diff options
Diffstat (limited to 'drivers/gpu/drm/i915/gt/intel_reset.c')
| -rw-r--r-- | drivers/gpu/drm/i915/gt/intel_reset.c | 189 |
1 files changed, 138 insertions, 51 deletions
diff --git a/drivers/gpu/drm/i915/gt/intel_reset.c b/drivers/gpu/drm/i915/gt/intel_reset.c index cc6bd21a3e51..41b5036dc538 100644 --- a/drivers/gpu/drm/i915/gt/intel_reset.c +++ b/drivers/gpu/drm/i915/gt/intel_reset.c @@ -9,30 +9,28 @@ #include "display/intel_display_reset.h" #include "display/intel_overlay.h" - #include "gem/i915_gem_context.h" - #include "gt/intel_gt_regs.h" - #include "gt/uc/intel_gsc_fw.h" +#include "uc/intel_guc.h" #include "i915_drv.h" #include "i915_file_private.h" #include "i915_gpu_error.h" #include "i915_irq.h" #include "i915_reg.h" +#include "i915_wait_util.h" #include "intel_breadcrumbs.h" #include "intel_engine_pm.h" #include "intel_engine_regs.h" #include "intel_gt.h" #include "intel_gt_pm.h" +#include "intel_gt_print.h" #include "intel_gt_requests.h" #include "intel_mchbar_regs.h" #include "intel_pci_config.h" #include "intel_reset.h" -#include "uc/intel_guc.h" - #define RESET_MAX_RETRIES 3 static void client_mark_guilty(struct i915_gem_context *ctx, bool banned) @@ -161,16 +159,16 @@ static int i915_do_reset(struct intel_gt *gt, struct pci_dev *pdev = to_pci_dev(gt->i915->drm.dev); int err; - /* Assert reset for at least 20 usec, and wait for acknowledgement. */ + /* Assert reset for at least 50 usec, and wait for acknowledgement. */ pci_write_config_byte(pdev, I915_GDRST, GRDOM_RESET_ENABLE); udelay(50); - err = wait_for_atomic(i915_in_reset(pdev), 50); + err = _wait_for_atomic(i915_in_reset(pdev), 50000, 0); /* Clear the reset request. */ pci_write_config_byte(pdev, I915_GDRST, 0); udelay(50); if (!err) - err = wait_for_atomic(!i915_in_reset(pdev), 50); + err = _wait_for_atomic(!i915_in_reset(pdev), 50000, 0); return err; } @@ -190,7 +188,7 @@ static int g33_do_reset(struct intel_gt *gt, struct pci_dev *pdev = to_pci_dev(gt->i915->drm.dev); pci_write_config_byte(pdev, I915_GDRST, GRDOM_RESET_ENABLE); - return wait_for_atomic(g4x_reset_complete(pdev), 50); + return _wait_for_atomic(g4x_reset_complete(pdev), 50000, 0); } static int g4x_do_reset(struct intel_gt *gt, @@ -207,7 +205,7 @@ static int g4x_do_reset(struct intel_gt *gt, pci_write_config_byte(pdev, I915_GDRST, GRDOM_MEDIA | GRDOM_RESET_ENABLE); - ret = wait_for_atomic(g4x_reset_complete(pdev), 50); + ret = _wait_for_atomic(g4x_reset_complete(pdev), 50000, 0); if (ret) { GT_TRACE(gt, "Wait for media reset failed\n"); goto out; @@ -215,7 +213,7 @@ static int g4x_do_reset(struct intel_gt *gt, pci_write_config_byte(pdev, I915_GDRST, GRDOM_RENDER | GRDOM_RESET_ENABLE); - ret = wait_for_atomic(g4x_reset_complete(pdev), 50); + ret = _wait_for_atomic(g4x_reset_complete(pdev), 50000, 0); if (ret) { GT_TRACE(gt, "Wait for render reset failed\n"); goto out; @@ -592,10 +590,10 @@ static int gen8_engine_reset_prepare(struct intel_engine_cs *engine) ret = __intel_wait_for_register_fw(uncore, reg, mask, ack, 700, 0, NULL); if (ret) - drm_err(&engine->i915->drm, - "%s reset request timed out: {request: %08x, RESET_CTL: %08x}\n", - engine->name, request, - intel_uncore_read_fw(uncore, reg)); + gt_err(engine->gt, + "%s reset request timed out: {request: %08x, RESET_CTL: %08x}\n", + engine->name, request, + intel_uncore_read_fw(uncore, reg)); return ret; } @@ -705,7 +703,7 @@ static int __reset_guc(struct intel_gt *gt) static bool needs_wa_14015076503(struct intel_gt *gt, intel_engine_mask_t engine_mask) { - if (!IS_METEORLAKE(gt->i915) || !HAS_ENGINE(gt, GSC0)) + if (MEDIA_VER_FULL(gt->i915) != IP_VER(13, 0) || !HAS_ENGINE(gt, GSC0)) return false; if (!__HAS_ENGINE(engine_mask, GSC0)) @@ -763,7 +761,7 @@ wa_14015076503_end(struct intel_gt *gt, intel_engine_mask_t engine_mask) HECI_H_GS1_ER_PREP, 0); } -int __intel_gt_reset(struct intel_gt *gt, intel_engine_mask_t engine_mask) +static int __intel_gt_reset(struct intel_gt *gt, intel_engine_mask_t engine_mask) { const int retries = engine_mask == ALL_ENGINES ? RESET_MAX_RETRIES : 1; reset_func reset; @@ -785,9 +783,7 @@ int __intel_gt_reset(struct intel_gt *gt, intel_engine_mask_t engine_mask) reset_mask = wa_14015076503_start(gt, engine_mask, !retry); GT_TRACE(gt, "engine_mask=%x\n", reset_mask); - preempt_disable(); ret = reset(gt, reset_mask, retry); - preempt_enable(); wa_14015076503_end(gt, reset_mask); } @@ -880,8 +876,17 @@ static intel_engine_mask_t reset_prepare(struct intel_gt *gt) intel_engine_mask_t awake = 0; enum intel_engine_id id; - /* For GuC mode, ensure submission is disabled before stopping ring */ - intel_uc_reset_prepare(>->uc); + /** + * For GuC mode with submission enabled, ensure submission + * is disabled before stopping ring. + * + * For GuC mode with submission disabled, ensure that GuC is not + * sanitized, do that after engine reset. reset_prepare() + * is followed by engine reset which in this mode requires GuC to + * process any CSB FIFO entries generated by the resets. + */ + if (intel_uc_uses_guc_submission(>->uc)) + intel_uc_reset_prepare(>->uc); for_each_engine(engine, gt, id) { if (intel_engine_pm_get_if_awake(engine)) @@ -978,8 +983,8 @@ static void __intel_gt_set_wedged(struct intel_gt *gt) awake = reset_prepare(gt); /* Even if the GPU reset fails, it should still stop the engines */ - if (!INTEL_INFO(gt->i915)->gpu_reset_clobbers_display) - __intel_gt_reset(gt, ALL_ENGINES); + if (!intel_gt_gpu_reset_clobbers_display(gt)) + intel_gt_reset_all_engines(gt); for_each_engine(engine, gt, id) engine->submit_request = nop_submit_request; @@ -1005,6 +1010,15 @@ static void __intel_gt_set_wedged(struct intel_gt *gt) GT_TRACE(gt, "end\n"); } +static void set_wedged_work(struct work_struct *w) +{ + struct intel_gt *gt = container_of(w, struct intel_gt, wedge); + intel_wakeref_t wf; + + with_intel_runtime_pm(gt->uncore->rpm, wf) + __intel_gt_set_wedged(gt); +} + void intel_gt_set_wedged(struct intel_gt *gt) { intel_wakeref_t wakeref; @@ -1016,7 +1030,8 @@ void intel_gt_set_wedged(struct intel_gt *gt) mutex_lock(>->reset.mutex); if (GEM_SHOW_DEBUG()) { - struct drm_printer p = drm_debug_printer(__func__); + struct drm_printer p = drm_dbg_printer(>->i915->drm, + DRM_UT_DRIVER, NULL); struct intel_engine_cs *engine; enum intel_engine_id id; @@ -1080,7 +1095,7 @@ static bool __intel_gt_unset_wedged(struct intel_gt *gt) dma_fence_default_wait(fence, false, MAX_SCHEDULE_TIMEOUT); dma_fence_put(fence); - /* Restart iteration after droping lock */ + /* Restart iteration after dropping lock */ spin_lock(&timelines->lock); tl = list_entry(&timelines->active_list, typeof(*tl), link); } @@ -1088,8 +1103,8 @@ static bool __intel_gt_unset_wedged(struct intel_gt *gt) /* We must reset pending GPU events before restoring our submission */ ok = !HAS_EXECLISTS(gt->i915); /* XXX better agnosticism desired */ - if (!INTEL_INFO(gt->i915)->gpu_reset_clobbers_display) - ok = __intel_gt_reset(gt, ALL_ENGINES) == 0; + if (!intel_gt_gpu_reset_clobbers_display(gt)) + ok = intel_gt_reset_all_engines(gt) == 0; if (!ok) { /* * Warn CI about the unrecoverable wedged condition. @@ -1133,10 +1148,10 @@ static int do_reset(struct intel_gt *gt, intel_engine_mask_t stalled_mask) { int err, i; - err = __intel_gt_reset(gt, ALL_ENGINES); + err = intel_gt_reset_all_engines(gt); for (i = 0; err && i < RESET_MAX_RETRIES; i++) { msleep(10 * (i + 1)); - err = __intel_gt_reset(gt, ALL_ENGINES); + err = intel_gt_reset_all_engines(gt); } if (err) return err; @@ -1159,6 +1174,13 @@ static int resume(struct intel_gt *gt) return 0; } +bool intel_gt_gpu_reset_clobbers_display(struct intel_gt *gt) +{ + struct drm_i915_private *i915 = gt->i915; + + return INTEL_INFO(i915)->gpu_reset_clobbers_display; +} + /** * intel_gt_reset - reset chip after a hang * @gt: #intel_gt to reset @@ -1180,6 +1202,7 @@ void intel_gt_reset(struct intel_gt *gt, intel_engine_mask_t stalled_mask, const char *reason) { + struct intel_display *display = gt->i915->display; intel_engine_mask_t awake; int ret; @@ -1201,33 +1224,35 @@ void intel_gt_reset(struct intel_gt *gt, goto unlock; if (reason) - drm_notice(>->i915->drm, - "Resetting chip for %s\n", reason); + gt_notice(gt, "Resetting chip for %s\n", reason); atomic_inc(>->i915->gpu_error.reset_count); awake = reset_prepare(gt); if (!intel_has_gpu_reset(gt)) { if (gt->i915->params.reset) - drm_err(>->i915->drm, "GPU reset not supported\n"); + gt_err(gt, "GPU reset not supported\n"); else - drm_dbg(>->i915->drm, "GPU reset disabled\n"); + gt_dbg(gt, "GPU reset disabled\n"); goto error; } - if (INTEL_INFO(gt->i915)->gpu_reset_clobbers_display) - intel_runtime_pm_disable_interrupts(gt->i915); + if (intel_gt_gpu_reset_clobbers_display(gt)) + intel_irq_suspend(gt->i915); if (do_reset(gt, stalled_mask)) { - drm_err(>->i915->drm, "Failed to reset chip\n"); + gt_err(gt, "Failed to reset chip\n"); goto taint; } - if (INTEL_INFO(gt->i915)->gpu_reset_clobbers_display) - intel_runtime_pm_enable_interrupts(gt->i915); + if (intel_gt_gpu_reset_clobbers_display(gt)) + intel_irq_resume(gt->i915); - intel_overlay_reset(gt->i915); + intel_overlay_reset(display); + /* sanitize uC after engine reset */ + if (!intel_uc_uses_guc_submission(>->uc)) + intel_uc_reset_prepare(>->uc); /* * Next we need to restore the context, but we don't use those * yet either... @@ -1238,9 +1263,7 @@ void intel_gt_reset(struct intel_gt *gt, */ ret = intel_gt_init_hw(gt); if (ret) { - drm_err(>->i915->drm, - "Failed to initialise HW following reset (%d)\n", - ret); + gt_err(gt, "Failed to initialise HW following reset (%d)\n", ret); goto taint; } @@ -1273,7 +1296,30 @@ error: goto finish; } -static int intel_gt_reset_engine(struct intel_engine_cs *engine) +/** + * intel_gt_reset_all_engines() - Reset all engines in the given gt. + * @gt: the GT to reset all engines for. + * + * This function resets all engines within the given gt. + * + * Returns: + * Zero on success, negative error code on failure. + */ +int intel_gt_reset_all_engines(struct intel_gt *gt) +{ + return __intel_gt_reset(gt, ALL_ENGINES); +} + +/** + * intel_gt_reset_engine() - Reset a specific engine within a gt. + * @engine: engine to be reset. + * + * This function resets the specified engine within a gt. + * + * Returns: + * Zero on success, negative error code on failure. + */ +int intel_gt_reset_engine(struct intel_engine_cs *engine) { return __intel_gt_reset(engine->gt, engine->mask); } @@ -1297,7 +1343,7 @@ int __intel_engine_reset_bh(struct intel_engine_cs *engine, const char *msg) if (msg) drm_notice(&engine->i915->drm, "Resetting %s for %s\n", engine->name, msg); - atomic_inc(&engine->i915->gpu_error.reset_engine_count[engine->uabi_class]); + i915_increase_reset_engine_count(&engine->i915->gpu_error, engine); ret = intel_gt_reset_engine(engine); if (ret) { @@ -1351,6 +1397,11 @@ int intel_engine_reset(struct intel_engine_cs *engine, const char *msg) return err; } +static void display_reset_modeset_stuck(void *gt) +{ + intel_gt_set_wedged(gt); +} + static void intel_gt_reset_global(struct intel_gt *gt, u32 engine_mask, const char *reason) @@ -1368,15 +1419,34 @@ static void intel_gt_reset_global(struct intel_gt *gt, /* Use a watchdog to ensure that our reset completes */ intel_wedge_on_timeout(&w, gt, 60 * HZ) { - intel_display_reset_prepare(gt->i915); + struct drm_i915_private *i915 = gt->i915; + struct intel_display *display = i915->display; + bool need_display_reset; + bool reset_display; + + need_display_reset = intel_gt_gpu_reset_clobbers_display(gt) && + intel_has_gpu_reset(gt); + + reset_display = intel_display_reset_test(display) || + need_display_reset; + + if (reset_display) + reset_display = intel_display_reset_prepare(display, + display_reset_modeset_stuck, + gt); intel_gt_reset(gt, engine_mask, reason); - intel_display_reset_finish(gt->i915); + if (reset_display) + intel_display_reset_finish(display, !need_display_reset); } if (!test_bit(I915_WEDGED, >->reset.flags)) kobject_uevent_env(kobj, KOBJ_CHANGE, reset_done_event); + else + drm_dev_wedged_event(>->i915->drm, + DRM_WEDGE_RECOVERY_REBIND | DRM_WEDGE_RECOVERY_BUS_RESET, + NULL); } /** @@ -1437,7 +1507,7 @@ void intel_gt_handle_error(struct intel_gt *gt, intel_has_reset_engine(gt) && !intel_gt_is_wedged(gt)) { local_bh_disable(); for_each_engine_masked(engine, gt, engine_mask, tmp) { - BUILD_BUG_ON(I915_RESET_MODESET >= I915_RESET_ENGINE); + BUILD_BUG_ON(I915_RESET_BACKOFF >= I915_RESET_ENGINE); if (test_and_set_bit(I915_RESET_ENGINE + engine->id, >->reset.flags)) continue; @@ -1582,6 +1652,7 @@ void intel_gt_init_reset(struct intel_gt *gt) init_waitqueue_head(>->reset.queue); mutex_init(>->reset.mutex); init_srcu_struct(>->reset.backoff_srcu); + INIT_WORK(>->wedge, set_wedged_work); /* * While undesirable to wait inside the shrinker, complain anyway. @@ -1607,10 +1678,8 @@ static void intel_wedge_me(struct work_struct *work) { struct intel_wedge_me *w = container_of(work, typeof(*w), work.work); - drm_err(&w->gt->i915->drm, - "%s timed out, cancelling all in-flight rendering.\n", - w->name); - intel_gt_set_wedged(w->gt); + gt_err(w->gt, "%s timed out, cancelling all in-flight rendering.\n", w->name); + set_wedged_work(&w->gt->wedge); } void __intel_init_wedge(struct intel_wedge_me *w, @@ -1632,6 +1701,24 @@ void __intel_fini_wedge(struct intel_wedge_me *w) w->gt = NULL; } +/* + * Wa_22011802037 requires that we (or the GuC) ensure that no command + * streamers are executing MI_FORCE_WAKE while an engine reset is initiated. + */ +bool intel_engine_reset_needs_wa_22011802037(struct intel_gt *gt) +{ + if (GRAPHICS_VER(gt->i915) < 11) + return false; + + if (IS_GFX_GT_IP_STEP(gt, IP_VER(12, 70), STEP_A0, STEP_B0)) + return true; + + if (GRAPHICS_VER_FULL(gt->i915) >= IP_VER(12, 70)) + return false; + + return true; +} + #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) #include "selftest_reset.c" #include "selftest_hangcheck.c" |