Message ID | 20230207124008.1788759-1-ycliang@andestech.com |
---|---|
State | Accepted |
Headers | show |
Series | [1/1] tst_pids.c: fix fork failure on small memory systems | expand |
Hi Leo, > Running syscalls/msgstress03 on a system with less than ~1 GB of RAM fails: > msgstress03 1 TFAIL : msgstress03.c:163: Fork failed (may be OK if under stress) > The reason is that besides /proc/sys/kernel/pid_max, /proc/sys/kernel/threads-max > is another factor determining how many processes a system could create. > Use the smaller number between pid_max and threads-max as the nprocs. Makes sense. Reviewed-by: Petr Vorel <pvorel@suse.cz> BTW IMHO it'd be better to rewrite f = popen("ps -eT | wc -l", "r"); to list count of /proc/[0-9]*/ in plain C (TODO for us). Kind regards, Petr
On Tue, Feb 7, 2023 at 2:42 PM Petr Vorel <pvorel@suse.cz> wrote: > > Hi Leo, > > > Running syscalls/msgstress03 on a system with less than ~1 GB of RAM fails: > > > msgstress03 1 TFAIL : msgstress03.c:163: Fork failed (may be OK if under stress) > > > The reason is that besides /proc/sys/kernel/pid_max, /proc/sys/kernel/threads-max > > is another factor determining how many processes a system could create. > > > Use the smaller number between pid_max and threads-max as the nprocs. > > Makes sense. > Reviewed-by: Petr Vorel <pvorel@suse.cz> > Acked-by: Jan Stancek <jstancek@redhat.com>
Hi! > lib/tst_pid.c | 5 ++++- > 1 file changed, 4 insertions(+), 1 deletion(-) > > diff --git a/lib/tst_pid.c b/lib/tst_pid.c > index 21cadef2a..3fb3f39ad 100644 > --- a/lib/tst_pid.c > +++ b/lib/tst_pid.c > @@ -31,6 +31,7 @@ > #include "tst_safe_macros.h" > > #define PID_MAX_PATH "/proc/sys/kernel/pid_max" > +#define THREADS_MAX_PATH "/proc/sys/kernel/threads-max" > #define CGROUPS_V1_SLICE_FMT "/sys/fs/cgroup/pids/user.slice/user-%d.slice/pids.max" > #define CGROUPS_V2_SLICE_FMT "/sys/fs/cgroup/user.slice/user-%d.slice/pids.max" > /* Leave some available processes for the OS */ > @@ -113,7 +114,7 @@ static int get_session_pids_limit(void (*cleanup_fn) (void)) > int tst_get_free_pids_(void (*cleanup_fn) (void)) > { > FILE *f; > - int rc, used_pids, max_pids, max_session_pids; > + int rc, used_pids, max_pids, max_session_pids, max_threads; > > f = popen("ps -eT | wc -l", "r"); > if (!f) { > @@ -129,6 +130,8 @@ int tst_get_free_pids_(void (*cleanup_fn) (void)) > } > > SAFE_FILE_SCANF(cleanup_fn, PID_MAX_PATH, "%d", &max_pids); > + SAFE_FILE_SCANF(cleanup_fn, THREADS_MAX_PATH, "%d", &max_threads); > + max_pids = (max_pids < max_threads) ? max_pids : max_threads; BTW: We do have a MIN() macro definition in tst_minmax.h header so we can just do max_pids = MIN(max_pids, max_threads); > max_session_pids = get_session_pids_limit(cleanup_fn); > if ((max_session_pids > 0) && (max_session_pids < max_pids)) Otherwise: Reviewed-by: Cyril Hrubis <chrubis@suse.cz>
Hi Petr, On Tue, Feb 07, 2023 at 02:42:22PM +0100, Petr Vorel wrote: > Hi Leo, > > > Running syscalls/msgstress03 on a system with less than ~1 GB of RAM fails: > > > msgstress03 1 TFAIL : msgstress03.c:163: Fork failed (may be OK if under stress) > > > The reason is that besides /proc/sys/kernel/pid_max, /proc/sys/kernel/threads-max > > is another factor determining how many processes a system could create. > > > Use the smaller number between pid_max and threads-max as the nprocs. > > Makes sense. > Reviewed-by: Petr Vorel <pvorel@suse.cz> > > BTW IMHO it'd be better to rewrite > f = popen("ps -eT | wc -l", "r"); > to list count of /proc/[0-9]*/ in plain C (TODO for us). > Thanks for the review! I could send a v2 to add this as well. Best regards, Leo > Kind regards, > Petr
Hi Cyril, On Thu, Feb 09, 2023 at 02:05:46PM +0100, Cyril Hrubis wrote: > Hi! > > lib/tst_pid.c | 5 ++++- > > 1 file changed, 4 insertions(+), 1 deletion(-) > > > > diff --git a/lib/tst_pid.c b/lib/tst_pid.c > > index 21cadef2a..3fb3f39ad 100644 > > --- a/lib/tst_pid.c > > +++ b/lib/tst_pid.c > > @@ -31,6 +31,7 @@ > > #include "tst_safe_macros.h" > > > > #define PID_MAX_PATH "/proc/sys/kernel/pid_max" > > +#define THREADS_MAX_PATH "/proc/sys/kernel/threads-max" > > #define CGROUPS_V1_SLICE_FMT "/sys/fs/cgroup/pids/user.slice/user-%d.slice/pids.max" > > #define CGROUPS_V2_SLICE_FMT "/sys/fs/cgroup/user.slice/user-%d.slice/pids.max" > > /* Leave some available processes for the OS */ > > @@ -113,7 +114,7 @@ static int get_session_pids_limit(void (*cleanup_fn) (void)) > > int tst_get_free_pids_(void (*cleanup_fn) (void)) > > { > > FILE *f; > > - int rc, used_pids, max_pids, max_session_pids; > > + int rc, used_pids, max_pids, max_session_pids, max_threads; > > > > f = popen("ps -eT | wc -l", "r"); > > if (!f) { > > @@ -129,6 +130,8 @@ int tst_get_free_pids_(void (*cleanup_fn) (void)) > > } > > > > SAFE_FILE_SCANF(cleanup_fn, PID_MAX_PATH, "%d", &max_pids); > > + SAFE_FILE_SCANF(cleanup_fn, THREADS_MAX_PATH, "%d", &max_threads); > > + max_pids = (max_pids < max_threads) ? max_pids : max_threads; > > BTW: We do have a MIN() macro definition in tst_minmax.h header so we > can just do max_pids = MIN(max_pids, max_threads); > Thanks for the review! Will send a v2 to fix this! Best regards, Leo > > max_session_pids = get_session_pids_limit(cleanup_fn); > > if ((max_session_pids > 0) && (max_session_pids < max_pids)) > > Otherwise: > > Reviewed-by: Cyril Hrubis <chrubis@suse.cz> > > -- > Cyril Hrubis > chrubis@suse.cz
Hi all, Thanks Leo, I merged it with MIN(max_pids, max_threads), thanks Cyril for catching it. > > BTW IMHO it'd be better to rewrite > > f = popen("ps -eT | wc -l", "r"); > > to list count of /proc/[0-9]*/ in plain C (TODO for us). > Thanks for the review! > I could send a v2 to add this as well. Thank you, if you have time to send another patch to fix. Kind regards, Petr > Best regards, > Leo
diff --git a/lib/tst_pid.c b/lib/tst_pid.c index 21cadef2a..3fb3f39ad 100644 --- a/lib/tst_pid.c +++ b/lib/tst_pid.c @@ -31,6 +31,7 @@ #include "tst_safe_macros.h" #define PID_MAX_PATH "/proc/sys/kernel/pid_max" +#define THREADS_MAX_PATH "/proc/sys/kernel/threads-max" #define CGROUPS_V1_SLICE_FMT "/sys/fs/cgroup/pids/user.slice/user-%d.slice/pids.max" #define CGROUPS_V2_SLICE_FMT "/sys/fs/cgroup/user.slice/user-%d.slice/pids.max" /* Leave some available processes for the OS */ @@ -113,7 +114,7 @@ static int get_session_pids_limit(void (*cleanup_fn) (void)) int tst_get_free_pids_(void (*cleanup_fn) (void)) { FILE *f; - int rc, used_pids, max_pids, max_session_pids; + int rc, used_pids, max_pids, max_session_pids, max_threads; f = popen("ps -eT | wc -l", "r"); if (!f) { @@ -129,6 +130,8 @@ int tst_get_free_pids_(void (*cleanup_fn) (void)) } SAFE_FILE_SCANF(cleanup_fn, PID_MAX_PATH, "%d", &max_pids); + SAFE_FILE_SCANF(cleanup_fn, THREADS_MAX_PATH, "%d", &max_threads); + max_pids = (max_pids < max_threads) ? max_pids : max_threads; max_session_pids = get_session_pids_limit(cleanup_fn); if ((max_session_pids > 0) && (max_session_pids < max_pids))
Running syscalls/msgstress03 on a system with less than ~1 GB of RAM fails: msgstress03 1 TFAIL : msgstress03.c:163: Fork failed (may be OK if under stress) The reason is that besides /proc/sys/kernel/pid_max, /proc/sys/kernel/threads-max is another factor determining how many processes a system could create. Use the smaller number between pid_max and threads-max as the nprocs. Signed-off-by: Leo Yu-Chi Liang <ycliang@andestech.com> --- lib/tst_pid.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)