Re: [PATCH] selftests/input: Introduce basic tests for evdev ioctls

From: Enric Balletbo Serra
Date: Tue May 30 2023 - 06:12:05 EST

Next message: irqchip-bot for Marc Zyngier: "[irqchip: irq/irqchip-fixes] irqchip/gic: Correctly validate OF quirk descriptors"
Previous message: patchwork-bot+netdevbpf: "Re: [PATCH V3,net] net: mana: Fix perf regression: remove rx_cqes, tx_cqes counters"
In reply to: Peter Hutterer: "Re: [PATCH] selftests/input: Introduce basic tests for evdev ioctls"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Again, without html format. Sorry for the noise.

Hi Peter,

Many thanks for taking a look at this and for the review comments.

> Missatge de Peter Hutterer <peter.hutterer@xxxxxxxxxx> del dia dt., 30 de maig 2023 a les 7:56:
> >
> > Hi Enric,
> >
> > On 26/5/23 18:22, eballetbo@xxxxxxxxxx wrote:
> > > From: Enric Balletbo i Serra <eballetbo@xxxxxxxxxx>
> > >
> > > This provides a basic infrastructure for the creation of tests for the evdev
> > > interface. Most of this is code is adapted from the libevdev wrapper
> > > library. While most of evdev ioctls are covered and tested using
> > > libevdev tests there are some evdev ioctls that aren't because are not
> > > supported (and will not be supported) by libevdev. So, adding, at least
> > > those tests, would make sense.
> >
> > for the archives: libevdev doesn't support setting EV_REP because it's
> > inherently racy - one libevdev context to set those values via the ioctl
> > would cause all other libevdev contexts on the same device to be out of
> > sync. Since we do not get notifications when the values changed,
> > libevdev's buffered values for EV_REP will remain whatever they were
> > initially.
> >
> > The only way around that would be setting it via udev on plug, but that
> > udev doesnt need/want the libevdev dependency.
> >
> > plus, afaik most or all libevdev's users ignore kernel key repeats anyway :)
> >
> > >
> > > The test creates an uinput device (and an evdev device) so you can
> > > call the wanted ioctl from userspace. So, to run those tests you need
> > > to have support for uinput and evdev as well.
> > >
> > > Signed-off-by: Enric Balletbo i Serra <eballetbo@xxxxxxxxxx>
> > > ---
> > > tools/testing/selftests/Makefile | 1 +
> > > tools/testing/selftests/input/.gitignore | 2 +
> > > tools/testing/selftests/input/Makefile | 5 +
> > > tools/testing/selftests/input/config | 3 +
> > > tools/testing/selftests/input/evioc-test.c | 282 +++++++++++++++++++++
> > > 5 files changed, 293 insertions(+)
> > > create mode 100644 tools/testing/selftests/input/.gitignore
> > > create mode 100644 tools/testing/selftests/input/Makefile
> > > create mode 100644 tools/testing/selftests/input/config
> > > create mode 100644 tools/testing/selftests/input/evioc-test.c
> > >
> > > diff --git a/tools/testing/selftests/Makefile b/tools/testing/selftests/Makefile
> > > index 90a62cf75008..29fc77168aa7 100644
> > > --- a/tools/testing/selftests/Makefile
> > > +++ b/tools/testing/selftests/Makefile
> > > @@ -28,6 +28,7 @@ TARGETS += futex
> > > TARGETS += gpio
> > > TARGETS += hid
> > > TARGETS += intel_pstate
> > > +TARGETS += input
> > > TARGETS += iommu
> > > TARGETS += ipc
> > > TARGETS += ir
> > > diff --git a/tools/testing/selftests/input/.gitignore b/tools/testing/selftests/input/.gitignore
> > > new file mode 100644
> > > index 000000000000..37f5dff3255b
> > > --- /dev/null
> > > +++ b/tools/testing/selftests/input/.gitignore
> > > @@ -0,0 +1,2 @@
> > > +# SPDX-License-Identifier: GPL-2.0
> > > +evioc-test
> > > diff --git a/tools/testing/selftests/input/Makefile b/tools/testing/selftests/input/Makefile
> > > new file mode 100644
> > > index 000000000000..031729be0628
> > > --- /dev/null
> > > +++ b/tools/testing/selftests/input/Makefile
> > > @@ -0,0 +1,5 @@
> > > +# SPDX-License-Identifier: GPL-2.0
> > > +CFLAGS += -D_GNU_SOURCE -std=gnu99 -Wl,-no-as-needed -Wall $(KHDR_INCLUDES)
> > > +
> > > +TEST_GEN_PROGS := evioc-test
> > > +include ../lib.mk
> > > diff --git a/tools/testing/selftests/input/config b/tools/testing/selftests/input/config
> > > new file mode 100644
> > > index 000000000000..b7512f3e6d8d
> > > --- /dev/null
> > > +++ b/tools/testing/selftests/input/config
> > > @@ -0,0 +1,3 @@
> > > +CONFIG_INPUT=y
> > > +CONFIG_INPUT_EVDEV=y
> > > +CONFIG_INPUT_UINPUT=m
> > > diff --git a/tools/testing/selftests/input/evioc-test.c b/tools/testing/selftests/input/evioc-test.c
> > > new file mode 100644
> > > index 000000000000..4c0c8ebed378
> > > --- /dev/null
> > > +++ b/tools/testing/selftests/input/evioc-test.c
> > > @@ -0,0 +1,282 @@
> > > +// SPDX-License-Identifier: MIT
> > > +/*
> > > + * Copyright © 2023 Red Hat, Inc.
> > > + *
> > > + * Part of the code in this file is inspired and copied from the libevdev wrapper library
> > > + * for evdev devices written by Peter Hutterer.
> > > + */
> > > +
> > > +#include <dirent.h>
> > > +#include <errno.h>
> > > +#include <fcntl.h>
> > > +#include <linux/uinput.h>
> > > +#include <poll.h>
> > > +#include <stdio.h>
> > > +#include <stdlib.h>
> > > +#include <string.h>
> > > +#include <sys/stat.h>
> > > +#include <time.h>
> > > +#include <unistd.h>
> > > +
> > > +#include "../kselftest_harness.h"
> > > +
> > > +#define TEST_DEVICE_NAME "selftest input device"
> > > +
> > > +struct selftest_uinput {
> > > + int uinput_fd; /** file descriptor to uinput */
> > > + int evdev_fd; /** file descriptor to evdev */
> > > + char *name; /** device name */
> > > + char *syspath; /** /sys path */
> > > + char *devnode; /** device node */
> > > +};
> > > +
> > > +static int is_event_device(const struct dirent *dent)
> > > +{
> > > + return strncmp("event", dent->d_name, 5) == 0;
> > > +}
> > > +
> > > +static char *fetch_device_node(const char *path)
> > > +{
> > > + struct dirent **namelist;
> > > + char *devnode = NULL;
> > > + int ndev, i;
> > > +
> > > + ndev = scandir(path, &namelist, is_event_device, alphasort);
> > > + if (ndev <= 0)
> > > + return NULL;
> > > +
> > > + /* ndev should only ever be 1 */
> > > +
> > > + for (i = 0; i < ndev; i++) {
> > > + if (!devnode && asprintf(&devnode, "/dev/input/%s",
> > > + namelist[i]->d_name) == -1)
> > > + devnode = NULL;
> > > + free(namelist[i]);
> > > + }
> > > +
> > > + free(namelist);
> > > +
> > > + return devnode;
> > > +}
> > > +
> > > +static int is_input_device(const struct dirent *dent)
> > > +{
> > > + return strncmp("input", dent->d_name, 5) == 0;
> > > +}
> > > +
> > > +/*
> > > + * Device node is guessed based on the sysfs path, the sysfs path contains a
> > > + * eventN file, that corresponds to our /dev/input/eventN number.
> > > + */
> > > +static int fetch_syspath_and_devnode(struct selftest_uinput *uidev)
> > > +{
> > > +#define SYS_INPUT_DIR "/sys/devices/virtual/input/"
> > > + struct dirent **namelist;
> > > + int ndev, i;
> > > + int rc;
> > > + char buf[sizeof(SYS_INPUT_DIR) + 64] = SYS_INPUT_DIR;
> > > +
> > > + rc = ioctl(uidev->uinput_fd,
> > > + UI_GET_SYSNAME(sizeof(buf) - strlen(SYS_INPUT_DIR)),
> > > + &buf[strlen(SYS_INPUT_DIR)]);
> > > + if (rc != -1) {
> > > + uidev->syspath = strdup(buf);
> > > + uidev->devnode = fetch_device_node(buf);
> > > + return 0;
> > > + }
> >
> > It's not obvious in the the code, but the above is pretty much all you
> > need. UI_GET_SYSNAME has been around 3.15 and if that fails you might as
> > well bail out in this case. All the code below is the fallback that
> > pre-dates that ioctl.
> >

I see now, thanks. So I think we can simplify all this and make the
assumption that a recent kernel version will be used, so we can remove
the fallback code.

> > > +
> > > + ndev = scandir(SYS_INPUT_DIR, &namelist, is_input_device, alphasort);
> > > + if (ndev <= 0)
> > > + return -1;
> > > +
> > > + for (i = 0; i < ndev; i++) {
> > > + int fd, len;
> > > +
> > > + rc = snprintf(buf, sizeof(buf), "%s%s/name", SYS_INPUT_DIR,
> > > + namelist[i]->d_name);
> > > + if (rc < 0 || (size_t)rc >= sizeof(buf))
> > > + continue;
> > > +
> > > + /* created within time frame */
> > > + fd = open(buf, O_RDONLY);
> > > + if (fd < 0)
> > > + continue;
> > > +
> > > + len = read(fd, buf, sizeof(buf));
> > > + close(fd);
> > > + if (len <= 0)
> > > + continue;
> > > +
> > > + buf[len - 1] = '\0'; /* file contains \n */
> > > + if (strcmp(buf, uidev->name) == 0) {
> > > + if (uidev->syspath) {
> > > + fprintf(stderr,
> > > + "multiple identical devices found. syspath is unreliable\n");
> > > + break;
> > > + }
> > > +
> > > + rc = snprintf(buf, sizeof(buf), "%s%s", SYS_INPUT_DIR,
> > > + namelist[i]->d_name);
> > > +
> > > + if (rc < 0 || (size_t)rc >= sizeof(buf)) {
> > > + fprintf(stderr,
> > > + "Invalid syspath, syspath is unreliable\n");
> > > + break;
> > > + }
> > > +
> > > + uidev->syspath = strdup(buf);
> > > + uidev->devnode = fetch_device_node(buf);
> > > + }
> > > + }
> > > +
> > > + for (i = 0; i < ndev; i++)
> > > + free(namelist[i]);
> > > + free(namelist);
> > > +
> > > + return uidev->devnode ? 0 : -1;
> > > +#undef SYS_INPUT_DIR
> > > +}
> > > +
> > > +void selftest_uinput_destroy(struct selftest_uinput *uidev)
> > > +{
> > > + if (!uidev)
> > > + return;
> > > +
> > > + if (uidev->uinput_fd >= 0)
> > > + ioctl(uidev->uinput_fd, UI_DEV_DESTROY, NULL);
> > > +
> > > + close(uidev->evdev_fd);
> > > + close(uidev->uinput_fd);
> > > +
> > > + free(uidev->syspath);
> > > + free(uidev->devnode);
> > > + free(uidev->name);
> > > + free(uidev);
> > > +}
> > > +
> > > +int selftest_uinput_create_device(struct selftest_uinput **uidev, ...)
> > > +{
> > > + struct selftest_uinput *new_device;
> > > + struct uinput_setup setup;
> > > + va_list args;
> > > + int rc, fd;
> > > + int type;
> > > +
> > > + new_device = calloc(1, sizeof(struct selftest_uinput));
> > > + if (!new_device)
> > > + return -ENOMEM;
> > > +
> > > + memset(&setup, 0, sizeof(setup));
> > > + strncpy(setup.name, TEST_DEVICE_NAME, UINPUT_MAX_NAME_SIZE - 1);
> > > + setup.id.vendor = 0x1234; /* sample vendor */
> > > + setup.id.product = 0x5678; /* sample product */
> > > + setup.id.bustype = BUS_USB;
> > > +
> > > + fd = open("/dev/uinput", O_WRONLY | O_NONBLOCK);
> > > + if (fd < 0) {
> > > + fprintf(stderr, "cannot open uinput (%d): %m\n", errno);
> > > + goto error;
> > > + }
> > > +
> > > + va_start(args, uidev);
> > > + rc = 0;
> > > + do {
> > > + type = va_arg(args, int);
> > > + if (type == -1)
> > > + break;
> > > + rc = ioctl(fd, UI_SET_EVBIT, type);
> > > + } while (rc == 0);
> > > + va_end(args);
> > > +
> > > + rc = ioctl(fd, UI_DEV_SETUP, &setup);
> > > + if (rc == -1)
> > > + goto error;
> > > +
> > > + rc = ioctl(fd, UI_DEV_CREATE, NULL);
> > > + if (rc == -1)
> > > + goto error;
> > > +
> > > + new_device->name = strdup(TEST_DEVICE_NAME);
> > > + new_device->uinput_fd = fd;
> > > +
> > > + if (fetch_syspath_and_devnode(new_device) == -1) {
> > > + fprintf(stderr, "unable to fetch syspath or device node.\n");
> > > + errno = ENODEV;
> > > + goto error;
> > > + }
> > > +
> > > + fd = open(new_device->devnode, O_RDONLY);
> > > + if (fd < 0) {
> > > + fprintf(stderr, "cannot open uinput (%d): %m\n", errno);
> > > + goto error;
> > > + }
> > > + new_device->evdev_fd = fd;
> > > +
> > > + *uidev = new_device;
> > > +
> > > + return 0;
> > > +
> > > +error:
> > > + rc = -errno;
> > > + selftest_uinput_destroy(new_device);
> > > + return rc;
> > > +}
> > > +
> > > +const char *selftest_uinput_get_devnode(struct selftest_uinput *uidev)
> > > +{
> > > + return uidev->devnode;
> > > +}
> > > +
> > > +TEST(eviocgname_get_device_name)
> > > +{
> > > + struct selftest_uinput *uidev;
> > > + char buf[256];
> > > + int rc;
> > > +
> > > + rc = selftest_uinput_create_device(&uidev);
> > > + ASSERT_EQ(0, rc);
> > > + ASSERT_NE(NULL, uidev);
> > > +
> > > + memset(buf, 0, sizeof(buf));
> > > + /* ioctl to get the name */
> > > + rc = ioctl(uidev->evdev_fd, EVIOCGNAME(sizeof(buf) - 1), buf);
> > > + ASSERT_GE(rc, 0);
> > > + ASSERT_STREQ(TEST_DEVICE_NAME, buf);
> > > +
> > > + selftest_uinput_destroy(uidev);
> > > +}
> > > +
> > > +TEST(eviocgrep_get_repeat_settings)
> > > +{
> > > + struct selftest_uinput *uidev;
> > > + int rep_values[2];
> > > + int rc;
> > > +
> > > + memset(rep_values, 0, sizeof(rep_values));
> > > +
> > > + rc = selftest_uinput_create_device(&uidev);
> > > + ASSERT_EQ(0, rc);
> > > + ASSERT_NE(NULL, uidev);
> > > +
> > > + /* ioctl to get the repeat rates values */
> > > + rc = ioctl(uidev->evdev_fd, EVIOCSREP, rep_values);
> >
> > EVIOCSREP *sets* the repeat values

My intention was to use EVIOCGREP here instead of 'S', so instead of
fixing the comment I'd fix the ioctl. Thanks.

> >
> >
> > > + /* should fail because EV_REP is not set */
> > > + ASSERT_EQ(-1, rc);
> > > +
> > > + selftest_uinput_destroy(uidev);
> > > +
> > > + rc = selftest_uinput_create_device(&uidev, EV_REP);
> > This needs a -1 terminator, otherwise you're va-argsing into oblivion.
> >

Right, I'd add in the next version.

> >
> > > + ASSERT_EQ(0, rc);
> > > + ASSERT_NE(NULL, uidev);
> > > +
> > > + /* ioctl to get the repeat rates values */
> > > + rc = ioctl(uidev->evdev_fd, EVIOCGREP, rep_values);
> > > + ASSERT_EQ(0, rc);
> > > + /* should get the default delay an period values set by the kernel */
> >
> > typo -> "and"
> >

Fixed in the next version.

> > > + ASSERT_EQ(rep_values[0], 250);
> > > + ASSERT_EQ(rep_values[1], 33);
> >
> > So, interestingly enough you don't actually test setting the value here.
> > The EV_REP bit is set during device creation and you check the default
> > values but you don't call EVIOCSREP to change the values - which AFAIU
> > was the initial intention of these tests?
> >
>

Yes, one of the main purposes is to test EVIOC**S**REP, but I skipped
and sent a first version only testing the EVIOC**G**REP test to catch
feedback from the mailing list. In the next version I'll include both
tests, for getting and setting the repeat values (I think it makes
sense).

Cheers,
Enric
>
> > Cheers,
> > PEter
> >
> > > +
> > > + selftest_uinput_destroy(uidev);
> > > +}
> > > +
> > > +TEST_HARNESS_MAIN

Next message: irqchip-bot for Marc Zyngier: "[irqchip: irq/irqchip-fixes] irqchip/gic: Correctly validate OF quirk descriptors"
Previous message: patchwork-bot+netdevbpf: "Re: [PATCH V3,net] net: mana: Fix perf regression: remove rx_cqes, tx_cqes counters"
In reply to: Peter Hutterer: "Re: [PATCH] selftests/input: Introduce basic tests for evdev ioctls"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]