@@ -132,6 +132,7 @@ gt-y += \
gt/intel_sseu.o \
gt/intel_sseu_debugfs.o \
gt/intel_timeline.o \
+ gt/intel_tlb.o \
gt/intel_wopcm.o \
gt/intel_workarounds.o \
gt/shmem_utils.o \
@@ -7,7 +7,7 @@
#include <drm/drm_cache.h>
#include "gt/intel_gt.h"
-#include "gt/intel_gt_pm.h"
+#include "gt/intel_tlb.h"
#include "i915_drv.h"
#include "i915_gem_object.h"
@@ -198,7 +198,7 @@ static void flush_tlb_invalidate(struct drm_i915_gem_object *obj)
if (!obj->mm.tlb)
return;
- intel_gt_invalidate_tlb(gt, obj->mm.tlb);
+ intel_gt_invalidate_tlb_full(gt, obj->mm.tlb);
obj->mm.tlb = 0;
}
@@ -33,6 +33,7 @@
#include "intel_rps.h"
#include "intel_sa_media.h"
#include "intel_gt_sysfs.h"
+#include "intel_tlb.h"
#include "intel_uncore.h"
#include "shmem_utils.h"
@@ -50,8 +51,7 @@ void intel_gt_common_init_early(struct intel_gt *gt)
intel_gt_init_reset(gt);
intel_gt_init_requests(gt);
intel_gt_init_timelines(gt);
- mutex_init(>->tlb.invalidate_lock);
- seqcount_mutex_init(>->tlb.seqno, >->tlb.invalidate_lock);
+ intel_gt_init_tlb(gt);
intel_gt_pm_init_early(gt);
intel_wopcm_init_early(>->wopcm);
@@ -846,7 +846,7 @@ void intel_gt_driver_late_release_all(struct drm_i915_private *i915)
intel_gt_fini_requests(gt);
intel_gt_fini_reset(gt);
intel_gt_fini_timelines(gt);
- mutex_destroy(>->tlb.invalidate_lock);
+ intel_gt_fini_tlb(gt);
intel_engines_free(gt);
}
}
@@ -1003,137 +1003,3 @@ void intel_gt_info_print(const struct intel_gt_info *info,
intel_sseu_dump(&info->sseu, p);
}
-
-/*
- * HW architecture suggest typical invalidation time at 40us,
- * with pessimistic cases up to 100us and a recommendation to
- * cap at 1ms. We go a bit higher just in case.
- */
-#define TLB_INVAL_TIMEOUT_US 100
-#define TLB_INVAL_TIMEOUT_MS 4
-
-/*
- * On Xe_HP the TLB invalidation registers are located at the same MMIO offsets
- * but are now considered MCR registers. Since they exist within a GAM range,
- * the primary instance of the register rolls up the status from each unit.
- */
-static int wait_for_invalidate(struct intel_engine_cs *engine)
-{
- if (engine->tlb_inv.mcr)
- return intel_gt_mcr_wait_for_reg(engine->gt,
- engine->tlb_inv.reg.mcr_reg,
- engine->tlb_inv.done,
- 0,
- TLB_INVAL_TIMEOUT_US,
- TLB_INVAL_TIMEOUT_MS);
- else
- return __intel_wait_for_register_fw(engine->gt->uncore,
- engine->tlb_inv.reg.reg,
- engine->tlb_inv.done,
- 0,
- TLB_INVAL_TIMEOUT_US,
- TLB_INVAL_TIMEOUT_MS,
- NULL);
-}
-
-static void mmio_invalidate_full(struct intel_gt *gt)
-{
- struct drm_i915_private *i915 = gt->i915;
- struct intel_uncore *uncore = gt->uncore;
- struct intel_engine_cs *engine;
- intel_engine_mask_t awake, tmp;
- enum intel_engine_id id;
- unsigned long flags;
-
- if (GRAPHICS_VER(i915) < 8)
- return;
-
- intel_uncore_forcewake_get(uncore, FORCEWAKE_ALL);
-
- intel_gt_mcr_lock(gt, &flags);
- spin_lock(&uncore->lock); /* serialise invalidate with GT reset */
-
- awake = 0;
- for_each_engine(engine, gt, id) {
- if (!intel_engine_pm_is_awake(engine))
- continue;
-
- if (engine->tlb_inv.mcr)
- intel_gt_mcr_multicast_write_fw(gt,
- engine->tlb_inv.reg.mcr_reg,
- engine->tlb_inv.request);
- else
- intel_uncore_write_fw(uncore,
- engine->tlb_inv.reg.reg,
- engine->tlb_inv.request);
-
- awake |= engine->mask;
- }
-
- GT_TRACE(gt, "invalidated engines %08x\n", awake);
-
- /* Wa_2207587034:tgl,dg1,rkl,adl-s,adl-p */
- if (awake &&
- (IS_TIGERLAKE(i915) ||
- IS_DG1(i915) ||
- IS_ROCKETLAKE(i915) ||
- IS_ALDERLAKE_S(i915) ||
- IS_ALDERLAKE_P(i915)))
- intel_uncore_write_fw(uncore, GEN12_OA_TLB_INV_CR, 1);
-
- spin_unlock(&uncore->lock);
- intel_gt_mcr_unlock(gt, flags);
-
- for_each_engine_masked(engine, gt, awake, tmp) {
- if (wait_for_invalidate(engine))
- gt_err_ratelimited(gt,
- "%s TLB invalidation did not complete in %ums!\n",
- engine->name, TLB_INVAL_TIMEOUT_MS);
- }
-
- /*
- * Use delayed put since a) we mostly expect a flurry of TLB
- * invalidations so it is good to avoid paying the forcewake cost and
- * b) it works around a bug in Icelake which cannot cope with too rapid
- * transitions.
- */
- intel_uncore_forcewake_put_delayed(uncore, FORCEWAKE_ALL);
-}
-
-static bool tlb_seqno_passed(const struct intel_gt *gt, u32 seqno)
-{
- u32 cur = intel_gt_tlb_seqno(gt);
-
- /* Only skip if a *full* TLB invalidate barrier has passed */
- return (s32)(cur - ALIGN(seqno, 2)) > 0;
-}
-
-void intel_gt_invalidate_tlb(struct intel_gt *gt, u32 seqno)
-{
- intel_wakeref_t wakeref;
-
- if (I915_SELFTEST_ONLY(gt->awake == -ENODEV))
- return;
-
- if (intel_gt_is_wedged(gt))
- return;
-
- if (tlb_seqno_passed(gt, seqno))
- return;
-
- with_intel_gt_pm_if_awake(gt, wakeref) {
- mutex_lock(>->tlb.invalidate_lock);
- if (tlb_seqno_passed(gt, seqno))
- goto unlock;
-
- mmio_invalidate_full(gt);
-
- write_seqcount_invalidate(>->tlb.seqno);
-unlock:
- mutex_unlock(>->tlb.invalidate_lock);
- }
-}
-
-#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
-#include "selftest_tlb.c"
-#endif
@@ -107,16 +107,4 @@ void intel_gt_info_print(const struct intel_gt_info *info,
void intel_gt_watchdog_work(struct work_struct *work);
-static inline u32 intel_gt_tlb_seqno(const struct intel_gt *gt)
-{
- return seqprop_sequence(>->tlb.seqno);
-}
-
-static inline u32 intel_gt_next_invalidate_tlb_full(const struct intel_gt *gt)
-{
- return intel_gt_tlb_seqno(gt) | 1;
-}
-
-void intel_gt_invalidate_tlb(struct intel_gt *gt, u32 seqno);
-
#endif /* __INTEL_GT_H__ */
new file mode 100644
@@ -0,0 +1,159 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright © 2023 Intel Corporation
+ */
+
+#include "i915_drv.h"
+#include "i915_perf_oa_regs.h"
+#include "intel_engine_pm.h"
+#include "intel_gt.h"
+#include "intel_gt_mcr.h"
+#include "intel_gt_pm.h"
+#include "intel_gt_print.h"
+#include "intel_gt_regs.h"
+#include "intel_tlb.h"
+
+/*
+ * HW architecture suggest typical invalidation time at 40us,
+ * with pessimistic cases up to 100us and a recommendation to
+ * cap at 1ms. We go a bit higher just in case.
+ */
+#define TLB_INVAL_TIMEOUT_US 100
+#define TLB_INVAL_TIMEOUT_MS 4
+
+/*
+ * On Xe_HP the TLB invalidation registers are located at the same MMIO offsets
+ * but are now considered MCR registers. Since they exist within a GAM range,
+ * the primary instance of the register rolls up the status from each unit.
+ */
+static int wait_for_invalidate(struct intel_engine_cs *engine)
+{
+ if (engine->tlb_inv.mcr)
+ return intel_gt_mcr_wait_for_reg(engine->gt,
+ engine->tlb_inv.reg.mcr_reg,
+ engine->tlb_inv.done,
+ 0,
+ TLB_INVAL_TIMEOUT_US,
+ TLB_INVAL_TIMEOUT_MS);
+ else
+ return __intel_wait_for_register_fw(engine->gt->uncore,
+ engine->tlb_inv.reg.reg,
+ engine->tlb_inv.done,
+ 0,
+ TLB_INVAL_TIMEOUT_US,
+ TLB_INVAL_TIMEOUT_MS,
+ NULL);
+}
+
+static void mmio_invalidate_full(struct intel_gt *gt)
+{
+ struct drm_i915_private *i915 = gt->i915;
+ struct intel_uncore *uncore = gt->uncore;
+ struct intel_engine_cs *engine;
+ intel_engine_mask_t awake, tmp;
+ enum intel_engine_id id;
+ unsigned long flags;
+
+ if (GRAPHICS_VER(i915) < 8)
+ return;
+
+ intel_uncore_forcewake_get(uncore, FORCEWAKE_ALL);
+
+ intel_gt_mcr_lock(gt, &flags);
+ spin_lock(&uncore->lock); /* serialise invalidate with GT reset */
+
+ awake = 0;
+ for_each_engine(engine, gt, id) {
+ if (!intel_engine_pm_is_awake(engine))
+ continue;
+
+ if (engine->tlb_inv.mcr)
+ intel_gt_mcr_multicast_write_fw(gt,
+ engine->tlb_inv.reg.mcr_reg,
+ engine->tlb_inv.request);
+ else
+ intel_uncore_write_fw(uncore,
+ engine->tlb_inv.reg.reg,
+ engine->tlb_inv.request);
+
+ awake |= engine->mask;
+ }
+
+ GT_TRACE(gt, "invalidated engines %08x\n", awake);
+
+ /* Wa_2207587034:tgl,dg1,rkl,adl-s,adl-p */
+ if (awake &&
+ (IS_TIGERLAKE(i915) ||
+ IS_DG1(i915) ||
+ IS_ROCKETLAKE(i915) ||
+ IS_ALDERLAKE_S(i915) ||
+ IS_ALDERLAKE_P(i915)))
+ intel_uncore_write_fw(uncore, GEN12_OA_TLB_INV_CR, 1);
+
+ spin_unlock(&uncore->lock);
+ intel_gt_mcr_unlock(gt, flags);
+
+ for_each_engine_masked(engine, gt, awake, tmp) {
+ if (wait_for_invalidate(engine))
+ gt_err_ratelimited(gt,
+ "%s TLB invalidation did not complete in %ums!\n",
+ engine->name, TLB_INVAL_TIMEOUT_MS);
+ }
+
+ /*
+ * Use delayed put since a) we mostly expect a flurry of TLB
+ * invalidations so it is good to avoid paying the forcewake cost and
+ * b) it works around a bug in Icelake which cannot cope with too rapid
+ * transitions.
+ */
+ intel_uncore_forcewake_put_delayed(uncore, FORCEWAKE_ALL);
+}
+
+static bool tlb_seqno_passed(const struct intel_gt *gt, u32 seqno)
+{
+ u32 cur = intel_gt_tlb_seqno(gt);
+
+ /* Only skip if a *full* TLB invalidate barrier has passed */
+ return (s32)(cur - ALIGN(seqno, 2)) > 0;
+}
+
+void intel_gt_invalidate_tlb_full(struct intel_gt *gt, u32 seqno)
+{
+ intel_wakeref_t wakeref;
+
+ if (I915_SELFTEST_ONLY(gt->awake == -ENODEV))
+ return;
+
+ if (intel_gt_is_wedged(gt))
+ return;
+
+ if (tlb_seqno_passed(gt, seqno))
+ return;
+
+ with_intel_gt_pm_if_awake(gt, wakeref) {
+ mutex_lock(>->tlb.invalidate_lock);
+ if (tlb_seqno_passed(gt, seqno))
+ goto unlock;
+
+ mmio_invalidate_full(gt);
+
+ write_seqcount_invalidate(>->tlb.seqno);
+unlock:
+ mutex_unlock(>->tlb.invalidate_lock);
+ }
+}
+
+void intel_gt_init_tlb(struct intel_gt *gt)
+{
+ mutex_init(>->tlb.invalidate_lock);
+ seqcount_mutex_init(>->tlb.seqno, >->tlb.invalidate_lock);
+}
+
+void intel_gt_fini_tlb(struct intel_gt *gt)
+{
+ mutex_destroy(>->tlb.invalidate_lock);
+}
+
+#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
+#include "selftest_tlb.c"
+#endif
new file mode 100644
@@ -0,0 +1,29 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright © 2023 Intel Corporation
+ */
+
+#ifndef INTEL_TLB_H
+#define INTEL_TLB_H
+
+#include <linux/seqlock.h>
+#include <linux/types.h>
+
+#include "intel_gt_types.h"
+
+void intel_gt_invalidate_tlb_full(struct intel_gt *gt, u32 seqno);
+
+void intel_gt_init_tlb(struct intel_gt *gt);
+void intel_gt_fini_tlb(struct intel_gt *gt);
+
+static inline u32 intel_gt_tlb_seqno(const struct intel_gt *gt)
+{
+ return seqprop_sequence(>->tlb.seqno);
+}
+
+static inline u32 intel_gt_next_invalidate_tlb_full(const struct intel_gt *gt)
+{
+ return intel_gt_tlb_seqno(gt) | 1;
+}
+
+#endif /* INTEL_TLB_H */
@@ -6,6 +6,7 @@
#include "i915_selftest.h"
#include "gem/i915_gem_internal.h"
+#include "gem/i915_gem_lmem.h"
#include "gem/i915_gem_region.h"
#include "gen8_engine_cs.h"
@@ -354,7 +355,7 @@ mem_tlbinv(struct intel_gt *gt,
static void tlbinv_full(struct i915_address_space *vm, u64 addr, u64 length)
{
- intel_gt_invalidate_tlb(vm->gt, intel_gt_tlb_seqno(vm->gt) | 1);
+ intel_gt_invalidate_tlb_full(vm->gt, intel_gt_tlb_seqno(vm->gt) | 1);
}
static int invalidate_full(void *arg)
@@ -34,6 +34,7 @@
#include "gt/intel_engine_heartbeat.h"
#include "gt/intel_gt.h"
#include "gt/intel_gt_requests.h"
+#include "gt/intel_tlb.h"
#include "i915_drv.h"
#include "i915_gem_evict.h"