Skip to contentSkip to conversation

lib/igt_aux: Export statistics of signal helper.

PATCH In review Maarten Lankhorst 4 participants 13 messages Aug 2, 2017 – Aug 8, 2017 Library code

lore
Share & tools
lore
#0 lore raw

PATCH 1/3 lib/igt_aux: Export statistics of signal helper.

Export 2 functions, igt_signal_helper_get_num and
igt_signal_helper_get_hz.

This will allow tests to measure how much time in a test was spent in a uninterruptible state, which is useful when testing whether certain ioctl's can be interrupted or not.

---
 lib/igt_aux.c | 30 +++++++++++++++++++++++++++---
 lib/igt_aux.h |  2 ++
 2 files changed, 29 insertions(+), 3 deletions(-)
lib/igt_aux.c +27 -3
index 86a213c2032f..265e43f399e7 100644
@@ -275,7 +275,7 @@ bool __igt_sigiter_continue(struct __igt_sigiter *iter, bool enable)
 }
 
 static struct igt_helper_process signal_helper;
-long long int sig_stat;
+static int64_t sig_stat;
 static void __attribute__((noreturn)) signal_helper_process(pid_t pid)
 {
 	/* Interrupt the parent process at 500Hz, just to be annoying */
@@ -314,6 +314,9 @@ void igt_fork_signal_helper(void)
 	if (igt_only_list_subtests())
 		return;
 
+	/* Reset number of signalscaught */
+	sig_stat = 0;
+
 	/* We pick SIGCONT as it is a "safe" signal - if we send SIGCONT to
 	 * an unexpecting process it spuriously wakes up and does nothing.
 	 * Most other signals (e.g. SIGUSR1) cause the process to die if they
@@ -348,8 +351,29 @@ void igt_stop_signal_helper(void)
 		return;
 
 	igt_stop_helper(&signal_helper);
+}
 
-	sig_stat = 0;
+/**
+ * igt_signal_helper_get_num:
+ *
+ * Return the amount of signals generated since the last time
+ * igt_fork_signal_helper() was called.
+ *
+ * This is reset to 0 on every call to igt_fork_signal_helper.
+ */
+int64_t igt_signal_helper_get_num(void)
+{
+	return sig_stat;
+}
+
+/**
+ * igt_signal_helper_get_hz:
+ *
+ * Return the approximate amount of signals generated per second.
+ */
+int igt_signal_helper_get_hz(void)
+{
+	return 50;
 }
 
 static struct igt_helper_process shrink_helper;
@@ -357,7 +381,7 @@ static void __attribute__((noreturn)) shrink_helper_process(int fd, pid_t pid)
 {
 	while (1) {
 		igt_drop_caches_set(fd, DROP_SHRINK_ALL);
-		usleep(1000 * 1000 / 50);
+		usleep(1000 * 1000 / igt_signal_helper_get_hz());
 		if (kill(pid, 0)) /* Parent has died, so must we. */
 			exit(0);
 	}
lib/igt_aux.h +2 -0
index 499a16796ebb..7e080089dcbc 100644
@@ -55,6 +55,8 @@ extern int num_trash_bos;
 /* generally useful helpers */
 void igt_fork_signal_helper(void);
 void igt_stop_signal_helper(void);
+int64_t igt_signal_helper_get_num(void);
+int igt_signal_helper_get_hz(void);
 
 void igt_fork_shrink_helper(int fd);
 void igt_stop_shrink_helper(void);
Signature · 6 lines
2.11.0

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#1 lore raw

PATCH 2/3 lib/igt_kms: Remove vblank wait after plane update.

With the conversion to atomic, this is already handled in the core.

---
 lib/igt_kms.c | 15 ---------------
 lib/igt_kms.h |  1 -
 2 files changed, 16 deletions(-)
lib/igt_kms.c +0 -15
index 6390229f1546..14e2701c3afd 100644
@@ -2374,8 +2374,6 @@ static int igt_primary_plane_commit_legacy(igt_plane_t *primary,
 
 	CHECK_RETURN(ret, fail_on_error);
 
-	primary->pipe->enabled = (fb_id != 0);
-
 	return 0;
 }
 
@@ -2413,10 +2411,8 @@ static int igt_pipe_commit(igt_pipe_t *pipe,
 			   enum igt_commit_style s,
 			   bool fail_on_error)
 {
-	igt_display_t *display = pipe->display;
 	int i;
 	int ret;
-	bool need_wait_for_vblank = false;
 
 	if (pipe->background_changed) {
 		igt_crtc_set_property(pipe, pipe->background_property,
@@ -2435,21 +2431,10 @@ static int igt_pipe_commit(igt_pipe_t *pipe,
 	for (i = 0; i < pipe->n_planes; i++) {
 		igt_plane_t *plane = &pipe->planes[i];
 
-		if (plane->fb_changed || plane->position_changed || plane->size_changed)
-			need_wait_for_vblank = true;
-
 		ret = igt_plane_commit(plane, pipe, s, fail_on_error);
 		CHECK_RETURN(ret, fail_on_error);
 	}
 
-	/*
-	 * If the crtc is enabled, wait until the next vblank before returning
-	 * if we made changes to any of the planes.
-	 */
-	if (need_wait_for_vblank && pipe->enabled) {
-		igt_wait_for_vblank(display->drm_fd, pipe->pipe);
-	}
-
 	return 0;
 }
 
lib/igt_kms.h +0 -1
index 35428f3e9675..d18a92600933 100644
@@ -323,7 +323,6 @@ typedef struct {
 struct igt_pipe {
 	igt_display_t *display;
 	enum pipe pipe;
-	bool enabled;
 
 	int n_planes;
 	int plane_cursor;
Signature · 6 lines
2.11.0

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#2 lore raw

Re: [PATCH i-g-t 2/3] lib/igt_kms: Remove vblank wait after plane update.

Quoted 17 lines · Maarten Lankhorst

With the conversion to atomic, this is already handled in the core.

---
 lib/igt_kms.c | 15 ---------------
 lib/igt_kms.h |  1 -

 2 files changed, 16 deletions(-)
lib/igt_kms.c +0 -0
index 6390229f1546..14e2701c3afd 100644
@@ -2374,8 +2374,6 @@ static int
 igt_primary_plane_commit_legacy(igt_plane_t *primary,


  CHECK_RETURN(ret, fail_on_error);

lib/igt_kms.c static int igt_primary_plane_commit_legacy(igt_plane_t *primary, · line 2378

Quoted 2 lines
lib/igt_kms.c +0 -2
-	primary->pipe->enabled = (fb_id != 0);
-

How was this related to vblank wait removal?

Quoted 52 lines

  return 0;
 }

lib/igt_kms.c +0 -2
@@ -2413,10 +2411,8 @@ static int igt_pipe_commit(igt_pipe_t *pipe,
  			   enum igt_commit_style s,
  			   bool fail_on_error)
  {
-	igt_display_t *display = pipe->display;
  	int i;
  	int ret;
-	bool need_wait_for_vblank = false;


  if (pipe->background_changed) {
  igt_crtc_set_property(pipe, pipe-

Quoted 1 line

background_property,

lib/igt_kms.c +0 -0
@@ -2435,21 +2431,10 @@ static int igt_pipe_commit(igt_pipe_t *pipe,
  	for (i = 0; i < pipe->n_planes; i++) {
  		igt_plane_t *plane = &pipe->planes[i];

lib/igt_kms.c +0 -1
-		if (plane->fb_changed || plane->position_changed ||

plane->size_changed)

lib/igt_kms.c +0 -2
-			need_wait_for_vblank = true;
-

  ret = igt_plane_commit(plane, pipe, s, fail_on_error);
  CHECK_RETURN(ret, fail_on_error); }

lib/igt_kms.c +0 -2
-	/*
-	 * If the crtc is enabled, wait until the next vblank before

returning

lib/igt_kms.c +0 -6
-	 * if we made changes to any of the planes.
-	 */
-	if (need_wait_for_vblank && pipe->enabled) {
-		igt_wait_for_vblank(display->drm_fd, pipe->pipe);
-	}
-

  return 0;
 }

lib/igt_kms.h +0 -1
index 35428f3e9675..d18a92600933 100644
@@ -323,7 +323,6 @@ typedef struct {
  struct igt_pipe {
  	igt_display_t *display;
  	enum pipe pipe;
-	bool enabled;


  int n_planes;
  int plane_cursor;

Signature · 6 lines
Mika Kahola - Intel OTC

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#3 lore raw

Re: [PATCH i-g-t 2/3] lib/igt_kms: Remove vblank wait after plane update.

Op 04-08-17 om 10:07 schreef Mika Kahola:

Quoted 16 lines · Maarten Lankhorst
Quoted 15 lines

With the conversion to atomic, this is already handled in the core.

---
 lib/igt_kms.c | 15 ---------------
 lib/igt_kms.h |  1 -
 2 files changed, 16 deletions(-)
lib/igt_kms.c +0 -0
index 6390229f1546..14e2701c3afd 100644
@@ -2374,8 +2374,6 @@ static int
 igt_primary_plane_commit_legacy(igt_plane_t *primary,
 
Quoted 5 lines

lib/igt_kms.c static int igt_primary_plane_commit_legacy(igt_plane_t *primary, · line 2378

Quoted 4 lines
lib/igt_kms.c +0 -2
 	CHECK_RETURN(ret, fail_on_error);
 
-	primary->pipe->enabled = (fb_id != 0);
-

How was this related to vblank wait removal?

There are no users of pipe->enabled left, so it's removed.

Quoted 50 lines
Quoted 50 lines
lib/igt_kms.c +0 -2
 	return 0;
 }
@@ -2413,10 +2411,8 @@ static int igt_pipe_commit(igt_pipe_t *pipe,
 			   enum igt_commit_style s,
 			   bool fail_on_error)
 {
-	igt_display_t *display = pipe->display;
 	int i;
 	int ret;
-	bool need_wait_for_vblank = false;
 
 	if (pipe->background_changed) {
 		igt_crtc_set_property(pipe, pipe-
Quoted 1 line

background_property,

lib/igt_kms.c +0 -5
@@ -2435,21 +2431,10 @@ static int igt_pipe_commit(igt_pipe_t *pipe,
 	for (i = 0; i < pipe->n_planes; i++) {
 		igt_plane_t *plane = &pipe->planes[i];
 
-		if (plane->fb_changed || plane->position_changed ||
 plane->size_changed)
-			need_wait_for_vblank = true;
-
 		ret = igt_plane_commit(plane, pipe, s,
 fail_on_error);
 		CHECK_RETURN(ret, fail_on_error);
 	}
 
-	/*
-	 * If the crtc is enabled, wait until the next vblank before

returning

lib/igt_kms.c +0 -6
-	 * if we made changes to any of the planes.
-	 */
-	if (need_wait_for_vblank && pipe->enabled) {
-		igt_wait_for_vblank(display->drm_fd, pipe->pipe);
-	}
-
 	return 0;
 }
lib/igt_kms.h +0 -1
index 35428f3e9675..d18a92600933 100644
@@ -323,7 +323,6 @@ typedef struct {
 struct igt_pipe {
 	igt_display_t *display;
 	enum pipe pipe;
-	bool enabled;
 
Quoted 2 lines

lib/igt_kms.h typedef struct { · line 328

Quoted 2 lines
lib/igt_kms.h +0 -0
 	int n_planes;
 	int plane_cursor;

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Reviewed-by
#4 lore raw

Re: [PATCH i-g-t 2/3] lib/igt_kms: Remove vblank wait after plane update.

Quoted 26 lines · Maarten Lankhorst

Op 04-08-17 om 10:07 schreef Mika Kahola:

Quoted 13 lines · Maarten Lankhorst
Quoted 11 lines

With the conversion to atomic, this is already handled in the core.

om>
---
 lib/igt_kms.c | 15 ---------------
 lib/igt_kms.h |  1 -

 2 files changed, 16 deletions(-)
Quoted 12 lines

lib/igt_kms.c static int igt_primary_plane_commit_legacy(igt_plane_t *primary, · line 2378

Quoted 11 lines
lib/igt_kms.c +0 -0
index 6390229f1546..14e2701c3afd 100644
@@ -2374,8 +2374,6 @@ static int
 igt_primary_plane_commit_legacy(igt_plane_t *primary,


  CHECK_RETURN(ret, fail_on_error);

lib/igt_kms.c +0 -2
-	primary->pipe->enabled = (fb_id != 0);
-

How was this related to vblank wait removal?

Quoted 1 line

There are no users of pipe->enabled left, so it's removed.

Ok. You can add my

Quoted 61 lines
Quoted 60 lines
Quoted 60 lines

  return 0;
 }

lib/igt_kms.c +0 -2
@@ -2413,10 +2411,8 @@ static int igt_pipe_commit(igt_pipe_t
 *pipe,
  			   enum igt_commit_style s,
  			   bool fail_on_error)
  {
-	igt_display_t *display = pipe->display;
  	int i;
  	int ret;
-	bool need_wait_for_vblank = false;


  if (pipe->background_changed) {
  igt_crtc_set_property(pipe, pipe-

Quoted 2 lines

background_property,

lib/igt_kms.c +0 -0
@@ -2435,21 +2431,10 @@ static int igt_pipe_commit(igt_pipe_t
 *pipe,
  	for (i = 0; i < pipe->n_planes; i++) {
  		igt_plane_t *plane = &pipe->planes[i];

lib/igt_kms.c +0 -1
-		if (plane->fb_changed || plane->position_changed 

||
plane->size_changed)

lib/igt_kms.c +0 -2
-			need_wait_for_vblank = true;
-

  ret = igt_plane_commit(plane, pipe, s, fail_on_error);
  CHECK_RETURN(ret, fail_on_error); }

lib/igt_kms.c +0 -2
-	/*
-	 * If the crtc is enabled, wait until the next vblank

before
returning

lib/igt_kms.c +0 -4
-	 * if we made changes to any of the planes.
-	 */
-	if (need_wait_for_vblank && pipe->enabled) {
-		igt_wait_for_vblank(display->drm_fd, pipe-
Quoted 1 line

pipe);

lib/igt_kms.c +0 -2
-	}
-

  return 0;
 }

lib/igt_kms.h +0 -1
index 35428f3e9675..d18a92600933 100644
@@ -323,7 +323,6 @@ typedef struct {
  struct igt_pipe {
  	igt_display_t *display;
  	enum pipe pipe;
-	bool enabled;


  int n_planes;
  int plane_cursor;

Signature · 6 lines
Mika Kahola - Intel OTC

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#5 lore raw

PATCH 3/3 tests: Add kms_atomic_interruptible test.

In reply to Maarten Lankhorst

To make sure that we have test exposure when allowing atomic ioctl's to fail interruptibly, we add a test that will fail to lock the mutexes until the fence is signaled.

If the locking is done interruptibly, our signal helper will interrupt often, and with the statistics we can ensure that we received enough interrupts to pass.

---
 tests/Makefile.sources           |   1 +
 tests/kms_atomic_interruptible.c | 193 +++++++++++++++++++++++++++++++++++++++
 2 files changed, 194 insertions(+)
 create mode 100644 tests/kms_atomic_interruptible.c
tests/Makefile.sources +1 -0
index 5b98a5a371b8..9a6a846ab78a 100644
@@ -171,6 +171,7 @@ TESTS_progs = \
 	kms_3d \
 	kms_addfb_basic \
 	kms_atomic \
+	kms_atomic_interruptible \
 	kms_atomic_transition \
 	kms_busy \
 	kms_ccs \
tests/kms_atomic_interruptible.c new file +193 -0

193 lines not on this page · raw message

Signature · 6 lines
2.11.0

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#6 lore raw

Re: [PATCH i-g-t 3/3] tests: Add kms_atomic_interruptible test.

Quoted 37 lines · Maarten Lankhorst

To make sure that we have test exposure when allowing atomic ioctl's to
fail interruptibly, we add a test that will fail to lock the mutexes until the fence is signaled.

If the locking is done interruptibly, our signal helper will interrupt
often, and with the statistics we can ensure that we received enough interrupts to pass.

---
 tests/Makefile.sources           |   1 +
 tests/kms_atomic_interruptible.c | 193
+++++++++++++++++++++++++++++++++++++++

 2 files changed, 194 insertions(+)

 create mode 100644 tests/kms_atomic_interruptible.c

tests/Makefile.sources +1 -0
index 5b98a5a371b8..9a6a846ab78a 100644
@@ -171,6 +171,7 @@ TESTS_progs = \
  	kms_3d \
  	kms_addfb_basic \
  	kms_atomic \
+	kms_atomic_interruptible \
  	kms_atomic_transition \
  	kms_busy \
  	kms_ccs \
tests/kms_atomic_interruptible.c new file +0 -0
new file mode 100644
index 000000000000..4015988c0e2c

tests/kms_atomic_interruptible.c line 2

Quoted 2 lines
tests/kms_atomic_interruptible.c new file +2 -0
@@ -0,0 +1,193 @@
+/*
+ * Copyright © 2016 Intel Corporation

Update a year here?

Quoted 117 lines
tests/kms_atomic_interruptible.c new file +2 -0
+ *
+ * Permission is hereby granted, free of charge, to any person

obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */

tests/kms_atomic_interruptible.c new file +13 -0
+
+#include "igt.h"
+#include "drmtest.h"
+#include "sw_sync.h"
+
+enum plane_test_type
+{
+	test_setmode,
+	test_setplane,
+	test_pageflip
+};
+
+static int block_plane(igt_display_t *display, igt_output_t *output,

enum plane_test_type test_type, igt_plane_t *plane)

tests/kms_atomic_interruptible.c new file +5 -0
+{
+	int timeline = sw_sync_timeline_create();
+
+	igt_fork(child, 1) {
+		/* Ignore the signal helper, we need to block

indefinitely on the fence. */

tests/kms_atomic_interruptible.c new file +7 -0
+		signal(SIGCONT, SIG_IGN);
+
+		if (test_type == test_setmode) {
+			igt_output_set_pipe(output, PIPE_NONE);
+			igt_plane_set_fb(plane, NULL);
+		}
+		igt_plane_set_fence_fd(plane,

sw_sync_timeline_create_fence(timeline, 1));

tests/kms_atomic_interruptible.c new file +14 -0
+
+		igt_display_commit2(display, COMMIT_ATOMIC);
+	}
+
+	return timeline;
+}
+
+static void unblock(int block)
+{
+	sw_sync_timeline_inc(block, 1);
+	close(block);
+}
+
+static void run_plane_test(igt_display_t *display, enum pipe pipe,

igt_output_t *output,

tests/kms_atomic_interruptible.c new file +1 -0
+			   enum plane_test_type test_type, unsigned

plane_type, enum igt_commit_style style)

tests/kms_atomic_interruptible.c new file +8 -0
+{
+	drmModeModeInfo *mode;
+	igt_fb_t fb, fb2;
+	igt_plane_t *primary, *plane;
+	int block;
+
+	/*
+	 * Make sure we start with everything disabled to force a

real modeset.

tests/kms_atomic_interruptible.c new file +1 -0
+	 * igt_display_init only sets sw state, and assumes the

first test doesn't care

tests/kms_atomic_interruptible.c new file +7 -0
+	 * about hw state.
+	 */
+	igt_display_commit2(display, COMMIT_ATOMIC);
+
+	igt_output_set_pipe(output, pipe);
+
+	primary = igt_output_get_plane_type(output,

DRM_PLANE_TYPE_PRIMARY);

tests/kms_atomic_interruptible.c new file +4 -0
+	plane = igt_output_get_plane_type(output, plane_type);
+	mode = igt_output_get_mode(output);
+
+	igt_create_fb(display->drm_fd, mode->hdisplay, mode-
Quoted 1 line

vdisplay,

tests/kms_atomic_interruptible.c new file +1 -0
+		      DRM_FORMAT_XRGB8888,

LOCAL_DRM_FORMAT_MOD_NONE, &fb);

tests/kms_atomic_interruptible.c new file +4 -0
+
+	switch (plane_type) {
+	case DRM_PLANE_TYPE_PRIMARY:
+		igt_create_fb(display->drm_fd, mode->hdisplay, mode-
Quoted 1 line

vdisplay,

tests/kms_atomic_interruptible.c new file +1 -0
+			      DRM_FORMAT_XRGB8888,

LOCAL_DRM_FORMAT_MOD_NONE, &fb2);

tests/kms_atomic_interruptible.c new file +4 -0
+		break;
+	case DRM_PLANE_TYPE_CURSOR:
+		igt_create_fb(display->drm_fd, 64, 64,
+		      DRM_FORMAT_ARGB8888,

LOCAL_DRM_FORMAT_MOD_NONE, &fb2);

tests/kms_atomic_interruptible.c

Quoted 2 lines
tests/kms_atomic_interruptible.c new file +2 -0
+		break;
+	}

We could squeeze this into a one line and get rid of the switch()? Anyway this is more matter of a taste and style than functionality.

Quoted 105 lines
tests/kms_atomic_interruptible.c new file +24 -0
+
+	if (test_type != test_setmode) {
+		igt_plane_set_fb(primary, &fb);
+		igt_display_commit2(display, COMMIT_ATOMIC);
+	}
+
+	igt_plane_set_fb(plane, &fb2);
+
+	block = block_plane(display, output, test_type, plane);
+	sleep(1);
+
+	igt_fork(child, 1) {
+		signal(SIGCONT, SIG_IGN);
+		igt_assert(sleep(5) == 0);
+
+		unblock(block);
+	}
+
+	/* run the test */
+	igt_fork_signal_helper();
+	if (test_type != test_pageflip)
+		igt_display_commit2(display, style);
+	else
+		do_or_die(drmModePageFlip(display->drm_fd, display-
Quoted 1 line

pipes[pipe].crtc_id, fb.fb_id, 0, NULL));

tests/kms_atomic_interruptible.c new file +4 -0
+	igt_stop_signal_helper();
+
+	/* Test whether we did interrupt the test? */
+	igt_assert_lt(2 * igt_signal_helper_get_hz(),

igt_signal_helper_get_num());

tests/kms_atomic_interruptible.c new file +32 -0
+
+	igt_waitchildren();
+
+	igt_plane_set_fb(primary, NULL);
+	igt_output_set_pipe(output, PIPE_NONE);
+	igt_remove_fb(display->drm_fd, &fb);
+}
+
+igt_main
+{
+	igt_display_t display;
+	igt_output_t *output;
+	enum pipe pipe;
+
+	igt_skip_on_simulation();
+
+	igt_fixture {
+		display.drm_fd = drm_open_driver_master(DRIVER_ANY);
+
+		kmstest_set_vt_graphics_mode();
+
+		igt_display_init(&display, display.drm_fd);
+
+		igt_require(display.is_atomic);
+
+		igt_display_require_output(&display);
+
+		igt_require_sw_sync();
+	}
+
+	igt_subtest("legacy-setmode")
+		for_each_pipe_with_valid_output(&display, pipe,

output) {

tests/kms_atomic_interruptible.c new file +1 -0
+			run_plane_test(&display, pipe, output,

test_setmode, DRM_PLANE_TYPE_PRIMARY, COMMIT_LEGACY);

tests/kms_atomic_interruptible.c new file +5 -0
+			break;
+		}
+
+	igt_subtest("atomic-setmode")
+		for_each_pipe_with_valid_output(&display, pipe,

output) {

tests/kms_atomic_interruptible.c new file +1 -0
+			run_plane_test(&display, pipe, output,

test_setmode, DRM_PLANE_TYPE_PRIMARY, COMMIT_ATOMIC);

tests/kms_atomic_interruptible.c new file +5 -0
+			break;
+		}
+
+	igt_subtest("legacy-pageflip")
+		for_each_pipe_with_valid_output(&display, pipe,

output) {

tests/kms_atomic_interruptible.c new file +1 -0
+			run_plane_test(&display, pipe, output,

test_pageflip, DRM_PLANE_TYPE_PRIMARY, COMMIT_LEGACY);

tests/kms_atomic_interruptible.c new file +5 -0
+			break;
+		}
+
+	igt_subtest("legacy-cursor")
+		for_each_pipe_with_valid_output(&display, pipe,

output) {

tests/kms_atomic_interruptible.c new file +1 -0
+			run_plane_test(&display, pipe, output,

test_setplane, DRM_PLANE_TYPE_CURSOR, COMMIT_LEGACY);

tests/kms_atomic_interruptible.c new file +5 -0
+			break;
+		}
+
+	igt_subtest("universal-setplane-primary")
+		for_each_pipe_with_valid_output(&display, pipe,

output) {

tests/kms_atomic_interruptible.c new file +1 -0
+			run_plane_test(&display, pipe, output,

test_setplane, DRM_PLANE_TYPE_PRIMARY, COMMIT_UNIVERSAL);

tests/kms_atomic_interruptible.c new file +5 -0
+			break;
+		}
+
+	igt_subtest("universal-setplane-cursor")
+		for_each_pipe_with_valid_output(&display, pipe,

output) {

tests/kms_atomic_interruptible.c new file +1 -0
+			run_plane_test(&display, pipe, output,

test_setplane, DRM_PLANE_TYPE_CURSOR, COMMIT_UNIVERSAL);

tests/kms_atomic_interruptible.c line 193

Quoted 8 lines
tests/kms_atomic_interruptible.c new file +8 -0
+			break;
+		}
+
+	/* TODO: DPMS, gamma_set, setprop, getprop */
+	igt_fixture {
+		igt_display_fini(&display);
+	}
+}

Otherwise looks good to me.

Signature · 6 lines
Mika Kahola - Intel OTC

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#7 lore raw

Re: [PATCH i-g-t 3/3] tests: Add kms_atomic_interruptible test.

Op 04-08-17 om 09:50 schreef Mika Kahola:

Quoted 151 lines · Maarten Lankhorst
Quoted 37 lines

To make sure that we have test exposure when allowing atomic ioctl's to
fail interruptibly, we add a test that will fail to lock the mutexes until the fence is signaled.

If the locking is done interruptibly, our signal helper will interrupt
often, and with the statistics we can ensure that we received enough interrupts to pass.

---
 tests/Makefile.sources           |   1 +
 tests/kms_atomic_interruptible.c | 193
+++++++++++++++++++++++++++++++++++++++
 2 files changed, 194 insertions(+)
 create mode 100644 tests/kms_atomic_interruptible.c
tests/Makefile.sources +1 -0
index 5b98a5a371b8..9a6a846ab78a 100644
@@ -171,6 +171,7 @@ TESTS_progs = \
 	kms_3d \
 	kms_addfb_basic \
 	kms_atomic \
+	kms_atomic_interruptible \
 	kms_atomic_transition \
 	kms_busy \
 	kms_ccs \
tests/kms_atomic_interruptible.c new file +0 -0
new file mode 100644
index 000000000000..4015988c0e2c

tests/kms_atomic_interruptible.c line 2

Quoted 2 lines
tests/kms_atomic_interruptible.c new file +2 -0
@@ -0,0 +1,193 @@
+/*
+ * Copyright © 2016 Intel Corporation

Update a year here?

Quoted 110 lines
tests/kms_atomic_interruptible.c new file +2 -0
+ *
+ * Permission is hereby granted, free of charge, to any person

obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */

tests/kms_atomic_interruptible.c new file +13 -0
+
+#include "igt.h"
+#include "drmtest.h"
+#include "sw_sync.h"
+
+enum plane_test_type
+{
+	test_setmode,
+	test_setplane,
+	test_pageflip
+};
+
+static int block_plane(igt_display_t *display, igt_output_t *output,

enum plane_test_type test_type, igt_plane_t *plane)

tests/kms_atomic_interruptible.c new file +5 -0
+{
+	int timeline = sw_sync_timeline_create();
+
+	igt_fork(child, 1) {
+		/* Ignore the signal helper, we need to block

indefinitely on the fence. */

tests/kms_atomic_interruptible.c new file +7 -0
+		signal(SIGCONT, SIG_IGN);
+
+		if (test_type == test_setmode) {
+			igt_output_set_pipe(output, PIPE_NONE);
+			igt_plane_set_fb(plane, NULL);
+		}
+		igt_plane_set_fence_fd(plane,

sw_sync_timeline_create_fence(timeline, 1));

tests/kms_atomic_interruptible.c new file +14 -0
+
+		igt_display_commit2(display, COMMIT_ATOMIC);
+	}
+
+	return timeline;
+}
+
+static void unblock(int block)
+{
+	sw_sync_timeline_inc(block, 1);
+	close(block);
+}
+
+static void run_plane_test(igt_display_t *display, enum pipe pipe,

igt_output_t *output,

tests/kms_atomic_interruptible.c new file +1 -0
+			   enum plane_test_type test_type, unsigned

plane_type, enum igt_commit_style style)

tests/kms_atomic_interruptible.c new file +8 -0
+{
+	drmModeModeInfo *mode;
+	igt_fb_t fb, fb2;
+	igt_plane_t *primary, *plane;
+	int block;
+
+	/*
+	 * Make sure we start with everything disabled to force a

real modeset.

tests/kms_atomic_interruptible.c new file +1 -0
+	 * igt_display_init only sets sw state, and assumes the

first test doesn't care

tests/kms_atomic_interruptible.c new file +7 -0
+	 * about hw state.
+	 */
+	igt_display_commit2(display, COMMIT_ATOMIC);
+
+	igt_output_set_pipe(output, pipe);
+
+	primary = igt_output_get_plane_type(output,

DRM_PLANE_TYPE_PRIMARY);

tests/kms_atomic_interruptible.c new file +4 -0
+	plane = igt_output_get_plane_type(output, plane_type);
+	mode = igt_output_get_mode(output);
+
+	igt_create_fb(display->drm_fd, mode->hdisplay, mode-
Quoted 1 line

vdisplay,

tests/kms_atomic_interruptible.c new file +1 -0
+		      DRM_FORMAT_XRGB8888,

LOCAL_DRM_FORMAT_MOD_NONE, &fb);

tests/kms_atomic_interruptible.c new file +4 -0
+
+	switch (plane_type) {
+	case DRM_PLANE_TYPE_PRIMARY:
+		igt_create_fb(display->drm_fd, mode->hdisplay, mode-
Quoted 1 line

vdisplay,

Quoted 11 lines

tests/kms_atomic_interruptible.c

Quoted 9 lines
tests/kms_atomic_interruptible.c new file +1 -0
+			      DRM_FORMAT_XRGB8888,

LOCAL_DRM_FORMAT_MOD_NONE, &fb2);

tests/kms_atomic_interruptible.c new file +4 -0
+		break;
+	case DRM_PLANE_TYPE_CURSOR:
+		igt_create_fb(display->drm_fd, 64, 64,
+		      DRM_FORMAT_ARGB8888,

LOCAL_DRM_FORMAT_MOD_NONE, &fb2);

tests/kms_atomic_interruptible.c new file +2 -0
+		break;
+	}

We could squeeze this into a one line and get rid of the switch()? Anyway this is more matter of a taste and style than functionality.

Was thinking of adding overlay, but that's probably overkill for now.

Quoted 106 lines
Quoted 106 lines
tests/kms_atomic_interruptible.c new file +24 -0
+
+	if (test_type != test_setmode) {
+		igt_plane_set_fb(primary, &fb);
+		igt_display_commit2(display, COMMIT_ATOMIC);
+	}
+
+	igt_plane_set_fb(plane, &fb2);
+
+	block = block_plane(display, output, test_type, plane);
+	sleep(1);
+
+	igt_fork(child, 1) {
+		signal(SIGCONT, SIG_IGN);
+		igt_assert(sleep(5) == 0);
+
+		unblock(block);
+	}
+
+	/* run the test */
+	igt_fork_signal_helper();
+	if (test_type != test_pageflip)
+		igt_display_commit2(display, style);
+	else
+		do_or_die(drmModePageFlip(display->drm_fd, display-
Quoted 1 line

pipes[pipe].crtc_id, fb.fb_id, 0, NULL));

tests/kms_atomic_interruptible.c new file +4 -0
+	igt_stop_signal_helper();
+
+	/* Test whether we did interrupt the test? */
+	igt_assert_lt(2 * igt_signal_helper_get_hz(),

igt_signal_helper_get_num());

tests/kms_atomic_interruptible.c new file +32 -0
+
+	igt_waitchildren();
+
+	igt_plane_set_fb(primary, NULL);
+	igt_output_set_pipe(output, PIPE_NONE);
+	igt_remove_fb(display->drm_fd, &fb);
+}
+
+igt_main
+{
+	igt_display_t display;
+	igt_output_t *output;
+	enum pipe pipe;
+
+	igt_skip_on_simulation();
+
+	igt_fixture {
+		display.drm_fd = drm_open_driver_master(DRIVER_ANY);
+
+		kmstest_set_vt_graphics_mode();
+
+		igt_display_init(&display, display.drm_fd);
+
+		igt_require(display.is_atomic);
+
+		igt_display_require_output(&display);
+
+		igt_require_sw_sync();
+	}
+
+	igt_subtest("legacy-setmode")
+		for_each_pipe_with_valid_output(&display, pipe,

output) {

tests/kms_atomic_interruptible.c new file +1 -0
+			run_plane_test(&display, pipe, output,

test_setmode, DRM_PLANE_TYPE_PRIMARY, COMMIT_LEGACY);

tests/kms_atomic_interruptible.c new file +5 -0
+			break;
+		}
+
+	igt_subtest("atomic-setmode")
+		for_each_pipe_with_valid_output(&display, pipe,

output) {

tests/kms_atomic_interruptible.c new file +1 -0
+			run_plane_test(&display, pipe, output,

test_setmode, DRM_PLANE_TYPE_PRIMARY, COMMIT_ATOMIC);

tests/kms_atomic_interruptible.c new file +5 -0
+			break;
+		}
+
+	igt_subtest("legacy-pageflip")
+		for_each_pipe_with_valid_output(&display, pipe,

output) {

tests/kms_atomic_interruptible.c new file +1 -0
+			run_plane_test(&display, pipe, output,

test_pageflip, DRM_PLANE_TYPE_PRIMARY, COMMIT_LEGACY);

tests/kms_atomic_interruptible.c new file +5 -0
+			break;
+		}
+
+	igt_subtest("legacy-cursor")
+		for_each_pipe_with_valid_output(&display, pipe,

output) {

tests/kms_atomic_interruptible.c new file +1 -0
+			run_plane_test(&display, pipe, output,

test_setplane, DRM_PLANE_TYPE_CURSOR, COMMIT_LEGACY);

tests/kms_atomic_interruptible.c new file +5 -0
+			break;
+		}
+
+	igt_subtest("universal-setplane-primary")
+		for_each_pipe_with_valid_output(&display, pipe,

output) {

tests/kms_atomic_interruptible.c new file +1 -0
+			run_plane_test(&display, pipe, output,

test_setplane, DRM_PLANE_TYPE_PRIMARY, COMMIT_UNIVERSAL);

tests/kms_atomic_interruptible.c new file +5 -0
+			break;
+		}
+
+	igt_subtest("universal-setplane-cursor")
+		for_each_pipe_with_valid_output(&display, pipe,

output) {

tests/kms_atomic_interruptible.c new file +1 -0
+			run_plane_test(&display, pipe, output,

test_setplane, DRM_PLANE_TYPE_CURSOR, COMMIT_UNIVERSAL);

tests/kms_atomic_interruptible.c new file +1 -0
+			break;
Quoted 9 lines

tests/kms_atomic_interruptible.c line 193

Quoted 7 lines
tests/kms_atomic_interruptible.c new file +7 -0
+		}
+
+	/* TODO: DPMS, gamma_set, setprop, getprop */
+	igt_fixture {
+		igt_display_fini(&display);
+	}
+}

Otherwise looks good to me.

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#8 lore raw
In reply to Maarten Lankhorst
Quoted 70 lines · Maarten Lankhorst

Export 2 functions, igt_signal_helper_get_num and
igt_signal_helper_get_hz.

This will allow tests to measure how much time in a test was spent in a uninterruptible state, which is useful when testing whether certain ioctl's can be interrupted or not.

---
 lib/igt_aux.c | 30 +++++++++++++++++++++++++++--- lib/igt_aux.h |  2 ++

 2 files changed, 29 insertions(+), 3 deletions(-)
lib/igt_aux.c +0 -0
index 86a213c2032f..265e43f399e7 100644
@@ -275,7 +275,7 @@ bool __igt_sigiter_continue(struct __igt_sigiter
 *iter, bool enable)
  }


 static struct igt_helper_process signal_helper;

lib/igt_aux.c +1 -1
-long long int sig_stat;
+static int64_t sig_stat;

 static void __attribute__((noreturn)) signal_helper_process(pid_t pid)
 {
  /* Interrupt the parent process at 500Hz, just to be annoying */

lib/igt_aux.c +0 -0
@@ -314,6 +314,9 @@ void igt_fork_signal_helper(void)
  	if (igt_only_list_subtests())
  		return;

lib/igt_aux.c +3 -0
+	/* Reset number of signalscaught */
+	sig_stat = 0;
+

  /* We pick SIGCONT as it is a "safe" signal - if we send SIGCONT to
   * an unexpecting process it spuriously wakes up and does nothing.
   * Most other signals (e.g. SIGUSR1) cause the process to die if they

lib/igt_aux.c +0 -0
@@ -348,8 +351,29 @@ void igt_stop_signal_helper(void)
  		return;


  igt_stop_helper(&signal_helper);
+}

lib/igt_aux.c +21 -1
-	sig_stat = 0;
+/**
+ * igt_signal_helper_get_num:
+ *
+ * Return the amount of signals generated since the last time
+ * igt_fork_signal_helper() was called.
+ *
+ * This is reset to 0 on every call to igt_fork_signal_helper.
+ */
+int64_t igt_signal_helper_get_num(void)
+{
+	return sig_stat;
+}
+
+/**
+ * igt_signal_helper_get_hz:
+ *
+ * Return the approximate amount of signals generated per second.
+ */
+int igt_signal_helper_get_hz(void)
+{
+	return 50;
Quoted 1 line

 }

I wonder if this is really necessary? The function returns hardcoded value and it is only used in one place i.e. third patch of this series.

Quoted 25 lines


 static struct igt_helper_process shrink_helper;

lib/igt_aux.c +1 -1
@@ -357,7 +381,7 @@ static void __attribute__((noreturn))
 shrink_helper_process(int fd, pid_t pid)
  {
  	while (1) {
  		igt_drop_caches_set(fd, DROP_SHRINK_ALL);
-		usleep(1000 * 1000 / 50);
+		usleep(1000 * 1000 / igt_signal_helper_get_hz());

  if (kill(pid, 0)) /* Parent has died, so must we. */ exit(0);
  }

lib/igt_aux.h +2 -0
index 499a16796ebb..7e080089dcbc 100644
@@ -55,6 +55,8 @@ extern int num_trash_bos;
  /* generally useful helpers */
  void igt_fork_signal_helper(void);
  void igt_stop_signal_helper(void);
+int64_t igt_signal_helper_get_num(void);
+int igt_signal_helper_get_hz(void);


 void igt_fork_shrink_helper(int fd);
 void igt_stop_shrink_helper(void);

Signature · 6 lines
Mika Kahola - Intel OTC

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#9 lore raw
In reply to Maarten Lankhorst

Quoting Maarten Lankhorst (2017-08-02 11:29:17)

Quoted 6 lines

Export 2 functions, igt_signal_helper_get_num and
igt_signal_helper_get_hz.

This will allow tests to measure how much time in a test was spent in a uninterruptible state, which is useful when testing whether certain ioctl's can be interrupted or not.

Use sig_ioctl, the purpose of that wrapper is to measure whether or not it was interrupted. Refine it to suit your purposes, it is the surgical scalpel compared to the shotgun of signal_helper.
-Chris
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#10 lore raw

Op 04-08-17 om 09:50 schreef Chris Wilson:

Quoted 11 lines

Quoting Maarten Lankhorst (2017-08-02 11:29:17)

Quoted 6 lines

Export 2 functions, igt_signal_helper_get_num and
igt_signal_helper_get_hz.

This will allow tests to measure how much time in a test was spent in a uninterruptible state, which is useful when testing whether certain ioctl's can be interrupted or not.

Use sig_ioctl, the purpose of that wrapper is to measure whether or not it was interrupted. Refine it to suit your purposes, it is the surgical scalpel compared to the shotgun of signal_helper.
-Chris

I've been using the igt display helpers now so I don't have to worry about the ioctl's. Using sig_ioctl would mean having to perform the ioctl's myself, and that makes the code a lot more verbose..

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#11 lore raw

Quoting Maarten Lankhorst (2017-08-07 10:45:38)

Quoted 8 lines

Op 04-08-17 om 09:50 schreef Chris Wilson:

Quoted 7 lines

Quoting Maarten Lankhorst (2017-08-02 11:29:17)

Quoted 6 lines

Export 2 functions, igt_signal_helper_get_num and
igt_signal_helper_get_hz.

This will allow tests to measure how much time in a test was spent in a uninterruptible state, which is useful when testing whether certain ioctl's can be interrupted or not.

Quoted 6 lines
Quoted 4 lines

Use sig_ioctl, the purpose of that wrapper is to measure whether or not it was interrupted. Refine it to suit your purposes, it is the surgical scalpel compared to the shotgun of signal_helper.
-Chris

I've been using the igt display helpers now so I don't have to worry about the ioctl's. Using sig_ioctl would mean having to perform the ioctl's myself, and that makes the code a lot more verbose..

You know that we igt uses igt_ioctl so that we can switch the ioctl wrapper. What you have here is sig_ioctl so rather than painting the shotgun signal helper a different colour, use the interface that actually supports what you have in mind.
-Chris

Quoted 1 line

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#12 lore raw
Quoted 14 lines · Chris Wilson

Quoting Maarten Lankhorst (2017-08-07 10:45:38)

Quoted 13 lines

Op 04-08-17 om 09:50 schreef Chris Wilson:

Quoted 11 lines

Quoting Maarten Lankhorst (2017-08-02 11:29:17)

Quoted 6 lines

Export 2 functions, igt_signal_helper_get_num and
igt_signal_helper_get_hz.

This will allow tests to measure how much time in a test was spent in a uninterruptible state, which is useful when testing whether certain ioctl's can be interrupted or not.

Use sig_ioctl, the purpose of that wrapper is to measure whether or not it was interrupted. Refine it to suit your purposes, it is the surgical scalpel compared to the shotgun of signal_helper.
-Chris

Quoted 6 lines
Quoted 1 line

I've been using the igt display helpers now so I don't have to worry about the ioctl's. Using sig_ioctl would mean having to perform the ioctl's myself, and that makes the code a lot more verbose..

You know that we igt uses igt_ioctl so that we can switch the ioctl wrapper. What you have here is sig_ioctl so rather than painting the shotgun signal helper a different colour, use the interface that actually supports what you have in mind.

Might be good to spicy the docs up a bit to make this clearer, maybe even with an example and some links between the different bits? Mika/Maarten? -Daniel

Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx