[igt-dev] [RFC, i-g-t, v2] tests/device_reset: Test device sysfs reset
Bernatowicz, Marcin
marcin.bernatowicz at intel.com
Fri Jul 3 09:05:30 UTC 2020
On Wed, 2020-07-01 at 10:08 +0200, Janusz Krzysztofik wrote:
> Hi Marcin,
>
> On Mon, 2020-06-29 at 11:23 +0200, Marcin Bernatowicz wrote:
> > Device reset is initiated by writing "1" to reset sysfs file,
> > which should initiate device FLR if supported by device.
> >
> > Test scenarios:
> > 1. unbind driver from device, initiate sysfs reset, rebind driver
> > to
> > device
> > 2. device reset with bound driver
> >
> > v2: removed unbind-rebind (duplicates core_hotunplug at unbind-rebind)
> > added healthcheck to each test (Janusz)
> >
> > Signed-off-by: Marcin Bernatowicz <marcin.bernatowicz at intel.com>
> > ---
> > tests/device_reset.c | 263
> > +++++++++++++++++++++++++++++++++++++++++++
> > tests/meson.build | 1 +
> > 2 files changed, 264 insertions(+)
> > create mode 100644 tests/device_reset.c
> >
> > diff --git a/tests/device_reset.c b/tests/device_reset.c
> > new file mode 100644
> > index 000000000..0e1e9e3aa
> > --- /dev/null
> > +++ b/tests/device_reset.c
> > @@ -0,0 +1,263 @@
> > +// SPDX-License-Identifier: MIT
> > +/*
> > + * Copyright(c) 2020 Intel Corporation. All rights reserved.
> > + */
> > +
> > +#include <fcntl.h>
> > +#include <sys/stat.h>
> > +#include "igt.h"
> > +#include "igt_device_scan.h"
> > +#include "igt_sysfs.h"
> > +
> > +IGT_TEST_DESCRIPTION("Examine behavior of a driver on device sysfs
> > reset");
> > +
> > +
> > +#define DEV_PATH_LEN 80
> > +#define DEV_BUS_ADDR_LEN 13 /* addr has form 0000:00:00.0 */
> > +
> > +/**
> > + * Helper structure containing sysfs file descriptors
> > + * and path related to tested device
> > + */
> > +struct sysfs_fds {
> > + struct {
> > + int dev;
>
> NIT: In your code, sysfs.fds.dev doesn't correspond to a sysfs node,
> only to a device special file under /dev. Then, 'sysfs_fds' as the
> name of the structure type and 'sysfs' as the name of its instance
> may
> be misleading in regard to the fds.dev member.
renamed to device_fds
>
> > + int dev_dir;
> > + int drv_dir;
> > + } fds;
> > + char dev_bus_addr[DEV_BUS_ADDR_LEN];
> > +};
> > +
> > +static int __open_sysfs_dir(int fd, const char* path)
> > +{
> > + int sysfs;
> > +
> > + sysfs = igt_sysfs_open(fd);
> > + if (sysfs < 0) {
> > + return -1;
> > + }
> > +
> > + return openat(sysfs, path, O_DIRECTORY);
>
> Can we take care of closing the 'sysfs' fd before returning?
thanks, corrected in rev3
>
> > +}
> > +
> > +static int open_device_sysfs_dir(int fd)
> > +{
> > + return __open_sysfs_dir(fd, "device");
> > +}
> > +
> > +static int open_driver_sysfs_dir(int fd)
> > +{
> > + return __open_sysfs_dir(fd, "device/driver");
> > +}
> > +
> > +/**
> > + * device_sysfs_path:
> > + * @fd: opened device file descriptor
> > + * @path: buffer to store sysfs path to device directory
> > + *
> > + * Returns:
> > + * On successfull path resolution sysfs path to device directory,
> > + * NULL otherwise
> > + */
> > +static char *device_sysfs_path(int fd, char *path)
> > +{
> > + char sysfs[DEV_PATH_LEN];
> > +
> > + if (!igt_sysfs_path(fd, sysfs, sizeof(sysfs)))
> > + return NULL;
> > +
> > + if (DEV_PATH_LEN <= (strlen(sysfs) + strlen("/device")))
> > + return NULL;
> > +
> > + strcat(sysfs, "/device");
> > +
> > + return realpath(sysfs, path);
> > +}
> > +
> > +static void init_sysfs_fds(struct sysfs_fds *sysfs)
> > +{
> > + char dev_path[PATH_MAX];
> > + char *addr_pos;
> > +
> > + igt_debug("open device\n");
> > + /**
> > + * As subtests must be able to close examined devices
> > + * completely, don't use drm_open_driver() as it keeps
> > + * a device file descriptor open for exit handler use.
> > + */
> > + sysfs->fds.dev = __drm_open_driver(DRIVER_ANY);
> > + igt_assert_fd(sysfs->fds.dev);
> > +
> > + igt_assert(device_sysfs_path(sysfs->fds.dev, dev_path));
> > + addr_pos = strrchr(dev_path, '/');
> > + snprintf(sysfs->dev_bus_addr, sizeof(sysfs->dev_bus_addr),
> > + "%s", (addr_pos ? addr_pos + 1 : ""));
>
> NIT: Couldn't we assert non-NULL addr_pos before using it instead of
> using this conditional expression?
> > + igt_assert(sysfs->dev_bus_addr[0]);
>
> NIT: Couldn't we assert correctness of return value of snprintf()
> instead?
> > +
> > + sysfs->fds.dev_dir = open_device_sysfs_dir(sysfs->fds.dev);
> > + igt_assert_fd(sysfs->fds.dev_dir);
> > +
> > + sysfs->fds.drv_dir = open_driver_sysfs_dir(sysfs->fds.dev);
> > + igt_assert_fd(sysfs->fds.drv_dir);
> > +}
> > +
> > +static int close_if_opened(int *fd)
> > +{
> > + int rc = 0;
>
> Could we have declarations separated here from the code with a blank
> line?
>
> > + if (fd && *fd != -1) {
> > + rc = close(*fd);
> > + *fd = -1;
> > + }
> > + return rc;
> > +}
> > +
> > +static void cleanup_sysfs_fds(struct sysfs_fds *sysfs)
> > +{
> > + close_if_opened(&sysfs->fds.dev);
> > + close_if_opened(&sysfs->fds.dev_dir);
> > + close_if_opened(&sysfs->fds.drv_dir);
>
> NIT: There is an igt_ignore_warn() helper which you may want to use
> for
> preventively suppressing potential compiler warnings on ignored
> return
> values of non-void functions.
>
> > + sysfs->dev_bus_addr[0] = '\0';
>
> Is cleanup of dev_bus_addr content really needed?
>
> > +}
> > +
> > +/**
> > + * is_sysfs_reset_supported:
> > + * @fd: opened device file descriptor
> > + *
> > + * Check if device supports reset based on sysfs file presence.
> > + *
> > + * Returns:
> > + * True if device supports reset, false otherwise.
> > + */
> > +static bool is_sysfs_reset_supported(int fd)
> > +{
> > + struct stat st;
> > + int rc;
> > + int sysfs;
> > + int reset_fd = -1;
> > +
> > + sysfs = igt_sysfs_open(fd);
> > +
> > + if (sysfs >= 0) {
> > + reset_fd = openat(sysfs, "device/reset", O_WRONLY);
> > + close(sysfs);
> > + }
> > +
> > + if (reset_fd < 0)
> > + return false;
> > +
> > + rc = fstat(reset_fd, &st);
> > + close(reset_fd);
> > +
> > + if (rc || !S_ISREG(st.st_mode))
> > + return false;
> > +
> > + return true;
> > +}
> > +
> > +/* Unbind the driver from the device */
> > +static void driver_unbind(struct sysfs_fds *sysfs)
> > +{
> > + igt_debug("unbind the driver from the device\n");
> > + igt_assert(igt_sysfs_set(sysfs->fds.drv_dir, "unbind",
> > + sysfs->dev_bus_addr));
> > +}
> > +
> > +/* Re-bind the driver to the device */
> > +static void driver_bind(struct sysfs_fds *sysfs)
> > +{
> > + igt_debug("rebind the driver to the device\n");
> > + igt_abort_on_f(!igt_sysfs_set(sysfs->fds.drv_dir, "bind",
> > + sysfs->dev_bus_addr), "driver rebind failed");
> > +}
> > +
> > +/* Initiate device reset */
> > +static void initiate_device_reset(struct sysfs_fds *sysfs)
> > +{
> > + igt_debug("reset device\n");
> > + igt_assert(igt_sysfs_set(sysfs->fds.dev_dir, "reset", "1"));
> > +}
> > +
> > +/**
> > + * healthcheck:
> > + * @sysfs: structure with device descriptor, if descriptor equals
> > -1
> > + * the device is reopened
> > + */
> > +static void healthcheck(struct sysfs_fds *sysfs)
> > +{
> > + if (sysfs->fds.dev == -1) {
> > + /* refresh device list */
> > + igt_devices_scan(true);
> > + igt_debug("reopen the device\n");
> > + sysfs->fds.dev = __drm_open_driver(DRIVER_ANY);
> > + }
> > + igt_assert_fd(sysfs->fds.dev);
> > +
> > + if (is_i915_device(sysfs->fds.dev))
> > + gem_test_engine(sysfs->fds.dev, ALL_ENGINES);
> > +}
> > +
> > +/**
> > + * set_device_filter:
> > + *
> > + * Sets device filter to ensure subtests always reopen the same
> > device
> > + *
> > + * @dev_path: path to device under tests
> > + */
> > +static void set_device_filter(const char* dev_path)
> > +{
> > + char filter[strlen("sys:") + strlen(dev_path) + 1];
>
> Can we avoiding variable-length arrays?
>
> Other than that, LGTM.
>
> Thanks,
> Janusz
>
I've corrected the issues and send updated version
Thanks,
Marcin
>
> > +
> > + snprintf(filter, sizeof(filter), "sys:%s", dev_path);
> > + igt_device_filter_free_all();
> > + igt_assert_eq(igt_device_filter_add(filter), 1);
> > +}
> > +
> > +static void unbind_reset_rebind(struct sysfs_fds *sysfs)
> > +{
> > + igt_debug("close the device\n");
> > + close_if_opened(&sysfs->fds.dev);
> > +
> > + driver_unbind(sysfs);
> > +
> > + initiate_device_reset(sysfs);
> > +
> > + driver_bind(sysfs);
> > +}
> > +
> > +igt_main
> > +{
> > + struct sysfs_fds sysfs;
> > +
> > + igt_fixture {
> > + char dev_path[PATH_MAX];
> > +
> > + igt_debug("opening device\n");
> > + init_sysfs_fds(&sysfs);
> > +
> > + /* Make sure subtests always reopen the same device */
> > + igt_assert(device_sysfs_path(sysfs.fds.dev, dev_path));
> > + set_device_filter(dev_path);
> > +
> > + igt_skip_on(!is_sysfs_reset_supported(sysfs.fds.dev));
> > +
> > + igt_set_timeout(60, "device reset tests timed out after
> > 60s");
> > + }
> > +
> > + igt_describe("Unbinds driver from device, initiates reset"
> > + " then rebinds driver to device");
> > + igt_subtest("unbind-reset-rebind") {
> > + unbind_reset_rebind(&sysfs);
> > + healthcheck(&sysfs);
> > + }
> > +
> > + igt_describe("Resets device with bound driver");
> > + igt_subtest("reset-bound") {
> > + initiate_device_reset(&sysfs);
> > + healthcheck(&sysfs);
> > + }
> > +
> > + igt_fixture {
> > + igt_reset_timeout();
> > + cleanup_sysfs_fds(&sysfs);
> > + }
> > +}
> > diff --git a/tests/meson.build b/tests/meson.build
> > index 28091794f..b0acdf7c0 100644
> > --- a/tests/meson.build
> > +++ b/tests/meson.build
> > @@ -8,6 +8,7 @@ test_progs = [
> > 'core_setmaster_vs_auth',
> > 'debugfs_test',
> > 'dmabuf',
> > + 'device_reset',
> > 'drm_import_export',
> > 'drm_mm',
> > 'drm_read',
>
>
---------------------------------------------------------------------
Intel Technology Poland sp. z o.o.
ul. Sowackiego 173 | 80-298 Gdask | Sd Rejonowy Gdask Pnoc | VII Wydzia Gospodarczy Krajowego Rejestru Sdowego - KRS 101882 | NIP 957-07-52-316 | Kapita zakadowy 200.000 PLN.
Ta wiadomo wraz z zacznikami jest przeznaczona dla okrelonego adresata i moe zawiera informacje poufne. W razie przypadkowego otrzymania tej wiadomoci, prosimy o powiadomienie nadawcy oraz trwae jej usunicie; jakiekolwiek przegldanie lub rozpowszechnianie jest zabronione.
This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). If you are not the intended recipient, please contact the sender and delete all copies; any review or distribution by others is strictly prohibited.
More information about the igt-dev
mailing list