* Re: Problem with gettimeofday [not found] <CALkMTNnMEujZy7H2tv+iUApHK+10b3Q3aNu4E+cxy=U-uE37ew@mail.gmail.com> @ 2014-05-02 2:52 ` Alexander Samilovskih 2014-05-02 8:26 ` Sami Kerola 0 siblings, 1 reply; 4+ messages in thread From: Alexander Samilovskih @ 2014-05-02 2:52 UTC (permalink / raw) To: util-linux Hi, There are several places where gettimeofday should be replaced by clock_gettime(CLOCK_MONOTONIC_RAW). According to manual pages gettimeofday affected by discontinues jumps made by system administrator, can be slewed by ntpd or kinda stuff. So when we subtract two different timestamps from gettimeofday we can get confusing results sys-utils/eject.c libmount/lock.c maybe some other places... diff --git a/include/timeutils.h b/include/timeutils.h index bcae613..7cf972d 100644 --- a/include/timeutils.h +++ b/include/timeutils.h @@ -51,5 +51,6 @@ typedef uint64_t nsec_t; #define FORMAT_TIMESPAN_MAX 64 int parse_timestamp(const char *t, usec_t *usec); +int gettime_monotonic(struct timeval *tv); #endif /* UTIL_LINUX_TIME_UTIL_H */ diff --git a/lib/timeutils.c b/lib/timeutils.c index 7fe6218..06ae597 100644 --- a/lib/timeutils.c +++ b/lib/timeutils.c @@ -336,3 +336,24 @@ int parse_timestamp(const char *t, usec_t *usec) return 0; } + +int gettime_monotonic(struct timeval *tv) +{ +#ifdef CLOCK_MONOTONIC + //Can slew only by ntp and adjtime + int ret; + struct timespec ts; +#ifdef CLOCK_MONOTONIC_RAW + //Linux specific, cant slew + if (!(ret = clock_gettime(CLOCK_MONOTONIC_RAW, &ts))) { +#else + if (!(ret = clock_gettime(CLOCK_MONOTONIC, &ts))) { +#endif + tv->tv_sec = ts.tv_sec; + tv->tv_usec = ts.tv_nsec / 1000; + } + return ret; +#else + return gettimeofday(tv, NULL); +#endif +} \ No newline at end of file diff --git a/libmount/src/lock.c b/libmount/src/lock.c index 8667db9..41f8ed1 100644 --- a/libmount/src/lock.c +++ b/libmount/src/lock.c @@ -25,6 +25,7 @@ #include "closestream.h" #include "pathnames.h" #include "mountP.h" +#include "timeutils.h" /* * lock handler @@ -258,7 +259,7 @@ static int mnt_wait_mtab_lock(struct libmnt_lock *ml, struct flock *fl, time_t m struct sigaction sa, osa; int ret = 0; - gettimeofday(&now, NULL); + gettime_monotonic(&now); if (now.tv_sec >= maxtime) return 1; /* timeout */ @@ -408,7 +409,7 @@ static int lock_mtab(struct libmnt_lock *ml) } close(i); - gettimeofday(&maxtime, NULL); + gettime_monotonic(&maxtime); maxtime.tv_sec += MOUNTLOCK_MAXTIME; waittime.tv_sec = 0; @@ -434,7 +435,7 @@ static int lock_mtab(struct libmnt_lock *ml) if (ml->lockfile_fd < 0) { /* Strange... Maybe the file was just deleted? */ int errsv = errno; - gettimeofday(&now, NULL); + gettime_monotonic(&now); if (errsv == ENOENT && now.tv_sec < maxtime.tv_sec) { ml->locked = 0; continue; diff --git a/sys-utils/eject.c b/sys-utils/eject.c index 03744c7..049cdb7 100644 --- a/sys-utils/eject.c +++ b/sys-utils/eject.c @@ -52,6 +52,7 @@ #include "xalloc.h" #include "pathnames.h" #include "sysfs.h" +#include "timeutils.h" /* * sg_io_hdr_t driver_status -- see kernel include/scsi/scsi.h @@ -460,7 +461,7 @@ static void toggle_tray(int fd) * needed. In my experience the function needs less than 0.05 * seconds if the tray was already open, and at least 1.5 seconds * if it was closed. */ - gettimeofday(&time_start, NULL); + gettime_monotonic(&time_start); /* Send the CDROMEJECT command to the device. */ if (!eject_cdrom(fd)) @@ -468,7 +469,7 @@ static void toggle_tray(int fd) /* Get the second timestamp, to measure the time needed to open * the tray. */ - gettimeofday(&time_stop, NULL); + gettime_monotonic(&time_stop); time_elapsed = (time_stop.tv_sec * 1000000 + time_stop.tv_usec) - (time_start.tv_sec * 1000000 + time_start.tv_usec); ^ permalink raw reply related [flat|nested] 4+ messages in thread
* Re: Problem with gettimeofday 2014-05-02 2:52 ` Problem with gettimeofday Alexander Samilovskih @ 2014-05-02 8:26 ` Sami Kerola 2014-05-02 12:12 ` Alexander Samilovskih 2014-11-18 13:43 ` Karel Zak 0 siblings, 2 replies; 4+ messages in thread From: Sami Kerola @ 2014-05-02 8:26 UTC (permalink / raw) To: Alexander Samilovskih; +Cc: util-linux On 2 May 2014 03:52, Alexander Samilovskih <alexsamilovskih@gmail.com> wrote: > There are several places where gettimeofday should be replaced by > clock_gettime(CLOCK_MONOTONIC_RAW). According to manual pages > gettimeofday affected by discontinues jumps made by system > administrator, can be slewed by ntpd or kinda stuff. So when we > subtract two different timestamps from gettimeofday we can get > confusing results > > sys-utils/eject.c > libmount/lock.c > maybe some other places... Hi Alexander, The diff you sent does not seem to work. fatal: corrupt patch at line 123 The format you get from $ git format-patch HEAD~1 would be appreciated. > diff --git a/include/timeutils.h b/include/timeutils.h > index bcae613..7cf972d 100644 > --- a/include/timeutils.h > +++ b/include/timeutils.h > @@ -51,5 +51,6 @@ typedef uint64_t nsec_t; > #define FORMAT_TIMESPAN_MAX 64 > > int parse_timestamp(const char *t, usec_t *usec); > +int gettime_monotonic(struct timeval *tv); > > #endif /* UTIL_LINUX_TIME_UTIL_H */ > diff --git a/lib/timeutils.c b/lib/timeutils.c > index 7fe6218..06ae597 100644 > --- a/lib/timeutils.c > +++ b/lib/timeutils.c > @@ -336,3 +336,24 @@ int parse_timestamp(const char *t, usec_t *usec) > > return 0; > } > + > +int gettime_monotonic(struct timeval *tv) > +{ > +#ifdef CLOCK_MONOTONIC > + //Can slew only by ntp and adjtime Could you use /* comment */ style rather than // from C++ world. > + int ret; > + struct timespec ts; > +#ifdef CLOCK_MONOTONIC_RAW > + //Linux specific, cant slew > + if (!(ret = clock_gettime(CLOCK_MONOTONIC_RAW, &ts))) { > +#else > + if (!(ret = clock_gettime(CLOCK_MONOTONIC, &ts))) { > +#endif > + tv->tv_sec = ts.tv_sec; > + tv->tv_usec = ts.tv_nsec / 1000; > + } > + return ret; > +#else A comment could ease understanding what is going on, such as #else /* CLOCK_MONOTONIC is not available */ > + return gettimeofday(tv, NULL); > +#endif > +} Th rest looks pretty ok. -- Sami Kerola http://www.iki.fi/kerolasa/ ^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: Problem with gettimeofday 2014-05-02 8:26 ` Sami Kerola @ 2014-05-02 12:12 ` Alexander Samilovskih 2014-11-18 13:43 ` Karel Zak 1 sibling, 0 replies; 4+ messages in thread From: Alexander Samilovskih @ 2014-05-02 12:12 UTC (permalink / raw) To: kerolasa; +Cc: util-linux 2014-05-02 8:26 GMT+00:00 Sami Kerola <kerolasa@iki.fi>: > On 2 May 2014 03:52, Alexander Samilovskih <alexsamilovskih@gmail.com> wrote: >> There are several places where gettimeofday should be replaced by >> clock_gettime(CLOCK_MONOTONIC_RAW). According to manual pages >> gettimeofday affected by discontinues jumps made by system >> administrator, can be slewed by ntpd or kinda stuff. So when we >> subtract two different timestamps from gettimeofday we can get >> confusing results >> >> sys-utils/eject.c >> libmount/lock.c >> maybe some other places... > > Hi Alexander, > > The diff you sent does not seem to work. > > fatal: corrupt patch at line 123 > > The format you get from > > $ git format-patch HEAD~1 > > would be appreciated. > >> diff --git a/include/timeutils.h b/include/timeutils.h >> index bcae613..7cf972d 100644 >> --- a/include/timeutils.h >> +++ b/include/timeutils.h >> @@ -51,5 +51,6 @@ typedef uint64_t nsec_t; >> #define FORMAT_TIMESPAN_MAX 64 >> >> int parse_timestamp(const char *t, usec_t *usec); >> +int gettime_monotonic(struct timeval *tv); >> >> #endif /* UTIL_LINUX_TIME_UTIL_H */ >> diff --git a/lib/timeutils.c b/lib/timeutils.c >> index 7fe6218..06ae597 100644 >> --- a/lib/timeutils.c >> +++ b/lib/timeutils.c >> @@ -336,3 +336,24 @@ int parse_timestamp(const char *t, usec_t *usec) >> >> return 0; >> } >> + >> +int gettime_monotonic(struct timeval *tv) >> +{ >> +#ifdef CLOCK_MONOTONIC >> + //Can slew only by ntp and adjtime > > Could you use /* comment */ style rather than // from C++ world. > >> + int ret; >> + struct timespec ts; >> +#ifdef CLOCK_MONOTONIC_RAW >> + //Linux specific, cant slew >> + if (!(ret = clock_gettime(CLOCK_MONOTONIC_RAW, &ts))) { >> +#else >> + if (!(ret = clock_gettime(CLOCK_MONOTONIC, &ts))) { >> +#endif >> + tv->tv_sec = ts.tv_sec; >> + tv->tv_usec = ts.tv_nsec / 1000; >> + } >> + return ret; >> +#else > > A comment could ease understanding what is going on, such as > > #else /* CLOCK_MONOTONIC is not available */ > >> + return gettimeofday(tv, NULL); >> +#endif >> +} > > Th rest looks pretty ok. > > -- > Sami Kerola > http://www.iki.fi/kerolasa/ Fixed, git format-patch HEAD~3 >From b641e268ab0aa1597e46f79accd169c388f06c7c Mon Sep 17 00:00:00 2001 From: Alex Samilovskih <alexsamilovskih@gmail.com> Date: Fri, 2 May 2014 18:15:12 +0000 Subject: [PATCH 1/3] Add support for monotonic clock --- include/timeutils.h | 1 + lib/timeutils.c | 24 ++++++++++++++++++++++++ 2 files changed, 25 insertions(+) diff --git a/include/timeutils.h b/include/timeutils.h index bcae613..7cf972d 100644 --- a/include/timeutils.h +++ b/include/timeutils.h @@ -51,5 +51,6 @@ typedef uint64_t nsec_t; #define FORMAT_TIMESPAN_MAX 64 int parse_timestamp(const char *t, usec_t *usec); +int gettime_monotonic(struct timeval *tv); #endif /* UTIL_LINUX_TIME_UTIL_H */ diff --git a/lib/timeutils.c b/lib/timeutils.c index 7fe6218..b3fb87b 100644 --- a/lib/timeutils.c +++ b/lib/timeutils.c @@ -336,3 +336,27 @@ int parse_timestamp(const char *t, usec_t *usec) return 0; } + +/* Get the current time that is not a subject to resetting and drifting */ +int gettime_monotonic(struct timeval *tv) +{ +#ifdef CLOCK_MONOTONIC + int ret; + struct timespec ts; + +#ifdef CLOCK_MONOTONIC_RAW + if (!(ret = clock_gettime(CLOCK_MONOTONIC_RAW, &ts))) { +#else /* CLOCK_MONOTONIC_RAW is not available */ + if (!(ret = clock_gettime(CLOCK_MONOTONIC, &ts))) { +#endif + tv->tv_sec = ts.tv_sec; + tv->tv_usec = ts.tv_nsec / 1000; + } + + return ret; + +#else /* CLOCK_MONOTONIC is not available */ + + return gettimeofday(tv, NULL); +#endif +} -- 1.9.2 >From b72b47b2f8e12c4cb53e817939bad464139e9042 Mon Sep 17 00:00:00 2001 From: Alex Samilovskih <alexsamilovskih@gmail.com> Date: Fri, 2 May 2014 18:21:20 +0000 Subject: [PATCH 2/3] Add support for monotonic clock in eject --- sys-utils/eject.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/sys-utils/eject.c b/sys-utils/eject.c index 03744c7..049cdb7 100644 --- a/sys-utils/eject.c +++ b/sys-utils/eject.c @@ -52,6 +52,7 @@ #include "xalloc.h" #include "pathnames.h" #include "sysfs.h" +#include "timeutils.h" /* * sg_io_hdr_t driver_status -- see kernel include/scsi/scsi.h @@ -460,7 +461,7 @@ static void toggle_tray(int fd) * needed. In my experience the function needs less than 0.05 * seconds if the tray was already open, and at least 1.5 seconds * if it was closed. */ - gettimeofday(&time_start, NULL); + gettime_monotonic(&time_start); /* Send the CDROMEJECT command to the device. */ if (!eject_cdrom(fd)) @@ -468,7 +469,7 @@ static void toggle_tray(int fd) /* Get the second timestamp, to measure the time needed to open * the tray. */ - gettimeofday(&time_stop, NULL); + gettime_monotonic(&time_stop); time_elapsed = (time_stop.tv_sec * 1000000 + time_stop.tv_usec) - (time_start.tv_sec * 1000000 + time_start.tv_usec); -- 1.9.2 >From b8ca74bfdb93034e9d1c52ab8702ff02dab8787c Mon Sep 17 00:00:00 2001 From: Alex Samilovskih <alexsamilovskih@gmail.com> Date: Fri, 2 May 2014 18:28:31 +0000 Subject: [PATCH 3/3] Add support for monotonic clock in libmount --- libmount/src/lock.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/libmount/src/lock.c b/libmount/src/lock.c index 8667db9..41f8ed1 100644 --- a/libmount/src/lock.c +++ b/libmount/src/lock.c @@ -25,6 +25,7 @@ #include "closestream.h" #include "pathnames.h" #include "mountP.h" +#include "timeutils.h" /* * lock handler @@ -258,7 +259,7 @@ static int mnt_wait_mtab_lock(struct libmnt_lock *ml, struct flock *fl, time_t m struct sigaction sa, osa; int ret = 0; - gettimeofday(&now, NULL); + gettime_monotonic(&now); if (now.tv_sec >= maxtime) return 1; /* timeout */ @@ -408,7 +409,7 @@ static int lock_mtab(struct libmnt_lock *ml) } close(i); - gettimeofday(&maxtime, NULL); + gettime_monotonic(&maxtime); maxtime.tv_sec += MOUNTLOCK_MAXTIME; waittime.tv_sec = 0; @@ -434,7 +435,7 @@ static int lock_mtab(struct libmnt_lock *ml) if (ml->lockfile_fd < 0) { /* Strange... Maybe the file was just deleted? */ int errsv = errno; - gettimeofday(&now, NULL); + gettime_monotonic(&now); if (errsv == ENOENT && now.tv_sec < maxtime.tv_sec) { ml->locked = 0; continue; @@ -646,7 +647,7 @@ int test_lock(struct libmnt_test *ts, int argc, char *argv[]) /* start the test in exactly defined time */ if (synctime) { - gettimeofday(&tv, NULL); + gettimeofday(&tv); if (synctime && synctime - tv.tv_sec > 1) { usecs = ((synctime - tv.tv_sec) * 1000000UL) - (1000000UL - tv.tv_usec); -- 1.9.2 ^ permalink raw reply related [flat|nested] 4+ messages in thread
* Re: Problem with gettimeofday 2014-05-02 8:26 ` Sami Kerola 2014-05-02 12:12 ` Alexander Samilovskih @ 2014-11-18 13:43 ` Karel Zak 1 sibling, 0 replies; 4+ messages in thread From: Karel Zak @ 2014-11-18 13:43 UTC (permalink / raw) To: kerolasa; +Cc: Alexander Samilovskih, util-linux On Fri, May 02, 2014 at 09:26:02AM +0100, Sami Kerola wrote: > On 2 May 2014 03:52, Alexander Samilovskih <alexsamilovskih@gmail.com> wrote: > > There are several places where gettimeofday should be replaced by > > clock_gettime(CLOCK_MONOTONIC_RAW). According to manual pages > > gettimeofday affected by discontinues jumps made by system > > administrator, can be slewed by ntpd or kinda stuff. So when we > > subtract two different timestamps from gettimeofday we can get > > confusing results > > > > sys-utils/eject.c > > libmount/lock.c > > maybe some other places... > > Hi Alexander, > > The diff you sent does not seem to work. > > fatal: corrupt patch at line 123 > > The format you get from > > $ git format-patch HEAD~1 > > would be appreciated. It seems Alexander does not care anymore, so I have fixed and applied the patch. Karel -- Karel Zak <kzak@redhat.com> http://karelzak.blogspot.com ^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2014-11-18 13:43 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
[not found] <CALkMTNnMEujZy7H2tv+iUApHK+10b3Q3aNu4E+cxy=U-uE37ew@mail.gmail.com>
2014-05-02 2:52 ` Problem with gettimeofday Alexander Samilovskih
2014-05-02 8:26 ` Sami Kerola
2014-05-02 12:12 ` Alexander Samilovskih
2014-11-18 13:43 ` Karel Zak
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox