[igt-dev] [i-g-t] [PATCH v1 1/1] tests/xe/xe_uevent: add new test for uevent gt reset failure

[Kumar, Janga Rahul]

LGTM,
Reviewed-by: Janga Rahul Kumar <janga.rahul.kumar at intel.com>

> -----Original Message-----
> From: Ghimiray, Himal Prasad <himal.prasad.ghimiray at intel.com>
> Sent: 23 June 2023 17:21
> To: igt-dev at lists.freedesktop.org
> Cc: Kamil Konieczny <kamil.konieczny at linux.intel.com>; Kumar, Janga Rahul
> <janga.rahul.kumar at intel.com>; Dugast, Francois
> <francois.dugast at intel.com>; Upadhyay, Tejas <tejas.upadhyay at intel.com>;
> Iddamsetty, Aravind <aravind.iddamsetty at intel.com>
> Subject: RE: [i-g-t] [PATCH v1 1/1] tests/xe/xe_uevent: add new test for uevent
> gt reset failure
> 
> 
> 
> > -----Original Message-----
> > From: Ghimiray, Himal Prasad <himal.prasad.ghimiray at intel.com>
> > Sent: 15 June 2023 18:38
> > To: igt-dev at lists.freedesktop.org
> > Cc: Ghimiray, Himal Prasad <himal.prasad.ghimiray at intel.com>; Kamil
> > Konieczny <kamil.konieczny at linux.intel.com>; Kumar, Janga Rahul
> > <janga.rahul.kumar at intel.com>; Dugast, Francois
> > <francois.dugast at intel.com>
> > Subject: [i-g-t] [PATCH v1 1/1] tests/xe/xe_uevent: add new test for
> > uevent gt reset failure
> >
> > This test is to cause the fake reset failure and capture the uevent
> > sent in case of gt reset failure. We are relying on debugfs to cause
> > fake reset failure because there is no hardware mechanism using which
> > we can force gt to fail its reset.
> >
> > The sending of Uevent is taken care by series:
> > drm/xe: Notify Userspace when engine/gt reset fails.
> > https://patchwork.freedesktop.org/series/117504/
> >
> > The infra to cause fake reset failure is in series:
> > drm/xe: Add a debugfs for faking gt reset failure.
> > https://patchwork.freedesktop.org/series/119375/
> >
> > v2:
> > - Make fake reset for all GT's in platform.
> > - Add test names in order (Rahul)
> > - Add headers in order.
> > - Instead of system call use igt helper for reading debugfs.
> > - Use igt_debug instead of prints.
> > - Use break instead of unnecessary goto.
> > - Change test name.
> > - Define values as constants. (Kamil)
> >
> > Cc: Kamil Konieczny <kamil.konieczny at linux.intel.com>
> > Cc: Janga Rahul Kumar <janga.rahul.kumar at intel.com>
> > Cc: Francois Dugast <francois.dugast at intel.com>
> > Signed-off-by: Himal Prasad Ghimiray <himal.prasad.ghimiray at intel.com>
> > ---
> >  tests/meson.build    |   1 +
> >  tests/xe/xe_uevent.c | 129
> > +++++++++++++++++++++++++++++++++++++++++++
> >  2 files changed, 130 insertions(+)
> >  create mode 100644 tests/xe/xe_uevent.c
> >
> > diff --git a/tests/meson.build b/tests/meson.build index
> > 61dcc076..27abc13f
> > 100644
> > --- a/tests/meson.build
> > +++ b/tests/meson.build
> > @@ -267,6 +267,7 @@ xe_progs = [
> >  	'xe_pm',
> >  	'xe_prime_self_import',
> >  	'xe_query',
> > +	'xe_uevent',
> >  	'xe_vm',
> >  	'xe_waitfence',
> >  ]
> > diff --git a/tests/xe/xe_uevent.c b/tests/xe/xe_uevent.c new file mode
> > 100644 index 00000000..6326896d
> > --- /dev/null
> > +++ b/tests/xe/xe_uevent.c
> > @@ -0,0 +1,129 @@
> > +// 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 <libudev.h>
> > +#include <string.h>
> > +#include <sys/stat.h>
> > +
> > +#include "igt.h"
> > +
> > +#include "xe_drm.h"
> > +#include "xe/xe_query.h"
> > +
> > +static bool xe_force_gt_fake_reset(int fd, int gt) {
> > +	char rst_fail[32];
> > +
> > +	snprintf(rst_fail, sizeof(rst_fail), "gt%d/fake_reset_failure", gt);
> > +	__igt_debugfs_read(fd, rst_fail, rst_fail, sizeof(rst_fail));
> > +
> > +	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")) {
> > +			igt_debug("%s = %s\n", name, val);
> > +			reset_failed = true;
> > +			continue;
> > +		}
> > +
> > +		if (!strcmp(name, "RESET_UNIT") && !strcmp(val, "gt")) {
> > +			igt_debug("%s = %s\n", name, val);
> > +			reset_unit_is_gt = true;
> > +			continue;
> > +		}
> > +
> > +		if (!strcmp(name, "RESET_ID") && (atoi(val) == gt_id)) {
> > +			igt_debug("%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_received = false;
> > +	bool event_sent = false;
> > +	const u32 listener_timeout = 5;
> > +	const u32 reset_completed = 50000;
> > +
> > +	/* create udev object */
> > +	udev = udev_new();
> > +	if (!udev)
> > +		igt_assert_f(false, "New udev object creation failed");
> > +
> > +	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(listener_timeout) {
> > +			dev = udev_monitor_receive_device(mon);
> > +
> > +			if (dev) {
> > +				event_received =
> > listen_reset_fail_uevent(dev, "kernel", gt_id);
> > +				udev_device_unref(dev);
> > +			}
> > +
> > +			if (event_sent)
> > +				break;
> > +
> > +		event_sent = xe_force_gt_fake_reset(fd, gt_id);
> > +		usleep(reset_completed);
> > +	}
> > +	udev_unref(udev);
> > +	igt_assert_f(event_received, "Event not received"); }
> > +
> > +igt_main
> > +{
> > +	int fd;
> > +	int gt;
> > +	const u32 settle_xe_load_uevents = 50000;
> > +
> > +	igt_fixture {
> > +		fd = drm_open_driver(DRIVER_XE);
> > +		xe_device_get(fd);
> > +	}
> > +
> > +	/* Ensures uevents triggered in case of driver
> > +	 * load are settled down.
> > +	 */
> > +	usleep(settle_xe_load_uevents);
> > +
> > +	igt_subtest("fake_reset_uevent_listener")
> > +		xe_for_each_gt(fd, gt) {
> > +			fake_reset_uevent_listener(fd, gt);
> > +		}
> > +
> > +	igt_fixture {
> > +		xe_device_put(fd);
> > +		close(fd);
> > +	}
> > +}
> > --
> > 2.25.1