAL;
>> @@ -3387,6 +3429,14 @@ static int i915_perf_release(struct inode *inode, struct file *file)
>>          struct i915_perf_stream *stream = file->private_data;
>>          struct i915_perf *perf = stream->perf;
>>   
>> +       /*
>> +        * User could have multiple vmas from multiple mmaps. We want to zap
>> +        * them all here. Note that a fresh fault cannot occur as the mmap holds
>> +        * a reference to the stream via the vma->vm_file, so before user's
>> +        * munmap, the stream cannot be destroyed.
>> +        */
>> +       unmap_mapping_range(file->f_mapping, 0, -1, 1);
>> +
>>          mutex_lock(&perf->lock);
>>          i915_perf_destroy_locked(stream);
>>          mutex_unlock(&perf->lock);
>> @@ -3397,6 +3447,75 @@ static int i915_perf_release(struct inode *inode, struct file *file)
>>          return 0;
>>   }
>>   
>> +static vm_fault_t vm_fault_oa(struct vm_fault *vmf)
>> +{
>> +       struct vm_area_struct *vma = vmf->vma;
>> +       struct i915_perf_stream *stream = vma->vm_private_data;
>> +       int err;
>> +
>> +       err = remap_io_sg(vma,
>> +                         vma->vm_start, vma->vm_end - vma->vm_start,
>> +                         stream->oa_buffer.vma->pages->sgl, -1);
>> +
>> +       return i915_error_to_vmf_fault(err);
>> +}
>> +
>> +static const struct vm_operations_struct vm_ops_oa = {
>> +       .fault = vm_fault_oa,
>> +};
>> +
>> +static int i915_perf_mmap(struct file *file, struct vm_area_struct *vma)
>> +{
>> +       struct i915_perf_stream *stream = file->private_data;
>> +
>> +       /* mmap-ing OA buffer to user space MUST absolutely be privileged */
>> +       if (i915_perf_stream_paranoid && !perfmon_capable()) {
>> +               DRM_DEBUG("Insufficient privileges to map OA buffer\n");
>> +               return -EACCES;
>> +       }
>> +
>> +       switch (vma->vm_pgoff) {
>> +       /*
>> +        * A non-zero offset ensures that we are mapping the right object. Also
>> +        * leaves room for future objects added to this implementation.
>> +        */
>> +       case I915_PERF_OA_BUFFER_MMAP_OFFSET:
>> +               if (!(stream->sample_flags & SAMPLE_OA_REPORT))
>> +                       return -EINVAL;
>> +
>> +               if (vma->vm_end - vma->vm_start > OA_BUFFER_SIZE)
>> +                       return -EINVAL;
>> +
>> +               /*
>> +                * Only support VM_READ. Enforce MAP_PRIVATE by checking for
>> +                * VM_MAYSHARE.
>> +                */
>> +               if (vma->vm_flags & (VM_WRITE | VM_EXEC |
>> +                                    VM_SHARED | VM_MAYSHARE))
>> +                       return -EINVAL;
>> +
>> +               vma->vm_flags &= ~(VM_MAYWRITE | VM_MAYEXEC);
>> +
>> +               /*
>> +                * If the privileged parent forks and child drops root
>> +                * privilege, we do not want the child to retain access to the
>> +                * mapped OA buffer. Explicitly set VM_DONTCOPY to avoid such
>> +                * cases.
>> +                */
>> +               vma->vm_flags |= VM_PFNMAP | VM_DONTEXPAND |
>> +                                VM_DONTDUMP | VM_DONTCOPY;
>> +               break;
>> +
>> +       default:
>> +               return -EINVAL;
>> +       }
>> +
>> +       vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
>> +       vma->vm_private_data = stream;
>> +       vma->vm_ops = &vm_ops_oa;
>> +
>> +       return 0;
>> +}
>>   
>>   static const struct file_operations fops = {
>>          .owner          = THIS_MODULE,
>> @@ -3409,6 +3528,7 @@ static const struct file_operations fops = {
>>           * to handle 32bits compatibility.
>>           */
>>          .compat_ioctl   = i915_perf_ioctl,
>> +       .mmap           = i915_perf_mmap,
>>   };
>>   
>>   
>> @@ -4559,8 +4679,12 @@ int i915_perf_ioctl_version(void)
>>           *
>>           *    - OA buffer head/tail/status/buffer registers for read only
>>           *    - OA counters A18, A19, A20 for read/write
>> +        *
>> +        * 8: Added an option to map oa buffer at umd driver level and trigger
>> +        *    oa reports within oa buffer from command buffer. See
>> +        *    I915_PERF_IOCTL_GET_OA_BUFFER_INFO.
>>           */
>> -       return 7;
>> +       return 8;
>>   }
>>   
>>   #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
>> diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
>> index fa1f3d62f9a6..cc1702ddc859 100644
>> --- a/include/uapi/drm/i915_drm.h
>> +++ b/include/uapi/drm/i915_drm.h
>> @@ -2101,6 +2101,39 @@ struct drm_i915_perf_open_param {
>>    */
>>   #define I915_PERF_IOCTL_CONFIG _IO('i', 0x2)
>>   
>> +/**
>> + * Returns OA buffer properties to be used with mmap.
>> + *
>> + * This ioctl is available in perf revision 8.
>> + */
>> +#define I915_PERF_IOCTL_GET_OA_BUFFER_INFO _IOWR('i', 0x3, struct drm_i915_perf_oa_buffer_info)
>> +
>> +/**
>> + * OA buffer size and offset.
>> + *
>> + * OA output buffer
>> + *   type: 0
>> + *   flags: mbz
>> + *
>> + *   After querying the info, pass (size,offset) to mmap(),
>> + *
>> + *   mmap(0, info.size, PROT_READ, MAP_PRIVATE, perf_fd, info.offset).
>> + *
>> + *   Note that only a private (not shared between processes, or across fork())
>> + *   read-only mmapping is allowed.
>> + *
>> + *   Userspace must treat the incoming data as tainted, but it conforms to the OA
>> + *   format as specified by user config. The buffer provides reports that have
>> + *   OA counters - A, B and C.
>> + */
>> +struct drm_i915_perf_oa_buffer_info {
>> +       __u32 type;   /* in */
>> +       __u32 flags;  /* in */
>> +       __u64 size;   /* out */
>> +       __u64 offset; /* out */
>> +       __u64 rsvd;   /* mbz */
>> +};
>> +
>>   /**
>>    * Common to all i915 perf records
>>    */
>> -- 
>> 2.20.1
>>
>> _______________________________________________
>> Intel-gfx mailing list
>> Intel-gfx@lists.freedesktop.org
>> https://lists.freedesktop.org/mailman/listinfo/intel-gfx


_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

^ permalink raw reply	[flat|nested] 18+ messages in thread
* Re: [Intel-gfx] [PATCH 8/8] drm/i915/perf: Map OA buffer to user space for gen12 performance query
  2020-11-18 11:57     ` Lionel Landwerlin
@ 2020-11-19 10:56       ` Maciejewski, Piotr
  0 siblings, 0 replies; 18+ messages in thread
From: Maciejewski, Piotr @ 2020-11-19 10:56 UTC (permalink / raw)
  To: Landwerlin, Lionel G, Joonas Lahtinen, Chris Wilson, intel-gfx,
	Nerlige Ramappa, Umesh

Hi,

Indeed, we use i915 perf changes introduced by Umesh within Metrics Library (a common library used to obtain performance counters from various operating systems and graphics api).

Metrics Library repo: https://github.com/intel/metrics-library  
Io controls usage: https://github.com/intel/metrics-library/blob/master/instrumentation/metrics_library/traits/kernel_interface/linux/ml_io_control.h 
Tbs usage: https://github.com/intel/metrics-library/blob/master/instrumentation/metrics_library/traits/kernel_interface/linux/ml_tbs_interface.h 

Piotr

-----Original Message-----
From: Lionel Landwerlin <lionel.g.landwerlin@intel.com> 
Sent: Wednesday, November 18, 2020 12:57 PM
To: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>; Chris Wilson <chris@chris-wilson.co.uk>; intel-gfx@lists.freedesktop.org; Nerlige Ramappa, Umesh <umesh.nerlige.ramappa@intel.com>
Cc: Maciejewski, Piotr <piotr.maciejewski@intel.com>
Subject: Re: [Intel-gfx] [PATCH 8/8] drm/i915/perf: Map OA buffer to user space for gen12 performance query

Tests are on the IGT ml : 
https://patchwork.freedesktop.org/series/82352/
<https://patchwork.freedesktop.org/series/82352/>

This feature was requested by the metrics-discovery team.
Pretty sure this was tested with it, maybe a branch can be shared by Piotr?

Cheers,

-Lionel

On 18/11/2020 13:41, Joonas Lahtinen wrote:
> + Umesh, Lionel
>
> Do we have a link to the userspace changes and IGT tests? Those are 
> absolutely needed before we can do a final review and merge.
>
> We should really test and review the kernel and userspace changes 
> together to make sure that we're coming up with a solid uAPI.
>
> Regards, Joonas
>
> Quoting Chris Wilson (2020-11-17 13:01:32)
>> From: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
>>
>> i915 used to support time based sampling mode which is good for 
>> overall system monitoring, but is not enough for query mode used to 
>> measure a single draw call or dispatch. Gen9-Gen11 are using current 
>> i915 perf implementation for query, but Gen12+ requires a new 
>> approach for query based on triggered reports within oa buffer.
>>
>> Triggering reports into the OA buffer is achieved by writing into a a 
>> trigger register. Optionally an unused counter/register is set with a 
>> marker value such that a triggered report can be identified in the OA 
>> buffer. Reports are usually triggered at the start and end of work 
>> that is measured.
>>
>> Since OA buffer is large and queries can be frequent, an efficient 
>> way to look for triggered reports is required. By knowing the current 
>> head and tail offsets into the OA buffer, it is easier to determine 
>> the locality of the reports of interest.
>>
>> Current perf OA interface does not expose head/tail information to 
>> the user and it filters out invalid reports before sending data to user.
>> Also considering limited size of user buffer used during a query, 
>> creating a 1:1 copy of the OA buffer at the user space added 
>> undesired complexity.
>>
>> The solution was to map the OA buffer to user space provided
>>
>> (1) that it is accessed from a privileged user.
>> (2) OA report filtering is not used.
>>
>> These 2 conditions would satisfy the safety criteria that the current 
>> perf interface addresses.
>>
>> To enable the query:
>> - Add an ioctl to expose head and tail to the user
>> - Add an ioctl to return size and offset of the OA buffer
>> - Map the OA buffer to the user space
>>
>> v2:
>> - Improve commit message (Chris)
>> - Do not mmap based on gem object filp. Instead, use perf_fd and support
>>    mmap syscall (Chris)
>> - Pass non-zero offset in mmap to enforce the right object is
>>    mapped (Chris)
>> - Do not expose gpu_address (Chris)
>> - Verify start and length of vma for page alignment (Lionel)
>> - Move SQNTL config out (Lionel)
>>
>> v3: (Chris)
>> - Omit redundant checks
>> - Return VM_FAULT_SIGBUS is old stream is closed
>> - Maintain reference counts to stream in vm_open and vm_close
>> - Use switch to identify object to be mapped
>>
>> v4: Call kref_put on closing perf fd (Chris)
>> v5:
>> - Strip access to OA buffer from unprivileged child of a privileged
>>    parent. Use VM_DONTCOPY
>> - Enforce MAP_PRIVATE by checking for VM_MAYSHARE
>>
>> v6:
>> (Chris)
>> - Use len of -1 in unmap_mapping_range
>> - Don't use stream->oa_buffer.vma->obj in vm_fault_oa
>> - Use kernel block comment style
>> - do_mmap gets a reference to the file and puts it in do_munmap, so
>>    no need to maintain a reference to i915_perf_stream. Hence, remove
>>    vm_open/vm_close and stream->closed hooks/checks.
>> (Umesh)
>> - Do not allow mmap if SAMPLE_OA_REPORT is not set during
>>    i915_perf_open_ioctl.
>> - Drop ioctl returning head/tail since this information is already
>>    whitelisted. Remove hooks to read head register.
>>
>> v7: (Chris)
>> - unmap before destroy
>> - change ioctl argument struct
>>
>> v8: Documentation and more checks (Chris)
>>
>> Signed-off-by: Piotr Maciejewski <piotr.maciejewski@intel.com>
>> Signed-off-by: Umesh Nerlige Ramappa 
>> <umesh.nerlige.ramappa@intel.com>
>> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
>> ---
>>   drivers/gpu/drm/i915/gem/i915_gem_mman.c |   2 +-
>>   drivers/gpu/drm/i915/gem/i915_gem_mman.h |   2 +
>>   drivers/gpu/drm/i915/i915_perf.c         | 126 ++++++++++++++++++++++-
>>   include/uapi/drm/i915_drm.h              |  33 ++++++
>>   4 files changed, 161 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_mman.c 
>> b/drivers/gpu/drm/i915/gem/i915_gem_mman.c
>> index 3d69e51f3e4d..2ab08b152b9d 100644
>> --- a/drivers/gpu/drm/i915/gem/i915_gem_mman.c
>> +++ b/drivers/gpu/drm/i915/gem/i915_gem_mman.c
>> @@ -204,7 +204,7 @@ compute_partial_view(const struct drm_i915_gem_object *obj,
>>          return view;
>>   }
>>   
>> -static vm_fault_t i915_error_to_vmf_fault(int err)
>> +vm_fault_t i915_error_to_vmf_fault(int err)
>>   {
>>          switch (err) {
>>          default:
>> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_mman.h 
>> b/drivers/gpu/drm/i915/gem/i915_gem_mman.h
>> index efee9e0d2508..1190a3a228ea 100644
>> --- a/drivers/gpu/drm/i915/gem/i915_gem_mman.h
>> +++ b/drivers/gpu/drm/i915/gem/i915_gem_mman.h
>> @@ -29,4 +29,6 @@ void i915_gem_object_release_mmap_gtt(struct 
>> drm_i915_gem_object *obj);
>>   
>>   void i915_gem_object_release_mmap_offset(struct drm_i915_gem_object 
>> *obj);
>>   
>> +vm_fault_t i915_error_to_vmf_fault(int err);
>> +
>>   #endif
>> diff --git a/drivers/gpu/drm/i915/i915_perf.c 
>> b/drivers/gpu/drm/i915/i915_perf.c
>> index c91f2da84189..6fd669b520d8 100644
>> --- a/drivers/gpu/drm/i915/i915_perf.c
>> +++ b/drivers/gpu/drm/i915/i915_perf.c
>> @@ -192,10 +192,12 @@
>>    */
>>   
>>   #include <linux/anon_inodes.h>
>> +#include <linux/mman.h>
>>   #include <linux/sizes.h>
>>   #include <linux/uuid.h>
>>   
>>   #include "gem/i915_gem_context.h"
>> +#include "gem/i915_gem_mman.h"
>>   #include "gt/intel_engine_pm.h"
>>   #include "gt/intel_engine_user.h"
>>   #include "gt/intel_gt.h"
>> @@ -3291,6 +3293,44 @@ static long i915_perf_config_locked(struct i915_perf_stream *stream,
>>          return ret;
>>   }
>>   
>> +#define I915_PERF_OA_BUFFER_MMAP_OFFSET 1
>> +
>> +/**
>> + * i915_perf_oa_buffer_info_locked - size and offset of the OA 
>> +buffer
>> + * @stream: i915 perf stream
>> + * @cmd: ioctl command
>> + * @arg: pointer to oa buffer info filled by this function.
>> + */
>> +static int i915_perf_oa_buffer_info_locked(struct i915_perf_stream *stream,
>> +                                          unsigned int cmd,
>> +                                          unsigned long arg) {
>> +       struct drm_i915_perf_oa_buffer_info info;
>> +       void __user *output = (void __user *)arg;
>> +
>> +       if (i915_perf_stream_paranoid && !perfmon_capable()) {
>> +               DRM_DEBUG("Insufficient privileges to access OA buffer info\n");
>> +               return -EACCES;
>> +       }
>> +
>> +       if (_IOC_SIZE(cmd) != sizeof(info))
>> +               return -EINVAL;
>> +
>> +       if (copy_from_user(&info, output, sizeof(info)))
>> +               return -EFAULT;
>> +
>> +       if (info.type || info.flags || info.rsvd)
>> +               return -EINVAL;
>> +
>> +       info.size = stream->oa_buffer.vma->size;
>> +       info.offset = I915_PERF_OA_BUFFER_MMAP_OFFSET * PAGE_SIZE;
>> +
>> +       if (copy_to_user(output, &info, sizeof(info)))
>> +               return -EFAULT;
>> +
>> +       return 0;
>> +}
>> +
>>   /**
>>    * i915_perf_ioctl_locked - support ioctl() usage with i915 perf stream FDs
>>    * @stream: An i915 perf stream
>> @@ -3316,6 +3356,8 @@ static long i915_perf_ioctl_locked(struct i915_perf_stream *stream,
>>                  return 0;
>>          case I915_PERF_IOCTL_CONFIG:
>>                  return i915_perf_config_locked(stream, arg);
>> +       case I915_PERF_IOCTL_GET_OA_BUFFER_INFO:
>> +               return i915_perf_oa_buffer_info_locked(stream, cmd, 
>> + arg);
>>          }
>>   
>>          return -EINVAL;
>> @@ -3387,6 +3429,14 @@ static int i915_perf_release(struct inode *inode, struct file *file)
>>          struct i915_perf_stream *stream = file->private_data;
>>          struct i915_perf *perf = stream->perf;
>>   
>> +       /*
>> +        * User could have multiple vmas from multiple mmaps. We want to zap
>> +        * them all here. Note that a fresh fault cannot occur as the mmap holds
>> +        * a reference to the stream via the vma->vm_file, so before user's
>> +        * munmap, the stream cannot be destroyed.
>> +        */
>> +       unmap_mapping_range(file->f_mapping, 0, -1, 1);
>> +
>>          mutex_lock(&perf->lock);
>>          i915_perf_destroy_locked(stream);
>>          mutex_unlock(&perf->lock);
>> @@ -3397,6 +3447,75 @@ static int i915_perf_release(struct inode *inode, struct file *file)
>>          return 0;
>>   }
>>   
>> +static vm_fault_t vm_fault_oa(struct vm_fault *vmf) {
>> +       struct vm_area_struct *vma = vmf->vma;
>> +       struct i915_perf_stream *stream = vma->vm_private_data;
>> +       int err;
>> +
>> +       err = remap_io_sg(vma,
>> +                         vma->vm_start, vma->vm_end - vma->vm_start,
>> +                         stream->oa_buffer.vma->pages->sgl, -1);
>> +
>> +       return i915_error_to_vmf_fault(err); }
>> +
>> +static const struct vm_operations_struct vm_ops_oa = {
>> +       .fault = vm_fault_oa,
>> +};
>> +
>> +static int i915_perf_mmap(struct file *file, struct vm_area_struct 
>> +*vma) {
>> +       struct i915_perf_stream *stream = file->private_data;
>> +
>> +       /* mmap-ing OA buffer to user space MUST absolutely be privileged */
>> +       if (i915_perf_stream_paranoid && !perfmon_capable()) {
>> +               DRM_DEBUG("Insufficient privileges to map OA buffer\n");
>> +               return -EACCES;
>> +       }
>> +
>> +       switch (vma->vm_pgoff) {
>> +       /*
>> +        * A non-zero offset ensures that we are mapping the right object. Also
>> +        * leaves room for future objects added to this implementation.
>> +        */
>> +       case I915_PERF_OA_BUFFER_MMAP_OFFSET:
>> +               if (!(stream->sample_flags & SAMPLE_OA_REPORT))
>> +                       return -EINVAL;
>> +
>> +               if (vma->vm_end - vma->vm_start > OA_BUFFER_SIZE)
>> +                       return -EINVAL;
>> +
>> +               /*
>> +                * Only support VM_READ. Enforce MAP_PRIVATE by checking for
>> +                * VM_MAYSHARE.
>> +                */
>> +               if (vma->vm_flags & (VM_WRITE | VM_EXEC |
>> +                                    VM_SHARED | VM_MAYSHARE))
>> +                       return -EINVAL;
>> +
>> +               vma->vm_flags &= ~(VM_MAYWRITE | VM_MAYEXEC);
>> +
>> +               /*
>> +                * If the privileged parent forks and child drops root
>> +                * privilege, we do not want the child to retain access to the
>> +                * mapped OA buffer. Explicitly set VM_DONTCOPY to avoid such
>> +                * cases.
>> +                */
>> +               vma->vm_flags |= VM_PFNMAP | VM_DONTEXPAND |
>> +                                VM_DONTDUMP | VM_DONTCOPY;
>> +               break;
>> +
>> +       default:
>> +               return -EINVAL;
>> +       }
>> +
>> +       vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
>> +       vma->vm_private_data = stream;
>> +       vma->vm_ops = &vm_ops_oa;
>> +
>> +       return 0;
>> +}
>>   
>>   static const struct file_operations fops = {
>>          .owner          = THIS_MODULE,
>> @@ -3409,6 +3528,7 @@ static const struct file_operations fops = {
>>           * to handle 32bits compatibility.
>>           */
>>          .compat_ioctl   = i915_perf_ioctl,
>> +       .mmap           = i915_perf_mmap,
>>   };
>>   
>>   
>> @@ -4559,8 +4679,12 @@ int i915_perf_ioctl_version(void)
>>           *
>>           *    - OA buffer head/tail/status/buffer registers for read only
>>           *    - OA counters A18, A19, A20 for read/write
>> +        *
>> +        * 8: Added an option to map oa buffer at umd driver level and trigger
>> +        *    oa reports within oa buffer from command buffer. See
>> +        *    I915_PERF_IOCTL_GET_OA_BUFFER_INFO.
>>           */
>> -       return 7;
>> +       return 8;
>>   }
>>   
>>   #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
>> diff --git a/include/uapi/drm/i915_drm.h 
>> b/include/uapi/drm/i915_drm.h index fa1f3d62f9a6..cc1702ddc859 100644
>> --- a/include/uapi/drm/i915_drm.h
>> +++ b/include/uapi/drm/i915_drm.h
>> @@ -2101,6 +2101,39 @@ struct drm_i915_perf_open_param {
>>    */
>>   #define I915_PERF_IOCTL_CONFIG _IO('i', 0x2)
>>   
>> +/**
>> + * Returns OA buffer properties to be used with mmap.
>> + *
>> + * This ioctl is available in perf revision 8.
>> + */
>> +#define I915_PERF_IOCTL_GET_OA_BUFFER_INFO _IOWR('i', 0x3, struct 
>> +drm_i915_perf_oa_buffer_info)
>> +
>> +/**
>> + * OA buffer size and offset.
>> + *
>> + * OA output buffer
>> + *   type: 0
>> + *   flags: mbz
>> + *
>> + *   After querying the info, pass (size,offset) to mmap(),
>> + *
>> + *   mmap(0, info.size, PROT_READ, MAP_PRIVATE, perf_fd, info.offset).
>> + *
>> + *   Note that only a private (not shared between processes, or across fork())
>> + *   read-only mmapping is allowed.
>> + *
>> + *   Userspace must treat the incoming data as tainted, but it conforms to the OA
>> + *   format as specified by user config. The buffer provides reports that have
>> + *   OA counters - A, B and C.
>> + */
>> +struct drm_i915_perf_oa_buffer_info {
>> +       __u32 type;   /* in */
>> +       __u32 flags;  /* in */
>> +       __u64 size;   /* out */
>> +       __u64 offset; /* out */
>> +       __u64 rsvd;   /* mbz */
>> +};
>> +
>>   /**
>>    * Common to all i915 perf records
>>    */
>> --
>> 2.20.1
>>
>> _______________________________________________
>> Intel-gfx mailing list
>> Intel-gfx@lists.freedesktop.org
>> https://lists.freedesktop.org/mailman/listinfo/intel-gfx


---------------------------------------------------------------------
Intel Technology Poland sp. z o.o.
ul. Sowackiego 173 | 80-298 Gdask | Sd Rejonowy Gdask Pnoc | VII Wydzia Gospodarczy Krajowego Rejestru Sdowego - KRS 101882 | NIP 957-07-52-316 | Kapita zakadowy 200.000 PLN.
Ta wiadomo wraz z zacznikami jest przeznaczona dla okrelonego adresata i moe zawiera informacje poufne. W razie przypadkowego otrzymania tej wiadomoci, prosimy o powiadomienie nadawcy oraz trwae jej usunicie; jakiekolwiek przegldanie lub rozpowszechnianie jest zabronione.
This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). If you are not the intended recipient, please contact the sender and delete all copies; any review or distribution by others is strictly prohibited.
 
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

^ permalink raw reply	[flat|nested] 18+ messages in thread
* Re: [Intel-gfx] [PATCH 2/8] drm/i915/gt: Refactor _wa_add to reuse wa_index and wa_list_grow
  2020-11-17 18:14   ` Umesh Nerlige Ramappa
@ 2021-02-27  6:58     ` Lucas De Marchi
  0 siblings, 0 replies; 18+ messages in thread
From: Lucas De Marchi @ 2021-02-27  6:58 UTC (permalink / raw)
  To: Umesh Nerlige Ramappa; +Cc: intel-gfx, Chris Wilson

On Tue, Nov 17, 2020 at 10:14:00AM -0800, Umesh Nerlige Ramappa wrote:
>This is quite different and much cleane from my original patch :), so 
>you should be the author. With that, this is
>
>Reviewed-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>

any chance to get at least patches 1 and 2 applied?

Lucas De Marchi

>
>Thanks,
>Umesh
>
>On Tue, Nov 17, 2020 at 11:01:26AM +0000, Chris Wilson wrote:
>>From: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
>>
>>Switch the search and grow code of the _wa_add to use _wa_index and
>>_wa_list_grow.
>>
>>Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
>>---
>>drivers/gpu/drm/i915/gt/intel_workarounds.c | 124 +++++++++++---------
>>1 file changed, 71 insertions(+), 53 deletions(-)
>>
>>diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.c b/drivers/gpu/drm/i915/gt/intel_workarounds.c
>>index c49083957074..e50c72d2b3f1 100644
>>--- a/drivers/gpu/drm/i915/gt/intel_workarounds.c
>>+++ b/drivers/gpu/drm/i915/gt/intel_workarounds.c
>>@@ -91,20 +91,19 @@ static void wa_init_start(struct i915_wa_list *wal, const char *name, const char
>>
>>#define WA_LIST_CHUNK (1 << 4)
>>
>>-static void wa_init_finish(struct i915_wa_list *wal)
>>+static void wa_trim(struct i915_wa_list *wal, gfp_t gfp)
>>{
>>+	struct i915_wa *list;
>>+
>>	/* Trim unused entries. */
>>-	if (!IS_ALIGNED(wal->count, WA_LIST_CHUNK)) {
>>-		struct i915_wa *list = kmemdup(wal->list,
>>-					       wal->count * sizeof(*list),
>>-					       GFP_KERNEL);
>>-
>>-		if (list) {
>>-			kfree(wal->list);
>>-			wal->list = list;
>>-		}
>>-	}
>>+	list = krealloc(wal->list, wal->count * sizeof(*list), gfp);
>>+	if (list)
>>+		wal->list = list;
>>+}
>>
>>+static void wa_init_finish(struct i915_wa_list *wal)
>>+{
>>+	wa_trim(wal, GFP_KERNEL);
>>	if (!wal->count)
>>		return;
>>
>>@@ -112,57 +111,60 @@ static void wa_init_finish(struct i915_wa_list *wal)
>>			 wal->wa_count, wal->name, wal->engine_name);
>>}
>>
>>-static void _wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
>>+static int wa_index(struct i915_wa_list *wal, i915_reg_t reg)
>>{
>>-	unsigned int addr = i915_mmio_reg_offset(wa->reg);
>>-	unsigned int start = 0, end = wal->count;
>>-	const unsigned int grow = WA_LIST_CHUNK;
>>-	struct i915_wa *wa_;
>>+	unsigned int addr = i915_mmio_reg_offset(reg);
>>+	int start = 0, end = wal->count;
>>
>>-	GEM_BUG_ON(!is_power_of_2(grow));
>>+	/* addr and wal->list[].reg, both include the R/W flags */
>>+	while (start < end) {
>>+		unsigned int mid = start + (end - start) / 2;
>>
>>-	if (IS_ALIGNED(wal->count, grow)) { /* Either uninitialized or full. */
>>-		struct i915_wa *list;
>>+		if (i915_mmio_reg_offset(wal->list[mid].reg) < addr)
>>+			start = mid + 1;
>>+		else if (i915_mmio_reg_offset(wal->list[mid].reg) > addr)
>>+			end = mid;
>>+		else
>>+			return mid;
>>+	}
>>
>>-		list = kmalloc_array(ALIGN(wal->count + 1, grow), sizeof(*wa),
>>-				     GFP_KERNEL);
>>-		if (!list) {
>>-			DRM_ERROR("No space for workaround init!\n");
>>-			return;
>>-		}
>>+	return -ENOENT;
>>+}
>>
>>-		if (wal->list) {
>>-			memcpy(list, wal->list, sizeof(*wa) * wal->count);
>>-			kfree(wal->list);
>>-		}
>>+static int wa_list_grow(struct i915_wa_list *wal, size_t count, gfp_t gfp)
>>+{
>>+	struct i915_wa *list;
>>
>>-		wal->list = list;
>>-	}
>>+	list = krealloc(wal->list, count * sizeof(*list), gfp);
>>+	if (!list)
>>+		return -ENOMEM;
>>
>>-	while (start < end) {
>>-		unsigned int mid = start + (end - start) / 2;
>>+	wal->list = list;
>>+	return 0;
>>+}
>>
>>-		if (i915_mmio_reg_offset(wal->list[mid].reg) < addr) {
>>-			start = mid + 1;
>>-		} else if (i915_mmio_reg_offset(wal->list[mid].reg) > addr) {
>>-			end = mid;
>>-		} else {
>>-			wa_ = &wal->list[mid];
>>-
>>-			if ((wa->clr | wa_->clr) && !(wa->clr & ~wa_->clr)) {
>>-				DRM_ERROR("Discarding overwritten w/a for reg %04x (clear: %08x, set: %08x)\n",
>>-					  i915_mmio_reg_offset(wa_->reg),
>>-					  wa_->clr, wa_->set);
>>-
>>-				wa_->set &= ~wa->clr;
>>-			}
>>-
>>-			wal->wa_count++;
>>-			wa_->set |= wa->set;
>>-			wa_->clr |= wa->clr;
>>-			wa_->read |= wa->read;
>>-			return;
>>+static void __wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
>>+{
>>+	struct i915_wa *wa_;
>>+	int index;
>>+
>>+	index = wa_index(wal, wa->reg);
>>+	if (index >= 0) {
>>+		wa_ = &wal->list[index];
>>+
>>+		if ((wa->clr | wa_->clr) && !(wa->clr & ~wa_->clr)) {
>>+			DRM_ERROR("Discarding overwritten w/a for reg %04x (clear: %08x, set: %08x)\n",
>>+				  i915_mmio_reg_offset(wa_->reg),
>>+				  wa_->clr, wa_->set);
>>+
>>+			wa_->set &= ~wa->clr;
>>		}
>>+
>>+		wal->wa_count++;
>>+		wa_->set |= wa->set;
>>+		wa_->clr |= wa->clr;
>>+		wa_->read |= wa->read;
>>+		return;
>>	}
>>
>>	wal->wa_count++;
>>@@ -180,6 +182,22 @@ static void _wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
>>	}
>>}
>>
>>+static void _wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
>>+{
>>+	const unsigned int grow = WA_LIST_CHUNK;
>>+
>>+	GEM_BUG_ON(!is_power_of_2(grow));
>>+
>>+	if (IS_ALIGNED(wal->count, grow) && /* Either uninitialized or full. */
>>+	    wa_list_grow(wal, ALIGN(wal->count + 1, grow), GFP_KERNEL)) {
>>+		DRM_ERROR("Unable to store w/a for reg %04x\n",
>>+			  i915_mmio_reg_offset(wa->reg));
>>+		return;
>>+	}
>>+
>>+	__wa_add(wal, wa);
>>+}
>>+
>>static void wa_add(struct i915_wa_list *wal, i915_reg_t reg,
>>		   u32 clear, u32 set, u32 read_mask)
>>{
>>-- 
>>2.20.1
>>
>_______________________________________________
>Intel-gfx mailing list
>Intel-gfx@lists.freedesktop.org
>https://lists.freedesktop.org/mailman/listinfo/intel-gfx
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

^ permalink raw reply	[flat|nested] 18+ messages in thread
* [Intel-gfx] [PATCH 2/8] drm/i915/gt: Refactor _wa_add to reuse wa_index and wa_list_grow
  2021-08-30 19:38 [Intel-gfx] [PATCH 0/8] Enable triggered perf query for Xe_HP Umesh Nerlige Ramappa
@ 2021-08-30 19:38 ` Umesh Nerlige Ramappa
  0 siblings, 0 replies; 18+ messages in thread
From: Umesh Nerlige Ramappa @ 2021-08-30 19:38 UTC (permalink / raw)
  To: intel-gfx, Lionel G Landwerlin, Ashutosh Dixit
  Cc: dri-devel, daniel.vetter, Joonas Lahtinen, jason

From: Chris Wilson <chris.p.wilson@intel.com>

Switch the search and grow code of the _wa_add to use _wa_index and
_wa_list_grow.

Signed-off-by: Chris Wilson <chris.p.wilson@intel.com>
Reviewed-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
---
 drivers/gpu/drm/i915/gt/intel_workarounds.c | 124 +++++++++++---------
 1 file changed, 71 insertions(+), 53 deletions(-)

diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.c b/drivers/gpu/drm/i915/gt/intel_workarounds.c
index 2a8cc0e2d1b1..6928f250cafe 100644
--- a/drivers/gpu/drm/i915/gt/intel_workarounds.c
+++ b/drivers/gpu/drm/i915/gt/intel_workarounds.c
@@ -60,20 +60,19 @@ static void wa_init_start(struct i915_wa_list *wal, const char *name, const char
 
 #define WA_LIST_CHUNK (1 << 4)
 
-static void wa_init_finish(struct i915_wa_list *wal)
+static void wa_trim(struct i915_wa_list *wal, gfp_t gfp)
 {
+	struct i915_wa *list;
+
 	/* Trim unused entries. */
-	if (!IS_ALIGNED(wal->count, WA_LIST_CHUNK)) {
-		struct i915_wa *list = kmemdup(wal->list,
-					       wal->count * sizeof(*list),
-					       GFP_KERNEL);
-
-		if (list) {
-			kfree(wal->list);
-			wal->list = list;
-		}
-	}
+	list = krealloc(wal->list, wal->count * sizeof(*list), gfp);
+	if (list)
+		wal->list = list;
+}
 
+static void wa_init_finish(struct i915_wa_list *wal)
+{
+	wa_trim(wal, GFP_KERNEL);
 	if (!wal->count)
 		return;
 
@@ -81,57 +80,60 @@ static void wa_init_finish(struct i915_wa_list *wal)
 			 wal->wa_count, wal->name, wal->engine_name);
 }
 
-static void _wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
+static int wa_index(struct i915_wa_list *wal, i915_reg_t reg)
 {
-	unsigned int addr = i915_mmio_reg_offset(wa->reg);
-	unsigned int start = 0, end = wal->count;
-	const unsigned int grow = WA_LIST_CHUNK;
-	struct i915_wa *wa_;
+	unsigned int addr = i915_mmio_reg_offset(reg);
+	int start = 0, end = wal->count;
 
-	GEM_BUG_ON(!is_power_of_2(grow));
+	/* addr and wal->list[].reg, both include the R/W flags */
+	while (start < end) {
+		unsigned int mid = start + (end - start) / 2;
 
-	if (IS_ALIGNED(wal->count, grow)) { /* Either uninitialized or full. */
-		struct i915_wa *list;
+		if (i915_mmio_reg_offset(wal->list[mid].reg) < addr)
+			start = mid + 1;
+		else if (i915_mmio_reg_offset(wal->list[mid].reg) > addr)
+			end = mid;
+		else
+			return mid;
+	}
 
-		list = kmalloc_array(ALIGN(wal->count + 1, grow), sizeof(*wa),
-				     GFP_KERNEL);
-		if (!list) {
-			DRM_ERROR("No space for workaround init!\n");
-			return;
-		}
+	return -ENOENT;
+}
 
-		if (wal->list) {
-			memcpy(list, wal->list, sizeof(*wa) * wal->count);
-			kfree(wal->list);
-		}
+static int wa_list_grow(struct i915_wa_list *wal, size_t count, gfp_t gfp)
+{
+	struct i915_wa *list;
 
-		wal->list = list;
-	}
+	list = krealloc(wal->list, count * sizeof(*list), gfp);
+	if (!list)
+		return -ENOMEM;
 
-	while (start < end) {
-		unsigned int mid = start + (end - start) / 2;
+	wal->list = list;
+	return 0;
+}
 
-		if (i915_mmio_reg_offset(wal->list[mid].reg) < addr) {
-			start = mid + 1;
-		} else if (i915_mmio_reg_offset(wal->list[mid].reg) > addr) {
-			end = mid;
-		} else {
-			wa_ = &wal->list[mid];
-
-			if ((wa->clr | wa_->clr) && !(wa->clr & ~wa_->clr)) {
-				DRM_ERROR("Discarding overwritten w/a for reg %04x (clear: %08x, set: %08x)\n",
-					  i915_mmio_reg_offset(wa_->reg),
-					  wa_->clr, wa_->set);
-
-				wa_->set &= ~wa->clr;
-			}
-
-			wal->wa_count++;
-			wa_->set |= wa->set;
-			wa_->clr |= wa->clr;
-			wa_->read |= wa->read;
-			return;
+static void __wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
+{
+	struct i915_wa *wa_;
+	int index;
+
+	index = wa_index(wal, wa->reg);
+	if (index >= 0) {
+		wa_ = &wal->list[index];
+
+		if ((wa->clr | wa_->clr) && !(wa->clr & ~wa_->clr)) {
+			DRM_ERROR("Discarding overwritten w/a for reg %04x (clear: %08x, set: %08x)\n",
+				  i915_mmio_reg_offset(wa_->reg),
+				  wa_->clr, wa_->set);
+
+			wa_->set &= ~wa->clr;
 		}
+
+		wal->wa_count++;
+		wa_->set |= wa->set;
+		wa_->clr |= wa->clr;
+		wa_->read |= wa->read;
+		return;
 	}
 
 	wal->wa_count++;
@@ -149,6 +151,22 @@ static void _wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
 	}
 }
 
+static void _wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
+{
+	const unsigned int grow = WA_LIST_CHUNK;
+
+	GEM_BUG_ON(!is_power_of_2(grow));
+
+	if (IS_ALIGNED(wal->count, grow) && /* Either uninitialized or full. */
+	    wa_list_grow(wal, ALIGN(wal->count + 1, grow), GFP_KERNEL)) {
+		DRM_ERROR("Unable to store w/a for reg %04x\n",
+			  i915_mmio_reg_offset(wa->reg));
+		return;
+	}
+
+	__wa_add(wal, wa);
+}
+
 static void wa_add(struct i915_wa_list *wal, i915_reg_t reg,
 		   u32 clear, u32 set, u32 read_mask, bool masked_reg)
 {
-- 
2.20.1


^ permalink raw reply related	[flat|nested] 18+ messages in thread
* [Intel-gfx] [PATCH 2/8] drm/i915/gt: Refactor _wa_add to reuse wa_index and wa_list_grow
  2021-08-03 20:13 [Intel-gfx] [PATCH 0/8] Enable triggered perf query for Xe_HP Umesh Nerlige Ramappa
@ 2021-08-03 20:13 ` Umesh Nerlige Ramappa
  0 siblings, 0 replies; 18+ messages in thread
From: Umesh Nerlige Ramappa @ 2021-08-03 20:13 UTC (permalink / raw)
  To: intel-gfx, Lionel G Landwerlin, Ashutosh Dixit, daniel; +Cc: dri-devel

From: Chris Wilson <chris.p.wilson@intel.com>

Switch the search and grow code of the _wa_add to use _wa_index and
_wa_list_grow.

Signed-off-by: Chris Wilson <chris.p.wilson@intel.com>
Reviewed-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
---
 drivers/gpu/drm/i915/gt/intel_workarounds.c | 124 +++++++++++---------
 1 file changed, 71 insertions(+), 53 deletions(-)

diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.c b/drivers/gpu/drm/i915/gt/intel_workarounds.c
index 011cde0bc38d..94540cdb90c4 100644
--- a/drivers/gpu/drm/i915/gt/intel_workarounds.c
+++ b/drivers/gpu/drm/i915/gt/intel_workarounds.c
@@ -60,20 +60,19 @@ static void wa_init_start(struct i915_wa_list *wal, const char *name, const char
 
 #define WA_LIST_CHUNK (1 << 4)
 
-static void wa_init_finish(struct i915_wa_list *wal)
+static void wa_trim(struct i915_wa_list *wal, gfp_t gfp)
 {
+	struct i915_wa *list;
+
 	/* Trim unused entries. */
-	if (!IS_ALIGNED(wal->count, WA_LIST_CHUNK)) {
-		struct i915_wa *list = kmemdup(wal->list,
-					       wal->count * sizeof(*list),
-					       GFP_KERNEL);
-
-		if (list) {
-			kfree(wal->list);
-			wal->list = list;
-		}
-	}
+	list = krealloc(wal->list, wal->count * sizeof(*list), gfp);
+	if (list)
+		wal->list = list;
+}
 
+static void wa_init_finish(struct i915_wa_list *wal)
+{
+	wa_trim(wal, GFP_KERNEL);
 	if (!wal->count)
 		return;
 
@@ -81,57 +80,60 @@ static void wa_init_finish(struct i915_wa_list *wal)
 			 wal->wa_count, wal->name, wal->engine_name);
 }
 
-static void _wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
+static int wa_index(struct i915_wa_list *wal, i915_reg_t reg)
 {
-	unsigned int addr = i915_mmio_reg_offset(wa->reg);
-	unsigned int start = 0, end = wal->count;
-	const unsigned int grow = WA_LIST_CHUNK;
-	struct i915_wa *wa_;
+	unsigned int addr = i915_mmio_reg_offset(reg);
+	int start = 0, end = wal->count;
 
-	GEM_BUG_ON(!is_power_of_2(grow));
+	/* addr and wal->list[].reg, both include the R/W flags */
+	while (start < end) {
+		unsigned int mid = start + (end - start) / 2;
 
-	if (IS_ALIGNED(wal->count, grow)) { /* Either uninitialized or full. */
-		struct i915_wa *list;
+		if (i915_mmio_reg_offset(wal->list[mid].reg) < addr)
+			start = mid + 1;
+		else if (i915_mmio_reg_offset(wal->list[mid].reg) > addr)
+			end = mid;
+		else
+			return mid;
+	}
 
-		list = kmalloc_array(ALIGN(wal->count + 1, grow), sizeof(*wa),
-				     GFP_KERNEL);
-		if (!list) {
-			DRM_ERROR("No space for workaround init!\n");
-			return;
-		}
+	return -ENOENT;
+}
 
-		if (wal->list) {
-			memcpy(list, wal->list, sizeof(*wa) * wal->count);
-			kfree(wal->list);
-		}
+static int wa_list_grow(struct i915_wa_list *wal, size_t count, gfp_t gfp)
+{
+	struct i915_wa *list;
 
-		wal->list = list;
-	}
+	list = krealloc(wal->list, count * sizeof(*list), gfp);
+	if (!list)
+		return -ENOMEM;
 
-	while (start < end) {
-		unsigned int mid = start + (end - start) / 2;
+	wal->list = list;
+	return 0;
+}
 
-		if (i915_mmio_reg_offset(wal->list[mid].reg) < addr) {
-			start = mid + 1;
-		} else if (i915_mmio_reg_offset(wal->list[mid].reg) > addr) {
-			end = mid;
-		} else {
-			wa_ = &wal->list[mid];
-
-			if ((wa->clr | wa_->clr) && !(wa->clr & ~wa_->clr)) {
-				DRM_ERROR("Discarding overwritten w/a for reg %04x (clear: %08x, set: %08x)\n",
-					  i915_mmio_reg_offset(wa_->reg),
-					  wa_->clr, wa_->set);
-
-				wa_->set &= ~wa->clr;
-			}
-
-			wal->wa_count++;
-			wa_->set |= wa->set;
-			wa_->clr |= wa->clr;
-			wa_->read |= wa->read;
-			return;
+static void __wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
+{
+	struct i915_wa *wa_;
+	int index;
+
+	index = wa_index(wal, wa->reg);
+	if (index >= 0) {
+		wa_ = &wal->list[index];
+
+		if ((wa->clr | wa_->clr) && !(wa->clr & ~wa_->clr)) {
+			DRM_ERROR("Discarding overwritten w/a for reg %04x (clear: %08x, set: %08x)\n",
+				  i915_mmio_reg_offset(wa_->reg),
+				  wa_->clr, wa_->set);
+
+			wa_->set &= ~wa->clr;
 		}
+
+		wal->wa_count++;
+		wa_->set |= wa->set;
+		wa_->clr |= wa->clr;
+		wa_->read |= wa->read;
+		return;
 	}
 
 	wal->wa_count++;
@@ -149,6 +151,22 @@ static void _wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
 	}
 }
 
+static void _wa_add(struct i915_wa_list *wal, const struct i915_wa *wa)
+{
+	const unsigned int grow = WA_LIST_CHUNK;
+
+	GEM_BUG_ON(!is_power_of_2(grow));
+
+	if (IS_ALIGNED(wal->count, grow) && /* Either uninitialized or full. */
+	    wa_list_grow(wal, ALIGN(wal->count + 1, grow), GFP_KERNEL)) {
+		DRM_ERROR("Unable to store w/a for reg %04x\n",
+			  i915_mmio_reg_offset(wa->reg));
+		return;
+	}
+
+	__wa_add(wal, wa);
+}
+
 static void wa_add(struct i915_wa_list *wal, i915_reg_t reg,
 		   u32 clear, u32 set, u32 read_mask, bool masked_reg)
 {
-- 
2.20.1


^ permalink raw reply related	[flat|nested] 18+ messages in thread
end of thread, other threads:[~2021-08-30 19:38 UTC | newest]

Thread overview: 18+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-11-17 11:01 [Intel-gfx] [PATCH 1/8] drm/i915/gt: Lock intel_engine_apply_whitelist with uncore->lock Chris Wilson
2020-11-17 11:01 ` [Intel-gfx] [PATCH 2/8] drm/i915/gt: Refactor _wa_add to reuse wa_index and wa_list_grow Chris Wilson
2020-11-17 18:14   ` Umesh Nerlige Ramappa
2021-02-27  6:58     ` Lucas De Marchi
2020-11-17 11:01 ` [Intel-gfx] [PATCH 3/8] drm/i915/gt: Check for conflicting RING_NONPRIV Chris Wilson
2020-11-17 18:08   ` Umesh Nerlige Ramappa
2020-11-17 11:01 ` [Intel-gfx] [PATCH 4/8] drm/i915/gt: Enable dynamic adjustment of RING_NONPRIV Chris Wilson
2020-11-17 20:34   ` Umesh Nerlige Ramappa
2020-11-17 11:01 ` [Intel-gfx] [PATCH 5/8] drm/i915/perf: Ensure observation logic is not clock gated Chris Wilson
2020-11-17 11:01 ` [Intel-gfx] [PATCH 6/8] drm/i915/perf: Whitelist OA report trigger registers Chris Wilson
2020-11-17 11:01 ` [Intel-gfx] [PATCH 7/8] drm/i915/perf: Whitelist OA counter and buffer registers Chris Wilson
2020-11-17 11:01 ` [Intel-gfx] [PATCH 8/8] drm/i915/perf: Map OA buffer to user space for gen12 performance query Chris Wilson
2020-11-18 11:41   ` Joonas Lahtinen
2020-11-18 11:57     ` Lionel Landwerlin
2020-11-19 10:56       ` Maciejewski, Piotr
2020-11-17 13:18 ` [Intel-gfx] ✗ Fi.CI.BAT: failure for series starting with [1/8] drm/i915/gt: Lock intel_engine_apply_whitelist with uncore->lock Patchwork
2021-08-03 20:13 [Intel-gfx] [PATCH 0/8] Enable triggered perf query for Xe_HP Umesh Nerlige Ramappa
2021-08-03 20:13 ` [Intel-gfx] [PATCH 2/8] drm/i915/gt: Refactor _wa_add to reuse wa_index and wa_list_grow Umesh Nerlige Ramappa
2021-08-30 19:38 [Intel-gfx] [PATCH 0/8] Enable triggered perf query for Xe_HP Umesh Nerlige Ramappa
2021-08-30 19:38 ` [Intel-gfx] [PATCH 2/8] drm/i915/gt: Refactor _wa_add to reuse wa_index and wa_list_grow Umesh Nerlige Ramappa

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).