[PATCH v2 0/2] Job cancel tests

[PATCH v2 0/2] Job cancel tests

[Matthew Brost]

Tests following series:
[1] https://patchwork.freedesktop.org/series/134593/

Matthew Brost (2):
  xe_exec_reset: Add job cancel tests
  xe_exec_reset: Add timeslice preempt test

 tests/intel/xe_exec_reset.c | 44 ++++++++++++++++++++++++++++++++++---
 1 file changed, 41 insertions(+), 3 deletions(-)

-- 
2.34.1

[PATCH v2 1/2] xe_exec_reset: Add job cancel tests

[Matthew Brost]

This test has pulled out when the ability to set the job timeout via
IOCTL was removed from the uAPI while upstreaming Xe. Add this coverage
back in but only run on a single hardware engine given by default jobs
take 5 seconds to timeout.

Signed-off-by: Matthew Brost <matthew.brost@intel.com>
---
 tests/intel/xe_exec_reset.c | 21 ++++++++++++++++++++-
 1 file changed, 20 insertions(+), 1 deletion(-)

diff --git a/tests/intel/xe_exec_reset.c b/tests/intel/xe_exec_reset.c
index 8d731e6893..5e9e48adda 100644
--- a/tests/intel/xe_exec_reset.c
+++ b/tests/intel/xe_exec_reset.c
@@ -99,6 +99,7 @@ static void test_spin(int fd, struct drm_xe_engine_class_instance *eci)
 #define VIRTUAL				(0x1 << 3)
 #define PARALLEL			(0x1 << 4)
 #define CAT_ERROR			(0x1 << 5)
+#define PREEMPT				(0x1 << 6)
 
 /**
  * SUBTEST: %s-cat-error
@@ -280,6 +281,12 @@ test_balancer(int fd, int gt, int class, int n_exec_queues, int n_execs,
  * SUBTEST: cat-error
  * Description: Test cat error
  *
+ * SUBTEST: cancel
+ * Description: Test job cancel
+ *
+ * SUBTEST: cancel-preempt
+ * Description: Test job cancel with a preemptable job
+ *
  * SUBTEST: gt-reset
  * Description: Test GT reset
  *
@@ -318,7 +325,7 @@ test_legacy_mode(int fd, struct drm_xe_engine_class_instance *eci,
 		uint64_t pad;
 		uint32_t data;
 	} *data;
-	struct xe_spin_opts spin_opts = { .preempt = false };
+	struct xe_spin_opts spin_opts = { .preempt = flags & PREEMPT };
 	int i, b;
 
 	igt_assert(n_exec_queues <= MAX_N_EXECQUEUES);
@@ -727,6 +734,18 @@ igt_main
 		xe_for_each_engine(fd, hwe)
 			test_legacy_mode(fd, hwe, 2, 2, CAT_ERROR);
 
+	igt_subtest("cancel")
+		xe_for_each_engine(fd, hwe) {
+			test_legacy_mode(fd, hwe, 1, 1, 0);
+			break;
+		}
+
+	igt_subtest("cancel-preempt")
+		xe_for_each_engine(fd, hwe) {
+			test_legacy_mode(fd, hwe, 1, 1, PREEMPT);
+			break;
+		}
+
 	igt_subtest("gt-reset")
 		xe_for_each_engine(fd, hwe)
 			test_legacy_mode(fd, hwe, 2, 2, GT_RESET);
-- 
2.34.1

[PATCH v2 2/2] xe_exec_reset: Add timeslice preempt test

[Matthew Brost]

Add tests in which submit multiple preemptable spinners to same hardware
engine ensuring a cumulative longer timeout and validating the restart
paths in the TDR.

Signed-off-by: Matthew Brost <matthew.brost@intel.com>
---
 tests/intel/xe_exec_reset.c | 23 +++++++++++++++++++++--
 1 file changed, 21 insertions(+), 2 deletions(-)

diff --git a/tests/intel/xe_exec_reset.c b/tests/intel/xe_exec_reset.c
index 5e9e48adda..144e3b3884 100644
--- a/tests/intel/xe_exec_reset.c
+++ b/tests/intel/xe_exec_reset.c
@@ -100,6 +100,7 @@ static void test_spin(int fd, struct drm_xe_engine_class_instance *eci)
 #define PARALLEL			(0x1 << 4)
 #define CAT_ERROR			(0x1 << 5)
 #define PREEMPT				(0x1 << 6)
+#define CANCEL				(0x1 << 7)
 
 /**
  * SUBTEST: %s-cat-error
@@ -287,6 +288,12 @@ test_balancer(int fd, int gt, int class, int n_exec_queues, int n_execs,
  * SUBTEST: cancel-preempt
  * Description: Test job cancel with a preemptable job
  *
+ * SUBTEST: cancel-timeslice-preempt
+ * Description: Test job cancel with 2 preemptable jobs
+ *
+ * SUBTEST: cancel-timeslice-many-preempt
+ * Description: Test job cancel with many preemptable jobs
+ *
  * SUBTEST: gt-reset
  * Description: Test GT reset
  *
@@ -361,7 +368,7 @@ test_legacy_mode(int fd, struct drm_xe_engine_class_instance *eci,
 		uint64_t exec_addr;
 		int e = i % n_exec_queues;
 
-		if (!i) {
+		if (!i || flags & CANCEL) {
 			spin_opts.addr = base_addr + spin_offset;
 			xe_spin_init(&data[i].spin, &spin_opts);
 			exec_addr = spin_opts.addr;
@@ -411,7 +418,7 @@ test_legacy_mode(int fd, struct drm_xe_engine_class_instance *eci,
 	xe_vm_unbind_async(fd, vm, 0, 0, addr, bo_size, sync, 1);
 	igt_assert(syncobj_wait(fd, &sync[0].handle, 1, INT64_MAX, 0, NULL));
 
-	if (!(flags & GT_RESET)) {
+	if (!(flags & (GT_RESET | CANCEL))) {
 		for (i = 1; i < n_execs; i++)
 			igt_assert_eq(data[i].data, 0xc0ffee);
 	}
@@ -746,6 +753,18 @@ igt_main
 			break;
 		}
 
+	igt_subtest("cancel-timeslice-preempt")
+		xe_for_each_engine(fd, hwe) {
+			test_legacy_mode(fd, hwe, 2, 2, CANCEL | PREEMPT);
+			break;
+		}
+
+	igt_subtest("cancel-timeslice-many-preempt")
+		xe_for_each_engine(fd, hwe) {
+			test_legacy_mode(fd, hwe, 4, 4, CANCEL | PREEMPT);
+			break;
+		}
+
 	igt_subtest("gt-reset")
 		xe_for_each_engine(fd, hwe)
 			test_legacy_mode(fd, hwe, 2, 2, GT_RESET);
-- 
2.34.1

Re: [PATCH v2 2/2] xe_exec_reset: Add timeslice preempt test

[Kamil Konieczny]

Hi Matthew,
On 2024-06-07 at 17:12:11 -0700, Matthew Brost wrote:

you did Cc to intel-xe list, in that case please add i-g-t
after PATCH:

[PATCH i-g-t v2 2/2] xe_exec_reset: Add timeslice preempt test

> Add tests in which submit multiple preemptable spinners to same hardware
> engine ensuring a cumulative longer timeout and validating the restart
> paths in the TDR.

Please explain what TDR means, like: in the TDR (T... D... R...).

Regards,
Kamil

> 
> Signed-off-by: Matthew Brost <matthew.brost@intel.com>
> ---
>  tests/intel/xe_exec_reset.c | 23 +++++++++++++++++++++--
>  1 file changed, 21 insertions(+), 2 deletions(-)
> 
> diff --git a/tests/intel/xe_exec_reset.c b/tests/intel/xe_exec_reset.c
> index 5e9e48adda..144e3b3884 100644
> --- a/tests/intel/xe_exec_reset.c
> +++ b/tests/intel/xe_exec_reset.c
> @@ -100,6 +100,7 @@ static void test_spin(int fd, struct drm_xe_engine_class_instance *eci)
>  #define PARALLEL			(0x1 << 4)
>  #define CAT_ERROR			(0x1 << 5)
>  #define PREEMPT				(0x1 << 6)
> +#define CANCEL				(0x1 << 7)
>  
>  /**
>   * SUBTEST: %s-cat-error
> @@ -287,6 +288,12 @@ test_balancer(int fd, int gt, int class, int n_exec_queues, int n_execs,
>   * SUBTEST: cancel-preempt
>   * Description: Test job cancel with a preemptable job
>   *
> + * SUBTEST: cancel-timeslice-preempt
> + * Description: Test job cancel with 2 preemptable jobs
> + *
> + * SUBTEST: cancel-timeslice-many-preempt
> + * Description: Test job cancel with many preemptable jobs
> + *
>   * SUBTEST: gt-reset
>   * Description: Test GT reset
>   *
> @@ -361,7 +368,7 @@ test_legacy_mode(int fd, struct drm_xe_engine_class_instance *eci,
>  		uint64_t exec_addr;
>  		int e = i % n_exec_queues;
>  
> -		if (!i) {
> +		if (!i || flags & CANCEL) {
>  			spin_opts.addr = base_addr + spin_offset;
>  			xe_spin_init(&data[i].spin, &spin_opts);
>  			exec_addr = spin_opts.addr;
> @@ -411,7 +418,7 @@ test_legacy_mode(int fd, struct drm_xe_engine_class_instance *eci,
>  	xe_vm_unbind_async(fd, vm, 0, 0, addr, bo_size, sync, 1);
>  	igt_assert(syncobj_wait(fd, &sync[0].handle, 1, INT64_MAX, 0, NULL));
>  
> -	if (!(flags & GT_RESET)) {
> +	if (!(flags & (GT_RESET | CANCEL))) {
>  		for (i = 1; i < n_execs; i++)
>  			igt_assert_eq(data[i].data, 0xc0ffee);
>  	}
> @@ -746,6 +753,18 @@ igt_main
>  			break;
>  		}
>  
> +	igt_subtest("cancel-timeslice-preempt")
> +		xe_for_each_engine(fd, hwe) {
> +			test_legacy_mode(fd, hwe, 2, 2, CANCEL | PREEMPT);
> +			break;
> +		}
> +
> +	igt_subtest("cancel-timeslice-many-preempt")
> +		xe_for_each_engine(fd, hwe) {
> +			test_legacy_mode(fd, hwe, 4, 4, CANCEL | PREEMPT);
> +			break;
> +		}
> +
>  	igt_subtest("gt-reset")
>  		xe_for_each_engine(fd, hwe)
>  			test_legacy_mode(fd, hwe, 2, 2, GT_RESET);
> -- 
> 2.34.1
> 

✗ CI.Patch_applied: failure for Job cancel tests

[Patchwork]

== Series Details ==

Series: Job cancel tests
URL   : https://patchwork.freedesktop.org/series/134641/
State : failure

== Summary ==

=== Applying kernel patches on branch 'drm-tip' with base: ===
Base commit: 2bea08bd3129 drm-tip: 2024y-06m-07d-19h-26m-01s UTC integration manifest
=== git am output follows ===
error: tests/intel/xe_exec_reset.c: does not exist in index
hint: Use 'git am --show-current-patch=diff' to see the failed patch
Applying: xe_exec_reset: Add job cancel tests
Patch failed at 0001 xe_exec_reset: Add job cancel tests
When you have resolved this problem, run "git am --continue".
If you prefer to skip this patch, run "git am --skip" instead.
To restore the original branch and stop patching, run "git am --abort".