linux-nvgpu/drivers/gpu/nvgpu/hal/rc/rc_gv11b.c

/*
 * Copyright (c) 2015-2021, NVIDIA CORPORATION.  All rights reserved.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in
 * all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
 * DEALINGS IN THE SOFTWARE.
 */

#include <nvgpu/bug.h>
#include <nvgpu/gmmu.h>
#include <nvgpu/soc.h>
#include <nvgpu/debug.h>
#include <nvgpu/mm.h>
#include <nvgpu/log2.h>
#include <nvgpu/io.h>
#include <nvgpu/utils.h>
#include <nvgpu/fifo.h>
#include <nvgpu/engines.h>
#include <nvgpu/rc.h>
#include <nvgpu/runlist.h>
#include <nvgpu/gk20a.h>
#include <nvgpu/channel.h>
#include <nvgpu/nvgpu_err.h>
#include <nvgpu/swprofile.h>
#include <nvgpu/fifo/swprofile.h>
#include <nvgpu/power_features/power_features.h>
#include <nvgpu/gr/fecs_trace.h>
#include <nvgpu/preempt.h>
#ifdef CONFIG_NVGPU_LS_PMU
#include <nvgpu/pmu/mutex.h>
#endif
#include <nvgpu/nvgpu_init.h>

#include "rc_gv11b.h"

#include <nvgpu/hw/gv11b/hw_fifo_gv11b.h>

static void gv11b_fifo_locked_abort_runlist_active_tsgs(struct gk20a *g,
			unsigned int rc_type,
			u32 runlists_mask)
{
	struct nvgpu_fifo *f = &g->fifo;
	struct nvgpu_tsg *tsg = NULL;
	unsigned long tsgid;
	struct nvgpu_runlist *runlist = NULL;
#ifdef CONFIG_NVGPU_LS_PMU
	u32 token = PMU_INVALID_MUTEX_OWNER_ID;
	int mutex_ret = 0;
#endif
	int err;
	u32 i;

	nvgpu_err(g, "abort active tsgs of runlists set in "
			"runlists_mask: 0x%08x", runlists_mask);
#ifdef CONFIG_NVGPU_LS_PMU
	/* runlist_lock  are locked by teardown */
	mutex_ret = nvgpu_pmu_lock_acquire(g, g->pmu,
			PMU_MUTEX_ID_FIFO, &token);
#endif
	for (i = 0U; i < f->num_runlists; i++) {
		runlist = &f->active_runlists[i];

		if ((runlists_mask & BIT32(runlist->id)) == 0U) {
			continue;
		}
		nvgpu_log(g, gpu_dbg_info, "abort runlist id %d",
				runlist->id);

		for_each_set_bit(tsgid, runlist->active_tsgs, f->num_channels) {
			tsg = &g->fifo.tsg[tsgid];

			if (!tsg->abortable) {
				nvgpu_log(g, gpu_dbg_info,
					  "tsg %lu is not abortable, skipping",
					  tsgid);
				continue;
			}
			nvgpu_log(g, gpu_dbg_info, "abort tsg id %lu", tsgid);

			g->ops.tsg.disable(tsg);

			nvgpu_tsg_reset_faulted_eng_pbdma(g, tsg, true, true);

#ifdef CONFIG_NVGPU_FECS_TRACE
			nvgpu_gr_fecs_trace_add_tsg_reset(g, tsg);
#endif
#ifdef CONFIG_NVGPU_DEBUGGER
			if (!g->fifo.deferred_reset_pending) {
#endif
				if (rc_type == RC_TYPE_MMU_FAULT) {
					nvgpu_tsg_set_ctx_mmu_error(g, tsg);
					/*
					 * Mark error (returned verbose flag is
					 * ignored since it is not needed here)
					 */
					(void) nvgpu_tsg_mark_error(g, tsg);
				}
#ifdef CONFIG_NVGPU_DEBUGGER
			}
#endif

			/*
			 * remove all entries from this runlist; don't wait for
			 * the update to finish on hw.
			 */
			err = nvgpu_runlist_update_locked(g,
				runlist->id, NULL, false, false);
			if (err != 0) {
				nvgpu_err(g, "runlist id %d is not cleaned up",
					runlist->id);
			}

			nvgpu_tsg_abort(g, tsg, false);

			nvgpu_log(g, gpu_dbg_info, "aborted tsg id %lu", tsgid);
		}
	}
#ifdef CONFIG_NVGPU_LS_PMU
	if (mutex_ret == 0) {
		err = nvgpu_pmu_lock_release(g, g->pmu, PMU_MUTEX_ID_FIFO,
				&token);
		if (err != 0) {
			nvgpu_err(g, "PMU_MUTEX_ID_FIFO not released err=%d",
					err);
		}
	}
#endif
}

void gv11b_fifo_recover(struct gk20a *g, u32 act_eng_bitmask,
			u32 id, unsigned int id_type, unsigned int rc_type,
			struct mmu_fault_info *mmufault)
{
	struct nvgpu_tsg *tsg = NULL;
	u32 runlists_mask, i;
	unsigned long bit;
	unsigned long bitmask;
	u32 pbdma_bitmask = 0U;
	struct nvgpu_runlist *runlist = NULL;
	u32 engine_id;
	struct nvgpu_fifo *f = &g->fifo;
	struct nvgpu_swprofiler *prof = &f->recovery_profiler;
#ifdef CONFIG_NVGPU_DEBUGGER
	u32 client_type = ~U32(0U);
	bool deferred_reset_pending = false;
#endif

	rec_dbg(g, "Recovery starting");
	rec_dbg(g, "  ID      = %u", id);
	rec_dbg(g, "  id_type = %s", nvgpu_id_type_to_str(id_type));
	rec_dbg(g, "  rc_type = %s", nvgpu_rc_type_to_str(rc_type));
	rec_dbg(g, "  Engine bitmask: 0x%x", act_eng_bitmask);

	nvgpu_swprofile_begin_sample(prof);

	rec_dbg(g, "Acquiring engines_reset_mutex");
	nvgpu_mutex_acquire(&f->engines_reset_mutex);

	/* acquire runlist_lock for num_runlists */
	rec_dbg(g, "Acquiring runlist_lock for active runlists");
	nvgpu_runlist_lock_active_runlists(g);

	nvgpu_swprofile_snapshot(prof, PROF_RECOVERY_ACQ_ACTIVE_RL);

	g->ops.fifo.intr_set_recover_mask(g);

	/* get tsg */
	if (id != INVAL_ID && id_type == ID_TYPE_TSG) {
		struct nvgpu_channel *c;
		tsg = &g->fifo.tsg[id];
		rec_dbg(g, "Channels bound to this TSG:");

		i = 0U;
		nvgpu_list_for_each_entry(c, &tsg->ch_list,
					  nvgpu_channel, ch_entry) {
			rec_dbg(g, " %2u | chid %u", i++, c->chid);
		}
	}

	/* Set unserviceable flag right at start of recovery to reduce
	 * the window of race between job submit and recovery on same
	 * TSG.
	 * The unserviceable flag is checked during job submit and
	 * prevent new jobs from being submitted to TSG which is headed
	 * for teardown.
	 */
	if (tsg != NULL) {
		/* Set error notifier before letting userspace
		 * know about faulty channel
		 * The unserviceable flag is moved early to
		 * disallow submits on the broken channel. If
		 * userspace checks the notifier code when a
		 * submit fails, we need it set to convey to
		 * userspace that channel is no longer usable.
		 */
		if (rc_type == RC_TYPE_MMU_FAULT) {
		/*
		 * If a debugger is attached and debugging is enabled,
		 * then do not set error notifier as it will cause the
		 * application to teardown the channels and debugger will
		 * not be able to collect any data.
		 */
#ifdef CONFIG_NVGPU_DEBUGGER
			if (!nvgpu_engine_should_defer_reset(g,
						mmufault->faulted_engine,
						mmufault->client_type, false)) {
#endif
				nvgpu_tsg_set_ctx_mmu_error(g, tsg);
#ifdef CONFIG_NVGPU_DEBUGGER
			}
#endif
		}
		nvgpu_tsg_set_unserviceable(g, tsg);
	}

	if (rc_type == RC_TYPE_MMU_FAULT && mmufault != NULL) {
		if (mmufault->faulted_pbdma != INVAL_ID) {
			pbdma_bitmask = BIT32(mmufault->faulted_pbdma);
		}
	}

	rec_dbg(g, "PBDMA   Bitmask: 0x%x", pbdma_bitmask);

	/* get runlists mask */
	runlists_mask = nvgpu_runlist_get_runlists_mask(g, id, id_type,
				act_eng_bitmask, pbdma_bitmask);

	rec_dbg(g, "Runlist Bitmask: 0x%x", runlists_mask);

	nvgpu_swprofile_snapshot(prof, PROF_RECOVERY_GET_RL_MASK);

	/*
	 * release runlist lock for the runlists that are not
	 * being recovered
	 */
	nvgpu_runlist_unlock_runlists(g, ~runlists_mask);


	/* Disable runlist scheduler */
	rec_dbg(g, "Disabling RL scheduler now");
	nvgpu_runlist_set_state(g, runlists_mask, RUNLIST_DISABLED);

	nvgpu_swprofile_snapshot(prof, PROF_RECOVERY_DISABLE_RL);

#ifdef CONFIG_NVGPU_NON_FUSA
	rec_dbg(g, "Disabling CG/PG now");
	if (nvgpu_cg_pg_disable(g) != 0) {
		nvgpu_warn(g, "fail to disable power mgmt");
	}
#endif

	if (rc_type == RC_TYPE_MMU_FAULT) {
		if (!nvgpu_swprofile_is_enabled(prof)) {
			gk20a_debug_dump(g);
		}
#ifdef CONFIG_NVGPU_DEBUGGER
		client_type = mmufault->client_type;
#endif
		rec_dbg(g, "Clearing PBDMA_FAULTED, ENG_FAULTED in CCSR register");
		nvgpu_tsg_reset_faulted_eng_pbdma(g, tsg, true, true);
	}

	if (tsg != NULL) {
		rec_dbg(g, "Disabling TSG");
		g->ops.tsg.disable(tsg);
	}

	nvgpu_swprofile_snapshot(prof, PROF_RECOVERY_DISABLE_TSG);

	/*
	 * Even though TSG preempt timed out, the RC sequence would by design
	 * require s/w to issue another preempt.
	 * If recovery includes an ENGINE_RESET, to not have race conditions,
	 * use RUNLIST_PREEMPT to kick all work off, and cancel any context
	 * load which may be pending. This is also needed to make sure
	 * that all PBDMAs serving the engine are not loaded when engine is
	 * reset.
	 */
	rec_dbg(g, "Preempting runlists for RC");
	nvgpu_fifo_preempt_runlists_for_rc(g, runlists_mask);

	nvgpu_swprofile_snapshot(prof, PROF_RECOVERY_PREEMPT_RL);

	/*
	 * For each PBDMA which serves the runlist, poll to verify the TSG is no
	 * longer on the PBDMA and the engine phase of the preempt has started.
	 */
	rec_dbg(g, "Polling for TSG to be off PBDMA");
	if (tsg != NULL && (nvgpu_preempt_poll_tsg_on_pbdma(g, tsg) != 0)) {
		nvgpu_err(g, "TSG preemption on PBDMA failed; "
			"PBDMA seems stuck; cannot recover stuck PBDMA.");
		/* Trigger Quiesce as recovery failed on hung PBDMA. */
		nvgpu_sw_quiesce(g);
		return;
	}
	rec_dbg(g, "  Done!");

	nvgpu_swprofile_snapshot(prof, PROF_RECOVERY_POLL_TSG_ON_PBDMA);

#ifdef CONFIG_NVGPU_DEBUGGER
	nvgpu_mutex_acquire(&f->deferred_reset_mutex);
	g->fifo.deferred_reset_pending = false;
	nvgpu_mutex_release(&f->deferred_reset_mutex);
#endif

	rec_dbg(g, "Resetting relevant engines");
	/* check if engine reset should be deferred */
	for (i = 0U; i < f->num_runlists; i++) {
		runlist = &f->active_runlists[i];

		if (((runlists_mask & BIT32(runlist->id)) == 0U) ||
		    (runlist->reset_eng_bitmask == 0U)) {
			continue;
		}

		bitmask = runlist->reset_eng_bitmask;
		rec_dbg(g, "  Engine bitmask for RL %u: 0x%lx",
		       runlist->id, bitmask);

		for_each_set_bit(bit, &bitmask, f->max_engines) {

			engine_id = U32(bit);
			rec_dbg(g, "  > Restting engine: ID=%u", engine_id);

#ifdef CONFIG_NVGPU_DEBUGGER
			if ((tsg != NULL) &&
			    nvgpu_engine_should_defer_reset(g, engine_id,
							    client_type, false)) {
				rec_dbg(g, "    (deferred)");

				f->deferred_fault_engines |= BIT64(engine_id);

				/* handled during channel free */
				nvgpu_mutex_acquire(&f->deferred_reset_mutex);
				f->deferred_reset_pending = true;
				nvgpu_mutex_release(&f->deferred_reset_mutex);

				deferred_reset_pending = true;

				nvgpu_log(g, gpu_dbg_intr | gpu_dbg_gpu_dbg,
					"sm debugger attached, deferring "
					"channel recovery to channel free");
			} else {
#endif
#ifdef CONFIG_NVGPU_ENGINE_RESET
				nvgpu_engine_reset(g, engine_id);
				rec_dbg(g, "    Done!");
#endif
#ifdef CONFIG_NVGPU_DEBUGGER
			}
#endif
		}
	}

	nvgpu_swprofile_snapshot(prof, PROF_RECOVERY_ENGINES_RESET);

#ifdef CONFIG_NVGPU_FECS_TRACE
	if (tsg != NULL)
		nvgpu_gr_fecs_trace_add_tsg_reset(g, tsg);
#endif
	if (tsg != NULL) {
#ifdef CONFIG_NVGPU_DEBUGGER
		if (deferred_reset_pending) {
			g->ops.tsg.disable(tsg);
		} else {
#endif
			nvgpu_tsg_wakeup_wqs(g, tsg);
			nvgpu_tsg_abort(g, tsg, false);
#ifdef CONFIG_NVGPU_DEBUGGER
		}
#endif
	} else {
		gv11b_fifo_locked_abort_runlist_active_tsgs(g, rc_type,
			runlists_mask);
	}

	rec_dbg(g, "Re-enabling runlists");
	nvgpu_runlist_set_state(g, runlists_mask, RUNLIST_ENABLED);

	nvgpu_swprofile_snapshot(prof, PROF_RECOVERY_ENABLE_RL);

#ifdef CONFIG_NVGPU_NON_FUSA
	rec_dbg(g, "Re-enabling CG/PG");
	if (nvgpu_cg_pg_enable(g) != 0) {
		nvgpu_warn(g, "fail to enable power mgmt");
	}
#endif

	g->ops.fifo.intr_unset_recover_mask(g);

	/* release runlist_lock for the recovered runlists */
	nvgpu_runlist_unlock_runlists(g, runlists_mask);

	rec_dbg(g, "Releasing engines reset mutex");
	nvgpu_mutex_release(&f->engines_reset_mutex);

	nvgpu_swprofile_snapshot(prof, PROF_RECOVERY_DONE);
}