On Thu, Jun 10, 2021 at 02:44:13PM -0700, Rob Clark wrote:
> From: Rob Clark <robdclark@xxxxxxxxxxxx>
>
> Wire up support to stall the SMMU on iova fault, and collect a devcore-
> dump snapshot for easier debugging of faults.
>
> Currently this is a6xx-only, but mostly only because so far it is the
> only one using adreno-smmu-priv.
Acked-by: Jordan Crouse <jordan@xxxxxxxxxxxxxxxxx>
> Signed-off-by: Rob Clark <robdclark@xxxxxxxxxxxx>
> ---
> drivers/gpu/drm/msm/adreno/a5xx_gpu.c | 19 +++++++-
> drivers/gpu/drm/msm/adreno/a6xx_gpu.c | 38 +++++++++++++++-
> drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c | 42 ++++++++++++++----
> drivers/gpu/drm/msm/adreno/adreno_gpu.c | 15 +++++++
> drivers/gpu/drm/msm/msm_gem.h | 1 +
> drivers/gpu/drm/msm/msm_gem_submit.c | 1 +
> drivers/gpu/drm/msm/msm_gpu.c | 48 +++++++++++++++++++++
> drivers/gpu/drm/msm/msm_gpu.h | 17 ++++++++
> drivers/gpu/drm/msm/msm_gpummu.c | 5 +++
> drivers/gpu/drm/msm/msm_iommu.c | 11 +++++
> drivers/gpu/drm/msm/msm_mmu.h | 1 +
> 11 files changed, 186 insertions(+), 12 deletions(-)
>
> diff --git a/drivers/gpu/drm/msm/adreno/a5xx_gpu.c b/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
> index eb030b00bff4..7a271de9a212 100644
> --- a/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
> +++ b/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
> @@ -1200,6 +1200,15 @@ static void a5xx_fault_detect_irq(struct msm_gpu *gpu)
> struct drm_device *dev = gpu->dev;
> struct msm_ringbuffer *ring = gpu->funcs->active_ring(gpu);
>
> + /*
> + * If stalled on SMMU fault, we could trip the GPU's hang detection,
> + * but the fault handler will trigger the devcore dump, and we want
> + * to otherwise resume normally rather than killing the submit, so
> + * just bail.
> + */
> + if (gpu_read(gpu, REG_A5XX_RBBM_STATUS3) & BIT(24))
> + return;
> +
> DRM_DEV_ERROR(dev->dev, "gpu fault ring %d fence %x status %8.8X rb %4.4x/%4.4x ib1 %16.16llX/%4.4x ib2 %16.16llX/%4.4x\n",
> ring ? ring->id : -1, ring ? ring->seqno : 0,
> gpu_read(gpu, REG_A5XX_RBBM_STATUS),
> @@ -1523,6 +1532,7 @@ static struct msm_gpu_state *a5xx_gpu_state_get(struct msm_gpu *gpu)
> {
> struct a5xx_gpu_state *a5xx_state = kzalloc(sizeof(*a5xx_state),
> GFP_KERNEL);
> + bool stalled = !!(gpu_read(gpu, REG_A5XX_RBBM_STATUS3) & BIT(24));
>
> if (!a5xx_state)
> return ERR_PTR(-ENOMEM);
> @@ -1535,8 +1545,13 @@ static struct msm_gpu_state *a5xx_gpu_state_get(struct msm_gpu *gpu)
>
> a5xx_state->base.rbbm_status = gpu_read(gpu, REG_A5XX_RBBM_STATUS);
>
> - /* Get the HLSQ regs with the help of the crashdumper */
> - a5xx_gpu_state_get_hlsq_regs(gpu, a5xx_state);
> + /*
> + * Get the HLSQ regs with the help of the crashdumper, but only if
> + * we are not stalled in an iommu fault (in which case the crashdumper
> + * would not have access to memory)
> + */
> + if (!stalled)
> + a5xx_gpu_state_get_hlsq_regs(gpu, a5xx_state);
>
> a5xx_set_hwcg(gpu, true);
>
> diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
> index fc19db10bff1..c3699408bd1f 100644
> --- a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
> +++ b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
> @@ -1081,6 +1081,16 @@ static int a6xx_fault_handler(void *arg, unsigned long iova, int flags, void *da
> struct msm_gpu *gpu = arg;
> struct adreno_smmu_fault_info *info = data;
> const char *type = "UNKNOWN";
> + const char *block;
> + bool do_devcoredump = info && !READ_ONCE(gpu->crashstate);
> +
> + /*
> + * If we aren't going to be resuming later from fault_worker, then do
> + * it now.
> + */
> + if (!do_devcoredump) {
> + gpu->aspace->mmu->funcs->resume_translation(gpu->aspace->mmu);
> + }
>
> /*
> * Print a default message if we couldn't get the data from the
> @@ -1104,15 +1114,30 @@ static int a6xx_fault_handler(void *arg, unsigned long iova, int flags, void *da
> else if (info->fsr & ARM_SMMU_FSR_EF)
> type = "EXTERNAL";
>
> + block = a6xx_fault_block(gpu, info->fsynr1 & 0xff);
> +
> pr_warn_ratelimited("*** gpu fault: ttbr0=%.16llx iova=%.16lx dir=%s type=%s source=%s (%u,%u,%u,%u)\n",
> info->ttbr0, iova,
> - flags & IOMMU_FAULT_WRITE ? "WRITE" : "READ", type,
> - a6xx_fault_block(gpu, info->fsynr1 & 0xff),
> + flags & IOMMU_FAULT_WRITE ? "WRITE" : "READ",
> + type, block,
> gpu_read(gpu, REG_A6XX_CP_SCRATCH_REG(4)),
> gpu_read(gpu, REG_A6XX_CP_SCRATCH_REG(5)),
> gpu_read(gpu, REG_A6XX_CP_SCRATCH_REG(6)),
> gpu_read(gpu, REG_A6XX_CP_SCRATCH_REG(7)));
>
> + if (do_devcoredump) {
> + /* Turn off the hangcheck timer to keep it from bothering us */
> + del_timer(&gpu->hangcheck_timer);
> +
> + gpu->fault_info.ttbr0 = info->ttbr0;
> + gpu->fault_info.iova = iova;
> + gpu->fault_info.flags = flags;
> + gpu->fault_info.type = type;
> + gpu->fault_info.block = block;
> +
> + kthread_queue_work(gpu->worker, &gpu->fault_work);
> + }
> +
> return 0;
> }
>
> @@ -1164,6 +1189,15 @@ static void a6xx_fault_detect_irq(struct msm_gpu *gpu)
> struct a6xx_gpu *a6xx_gpu = to_a6xx_gpu(adreno_gpu);
> struct msm_ringbuffer *ring = gpu->funcs->active_ring(gpu);
>
> + /*
> + * If stalled on SMMU fault, we could trip the GPU's hang detection,
> + * but the fault handler will trigger the devcore dump, and we want
> + * to otherwise resume normally rather than killing the submit, so
> + * just bail.
> + */
> + if (gpu_read(gpu, REG_A6XX_RBBM_STATUS3) & A6XX_RBBM_STATUS3_SMMU_STALLED_ON_FAULT)
> + return;
> +
> /*
> * Force the GPU to stay on until after we finish
> * collecting information
> diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c b/drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c
> index 21c49c5b4519..ad4ea0ed5d99 100644
> --- a/drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c
> +++ b/drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c
> @@ -832,6 +832,20 @@ static void a6xx_get_registers(struct msm_gpu *gpu,
> a6xx_get_ahb_gpu_registers(gpu,
> a6xx_state, &a6xx_vbif_reglist,
> &a6xx_state->registers[index++]);
> + if (!dumper) {
> + /*
> + * We can't use the crashdumper when the SMMU is stalled,
> + * because the GPU has no memory access until we resume
> + * translation (but we don't want to do that until after
> + * we have captured as much useful GPU state as possible).
> + * So instead collect registers via the CPU:
> + */
> + for (i = 0; i < ARRAY_SIZE(a6xx_reglist); i++)
> + a6xx_get_ahb_gpu_registers(gpu,
> + a6xx_state, &a6xx_reglist[i],
> + &a6xx_state->registers[index++]);
> + return;
> + }
>
> for (i = 0; i < ARRAY_SIZE(a6xx_reglist); i++)
> a6xx_get_crashdumper_registers(gpu,
> @@ -905,11 +919,13 @@ static void a6xx_get_indexed_registers(struct msm_gpu *gpu,
>
> struct msm_gpu_state *a6xx_gpu_state_get(struct msm_gpu *gpu)
> {
> - struct a6xx_crashdumper dumper = { 0 };
> + struct a6xx_crashdumper _dumper = { 0 }, *dumper = NULL;
> struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
> struct a6xx_gpu *a6xx_gpu = to_a6xx_gpu(adreno_gpu);
> struct a6xx_gpu_state *a6xx_state = kzalloc(sizeof(*a6xx_state),
> GFP_KERNEL);
> + bool stalled = !!(gpu_read(gpu, REG_A6XX_RBBM_STATUS3) &
> + A6XX_RBBM_STATUS3_SMMU_STALLED_ON_FAULT);
>
> if (!a6xx_state)
> return ERR_PTR(-ENOMEM);
> @@ -928,14 +944,24 @@ struct msm_gpu_state *a6xx_gpu_state_get(struct msm_gpu *gpu)
> /* Get the banks of indexed registers */
> a6xx_get_indexed_registers(gpu, a6xx_state);
>
> - /* Try to initialize the crashdumper */
> - if (!a6xx_crashdumper_init(gpu, &dumper)) {
> - a6xx_get_registers(gpu, a6xx_state, &dumper);
> - a6xx_get_shaders(gpu, a6xx_state, &dumper);
> - a6xx_get_clusters(gpu, a6xx_state, &dumper);
> - a6xx_get_dbgahb_clusters(gpu, a6xx_state, &dumper);
> + /*
> + * Try to initialize the crashdumper, if we are not dumping state
> + * with the SMMU stalled. The crashdumper needs memory access to
> + * write out GPU state, so we need to skip this when the SMMU is
> + * stalled in response to an iova fault
> + */
> + if (!stalled && !a6xx_crashdumper_init(gpu, &_dumper)) {
> + dumper = &_dumper;
> + }
> +
> + a6xx_get_registers(gpu, a6xx_state, dumper);
> +
> + if (dumper) {
> + a6xx_get_shaders(gpu, a6xx_state, dumper);
> + a6xx_get_clusters(gpu, a6xx_state, dumper);
> + a6xx_get_dbgahb_clusters(gpu, a6xx_state, dumper);
>
> - msm_gem_kernel_put(dumper.bo, gpu->aspace, true);
> + msm_gem_kernel_put(dumper->bo, gpu->aspace, true);
> }
>
> if (snapshot_debugbus)
> diff --git a/drivers/gpu/drm/msm/adreno/adreno_gpu.c b/drivers/gpu/drm/msm/adreno/adreno_gpu.c
> index c1b02f790804..2bfe014995c7 100644
> --- a/drivers/gpu/drm/msm/adreno/adreno_gpu.c
> +++ b/drivers/gpu/drm/msm/adreno/adreno_gpu.c
> @@ -684,6 +684,21 @@ void adreno_show(struct msm_gpu *gpu, struct msm_gpu_state *state,
> adreno_gpu->info->revn, adreno_gpu->rev.core,
> adreno_gpu->rev.major, adreno_gpu->rev.minor,
> adreno_gpu->rev.patchid);
> + /*
> + * If this is state collected due to iova fault, so fault related info
> + *
> + * TTBR0 would not be zero, so this is a good way to distinguish
> + */
> + if (state->fault_info.ttbr0) {
> + const struct msm_gpu_fault_info *info = &state->fault_info;
> +
> + drm_puts(p, "fault-info:\n");
> + drm_printf(p, " - ttbr0=%.16llx\n", info->ttbr0);
> + drm_printf(p, " - iova=%.16lx\n", info->iova);
> + drm_printf(p, " - dir=%s\n", info->flags & IOMMU_FAULT_WRITE ? "WRITE" : "READ");
> + drm_printf(p, " - type=%s\n", info->type);
> + drm_printf(p, " - source=%s\n", info->block);
> + }
>
> drm_printf(p, "rbbm-status: 0x%08x\n", state->rbbm_status);
>
> diff --git a/drivers/gpu/drm/msm/msm_gem.h b/drivers/gpu/drm/msm/msm_gem.h
> index 03e2cc2a2ce1..405f8411e395 100644
> --- a/drivers/gpu/drm/msm/msm_gem.h
> +++ b/drivers/gpu/drm/msm/msm_gem.h
> @@ -328,6 +328,7 @@ struct msm_gem_submit {
> struct dma_fence *fence;
> struct msm_gpu_submitqueue *queue;
> struct pid *pid; /* submitting process */
> + bool fault_dumped; /* Limit devcoredump dumping to one per submit */
> bool valid; /* true if no cmdstream patching needed */
> bool in_rb; /* "sudo" mode, copy cmds into RB */
> struct msm_ringbuffer *ring;
> diff --git a/drivers/gpu/drm/msm/msm_gem_submit.c b/drivers/gpu/drm/msm/msm_gem_submit.c
> index 5480852bdeda..44f84bfd0c0e 100644
> --- a/drivers/gpu/drm/msm/msm_gem_submit.c
> +++ b/drivers/gpu/drm/msm/msm_gem_submit.c
> @@ -50,6 +50,7 @@ static struct msm_gem_submit *submit_create(struct drm_device *dev,
> submit->cmd = (void *)&submit->bos[nr_bos];
> submit->queue = queue;
> submit->ring = gpu->rb[queue->prio];
> + submit->fault_dumped = false;
>
> /* initially, until copy_from_user() and bo lookup succeeds: */
> submit->nr_bos = 0;
> diff --git a/drivers/gpu/drm/msm/msm_gpu.c b/drivers/gpu/drm/msm/msm_gpu.c
> index fa7691cb4614..414ba2dd34e5 100644
> --- a/drivers/gpu/drm/msm/msm_gpu.c
> +++ b/drivers/gpu/drm/msm/msm_gpu.c
> @@ -400,6 +400,7 @@ static void msm_gpu_crashstate_capture(struct msm_gpu *gpu,
> /* Fill in the additional crash state information */
> state->comm = kstrdup(comm, GFP_KERNEL);
> state->cmd = kstrdup(cmd, GFP_KERNEL);
> + state->fault_info = gpu->fault_info;
>
> if (submit) {
> int i, nr = 0;
> @@ -572,6 +573,52 @@ static void recover_worker(struct kthread_work *work)
> msm_gpu_retire(gpu);
> }
>
> +static void fault_worker(struct kthread_work *work)
> +{
> + struct msm_gpu *gpu = container_of(work, struct msm_gpu, fault_work);
> + struct drm_device *dev = gpu->dev;
> + struct msm_gem_submit *submit;
> + struct msm_ringbuffer *cur_ring = gpu->funcs->active_ring(gpu);
> + char *comm = NULL, *cmd = NULL;
> +
> + mutex_lock(&dev->struct_mutex);
> +
> + submit = find_submit(cur_ring, cur_ring->memptrs->fence + 1);
> + if (submit && submit->fault_dumped)
> + goto resume_smmu;
> +
> + if (submit) {
> + struct task_struct *task;
> +
> + task = get_pid_task(submit->pid, PIDTYPE_PID);
> + if (task) {
> + comm = kstrdup(task->comm, GFP_KERNEL);
> + cmd = kstrdup_quotable_cmdline(task, GFP_KERNEL);
> + put_task_struct(task);
> + }
> +
> + /*
> + * When we get GPU iova faults, we can get 1000s of them,
> + * but we really only want to log the first one.
> + */
> + submit->fault_dumped = true;
> + }
> +
> + /* Record the crash state */
> + pm_runtime_get_sync(&gpu->pdev->dev);
> + msm_gpu_crashstate_capture(gpu, submit, comm, cmd);
> + pm_runtime_put_sync(&gpu->pdev->dev);
> +
> + kfree(cmd);
> + kfree(comm);
> +
> +resume_smmu:
> + memset(&gpu->fault_info, 0, sizeof(gpu->fault_info));
> + gpu->aspace->mmu->funcs->resume_translation(gpu->aspace->mmu);
> +
> + mutex_unlock(&dev->struct_mutex);
> +}
> +
> static void hangcheck_timer_reset(struct msm_gpu *gpu)
> {
> mod_timer(&gpu->hangcheck_timer,
> @@ -948,6 +995,7 @@ int msm_gpu_init(struct drm_device *drm, struct platform_device *pdev,
> INIT_LIST_HEAD(&gpu->active_list);
> kthread_init_work(&gpu->retire_work, retire_worker);
> kthread_init_work(&gpu->recover_work, recover_worker);
> + kthread_init_work(&gpu->fault_work, fault_worker);
>
> timer_setup(&gpu->hangcheck_timer, hangcheck_handler, 0);
>
> diff --git a/drivers/gpu/drm/msm/msm_gpu.h b/drivers/gpu/drm/msm/msm_gpu.h
> index 7a082a12d98f..8eefb3aeca10 100644
> --- a/drivers/gpu/drm/msm/msm_gpu.h
> +++ b/drivers/gpu/drm/msm/msm_gpu.h
> @@ -71,6 +71,15 @@ struct msm_gpu_funcs {
> uint32_t (*get_rptr)(struct msm_gpu *gpu, struct msm_ringbuffer *ring);
> };
>
> +/* Additional state for iommu faults: */
> +struct msm_gpu_fault_info {
> + u64 ttbr0;
> + unsigned long iova;
> + int flags;
> + const char *type;
> + const char *block;
> +};
> +
> struct msm_gpu {
> const char *name;
> struct drm_device *dev;
> @@ -135,6 +144,12 @@ struct msm_gpu {
> #define DRM_MSM_HANGCHECK_JIFFIES msecs_to_jiffies(DRM_MSM_HANGCHECK_PERIOD)
> struct timer_list hangcheck_timer;
>
> + /* Fault info for most recent iova fault: */
> + struct msm_gpu_fault_info fault_info;
> +
> + /* work for handling GPU ioval faults: */
> + struct kthread_work fault_work;
> +
> /* work for handling GPU recovery: */
> struct kthread_work recover_work;
>
> @@ -243,6 +258,8 @@ struct msm_gpu_state {
> char *comm;
> char *cmd;
>
> + struct msm_gpu_fault_info fault_info;
> +
> int nr_bos;
> struct msm_gpu_state_bo *bos;
> };
> diff --git a/drivers/gpu/drm/msm/msm_gpummu.c b/drivers/gpu/drm/msm/msm_gpummu.c
> index 379496186c7f..f7d1945e0c9f 100644
> --- a/drivers/gpu/drm/msm/msm_gpummu.c
> +++ b/drivers/gpu/drm/msm/msm_gpummu.c
> @@ -68,6 +68,10 @@ static int msm_gpummu_unmap(struct msm_mmu *mmu, uint64_t iova, size_t len)
> return 0;
> }
>
> +static void msm_gpummu_resume_translation(struct msm_mmu *mmu)
> +{
> +}
> +
> static void msm_gpummu_destroy(struct msm_mmu *mmu)
> {
> struct msm_gpummu *gpummu = to_msm_gpummu(mmu);
> @@ -83,6 +87,7 @@ static const struct msm_mmu_funcs funcs = {
> .map = msm_gpummu_map,
> .unmap = msm_gpummu_unmap,
> .destroy = msm_gpummu_destroy,
> + .resume_translation = msm_gpummu_resume_translation,
> };
>
> struct msm_mmu *msm_gpummu_new(struct device *dev, struct msm_gpu *gpu)
> diff --git a/drivers/gpu/drm/msm/msm_iommu.c b/drivers/gpu/drm/msm/msm_iommu.c
> index 6975b95c3c29..eed2a762e9dd 100644
> --- a/drivers/gpu/drm/msm/msm_iommu.c
> +++ b/drivers/gpu/drm/msm/msm_iommu.c
> @@ -184,6 +184,9 @@ struct msm_mmu *msm_iommu_pagetable_create(struct msm_mmu *parent)
> * the arm-smmu driver as a trigger to set up TTBR0
> */
> if (atomic_inc_return(&iommu->pagetables) == 1) {
> + /* Enable stall on iommu fault: */
> + adreno_smmu->set_stall(adreno_smmu->cookie, true);
> +
> ret = adreno_smmu->set_ttbr0_cfg(adreno_smmu->cookie, &ttbr0_cfg);
> if (ret) {
> free_io_pgtable_ops(pagetable->pgtbl_ops);
> @@ -226,6 +229,13 @@ static int msm_fault_handler(struct iommu_domain *domain, struct device *dev,
> return 0;
> }
>
> +static void msm_iommu_resume_translation(struct msm_mmu *mmu)
> +{
> + struct adreno_smmu_priv *adreno_smmu = dev_get_drvdata(mmu->dev);
> +
> + adreno_smmu->resume_translation(adreno_smmu->cookie, true);
> +}
> +
> static void msm_iommu_detach(struct msm_mmu *mmu)
> {
> struct msm_iommu *iommu = to_msm_iommu(mmu);
> @@ -273,6 +283,7 @@ static const struct msm_mmu_funcs funcs = {
> .map = msm_iommu_map,
> .unmap = msm_iommu_unmap,
> .destroy = msm_iommu_destroy,
> + .resume_translation = msm_iommu_resume_translation,
> };
>
> struct msm_mmu *msm_iommu_new(struct device *dev, struct iommu_domain *domain)
> diff --git a/drivers/gpu/drm/msm/msm_mmu.h b/drivers/gpu/drm/msm/msm_mmu.h
> index a88f44c3268d..de158e1bf765 100644
> --- a/drivers/gpu/drm/msm/msm_mmu.h
> +++ b/drivers/gpu/drm/msm/msm_mmu.h
> @@ -15,6 +15,7 @@ struct msm_mmu_funcs {
> size_t len, int prot);
> int (*unmap)(struct msm_mmu *mmu, uint64_t iova, size_t len);
> void (*destroy)(struct msm_mmu *mmu);
> + void (*resume_translation)(struct msm_mmu *mmu);
> };
>
> enum msm_mmu_type {
> --
> 2.31.1
>
