[RFC PATCH 7/8] drm/tegra: Support context isolation

[Mikko Perttunen]

For engines that support context isolation, allocate a context when
opening a channel, and set up stream ID offset and context fields
when submitting a job.

Signed-off-by: Mikko Perttunen <mperttunen at nvidia.com>
---
 drivers/gpu/drm/tegra/drm.h         |  1 +
 drivers/gpu/drm/tegra/uapi.h        |  1 +
 drivers/gpu/drm/tegra/uapi/submit.c | 13 +++++++++++
 drivers/gpu/drm/tegra/uapi/uapi.c   | 34 +++++++++++++++++++++++++++--
 4 files changed, 47 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/tegra/drm.h b/drivers/gpu/drm/tegra/drm.h
index 1af57c2016eb..5d3903372041 100644
--- a/drivers/gpu/drm/tegra/drm.h
+++ b/drivers/gpu/drm/tegra/drm.h
@@ -81,6 +81,7 @@ struct tegra_drm_client_ops {
 	int (*submit)(struct tegra_drm_context *context,
 		      struct drm_tegra_submit *args, struct drm_device *drm,
 		      struct drm_file *file);
+	int (*get_streamid_offset)(struct tegra_drm_client *client);
 };
 
 int tegra_drm_submit(struct tegra_drm_context *context,
diff --git a/drivers/gpu/drm/tegra/uapi.h b/drivers/gpu/drm/tegra/uapi.h
index 5c422607e8fa..8efb30bbdb4b 100644
--- a/drivers/gpu/drm/tegra/uapi.h
+++ b/drivers/gpu/drm/tegra/uapi.h
@@ -26,6 +26,7 @@ struct tegra_drm_file {
 struct tegra_drm_channel_ctx {
 	struct tegra_drm_client *client;
 	struct host1x_channel *channel;
+	struct host1x_context *context;
 	struct xarray mappings;
 };
 
diff --git a/drivers/gpu/drm/tegra/uapi/submit.c b/drivers/gpu/drm/tegra/uapi/submit.c
index 8633844ae3d7..1d0dd28af99f 100644
--- a/drivers/gpu/drm/tegra/uapi/submit.c
+++ b/drivers/gpu/drm/tegra/uapi/submit.c
@@ -337,6 +337,9 @@ static void release_job(struct host1x_job *job)
 	struct tegra_drm_submit_data *job_data = job->user_data;
 	u32 i;
 
+	if (job->context)
+		host1x_context_put(job->context);
+
 	for (i = 0; i < job_data->num_used_mappings; i++)
 		tegra_drm_mapping_put(job_data->used_mappings[i].mapping);
 
@@ -398,6 +401,16 @@ int tegra_drm_ioctl_channel_submit(struct drm_device *drm, void *data,
 	job->release = release_job;
 	job->timeout = 10000;
 
+	if (ctx->context && ctx->client->ops->get_streamid_offset) {
+		int offset = ctx->client->ops->get_streamid_offset(ctx->client);
+
+		if (offset >= 0) {
+			job->context = ctx->context;
+			job->engine_streamid_offset = offset;
+			host1x_context_get(job->context);
+		}
+	}
+
 	/*
 	 * job_data is now part of job reference counting, so don't release
 	 * it from here.
diff --git a/drivers/gpu/drm/tegra/uapi/uapi.c b/drivers/gpu/drm/tegra/uapi/uapi.c
index d503b5e817c4..e2e0c88b858d 100644
--- a/drivers/gpu/drm/tegra/uapi/uapi.c
+++ b/drivers/gpu/drm/tegra/uapi/uapi.c
@@ -49,6 +49,9 @@ static void tegra_drm_channel_ctx_close(struct tegra_drm_channel_ctx *ctx)
 	unsigned long mapping_id;
 	struct tegra_drm_mapping *mapping;
 
+	if (ctx->context)
+		host1x_context_put(ctx->context);
+
 	xa_for_each(&ctx->mappings, mapping_id, mapping)
 		tegra_drm_mapping_put(mapping);
 
@@ -82,6 +85,7 @@ void tegra_drm_uapi_close_file(struct tegra_drm_file *file)
 int tegra_drm_ioctl_channel_open(struct drm_device *drm, void *data,
 				 struct drm_file *file)
 {
+	struct host1x *host = tegra_drm_to_host1x(drm->dev_private);
 	struct tegra_drm_file *fpriv = file->driver_priv;
 	struct tegra_drm *tegra = drm->dev_private;
 	struct drm_tegra_channel_open *args = data;
@@ -116,10 +120,29 @@ int tegra_drm_ioctl_channel_open(struct drm_device *drm, void *data,
 		}
 	}
 
+	/* Only allocate context if the engine supports context isolation. */
+	if (client->ops->get_streamid_offset &&
+	    client->ops->get_streamid_offset(client) >= 0) {
+		ctx->context =
+			host1x_context_alloc(host, get_task_pid(current, PIDTYPE_TGID));
+		if (IS_ERR(ctx->context)) {
+			if (PTR_ERR(ctx->context) != -EOPNOTSUPP) {
+				err = PTR_ERR(ctx->context);
+				goto put_channel;
+			} else {
+				/*
+				 * OK, HW does not support contexts or contexts
+				 * are disabled.
+				 */
+				ctx->context = NULL;
+			}
+		}
+	}
+
 	err = xa_alloc(&fpriv->contexts, &args->channel_ctx, ctx,
 		       XA_LIMIT(1, U32_MAX), GFP_KERNEL);
 	if (err < 0)
-		goto put_channel;
+		goto put_context;
 
 	ctx->client = client;
 	xa_init_flags(&ctx->mappings, XA_FLAGS_ALLOC1);
@@ -128,6 +151,9 @@ int tegra_drm_ioctl_channel_open(struct drm_device *drm, void *data,
 
 	return 0;
 
+put_context:
+	if (ctx->context)
+		host1x_context_put(ctx->context);
 put_channel:
 	host1x_channel_put(ctx->channel);
 free_ctx:
@@ -188,7 +214,11 @@ int tegra_drm_ioctl_channel_map(struct drm_device *drm, void *data,
 		goto unlock;
 	}
 
-	mapping->dev = ctx->client->base.dev;
+	if (ctx->context)
+		mapping->dev = &ctx->context->dev;
+	else
+		mapping->dev = ctx->client->base.dev;
+
 	mapping->bo = &container_of(gem, struct tegra_bo, gem)->base;
 
 	if (!iommu_get_domain_for_dev(mapping->dev) ||
-- 
2.30.0