[PATCH i-g-t v5] runner/executor: Abort if dmesg is flooded
vitaly prosyak
vprosyak at amd.com
Thu Aug 28 02:02:41 UTC 2025
The change looks good to me. Reviewed-by Vitaly Prosyak <vitaly.prosyak at amd.com>
Just small correction below for edge case.
On 2025-08-26 08:53, Kamil Konieczny wrote:
> Current disk limit triggers once when it is exceeded during test
> monitoring. After that happens executor no longer checks if
> kernel is still printing plenty of messages.
>
> Create a way to abort test in such scenarios with the help of
> measuring kernel kmsg activity before first test is executed
> and then also check it after exceeding disk limit.
>
> Also print info about it after each test ends.
>
> v2: fix error when kmsg open fails, fix reading proc (Kamil)
> changed calculation of max bps (Peter)
> v3: abort only when disk limit was actually exceeded,
> print info instead of aborting after every test (Kamil)
> v4: added abort at main monitoring loop, also cleaning up
> printing messages (Kamil)
> v5: fix typo, use less vars in a loop (Krzysztof)
> reused existing dmesg reading function (Kamil)
>
> Cc: Ewelina Musial <ewelina.musial at intel.com>
> Cc: Karol Krol <karol.krol at intel.com>
> Cc: Krzysztof Karas <krzysztof.karas at intel.com>
> Cc: Petri Latvala <adrinael at adrinael.net>
> Cc: Peter Senna Tschudin <peter.senna at linux.intel.com>
> Cc: Ryszard Knop <ryszard.knop at intel.com>
> Cc: Vitaly Prosyak <vitaly.prosyak at amd.com>
> Cc: "Zbigniew Kempczyński" <zbigniew.kempczynski at intel.com>
> Signed-off-by: Kamil Konieczny <kamil.konieczny at linux.intel.com>
> ---
> runner/executor.c | 143 +++++++++++++++++++++++++++++++++++++++++-----
> 1 file changed, 128 insertions(+), 15 deletions(-)
>
> diff --git a/runner/executor.c b/runner/executor.c
> index 847abe481..b11381f39 100644
> --- a/runner/executor.c
> +++ b/runner/executor.c
> @@ -637,6 +637,11 @@ const char *get_out_filename(int fid)
> return "output-filename-index-error";
> }
>
> +static int open_kmsg_rdonly(void)
> +{
> + return open("/dev/kmsg", O_RDONLY | O_CLOEXEC | O_NONBLOCK);
> +}
> +
> /* Returns the number of bytes written to disk, or a negative number on error */
> static long dump_dmesg(int kmsgfd, int outfd, ssize_t size)
> {
> @@ -646,6 +651,8 @@ static long dump_dmesg(int kmsgfd, int outfd, ssize_t size)
> * /dev/kmsg doesn't support seeking to -1 from SEEK_END
> * so we need to use a second fd to read a message to
> * match against, or stop when we reach EAGAIN.
> + * If outfd==0 it will not write anything but still count
> + * how many bytes was readed from dmesg.
> */
>
> int comparefd;
> @@ -711,7 +718,9 @@ static long dump_dmesg(int kmsgfd, int outfd, ssize_t size)
> return written;
> }
>
> - write(outfd, buf, r);
> + if (outfd)
> + write(outfd, buf, r);
> +
> written += r;
>
> if (comparefd < 0 && sscanf(buf, "%u,%llu,%llu,%c;",
> @@ -948,6 +957,76 @@ static size_t calc_last_dmesg_chunk(size_t limit, size_t disk_usage)
> return dt != 0 ? dt : -1;
> }
>
> +/*
> + * Tries to read from dmesg for 0.1 seconds
> + *
> + * Returns:
> + * =>0.0 - Success, measured kmsg activity in bytes/second
> + * -1.0 - Failure
> + */
> +static double measure_dmesg_bytes_per_sec(int wrfd)
> +{
> + struct timespec time_beg, time_now, nsec_sleep;
> + unsigned long long dmsg_read, rnow;
> + double time;
> + int kmsgfd;
> +
> + if ((kmsgfd = open_kmsg_rdonly()) < 0) {
> + errf("Warning: Cannot open /dev/kmsg\n");
> +
> + return -1.0;
> + }
> +
> + lseek(kmsgfd, 0, SEEK_END);
> + nsec_sleep.tv_sec = 0;
> + nsec_sleep.tv_nsec = 10ULL * 1000ULL * 1000ULL; /* 10^7 nanoseconds = 10^-2 sec */
> + runner_gettime(&time_beg);
> + time = 0.0;
> + dmsg_read = 0.0;
> + while (1) {
> + rnow = dump_dmesg(kmsgfd, wrfd, 64 * 4096); /* 64KB max */
> + if (rnow > 0)
> + dmsg_read += rnow;
> +
> + runner_gettime(&time_now);
> + time = igt_time_elapsed(&time_beg, &time_now);
> + if (time <= 0.0) {
> + errf("Warning: Time underflow\n");
> + return -1.0;
> + }
> +
> + if (time >= 0.1)
> + break;
> +
> + if (rnow == 0)
> + nanosleep(&nsec_sleep, NULL);
> + }
> +
> + runner_gettime(&time_now);
> + time = igt_time_elapsed(&time_beg, &time_now);
> + if (time <= 0.0) {
> + errf("Warning: Time underflow\n");
> +
> + return -1.0;
> + }
> +
> + return (double)dmsg_read / time;
> +}
> +
> +/* Returns: true if ratio lower or equal than maxratio */
> +static bool check_dmesg_ratio(double maxratio, int log_level, bool sync, int wrfd)
> +{
> + double new_dmesg_bps = measure_dmesg_bytes_per_sec(wrfd);
> +
In the current patch, if /dev/kmsg can’t be opened, measure_dmesg_bytes_per_sec() returns -1.0, but later code still compares it against max_dmesg_ratio. That means a negative value could accidentally be treated as “ok” and cause misleading results.
The safe fix is to short-circuit inside check_dmesg_ratio() whenever measure_dmesg_bytes_per_sec() returns a negative number. For example:
static bool check_dmesg_ratio(double maxratio, int log_level, bool sync, int wrfd)
{
double new_dmesg_bps = measure_dmesg_bytes_per_sec(wrfd);
if (new_dmesg_bps < 0.0) {
if (log_level >= LOG_LEVEL_NORMAL) {
outf("Dmesg ratio: unavailable (no /dev/kmsg)\n");
if (sync)
fflush(stdout);
}
return true; /* don’t gate on missing measurement */
}
if (log_level >= LOG_LEVEL_NORMAL) {
outf("Dmesg ratio %0.3fKB/s\n", new_dmesg_bps / 1024);
if (sync)
fflush(stdout);
}
return new_dmesg_bps <= maxratio;
}
> + if (log_level >= LOG_LEVEL_NORMAL) {
> + outf("Dmesg ratio %0.3fKB/s\n", new_dmesg_bps / 1024);
> + if (sync)
> + fflush(stdout);
> + }
> +
> + return new_dmesg_bps <= maxratio;
> +}
> +
> /*
> * Returns:
> * =0 - Success
> @@ -960,6 +1039,7 @@ static int monitor_output(pid_t child,
> int *outputs,
> double *time_spent,
> struct settings *settings,
> + double max_dmesg_ratio,
> char **abortreason,
> bool *abort_already_written)
> {
> @@ -1463,29 +1543,31 @@ static int monitor_output(pid_t child,
> * exceeded the disk usage limit.
> */
> if (killed && disk_usage_limit_exceeded(settings, disk_usage)) {
> + char killmsg[256];
> +
> timeoutresult = false;
> + snprintf(killmsg, sizeof(killmsg),
> + "runner: This test was killed due to exceeding disk usage limit. "
> + "(Used %zd bytes, limit %zd)\n",
> + disk_usage,
> + settings->disk_usage_limit);
>
> if (socket_comms_used) {
> struct runnerpacket *message;
> - char killmsg[256];
>
> - snprintf(killmsg, sizeof(killmsg),
> - "runner: This test was killed due to exceeding disk usage limit. "
> - "(Used %zd bytes, limit %zd)\n",
> - disk_usage,
> - settings->disk_usage_limit);
> message = runnerpacket_log(STDOUT_FILENO, killmsg);
> write_packet_with_canary(outputs[_F_SOCKET], message, settings->sync);
> free(message);
> } else {
> - dprintf(outputs[_F_OUT],
> - "\nrunner: This test was killed due to exceeding disk usage limit. "
> - "(Used %zd bytes, limit %zd)\n",
> - disk_usage,
> - settings->disk_usage_limit);
> + dprintf(outputs[_F_OUT], "%s", killmsg);
> if (settings->sync)
> fdatasync(outputs[_F_OUT]);
> }
> +
> + if (!check_dmesg_ratio(max_dmesg_ratio, settings->log_level, settings->sync, outputs[_F_DMESG])) {
> + asprintf(abortreason, "Dmesg ratio exceeded during test run.");
> + aborting = true;
> + }
> }
>
> if (socket_comms_used) {
> @@ -1587,7 +1669,7 @@ static int monitor_output(pid_t child,
> dmesgwritten = dump_dmesg(kmsgfd, outputs[_F_DMESG], dmsg_chunk_size);
> if (settings->sync)
> fdatasync(outputs[_F_DMESG]);
> - if (dmesgwritten > 0) {
> + if (dmesgwritten > 0 && !aborting) {
> disk_usage += dmesgwritten;
> if (settings->disk_usage_limit && disk_usage > settings->disk_usage_limit) {
> char disk[1024];
> @@ -1599,6 +1681,11 @@ static int monitor_output(pid_t child,
> } else if (killed) {
> errf("%s", disk);
> }
> +
> + if (!check_dmesg_ratio(max_dmesg_ratio, settings->log_level, settings->sync, outputs[_F_DMESG])) {
> + asprintf(abortreason, "Dmesg ratio exceeded after test ends.");
> + aborting = true;
> + }
> }
> }
>
> @@ -1766,6 +1853,7 @@ static int execute_next_entry(struct execute_state *state,
> size_t total,
> double *time_spent,
> struct settings *settings,
> + double max_dmesg_ratio,
> struct job_list_entry *entry,
> int testdirfd, int resdirfd,
> int sigfd, sigset_t *sigmask,
> @@ -1814,7 +1902,7 @@ static int execute_next_entry(struct execute_state *state,
> goto out_pipe;
> }
>
> - if ((kmsgfd = open("/dev/kmsg", O_RDONLY | O_CLOEXEC | O_NONBLOCK)) < 0) {
> + if ((kmsgfd = open_kmsg_rdonly()) < 0) {
> errf("Warning: Cannot open /dev/kmsg\n");
> } else {
> /* TODO: Checking of abort conditions in pre-execute dmesg */
> @@ -1885,7 +1973,7 @@ static int execute_next_entry(struct execute_state *state,
>
> result = monitor_output(child, outfd, errfd, socketfd,
> kmsgfd, sigfd,
> - outputs, time_spent, settings,
> + outputs, time_spent, settings, max_dmesg_ratio,
> abortreason, abort_already_written);
>
> out_kmsgfd:
> @@ -2418,6 +2506,8 @@ bool execute(struct execute_state *state,
> struct settings *settings,
> struct job_list *job_list)
> {
> + static double dmesg_bps = -1.0;
> + static double max_dmesg_bps = -1.0;
> int resdirfd, testdirfd, unamefd, timefd, sigfd;
> struct environment_variable *env_var;
> struct utsname unamebuf;
> @@ -2560,6 +2650,26 @@ bool execute(struct execute_state *state,
> }
> }
>
> + if (max_dmesg_bps < 0.0) {
> + double ncpu_bps = 4 * 1024 * max_t(size_t, sysconf(_SC_NPROCESSORS_ONLN), 4);
> + double set_bps = 0.0;
> +
> + dmesg_bps = measure_dmesg_bytes_per_sec(0);
> +
> + if (settings->disk_usage_limit > 0 && settings->per_test_timeout > 0)
> + set_bps = (double)settings->disk_usage_limit / (double)settings->per_test_timeout;
> +
> + outf("Dmesg KB/s ratio settings:%0.3f ncpu:%0.3f current:%0.3f\n",
> + set_bps / 1024, ncpu_bps / 1024, dmesg_bps / 1024);
> +
> + if (set_bps > 0.0)
> + max_dmesg_bps = set_bps;
> + else
> + max_dmesg_bps = dmesg_bps > ncpu_bps ? dmesg_bps : ncpu_bps;
> +
> + outf("Using max level dmesg ratio %0.3fKB/s\n", max_dmesg_bps / 1024);
> + }
> +
> for (; state->next < job_list->size;
> state->next++) {
> char *reason = NULL;
> @@ -2596,6 +2706,7 @@ bool execute(struct execute_state *state,
> job_list->size,
> &time_spent,
> settings,
> + max_dmesg_bps,
> &job_list->entries[state->next],
> testdirfd, resdirfd,
> sigfd, &sigmask,
> @@ -2651,6 +2762,8 @@ bool execute(struct execute_state *state,
> break;
> }
>
> + check_dmesg_ratio(max_dmesg_bps, settings->log_level, settings->sync, 0);
> +
> if (result > 0) {
> double time_left = state->time_left;
>
More information about the igt-dev
mailing list