drm/i915/gem: Require per-engine reset support for non-persistent contexts

To enable non-persistent contexts, we require a means of cancelling any
inflight work from that context. This is first done "gracefully" by
using preemption to kick the active context off the engine, and then
forcefully by resetting the engine if it is active. If we are unable to
reset the engine to remove hostile userspace, we should not allow
userspace to opt into using non-persistent contexts.

If the per-engine reset fails, we still do a full GPU reset, but that is
rare and usually indicative of much deeper issues. The damage is already
done. However, the goal of the interface to allow long running compute
jobs without causing collateral damage elsewhere, and if we are unable
to support that we should make that known by not providing the
interface (and falsely pretending we can).

Fixes: a0e047156cde ("drm/i915/gem: Make context persistence optional")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Jon Bloomfield <jon.bloomfield@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200130164553.1937718-1-chris@chris-wilson.co.uk
(cherry picked from commit d1b9b5f127bc3797fc274cfa4f363e039f045c3a)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

1 files changed, 16 insertions, 0 deletions

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
index a2e57e62af30..151a1e8ae36a 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
@@ -565,6 +565,22 @@ static int __context_set_persistence(struct i915_gem_context *ctx, bool state)
 		if (!(ctx->i915->caps.scheduler & I915_SCHEDULER_CAP_PREEMPTION))
 			return -ENODEV;
 
+		/*
+		 * If the cancel fails, we then need to reset, cleanly!
+		 *
+		 * If the per-engine reset fails, all hope is lost! We resort
+		 * to a full GPU reset in that unlikely case, but realistically
+		 * if the engine could not reset, the full reset does not fare
+		 * much better. The damage has been done.
+		 *
+		 * However, if we cannot reset an engine by itself, we cannot
+		 * cleanup a hanging persistent context without causing
+		 * colateral damage, and we should not pretend we can by
+		 * exposing the interface.
+		 */
+		if (!intel_has_reset_engine(&ctx->i915->gt))
+			return -ENODEV;
+
 		i915_gem_context_clear_persistence(ctx);
 	}