libinput/tools/libinput-record.c

1056 lines
22 KiB
C
Raw Normal View History

tools: add a libinput-record tool This is a tool that does effectively the same job as evemu-record. evemu has two disadvantages: its API is clunky and hard to extend even for simple features. And it has a custom file format that requires special processing but is difficult to extend and hard to write manually. e.g. the bitmasks require keeping a line number state to know which bit an entry refers to. libinput-record records the same data but the output is YAML. That can be processed easier and extended in the future without breaking the parsing. We can (in the future) also interleave the evemu output with libinput's debug output, thus having a single file where the events can be compared and analysed without the need for replaying. Likewise, we can easily annotate the file with parsable bits of information without having to shove all that into a comment (like version numbers of libinput, kernel, etc). User-visible differences to evemu-record: * the output file requires an explicit -o or --output-file argument * no evemu-describe equivalent, if you just want the description simply cancel before any events are sent * to see key codes, a --show-keycodes flag must be supplied, otherwise all 'normal' keys end up as KEY_A. This protects against inadvertent information leakage * supports a --multiple option to record multiple devices simultaneously. All recordings have the same time offset, it is thus possible to reproduce bugs that depend on the interaction of more than one device. And to answer the question of: why a printf-approach to writing out yaml instead of a library, it's simply that we want to be able to have real-time output of the recording. Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
2017-11-23 11:01:45 +10:00
/*
* Copyright © 2018 Red Hat, Inc.
*
* 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.
*/
#include "config.h"
#include <errno.h>
#include <linux/input.h>
#include <libevdev/libevdev.h>
#include <sys/signalfd.h>
#include <sys/utsname.h>
#include <string.h>
#include <dirent.h>
#include <fcntl.h>
#include <getopt.h>
#include <poll.h>
#include <signal.h>
#include <stdbool.h>
#include "libinput-util.h"
#include "libinput-version.h"
static const int FILE_VERSION_NUMBER = 1;
struct record_device {
struct list link;
char *devnode; /* device node of the source device */
struct libevdev *evdev;
struct input_event *events;
size_t nevents;
size_t events_sz;
};
struct record_context {
int timeout;
bool show_keycodes;
uint64_t offset;
struct list devices;
int ndevices;
char *outfile; /* file name given on cmdline */
char *output_file; /* full file name with suffix */
int out_fd;
unsigned int indent;
};
static inline bool
obfuscate_keycode(struct input_event *ev)
{
switch (ev->type) {
case EV_KEY:
if (ev->code >= KEY_ESC && ev->code < KEY_ZENKAKUHANKAKU) {
ev->code = KEY_A;
return true;
}
break;
case EV_MSC:
if (ev->code == MSC_SCAN) {
ev->value = 30; /* KEY_A scancode */
return true;
}
break;
}
return false;
}
static inline void
indent_push(struct record_context *ctx)
{
ctx->indent += 2;
}
static inline void
indent_pop(struct record_context *ctx)
{
assert(ctx->indent >= 2);
ctx->indent -= 2;
}
/**
* Indented dprintf, indentation is given as second parameter.
*/
static inline void
iprintf(const struct record_context *ctx, const char *format, ...)
{
va_list args;
char fmt[1024];
static const char space[] = " ";
static const size_t len = sizeof(space);
unsigned int indent = ctx->indent;
int rc;
assert(indent < len);
assert(strlen(format) > 1);
/* Special case: if we're printing a new list item, we want less
* indentation because the '- ' takes up one level of indentation
*
* This is only needed because I don't want to deal with open/close
* lists statements.
*/
if (format[0] == '-')
indent -= 2;
snprintf(fmt, sizeof(fmt), "%s%s", &space[len - indent - 1], format);
va_start(args, format);
rc = vdprintf(ctx->out_fd, fmt, args);
va_end(args);
assert(rc != -1 && (unsigned int)rc > indent);
}
/**
* Normal printf, just wrapped for the context
*/
static inline void
noiprintf(const struct record_context *ctx, const char *format, ...)
{
va_list args;
int rc;
va_start(args, format);
rc = vdprintf(ctx->out_fd, format, args);
va_end(args);
assert(rc != -1 && (unsigned int)rc > 0);
}
static inline void
print_evdev_event(struct record_context *ctx, struct input_event *ev)
{
const char *cname;
bool was_modified = false;
char desc[1024];
if (ctx->offset == 0)
ctx->offset = tv2us(&ev->time);
ev->time = us2tv(tv2us(&ev->time) - ctx->offset);
/* Don't leak passwords unless the user wants to */
if (!ctx->show_keycodes)
was_modified = obfuscate_keycode(ev);
cname = libevdev_event_code_get_name(ev->type, ev->code);
if (ev->type == EV_SYN && ev->code == SYN_MT_REPORT) {
snprintf(desc,
sizeof(desc),
"++++++++++++ %s (%d) ++++++++++",
cname,
ev->value);
} else if (ev->type == EV_SYN) {
static unsigned long last_ms = 0;
unsigned long time, dt;
time = us2ms(tv2us(&ev->time));
if (last_ms == 0)
last_ms = time;
dt = time - last_ms;
last_ms = time;
snprintf(desc,
sizeof(desc),
"------------ %s (%d) ---------- %+ldms",
cname,
ev->value,
dt);
} else {
const char *tname = libevdev_event_type_get_name(ev->type);
snprintf(desc,
sizeof(desc),
"%s / %-20s %4d%s",
tname,
cname,
ev->value,
was_modified ? " (obfuscated)" : "");
}
iprintf(ctx,
"- [%3lu, %6u, %3d, %3d, %5d] # %s\n",
ev->time.tv_sec,
(unsigned int)ev->time.tv_usec,
ev->type,
ev->code,
ev->value,
desc);
}
static inline void
print_evdev_events(struct record_context *ctx, struct input_event *e, size_t nevents)
{
iprintf(ctx, "- evdev:\n");
indent_push(ctx);
for (size_t i = 0; i < nevents; i++)
print_evdev_event(ctx, &e[i]);
indent_pop(ctx);
}
static inline size_t
handle_frame(struct record_context *ctx, struct record_device *d, bool print)
{
struct libevdev *evdev = d->evdev;
struct input_event e;
size_t count = 0;
while (libevdev_next_event(evdev,
LIBEVDEV_READ_FLAG_NORMAL,
&e) == LIBEVDEV_READ_STATUS_SUCCESS) {
if (d->nevents == d->events_sz) {
void *tmp;
d->events_sz += 1000;
tmp = realloc(d->events, d->events_sz * sizeof(*d->events));
assert(tmp);
d->events = tmp;
}
d->events[d->nevents++] = e;
count++;
if (e.type == EV_SYN && e.code == SYN_REPORT)
break;
}
return count;
}
static inline void
handle_events(struct record_context *ctx, struct record_device *d, bool print)
{
while(true) {
size_t first_idx = d->nevents;
size_t evcount;
evcount = handle_frame(ctx, d, print);
if (evcount == 0)
break;
if (!print)
continue;
print_evdev_events(ctx,
&d->events[first_idx],
evcount);
}
}
static inline void
print_libinput_header(struct record_context *ctx)
{
iprintf(ctx, "libinput:\n");
indent_push(ctx);
iprintf(ctx, "version: \"%s\"\n", LIBINPUT_VERSION);
if (ctx->timeout > 0)
iprintf(ctx, "autorestart: %d\n", ctx->timeout);
indent_pop(ctx);
}
static inline void
print_system_header(struct record_context *ctx)
{
struct utsname u;
const char *kernel = "unknown";
FILE *dmi;
char modalias[2048] = "unknown";
if (uname(&u) != -1)
kernel = u.release;
dmi = fopen("/sys/class/dmi/id/modalias", "r");
if (dmi) {
if (fgets(modalias, sizeof(modalias), dmi)) {
modalias[strlen(modalias) - 1] = '\0'; /* linebreak */
} else {
sprintf(modalias, "unknown");
}
fclose(dmi);
}
iprintf(ctx, "system:\n");
indent_push(ctx);
iprintf(ctx, "kernel: \"%s\"\n", kernel);
iprintf(ctx, "dmi: \"%s\"\n", modalias);
indent_pop(ctx);
}
static inline void
print_header(struct record_context *ctx)
{
iprintf(ctx, "version: %d\n", FILE_VERSION_NUMBER);
iprintf(ctx, "ndevices: %d\n", ctx->ndevices);
print_libinput_header(ctx);
print_system_header(ctx);
}
static inline void
print_description_abs(struct record_context *ctx,
struct libevdev *dev,
unsigned int code)
{
const struct input_absinfo *abs;
abs = libevdev_get_abs_info(dev, code);
assert(abs);
iprintf(ctx, "# Value %6d\n", abs->value);
iprintf(ctx, "# Min %6d\n", abs->minimum);
iprintf(ctx, "# Max %6d\n", abs->maximum);
iprintf(ctx, "# Fuzz %6d\n", abs->fuzz);
iprintf(ctx, "# Flat %6d\n", abs->flat);
iprintf(ctx, "# Resolution %6d\n", abs->resolution);
}
static inline void
print_description_state(struct record_context *ctx,
struct libevdev *dev,
unsigned int type,
unsigned int code)
{
int state = libevdev_get_event_value(dev, type, code);
iprintf(ctx, "# State %d\n", state);
}
static inline void
print_description_codes(struct record_context *ctx,
struct libevdev *dev,
unsigned int type)
{
int max;
max = libevdev_event_type_get_max(type);
if (max == -1)
return;
iprintf(ctx,
"# Event type %d (%s)\n",
type,
libevdev_event_type_get_name(type));
if (type == EV_SYN)
return;
for (unsigned int code = 0; code <= (unsigned int)max; code++) {
if (!libevdev_has_event_code(dev, type, code))
continue;
iprintf(ctx,
"# Event code %d (%s)\n",
code,
libevdev_event_code_get_name(type,
code));
switch (type) {
case EV_ABS:
print_description_abs(ctx, dev, code);
break;
case EV_LED:
case EV_SW:
print_description_state(ctx, dev, type, code);
break;
}
}
}
static inline void
print_description(struct record_context *ctx, struct libevdev *dev)
{
const struct input_absinfo *x, *y;
iprintf(ctx, "# Name: %s\n", libevdev_get_name(dev));
iprintf(ctx,
"# ID: bus %#02x vendor %#02x product %#02x version %#02x\n",
libevdev_get_id_bustype(dev),
libevdev_get_id_vendor(dev),
libevdev_get_id_product(dev),
libevdev_get_id_version(dev));
x = libevdev_get_abs_info(dev, ABS_X);
y = libevdev_get_abs_info(dev, ABS_Y);
if (x && y) {
if (x->resolution && y->resolution) {
tools: add a libinput-record tool This is a tool that does effectively the same job as evemu-record. evemu has two disadvantages: its API is clunky and hard to extend even for simple features. And it has a custom file format that requires special processing but is difficult to extend and hard to write manually. e.g. the bitmasks require keeping a line number state to know which bit an entry refers to. libinput-record records the same data but the output is YAML. That can be processed easier and extended in the future without breaking the parsing. We can (in the future) also interleave the evemu output with libinput's debug output, thus having a single file where the events can be compared and analysed without the need for replaying. Likewise, we can easily annotate the file with parsable bits of information without having to shove all that into a comment (like version numbers of libinput, kernel, etc). User-visible differences to evemu-record: * the output file requires an explicit -o or --output-file argument * no evemu-describe equivalent, if you just want the description simply cancel before any events are sent * to see key codes, a --show-keycodes flag must be supplied, otherwise all 'normal' keys end up as KEY_A. This protects against inadvertent information leakage * supports a --multiple option to record multiple devices simultaneously. All recordings have the same time offset, it is thus possible to reproduce bugs that depend on the interaction of more than one device. And to answer the question of: why a printf-approach to writing out yaml instead of a library, it's simply that we want to be able to have real-time output of the recording. Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
2017-11-23 11:01:45 +10:00
int w, h;
w = (x->maximum - x->minimum)/x->resolution;
h = (y->maximum - y->minimum)/y->resolution;
iprintf(ctx, "# Size in mm: %dx%d\n", w, h);
} else {
iprintf(ctx,
"# Size in mm: unknown, missing resolution\n");
}
}
iprintf(ctx, "# Supported Events:\n");
for (unsigned int type = 0; type < EV_CNT; type++) {
if (!libevdev_has_event_type(dev, type))
continue;
print_description_codes(ctx, dev, type);
}
iprintf(ctx, "# Properties:\n");
for (unsigned int prop = 0; prop < INPUT_PROP_CNT; prop++) {
if (libevdev_has_property(dev, prop)) {
iprintf(ctx,
"# Property %d (%s)\n",
prop,
libevdev_property_get_name(prop));
}
}
}
static inline void
print_bits_info(struct record_context *ctx, struct libevdev *dev)
{
iprintf(ctx, "name: \"%s\"\n", libevdev_get_name(dev));
iprintf(ctx,
"id: [%d, %d, %d, %d]\n",
libevdev_get_id_bustype(dev),
libevdev_get_id_vendor(dev),
libevdev_get_id_product(dev),
libevdev_get_id_version(dev));
}
static inline void
print_bits_absinfo(struct record_context *ctx, struct libevdev *dev)
{
const struct input_absinfo *abs;
if (!libevdev_has_event_type(dev, EV_ABS))
return;
iprintf(ctx, "absinfo:\n");
indent_push(ctx);
for (unsigned int code = 0; code < ABS_CNT; code++) {
abs = libevdev_get_abs_info(dev, code);
if (!abs)
continue;
iprintf(ctx,
"%d: [%d, %d, %d, %d, %d]\n",
code,
abs->minimum,
abs->maximum,
abs->fuzz,
abs->flat,
abs->resolution);
}
indent_pop(ctx);
}
static inline void
print_bits_codes(struct record_context *ctx,
struct libevdev *dev,
unsigned int type)
{
int max;
bool first = true;
max = libevdev_event_type_get_max(type);
if (max == -1)
return;
iprintf(ctx, "%d: [", type);
for (unsigned int code = 0; code <= (unsigned int)max; code++) {
if (!libevdev_has_event_code(dev, type, code))
continue;
noiprintf(ctx, "%s%d", first ? "" : ", ", code);
first = false;
}
noiprintf(ctx, "] # %s\n", libevdev_event_type_get_name(type));
}
static inline void
print_bits_types(struct record_context *ctx, struct libevdev *dev)
{
iprintf(ctx, "codes:\n");
indent_push(ctx);
for (unsigned int type = 0; type < EV_CNT; type++) {
if (!libevdev_has_event_type(dev, type))
continue;
print_bits_codes(ctx, dev, type);
}
indent_pop(ctx);
}
static inline void
print_bits_props(struct record_context *ctx, struct libevdev *dev)
{
bool first = true;
iprintf(ctx, "properties: [");
for (unsigned int prop = 0; prop < INPUT_PROP_CNT; prop++) {
if (libevdev_has_property(dev, prop)) {
noiprintf(ctx, "%s%d", first ? "" : ", ", prop);
first = false;
}
}
noiprintf(ctx, "]\n"); /* last entry, no comma */
}
static inline void
print_evdev_description(struct record_context *ctx, struct record_device *dev)
{
struct libevdev *evdev = dev->evdev;
iprintf(ctx, "evdev:\n");
indent_push(ctx);
print_description(ctx, evdev);
print_bits_info(ctx, evdev);
print_bits_types(ctx, evdev);
print_bits_absinfo(ctx, evdev);
print_bits_props(ctx, evdev);
indent_pop(ctx);
}
static inline void
print_device_description(struct record_context *ctx, struct record_device *dev)
{
iprintf(ctx, "- node: %s\n", dev->devnode);
print_evdev_description(ctx, dev);
}
static int is_event_node(const struct dirent *dir) {
return strneq(dir->d_name, "event", 5);
}
static inline char *
select_device(void)
{
struct dirent **namelist;
int ndev, selected_device;
int rc;
char *device_path;
ndev = scandir("/dev/input", &namelist, is_event_node, versionsort);
if (ndev <= 0)
return NULL;
fprintf(stderr, "Available devices:\n");
for (int i = 0; i < ndev; i++) {
struct libevdev *device;
char path[PATH_MAX];
int fd = -1;
snprintf(path,
sizeof(path),
"/dev/input/%s",
namelist[i]->d_name);
fd = open(path, O_RDONLY);
if (fd < 0)
continue;
rc = libevdev_new_from_fd(fd, &device);
close(fd);
if (rc != 0)
continue;
fprintf(stderr, "%s: %s\n", path, libevdev_get_name(device));
libevdev_free(device);
}
for (int i = 0; i < ndev; i++)
free(namelist[i]);
free(namelist);
fprintf(stderr, "Select the device event number: ");
rc = scanf("%d", &selected_device);
if (rc != 1 || selected_device < 0)
return NULL;
rc = xasprintf(&device_path, "/dev/input/event%d", selected_device);
if (rc == -1)
return NULL;
return device_path;
}
static inline char **
all_devices(void)
{
struct dirent **namelist;
int ndev;
int rc;
char **devices = NULL;
ndev = scandir("/dev/input", &namelist, is_event_node, versionsort);
if (ndev <= 0)
return NULL;
devices = zalloc((ndev + 1)* sizeof *devices); /* NULL-terminated */
for (int i = 0; i < ndev; i++) {
char *device_path;
rc = xasprintf(&device_path,
"/dev/input/%s",
namelist[i]->d_name);
if (rc == -1)
goto error;
devices[i] = device_path;
}
return devices;
error:
if (devices)
strv_free(devices);
return NULL;
}
tools: add a libinput-record tool This is a tool that does effectively the same job as evemu-record. evemu has two disadvantages: its API is clunky and hard to extend even for simple features. And it has a custom file format that requires special processing but is difficult to extend and hard to write manually. e.g. the bitmasks require keeping a line number state to know which bit an entry refers to. libinput-record records the same data but the output is YAML. That can be processed easier and extended in the future without breaking the parsing. We can (in the future) also interleave the evemu output with libinput's debug output, thus having a single file where the events can be compared and analysed without the need for replaying. Likewise, we can easily annotate the file with parsable bits of information without having to shove all that into a comment (like version numbers of libinput, kernel, etc). User-visible differences to evemu-record: * the output file requires an explicit -o or --output-file argument * no evemu-describe equivalent, if you just want the description simply cancel before any events are sent * to see key codes, a --show-keycodes flag must be supplied, otherwise all 'normal' keys end up as KEY_A. This protects against inadvertent information leakage * supports a --multiple option to record multiple devices simultaneously. All recordings have the same time offset, it is thus possible to reproduce bugs that depend on the interaction of more than one device. And to answer the question of: why a printf-approach to writing out yaml instead of a library, it's simply that we want to be able to have real-time output of the recording. Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
2017-11-23 11:01:45 +10:00
static char *
init_output_file(const char *file, bool is_prefix)
{
char name[PATH_MAX];
assert(file != NULL);
if (is_prefix) {
struct tm *tm;
time_t t;
char suffix[64];
t = time(NULL);
tm = localtime(&t);
strftime(suffix, sizeof(suffix), "%F-%T", tm);
snprintf(name,
sizeof(name),
"%s.%s",
file,
suffix);
} else {
snprintf(name, sizeof(name), "%s", file);
}
return strdup(name);
}
static bool
open_output_file(struct record_context *ctx, bool is_prefix)
{
int out_fd;
if (ctx->outfile) {
char *fname = init_output_file(ctx->outfile, is_prefix);
ctx->output_file = fname;
out_fd = open(fname, O_WRONLY|O_CREAT|O_TRUNC, 0666);
if (out_fd < 0)
return false;
} else {
ctx->output_file = safe_strdup("stdout");
out_fd = STDOUT_FILENO;
}
ctx->out_fd = out_fd;
return true;
}
static int
mainloop(struct record_context *ctx)
{
bool autorestart = (ctx->timeout > 0);
struct pollfd fds[ctx->ndevices + 1];
struct record_device *d = NULL;
struct timespec ts;
sigset_t mask;
int idx;
assert(ctx->timeout != 0);
assert(!list_empty(&ctx->devices));
sigemptyset(&mask);
sigaddset(&mask, SIGINT);
sigaddset(&mask, SIGQUIT);
sigprocmask(SIG_BLOCK, &mask, NULL);
fds[0].fd = signalfd(-1, &mask, SFD_NONBLOCK);
fds[0].events = POLLIN;
fds[0].revents = 0;
assert(fds[0].fd != -1);
idx = 1;
list_for_each(d, &ctx->devices, link) {
fds[idx].fd = libevdev_get_fd(d->evdev);
fds[idx].events = POLLIN;
fds[idx].revents = 0;
assert(fds[idx].fd != -1);
idx++;
}
/* If we have more than one device, the time starts at recording
* start time. Otherwise, the first event starts the recording time.
*/
if (ctx->ndevices > 1) {
clock_gettime(CLOCK_MONOTONIC, &ts);
ctx->offset = s2us(ts.tv_sec) + ns2us(ts.tv_nsec);
}
do {
int rc;
bool had_events = false; /* we delete files without events */
if (!open_output_file(ctx, autorestart)) {
fprintf(stderr,
"Failed to open '%s'\n",
ctx->output_file);
break;
}
fprintf(stderr, "recording to '%s'\n", ctx->output_file);
print_header(ctx);
if (autorestart)
iprintf(ctx,
"# Autorestart timeout: %d\n",
ctx->timeout);
iprintf(ctx, "devices:\n");
indent_push(ctx);
/* we only print the first device's description, the
* rest is assembled after CTRL+C */
d = list_first_entry(&ctx->devices, d, link);
print_device_description(ctx, d);
iprintf(ctx, "events:\n");
indent_push(ctx);
while (true) {
rc = poll(fds, ARRAY_LENGTH(fds), ctx->timeout);
if (rc == -1) { /* error */
fprintf(stderr, "Error: %m\n");
autorestart = false;
break;
} else if (rc == 0) {
fprintf(stderr,
" ... timeout%s\n",
had_events ? "" : " (file is empty)");
break;
} else if (fds[0].revents != 0) { /* signal */
autorestart = false;
break;
} else { /* events */
int is_first = true;
had_events = true;
list_for_each(d, &ctx->devices, link) {
handle_events(ctx, d, is_first);
is_first = false;
}
}
}
indent_pop(ctx); /* events: */
if (autorestart) {
noiprintf(ctx,
"# Closing after %ds inactivity",
ctx->timeout/1000);
}
/* First device is printed, now append all the data from the
* other devices, if any */
list_for_each(d, &ctx->devices, link) {
if (d == list_first_entry(&ctx->devices, d, link))
continue;
print_device_description(ctx, d);
iprintf(ctx, "events:\n");
indent_push(ctx);
print_evdev_events(ctx, d->events, d->nevents);
indent_pop(ctx);
}
indent_pop(ctx); /* devices: */
assert(ctx->indent == 0);
fsync(ctx->out_fd);
/* If we didn't have events, delete the file. */
if (!isatty(ctx->out_fd)) {
if (!had_events && ctx->output_file) {
perror("");
fprintf(stderr, "No events recorded, deleting '%s'\n", ctx->output_file);
unlink(ctx->output_file);
}
close(ctx->out_fd);
ctx->out_fd = -1;
}
free(ctx->output_file);
ctx->output_file = NULL;
} while (autorestart);
close(fds[0].fd);
sigprocmask(SIG_UNBLOCK, &mask, NULL);
return 0;
}
static inline bool
init_device(struct record_context *ctx, char *path)
{
struct record_device *d;
int fd, rc;
d = zalloc(sizeof(*d));
d->devnode = path;
d->nevents = 0;
d->events_sz = 5000;
d->events = zalloc(d->events_sz * sizeof(*d->events));
fd = open(d->devnode, O_RDONLY|O_NONBLOCK);
if (fd < 0) {
fprintf(stderr,
"Failed to open device %s (%m)\n",
d->devnode);
return false;
}
rc = libevdev_new_from_fd(fd, &d->evdev);
if (rc != 0) {
fprintf(stderr,
"Failed to create context for %s (%s)\n",
d->devnode,
strerror(-rc));
close(fd);
return false;
}
libevdev_set_clock_id(d->evdev, CLOCK_MONOTONIC);
list_insert(&ctx->devices, &d->link);
ctx->ndevices++;
return true;
}
static inline void
usage(void)
{
printf("Usage: %s [--help] [--multiple|--all] [--autorestart] [--output-file filename] [/dev/input/event0] [...]\n"
tools: add a libinput-record tool This is a tool that does effectively the same job as evemu-record. evemu has two disadvantages: its API is clunky and hard to extend even for simple features. And it has a custom file format that requires special processing but is difficult to extend and hard to write manually. e.g. the bitmasks require keeping a line number state to know which bit an entry refers to. libinput-record records the same data but the output is YAML. That can be processed easier and extended in the future without breaking the parsing. We can (in the future) also interleave the evemu output with libinput's debug output, thus having a single file where the events can be compared and analysed without the need for replaying. Likewise, we can easily annotate the file with parsable bits of information without having to shove all that into a comment (like version numbers of libinput, kernel, etc). User-visible differences to evemu-record: * the output file requires an explicit -o or --output-file argument * no evemu-describe equivalent, if you just want the description simply cancel before any events are sent * to see key codes, a --show-keycodes flag must be supplied, otherwise all 'normal' keys end up as KEY_A. This protects against inadvertent information leakage * supports a --multiple option to record multiple devices simultaneously. All recordings have the same time offset, it is thus possible to reproduce bugs that depend on the interaction of more than one device. And to answer the question of: why a printf-approach to writing out yaml instead of a library, it's simply that we want to be able to have real-time output of the recording. Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
2017-11-23 11:01:45 +10:00
"Common use-cases:\n"
"\n"
" sudo %s -o recording.yml\n"
" Then select the device to record and it Ctrl+C to stop.\n"
" The recorded data is in recording.yml and can be attached to a bug report.\n"
"\n"
" sudo %s -o recording.yml --autorestart 2\n"
" As above, but restarts after 2s of inactivity on the device.\n"
" Note, the output file is only the prefix.\n"
"\n"
" sudo %s --multiple -o recording.yml /dev/input/event3 /dev/input/event4\n"
" Records the two devices into the same recordings file.\n"
"\n"
"For more information, see the %s(1) man page\n",
program_invocation_short_name,
program_invocation_short_name,
program_invocation_short_name,
program_invocation_short_name,
program_invocation_short_name);
}
enum options {
OPT_AUTORESTART,
OPT_HELP,
OPT_OUTFILE,
OPT_KEYCODES,
OPT_MULTIPLE,
OPT_ALL,
tools: add a libinput-record tool This is a tool that does effectively the same job as evemu-record. evemu has two disadvantages: its API is clunky and hard to extend even for simple features. And it has a custom file format that requires special processing but is difficult to extend and hard to write manually. e.g. the bitmasks require keeping a line number state to know which bit an entry refers to. libinput-record records the same data but the output is YAML. That can be processed easier and extended in the future without breaking the parsing. We can (in the future) also interleave the evemu output with libinput's debug output, thus having a single file where the events can be compared and analysed without the need for replaying. Likewise, we can easily annotate the file with parsable bits of information without having to shove all that into a comment (like version numbers of libinput, kernel, etc). User-visible differences to evemu-record: * the output file requires an explicit -o or --output-file argument * no evemu-describe equivalent, if you just want the description simply cancel before any events are sent * to see key codes, a --show-keycodes flag must be supplied, otherwise all 'normal' keys end up as KEY_A. This protects against inadvertent information leakage * supports a --multiple option to record multiple devices simultaneously. All recordings have the same time offset, it is thus possible to reproduce bugs that depend on the interaction of more than one device. And to answer the question of: why a printf-approach to writing out yaml instead of a library, it's simply that we want to be able to have real-time output of the recording. Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
2017-11-23 11:01:45 +10:00
};
int
main(int argc, char **argv)
{
struct record_context ctx = {
.timeout = -1,
.show_keycodes = false,
};
struct option opts[] = {
{ "autorestart", required_argument, 0, OPT_AUTORESTART },
{ "output-file", required_argument, 0, OPT_OUTFILE },
{ "show-keycodes", no_argument, 0, OPT_KEYCODES },
{ "multiple", no_argument, 0, OPT_MULTIPLE },
{ "all", no_argument, 0, OPT_ALL },
tools: add a libinput-record tool This is a tool that does effectively the same job as evemu-record. evemu has two disadvantages: its API is clunky and hard to extend even for simple features. And it has a custom file format that requires special processing but is difficult to extend and hard to write manually. e.g. the bitmasks require keeping a line number state to know which bit an entry refers to. libinput-record records the same data but the output is YAML. That can be processed easier and extended in the future without breaking the parsing. We can (in the future) also interleave the evemu output with libinput's debug output, thus having a single file where the events can be compared and analysed without the need for replaying. Likewise, we can easily annotate the file with parsable bits of information without having to shove all that into a comment (like version numbers of libinput, kernel, etc). User-visible differences to evemu-record: * the output file requires an explicit -o or --output-file argument * no evemu-describe equivalent, if you just want the description simply cancel before any events are sent * to see key codes, a --show-keycodes flag must be supplied, otherwise all 'normal' keys end up as KEY_A. This protects against inadvertent information leakage * supports a --multiple option to record multiple devices simultaneously. All recordings have the same time offset, it is thus possible to reproduce bugs that depend on the interaction of more than one device. And to answer the question of: why a printf-approach to writing out yaml instead of a library, it's simply that we want to be able to have real-time output of the recording. Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
2017-11-23 11:01:45 +10:00
{ "help", no_argument, 0, OPT_HELP },
{ 0, 0, 0, 0 },
};
struct record_device *d, *tmp;
const char *output_arg = NULL;
bool multiple = false, all = false;
tools: add a libinput-record tool This is a tool that does effectively the same job as evemu-record. evemu has two disadvantages: its API is clunky and hard to extend even for simple features. And it has a custom file format that requires special processing but is difficult to extend and hard to write manually. e.g. the bitmasks require keeping a line number state to know which bit an entry refers to. libinput-record records the same data but the output is YAML. That can be processed easier and extended in the future without breaking the parsing. We can (in the future) also interleave the evemu output with libinput's debug output, thus having a single file where the events can be compared and analysed without the need for replaying. Likewise, we can easily annotate the file with parsable bits of information without having to shove all that into a comment (like version numbers of libinput, kernel, etc). User-visible differences to evemu-record: * the output file requires an explicit -o or --output-file argument * no evemu-describe equivalent, if you just want the description simply cancel before any events are sent * to see key codes, a --show-keycodes flag must be supplied, otherwise all 'normal' keys end up as KEY_A. This protects against inadvertent information leakage * supports a --multiple option to record multiple devices simultaneously. All recordings have the same time offset, it is thus possible to reproduce bugs that depend on the interaction of more than one device. And to answer the question of: why a printf-approach to writing out yaml instead of a library, it's simply that we want to be able to have real-time output of the recording. Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
2017-11-23 11:01:45 +10:00
int ndevices;
int rc = 1;
list_init(&ctx.devices);
while (1) {
int c;
int option_index = 0;
c = getopt_long(argc, argv, "ho:", opts, &option_index);
if (c == -1)
break;
switch (c) {
case 'h':
case OPT_HELP:
usage();
rc = 0;
goto out;
case OPT_AUTORESTART:
if (!safe_atoi(optarg, &ctx.timeout) ||
ctx.timeout <= 0) {
usage();
goto out;
}
ctx.timeout = ctx.timeout * 1000;
break;
case 'o':
case OPT_OUTFILE:
output_arg = optarg;
break;
case OPT_KEYCODES:
ctx.show_keycodes = true;
break;
case OPT_MULTIPLE:
multiple = true;
break;
case OPT_ALL:
all = true;
break;
tools: add a libinput-record tool This is a tool that does effectively the same job as evemu-record. evemu has two disadvantages: its API is clunky and hard to extend even for simple features. And it has a custom file format that requires special processing but is difficult to extend and hard to write manually. e.g. the bitmasks require keeping a line number state to know which bit an entry refers to. libinput-record records the same data but the output is YAML. That can be processed easier and extended in the future without breaking the parsing. We can (in the future) also interleave the evemu output with libinput's debug output, thus having a single file where the events can be compared and analysed without the need for replaying. Likewise, we can easily annotate the file with parsable bits of information without having to shove all that into a comment (like version numbers of libinput, kernel, etc). User-visible differences to evemu-record: * the output file requires an explicit -o or --output-file argument * no evemu-describe equivalent, if you just want the description simply cancel before any events are sent * to see key codes, a --show-keycodes flag must be supplied, otherwise all 'normal' keys end up as KEY_A. This protects against inadvertent information leakage * supports a --multiple option to record multiple devices simultaneously. All recordings have the same time offset, it is thus possible to reproduce bugs that depend on the interaction of more than one device. And to answer the question of: why a printf-approach to writing out yaml instead of a library, it's simply that we want to be able to have real-time output of the recording. Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
2017-11-23 11:01:45 +10:00
}
}
if (all && multiple) {
fprintf(stderr,
"Only one of --multiple and --all allowed.\n");
goto out;
}
tools: add a libinput-record tool This is a tool that does effectively the same job as evemu-record. evemu has two disadvantages: its API is clunky and hard to extend even for simple features. And it has a custom file format that requires special processing but is difficult to extend and hard to write manually. e.g. the bitmasks require keeping a line number state to know which bit an entry refers to. libinput-record records the same data but the output is YAML. That can be processed easier and extended in the future without breaking the parsing. We can (in the future) also interleave the evemu output with libinput's debug output, thus having a single file where the events can be compared and analysed without the need for replaying. Likewise, we can easily annotate the file with parsable bits of information without having to shove all that into a comment (like version numbers of libinput, kernel, etc). User-visible differences to evemu-record: * the output file requires an explicit -o or --output-file argument * no evemu-describe equivalent, if you just want the description simply cancel before any events are sent * to see key codes, a --show-keycodes flag must be supplied, otherwise all 'normal' keys end up as KEY_A. This protects against inadvertent information leakage * supports a --multiple option to record multiple devices simultaneously. All recordings have the same time offset, it is thus possible to reproduce bugs that depend on the interaction of more than one device. And to answer the question of: why a printf-approach to writing out yaml instead of a library, it's simply that we want to be able to have real-time output of the recording. Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
2017-11-23 11:01:45 +10:00
if (ctx.timeout > 0 && output_arg == NULL) {
fprintf(stderr,
"Option --autorestart requires --output-file\n");
goto out;
}
ctx.outfile = safe_strdup(output_arg);
ndevices = argc - optind;
if (multiple) {
if (output_arg == NULL) {
fprintf(stderr,
"Option --multiple requires --output-file\n");
goto out;
}
if (ndevices <= 1) {
fprintf(stderr,
"Option --multiple requires all device nodes on the commandline\n");
goto out;
}
for (int i = ndevices; i > 0; i -= 1) {
char *devnode = safe_strdup(argv[optind + i - 1]);
if (!init_device(&ctx, devnode))
goto out;
}
} else if (all) {
char **devices; /* NULL-terminated */
char **d;
if (output_arg == NULL) {
fprintf(stderr,
"Option --all requires --output-file\n");
goto out;
}
devices = all_devices();
d = devices;
while (*d) {
if (!init_device(&ctx, safe_strdup(*d))) {
strv_free(devices);
goto out;
}
d++;
}
strv_free(devices);
tools: add a libinput-record tool This is a tool that does effectively the same job as evemu-record. evemu has two disadvantages: its API is clunky and hard to extend even for simple features. And it has a custom file format that requires special processing but is difficult to extend and hard to write manually. e.g. the bitmasks require keeping a line number state to know which bit an entry refers to. libinput-record records the same data but the output is YAML. That can be processed easier and extended in the future without breaking the parsing. We can (in the future) also interleave the evemu output with libinput's debug output, thus having a single file where the events can be compared and analysed without the need for replaying. Likewise, we can easily annotate the file with parsable bits of information without having to shove all that into a comment (like version numbers of libinput, kernel, etc). User-visible differences to evemu-record: * the output file requires an explicit -o or --output-file argument * no evemu-describe equivalent, if you just want the description simply cancel before any events are sent * to see key codes, a --show-keycodes flag must be supplied, otherwise all 'normal' keys end up as KEY_A. This protects against inadvertent information leakage * supports a --multiple option to record multiple devices simultaneously. All recordings have the same time offset, it is thus possible to reproduce bugs that depend on the interaction of more than one device. And to answer the question of: why a printf-approach to writing out yaml instead of a library, it's simply that we want to be able to have real-time output of the recording. Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
2017-11-23 11:01:45 +10:00
} else {
char *path;
if (ndevices > 1) {
fprintf(stderr, "More than one device, do you want --multiple?\n");
goto out;
}
path = ndevices <= 0 ? select_device() : safe_strdup(argv[optind++]);
if (path == NULL) {
fprintf(stderr, "Invalid device path\n");
goto out;
}
if (!init_device(&ctx, path))
goto out;
}
rc = mainloop(&ctx);
out:
list_for_each_safe(d, tmp, &ctx.devices, link) {
free(d->devnode);
libevdev_free(d->evdev);
}
return rc;
}