Re: [PATCH v7 13/15] drm/tegra: Implement job submission part of new UAPI

[Jon Hunter <jonathanh@nvidia.com>]

On 10/06/2021 12:04, Mikko Perttunen wrote:
> Implement the job submission IOCTL with a minimum feature set.
> 
> Signed-off-by: Mikko Perttunen <mperttunen@nvidia.com>
> ---
> v7:
> * Allocate gather BO with DMA API to get page-aligned
>   memory
> * Add error prints to a few places where they were missing
> v6:
> * Remove sgt bypass path in gather_bo - this would cause
>   cache maintenance to be skipped and is unnecessary in
>   general.
> * Changes related to moving to using syncpoint IDs
> * Add syncobj related code
> * Print warning on submit failure describing the issue
> * Use ARCH_DMA_ADDR_T_64BIT to check if that is indeed
>   the case
> * Add support for relative syncpoint wait
> * Use pm_runtime_resume_and_get
> * Only try to resume engines that support runtime PM
> * Removed uapi subdirectory
> * Don't use "copy_err" variables for copy_from_user
>   return value
> * Fix setting of blocklinear flag
> v5:
> * Add 16K size limit to copies from userspace.
> * Guard RELOC_BLOCKLINEAR flag handling to only exist in ARM64
>   to prevent oversized shift on 32-bit platforms.
> v4:
> * Remove all features that are not strictly necessary.
> * Split into two patches.
> v3:
> * Remove WRITE_RELOC. Relocations are now patched implicitly
>   when patching is needed.
> * Directly call PM runtime APIs on devices instead of using
>   power_on/power_off callbacks.
> * Remove incorrect mutex unlock in tegra_drm_ioctl_channel_open
> * Use XA_FLAGS_ALLOC1 instead of XA_FLAGS_ALLOC
> * Accommodate for removal of timeout field and inlining of
>   syncpt_incrs array.
> * Copy entire user arrays at a time instead of going through
>   elements one-by-one.
> * Implement waiting of DMA reservations.
> * Split out gather_bo implementation into a separate file.
> * Fix length parameter passed to sg_init_one in gather_bo
> * Cosmetic cleanup.
> ---
>  drivers/gpu/drm/tegra/Makefile    |   2 +
>  drivers/gpu/drm/tegra/drm.c       |   4 +-
>  drivers/gpu/drm/tegra/gather_bo.c |  82 +++++
>  drivers/gpu/drm/tegra/gather_bo.h |  24 ++
>  drivers/gpu/drm/tegra/submit.c    | 549 ++++++++++++++++++++++++++++++
>  drivers/gpu/drm/tegra/submit.h    |  17 +
>  6 files changed, 677 insertions(+), 1 deletion(-)
>  create mode 100644 drivers/gpu/drm/tegra/gather_bo.c
>  create mode 100644 drivers/gpu/drm/tegra/gather_bo.h
>  create mode 100644 drivers/gpu/drm/tegra/submit.c
>  create mode 100644 drivers/gpu/drm/tegra/submit.h

...

> +int tegra_drm_ioctl_channel_submit(struct drm_device *drm, void *data,
> +				   struct drm_file *file)
> +{
> +	struct tegra_drm_file *fpriv = file->driver_priv;
> +	struct drm_tegra_channel_submit *args = data;
> +	struct tegra_drm_submit_data *job_data;
> +	struct drm_syncobj *syncobj = NULL;
> +	struct tegra_drm_context *ctx;
> +	struct host1x_job *job;
> +	struct gather_bo *bo;
> +	u32 i;
> +	int err;
> +
> +	mutex_lock(&fpriv->lock);
> +	ctx = xa_load(&fpriv->contexts, args->channel_ctx);
> +	if (!ctx) {
> +		mutex_unlock(&fpriv->lock);
> +		pr_err_ratelimited("%s: %s: invalid channel_ctx '%d'", __func__,
> +			current->comm, args->channel_ctx);
> +		return -EINVAL;
> +	}
> +
> +	if (args->syncobj_in) {
> +		struct dma_fence *fence;
> +
> +		err = drm_syncobj_find_fence(file, args->syncobj_in, 0, 0, &fence);
> +		if (err) {
> +			SUBMIT_ERR(ctx, "invalid syncobj_in '%d'", args->syncobj_in);
> +			goto unlock;
> +		}
> +
> +		err = dma_fence_wait_timeout(fence, true, msecs_to_jiffies(10000));
> +		dma_fence_put(fence);
> +		if (err) {
> +			SUBMIT_ERR(ctx, "wait for syncobj_in timed out");
> +			goto unlock;
> +		}
> +	}
> +
> +	if (args->syncobj_out) {
> +		syncobj = drm_syncobj_find(file, args->syncobj_out);
> +		if (!syncobj) {
> +			SUBMIT_ERR(ctx, "invalid syncobj_out '%d'", args->syncobj_out);
> +			err = -ENOENT;
> +			goto unlock;
> +		}
> +	}
> +
> +	/* Allocate gather BO and copy gather words in. */
> +	err = submit_copy_gather_data(&bo, drm->dev, ctx, args);
> +	if (err)
> +		goto unlock;
> +
> +	job_data = kzalloc(sizeof(*job_data), GFP_KERNEL);
> +	if (!job_data) {
> +		SUBMIT_ERR(ctx, "failed to allocate memory for job data");
> +		err = -ENOMEM;
> +		goto put_bo;
> +	}
> +
> +	/* Get data buffer mappings and do relocation patching. */
> +	err = submit_process_bufs(ctx, bo, args, job_data);
> +	if (err)
> +		goto free_job_data;
> +
> +	/* Allocate host1x_job and add gathers and waits to it. */
> +	err = submit_create_job(&job, ctx, bo, args, job_data,
> +				&fpriv->syncpoints);
> +	if (err)
> +		goto free_job_data;
> +
> +	/* Map gather data for Host1x. */
> +	err = host1x_job_pin(job, ctx->client->base.dev);
> +	if (err) {
> +		SUBMIT_ERR(ctx, "failed to pin job: %d", err);
> +		goto put_job;
> +	}
> +
> +	/* Boot engine. */
> +	if (pm_runtime_enabled(ctx->client->base.dev)) {
> +		err = pm_runtime_resume_and_get(ctx->client->base.dev);
> +		if (err < 0) {
> +			SUBMIT_ERR(ctx, "could not power up engine: %d", err);
> +			goto unpin_job;
> +		}
> +	}
> +
> +	job->user_data = job_data;
> +	job->release = release_job;
> +	job->timeout = 10000;
> +
> +	/*
> +	 * job_data is now part of job reference counting, so don't release
> +	 * it from here.
> +	 */
> +	job_data = NULL;
> +
> +	/* Submit job to hardware. */
> +	err = host1x_job_submit(job);
> +	if (err) {
> +		SUBMIT_ERR(ctx, "host1x job submission failed: %d", err);
> +		goto unpin_job;
> +	}


If we fail here, it appears that we may leave rpm enabled. Should we be
calling pm_runtime_put() for any failures from here on?

Cheers
Jon

-- 
nvpublic