[igt-dev] [i-g-t, 1/1] Uevent listener for fake gt reset failure.

[Ghimiray, Himal Prasad]

> -----Original Message-----
> From: Kumar, Janga Rahul <janga.rahul.kumar at intel.com>
> Sent: 09 June 2023 14:35
> To: Ghimiray, Himal Prasad <himal.prasad.ghimiray at intel.com>; igt-
> dev at lists.freedesktop.org
> Cc: Ghimiray, Himal Prasad <himal.prasad.ghimiray at intel.com>; Gandi,
> Ramadevi <ramadevi.gandi at intel.com>
> Subject: RE: [igt-dev] [i-g-t, 1/1] Uevent listener for fake gt reset failure.
> 
> 
> 
> > -----Original Message-----
> > From: igt-dev <igt-dev-bounces at lists.freedesktop.org> On Behalf Of
> > Himal Prasad Ghimiray
> > Sent: 05 June 2023 11:21
> > To: igt-dev at lists.freedesktop.org
> > Cc: Ghimiray, Himal Prasad <himal.prasad.ghimiray at intel.com>
> > Subject: [igt-dev] [i-g-t, 1/1] Uevent listener for fake gt reset failure.
> >
> > This test is to cause the fake reset failure and capture the uevent
> > sent in case of gt reset failure.
> >
> > The sending of Uevent is taken care by series:
> > drm/xe: Notify Userspace when engine/gt reset fails.
> >
> > The infra to cause fake reset failure is in series:
> > drm/xe: Add a debugfs for faking gt reset failure.
> >
> > Signed-off-by: Himal Prasad Ghimiray <himal.prasad.ghimiray at intel.com>
> > ---
> >  tests/meson.build                             |   1 +
> >  .../xe_uevent_listener_fake_reset_failure.c   | 128 ++++++++++++++++++
> >  2 files changed, 129 insertions(+)
> >  create mode 100644 tests/xe/xe_uevent_listener_fake_reset_failure.c
> >
> > diff --git a/tests/meson.build b/tests/meson.build index
> > f71be1db..cc5843d7
> > 100644
> > --- a/tests/meson.build
> > +++ b/tests/meson.build
> > @@ -268,6 +268,7 @@ xe_progs = [
> >  	'xe_query',
> >  	'xe_vm',
> >  	'xe_waitfence',
> > +	'xe_uevent_listener_fake_reset_failure',
> Please maintain above list in alphabetically sorted order
Sure. Will do in next patch. 
> >  ]
> >
> >  msm_progs = [
> > diff --git a/tests/xe/xe_uevent_listener_fake_reset_failure.c
> > b/tests/xe/xe_uevent_listener_fake_reset_failure.c
> > new file mode 100644
> > index 00000000..dd22e4c2
> > --- /dev/null
> > +++ b/tests/xe/xe_uevent_listener_fake_reset_failure.c
> > @@ -0,0 +1,128 @@
> > +// SPDX-License-Identifier: MIT
> > +/*
> > + * Copyright © 2023 Intel Corporation  */
> > +
> > +/**
> > + * TEST: cause fake gt reset failure and listen uevent from KMD
> > + * SUBTEST:fake_reset_uevent_listener
> > + * Description:
> > + *		Test creates uevent listener and causes fake reset failure for
> gt0
> > + *		and returns success if uevent is sent by driver and listened by
> > listener.
> > + */
> > +
> > +#include "igt.h"
> > +#include "xe_drm.h"
> > +#include "xe/xe_query.h"
> > +#include <string.h>
> > +#include <libudev.h>
> > +#include <sys/stat.h>
> > +
> > +static bool xe_force_gt_fake_reset(int fd, int gt) {
> Move this function to lib

This function shouldn't be used by any other test cases.
If in future we find any reasons to cause fake reset failure will move it to lib.
> > +	char reset_string[128];
> > +	struct stat st;
> > +
> > +	igt_assert_eq(fstat(fd, &st), 0);
> > +	snprintf(reset_string, sizeof(reset_string),
> > +		 "cat /sys/kernel/debug/dri/%d/gt%d/fake_reset_failure",
> > +minor(st.st_rdev), gt);
> > +
> > +	if (system(reset_string))
> > +		return false;
> > +
> > +	return true;
> > +}
> > +
> > +static bool listen_reset_fail_uevent(struct udev_device *device,
> > +const char *source, int gt_id) {
> > +	struct udev_list_entry *list_entry;
> > +	bool reset_failed = false;
> > +	bool reset_unit_is_gt = false;
> > +	bool gt_id_matches = false;
> > +	const char *name, *val;
> > +
> > +	udev_list_entry_foreach(list_entry,
> > udev_device_get_properties_list_entry(device))
> > +	{
> > +		name = udev_list_entry_get_name(list_entry);
> > +		val = udev_list_entry_get_value(list_entry);
> > +
> > +		if (!strcmp(name, "RESET_FAILED") && !strcmp(val, "1")) {
> > +			printf("%s = %s\n", name, val);
> > +			reset_failed = true;
> > +			continue;
> > +		}
> > +
> > +		if (!strcmp(name, "RESET_UNIT") && !strcmp(val, "gt")) {
> > +			printf("%s = %s\n", name, val);
> > +			reset_unit_is_gt = true;
> > +			continue;
> > +		}
> > +
> > +		if (!strcmp(name, "RESET_ID") && (atoi(val) == gt_id)) {
> > +			printf("%s = %s\n", name, val);
> > +			gt_id_matches = true;
> > +			continue;
> > +		}
> > +	}
> > +
> > +	return (reset_failed && reset_unit_is_gt && gt_id_matches); }
> > +
> > +static void fake_reset_uevent_listener(int fd, int gt_id) {
> > +	struct udev *udev;
> > +	struct udev_device *dev;
> > +	struct udev_monitor *mon;
> > +	bool event_recieved = false;
> > +	bool event_sent = false;
> > +
> > +	/* create udev object */
> > +	udev = udev_new();
> > +	if (!udev) {
> > +		fprintf(stderr, "Can't create udev\n");
> > +		goto no_udev;
> > +	}
> > +
> > +	mon = udev_monitor_new_from_netlink(udev, "kernel");
> > +	udev_monitor_filter_add_match_subsystem_devtype(mon, "drm",
> > "drm_minor");
> > +	udev_monitor_enable_receiving(mon);
> > +	igt_until_timeout(5) {
> > +			dev = udev_monitor_receive_device(mon);
> > +
> > +			if (dev) {
> > +				event_recieved =
> listen_reset_fail_uevent(dev,
> > "kernel", gt_id);
> > +				udev_device_unref(dev);
> > +			}
> > +
> > +			if (event_sent)
> > +				goto free_udev;
> > +
> > +		event_sent = xe_force_gt_fake_reset(fd, gt_id);
> > +		usleep(50 * 1000);
> > +	}
> > +
> > +free_udev:
> > +	udev_unref(udev);
> > +no_udev:
> > +	igt_assert(event_recieved);
> > +}
> > +
> > +igt_main
> > +{
> > +	int fd;
> > +
> > +	igt_fixture {
> > +		fd = drm_open_driver(DRIVER_XE);
> > +		xe_device_get(fd);
> > +	}
> > +
> > +	usleep(50 * 1000);
> > +
> > +	igt_subtest("fake_reset_uevent_listener")
> > +		fake_reset_uevent_listener(fd, 0);
> Check for all the available GT's available on the testing platform.
IMO requirement is to test whether uevents are coming from KMD in case of gt reset failure or not.
So testing one of the GT should be sufficient. 
But if recommendation is to do it for all gt, that can be done. 
> 
> 
> Thanks,
> Rahul
> > +
> > +	igt_fixture {
> > +		xe_device_put(fd);
> > +		close(fd);
> > +	}
> > +}
> > --
> > 2.25.1