For review: pthread_setschedparam.3

For review: pthread_setschedparam.3

[Michael Kerrisk]

Here's another pthreads page looking for reviewers (and testers for
the example program)

thanks,

Michael

.\" Copyright (c) 2008 Linux Foundation, written by Michael Kerrisk
.\"     <mtk.manpages-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
.\"
.\" Permission is granted to make and distribute verbatim copies of this
.\" manual provided the copyright notice and this permission notice are
.\" preserved on all copies.
.\"
.\" Permission is granted to copy and distribute modified versions of this
.\" manual under the conditions for verbatim copying, provided that the
.\" entire resulting derived work is distributed under the terms of a
.\" permission notice identical to this one.
.\"
.\" Since the Linux kernel and libraries are constantly changing, this
.\" manual page may be incorrect or out-of-date.  The author(s) assume no
.\" responsibility for errors or omissions, or for damages resulting from
.\" the use of the information contained herein.  The author(s) may not
.\" have taken the same level of care in the production of this manual,
.\" which is licensed free of charge, as they might when working
.\" professionally.
.\"
.\" Formatted or processed versions of this manual, if unaccompanied by
.\" the source, must acknowledge the copyright and authors of this work.
.\"
.TH PTHREAD_SETSCHEDPARAM 3 2008-11-07 "Linux" "Linux Programmer's Manual"
.SH NAME
pthread_setschedparam, pthread_setschedparam \- set/get
scheduling policy and parameters of a thread
.SH SYNOPSIS
.nf
.B #include <pthread.h>

.BI "pthread_setschedparam(pthread_t " thread ", int " policy ,
.BI "                      const struct sched_param *" param );
.BI "pthread_getschedparam(pthread_t " thread ", int *" policy ,
.BI "                      struct sched_param *" param );
.sp
Compile and link with \fI\-pthread\fP.
.SH DESCRIPTION
The
.BR pthread_setschedparam ()
function sets the scheduling policy and parameters of the thread
.IR thread .

.I policy
specifies the new scheduling policy for
.IR thread .
The supported values for
.IR policy ,
and their semantics, are described in
.BR sched_setscheduler (2).
.\" FIXME . pthread_setschedparam() places no restriction on the policy,
.\" but pthread_attr_setschedpolicy() restricts policy to RR/FIFO/OTHER
.\" http://sourceware.org/bugzilla/show_bug.cgi?id=7013

The structure pointed to by
.I param
specifies the new scheduling parameters for
.IR thread .
Scheduling parameters are maintained in the following structure:

.in +4n
.nf
struct sched_param {
    int sched_priority;     /* Scheduling priority */
};
.fi
.in

As can be seen, only one scheduling parameter is supported
(this is the only parameter specified by POSIX.1-2001.)
For details of the permitted ranges for scheduling priorities
in each scheduling policy, see
.BR sched_setscheduler (2).

The
.BR pthread_getschedparam ()
function returns the scheduling policy and parameters of the thread
.IR thread ,
in the buffers pointed to by
.I policy
and
.IR param ,
respectively.
The returned priority value is that set by the most recent
.BR pthread_setschedparam (),
.BR pthread_setschedprio (3),
or
.BR pthread_create (3)
call that affected
.IR thread .
The returned priority does not reflect any temporary priority adjustments
as a result of calls to any priority inheritance or
priority ceiling functions (see, for example,
.BR pthread_mutexattr_setprioceiling (3)
and
.BR pthread_mutexattr_setprotocol (3)).
.\" FIXME . nptl/pthread_setschedparam.c has the following
.\"   /* If the thread should have higher priority because of some
.\"      PTHREAD_PRIO_PROTECT mutexes it holds, adjust the priority. */
.\" Eventually (perhaps after writing the mutexattr pages), we
.\" may want to add something on the topic to this page.
.SH RETURN VALUE
On success, these functions return 0;
on error, they return a non-zero error number.
If
.BR pthread_setschedparam ()
fails, the scheduling policy and parameters of
.I thread
are not changed.
.SH ERRORS
Both of these functions may fail with the following error:
.TP
.B ESRCH
No thread with the ID
.I thread
could be found.
.PP
.BR pthread_setschedparam ()
may additionally fail with the following errors:
.TP
.B EINVAL
.I policy
is not a recognized policy, or
.I param
does not make sense for the
.IR policy .
.TP
.B EPERM
The caller does not have appropriate privileges
.RB ( CAP_SYS_NICE )
to set the specified scheduling policy and parameters.
See
.BR sched_setscheduler (2)
for more details on what scheduling changes require privileges.
.PP
POSIX.1-2001 also documents an
.B ENOTSUP
("attempt was made to set the policy or scheduling parameters
to an unsupported value") error for
.BR pthread_setschedparam ().
.\" .SH VERSIONS
.\" Available since glibc 2.0
.SH CONFORMING TO
POSIX.1-2001.
.SH NOTES
See
.BR sched_setscheduler (2)
for a description of the effect of changing a thread's
scheduling policy and priority.
.SH EXAMPLE
The program below demonstrates the use of
.BR pthread_setschedparam ()
and
.BR pthread_getschedparam (),
as well as the use of a number of other scheduling-related
pthreads functions.

In the following run, the main thread sets its scheduling policy to
.BR SCHED_FIFO
with a priority of 10,
and initializes a thread attributes object with
a scheduling policy attribute of
.BR SCHED_RR
and a scheduling priority attribute of 20.
The program then sets (using
.BR pthread_attr_setinheritsched (3))
the inherit scheduler attribute of the thread attributes object to
.BR PTHREAD_EXPLICIT_SCHED ,
meaning that threads created using this attributes object should
take their scheduling attributes from the thread attributes object.
The program then creates a thread using the thread attributes object,
and that thread displays its scheduling policy and priority.
.in +4n
.nf

$ \fBsu\fP      # Need privilege to set real-time scheduling policies
Password:
# \fB./a.out \-mf10 \-ar20 \-i e\fP
Scheduler settings of main thread
    policy=SCHED_FIFO, priority=10

Scheduler settings in \(aqattr\(aq
    policy=SCHED_RR, priority=20
    inheritsched is EXPLICIT

Scheduler attributes of new thread
    policy=SCHED_RR, priority=20
.fi
.in

In the above output, one can see that the scheduling policy and priority
were taken from the values specified in the thread attributes object.

The next run is the same as the previous,
except that the inherit scheduler attribute is set to
.BR PTHREAD_INHERIT_SCHED ,
meaning that threads created using the thread attributes object should
ignore the scheduling attributes specified in the attributes object
and instead take their scheduling attributes from the creating thread.

.in +4n
.nf
# \fB./a.out \-mf10 \-ar20 \-i i\fP
Scheduler settings of main thread
    policy=SCHED_FIFO, priority=10

Scheduler settings in \(aqattr\(aq
    policy=SCHED_RR, priority=20
    inheritsched is INHERIT

Scheduler attributes of new thread
    policy=SCHED_FIFO, priority=10
.fi
.in

In the above output, one can see that the scheduling policy and priority
were taken from the creating thread,
rather than the thread attributes object.
.SS Program source
\&
.nf
/* pthreads_sched_test.c */

#include <pthread.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <errno.h>

/* Simple error handling functions */

#define errExit(msg)            { perror(msg); exit(EXIT_FAILURE); }

#define errExitEN(en, msg)      { errno = en; perror(msg); \\
                                  exit(EXIT_FAILURE); }

static void
usage(char *prog_name, char *msg)
{
    if (msg != NULL)
        fputs(msg, stderr);

    fprintf(stderr, "Usage: %s [options]\\n", prog_name);
    fprintf(stderr, "Options are:\\n");
#define fpe(msg) fprintf(stderr, "\\t%s", msg);          /* Shorter */
    fpe("\-a<policy><prio> Set scheduling policy and priority in\\n");
    fpe("                 thread attributes object\\n");
    fpe("                 <policy> can be\\n");
    fpe("                     f  SCHED_FIFO\\n");
    fpe("                     r  SCHED_RR\\n");
    fpe("                     o  SCHED_OTHER\\n");
    fpe("\-A               Use default thread attributes object\\n");
    fpe("\-i {e|s}         Set inherit scheduler attribute to\\n");
    fpe("                 \(aqexplicit\(aq or \(aqinherit\(aq\\n");
    fpe("\-m<policy><prio> Set scheduling policy and priority on\\n");
    fpe("                 main thread before pthread_create() call\\n");
    exit(EXIT_FAILURE);
} /* usage */

static int
get_policy(char p, int *policy)
{
    switch (p) {
    case \(aqf\(aq: *policy = SCHED_FIFO;     return 1;
    case \(aqr\(aq: *policy = SCHED_RR;       return 1;
    case \(aqo\(aq: *policy = SCHED_OTHER;    return 1;
    default:  return 0;
    }
} /* get_policy */

static void
display_sched_attr(int policy, struct sched_param *param)
{
    printf("    policy=%s, priority=%d\\n",
            (policy == SCHED_FIFO)  ? "SCHED_FIFO" :
            (policy == SCHED_RR)    ? "SCHED_RR" :
            (policy == SCHED_OTHER) ? "SCHED_OTHER" :
            "???",
            param\->sched_priority);
} /* display_sched_attr */

static void
display_thread_sched_attr(char *msg)
{
    int policy, s;
    struct sched_param param;

    s = pthread_getschedparam(pthread_self(), &policy, &param);
    if (s != 0)
        errExitEN(s, "pthread_getschedparam");

    printf("%s\\n", msg);
    display_sched_attr(policy, &param);
} /* display_thread_sched_attr */

static void *
thread_start(void *arg)
{
    display_thread_sched_attr("Scheduler attributes of new thread");

    return NULL;
} /* thread_start */

int
main(int argc, char *argv[])
{
    int s, opt, inheritsched, use_null_attrib, policy;
    pthread_t thread;
    pthread_attr_t attr;
    pthread_attr_t *attrp;
    char *attr_sched_str, *main_sched_str, *inheritsched_str;
    struct sched_param param;

    /* Process command\-line options */

    use_null_attrib = 0;
    attr_sched_str = NULL;
    main_sched_str = NULL;
    inheritsched_str = NULL;

    while ((opt = getopt(argc, argv, "a:Ai:m:")) != \-1) {
        switch (opt) {
        case \(aqa\(aq: attr_sched_str = optarg;      break;
        case \(aqA\(aq: use_null_attrib = 1;          break;
        case \(aqi\(aq: inheritsched_str = optarg;    break;
        case \(aqm\(aq: main_sched_str = optarg;      break;
        default:  usage(argv[0], "Unrecognized option\\n");
        }
    }

    if (use_null_attrib &&
            (inheritsched_str != NULL || attr_sched_str != NULL))
        usage(argv[0], "Can\(aqt specify \-A with \-i or \-a\\n");

    /* Optionally set scheduling attributes of main thread,
       and display the attributes */

    if (main_sched_str != NULL) {
        if (!get_policy(main_sched_str[0], &policy))
            usage(argv[0], "Bad policy for main thread (\-s)\\n");
        param.sched_priority = strtol(&main_sched_str[1], NULL, 0);

        s = pthread_setschedparam(pthread_self(), policy, &param);
        if (s != 0)
            errExitEN(s, "pthread_setschedparam");
    }

    display_thread_sched_attr("Scheduler settings of main thread");
    printf("\\n");

    /* Initialize thread attributes object according to options */

    attrp = NULL;

    if (!use_null_attrib) {
        s = pthread_attr_init(&attr);
        if (s != 0)
            errExitEN(s, "pthread_attr_init");
        attrp = &attr;
    }

    if (inheritsched_str != NULL) {
        if (inheritsched_str[0] == \(aqe\(aq)
            inheritsched = PTHREAD_EXPLICIT_SCHED;
        else if (inheritsched_str[0] == \(aqi\(aq)
            inheritsched = PTHREAD_INHERIT_SCHED;
        else
            usage(argv[0], "Value for \-i must be \(aqe\(aq or \(aqi\(aq\\n");

        s = pthread_attr_setinheritsched(&attr, inheritsched);
        if (s != 0)
            errExitEN(s, "pthread_attr_setinheritsched");
    }

    if (attr_sched_str != NULL) {
        if (!get_policy(attr_sched_str[0], &policy))
            usage(argv[0],
                    "Bad policy for \(aqattr\(aq (\-a)\\n");
        param.sched_priority = strtol(&attr_sched_str[1], NULL, 0);

        s = pthread_attr_setschedpolicy(&attr, policy);
        if (s != 0)
            errExitEN(s, "pthread_attr_setschedpolicy");
        s = pthread_attr_setschedparam(&attr, &param);
        if (s != 0)
            errExitEN(s, "pthread_attr_setschedparam");
    }

    /* If we initialized a thread attributes object, display
       the scheduling attributes that were set in the object */

    if (attrp != NULL) {
        s = pthread_attr_getschedparam(&attr, &param);
        if (s != 0)
            errExitEN(s, "pthread_attr_getschedparam");
        s = pthread_attr_getschedpolicy(&attr, &policy);
        if (s != 0)
            errExitEN(s, "pthread_attr_getschedpolicy");

        printf("Scheduler settings in \(aqattr\(aq\\n");
        display_sched_attr(policy, &param);

        s = pthread_attr_getinheritsched(&attr, &inheritsched);
        printf("    inheritsched is %s\\n",
                (inheritsched == PTHREAD_INHERIT_SCHED)  ? "INHERIT" :
                (inheritsched == PTHREAD_EXPLICIT_SCHED) ? "EXPLICIT" :
                "???");
        printf("\\n");
    }

    /* Create a thread that will display its scheduling attributes */

    s = pthread_create(&thread, attrp, &thread_start, NULL);
    if (s != 0)
        errExitEN(s, "pthread_create");

    /* Destroy unneeded thread attributes object */

    s = pthread_attr_destroy(&attr);
    if (s != 0)
        errExitEN(s, "pthread_attr_destroy");

    s = pthread_join(thread, NULL);
    if (s != 0)
        errExitEN(s, "pthread_join");

    exit(EXIT_SUCCESS);
} /* main */
.fi
.SH SEE ALSO
.BR sched_get_priority_min (2),
.BR sched_setscheduler (2),
.BR pthread_attr_init (3),
.BR pthread_attr_setinheritsched (3),
.BR pthread_attr_setschedparam (3),
.BR pthread_attr_setschedpolicy (3),
.BR pthread_setschedprio (3),
.BR pthread_create (3),
.BR pthread_self (3),
.BR pthreads (7)
--
To unsubscribe from this list: send the line "unsubscribe linux-man" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: For review: pthread_setschedparam.3

[Loic Domaigne]

Hallo Michael,

enclosed my review comments.

Regards,
Loïc.
--

> .\" Copyright (c) 2008 Linux Foundation, written by Michael Kerrisk
> .\"     <mtk.manpages-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
> .\"
> .\" Permission is granted to make and distribute verbatim copies of this
> .\" manual provided the copyright notice and this permission notice are
> .\" preserved on all copies.
> .\"
> .\" Permission is granted to copy and distribute modified versions of this
> .\" manual under the conditions for verbatim copying, provided that the
> .\" entire resulting derived work is distributed under the terms of a
> .\" permission notice identical to this one.
> .\"
> .\" Since the Linux kernel and libraries are constantly changing, this
> .\" manual page may be incorrect or out-of-date.  The author(s) assume no
> .\" responsibility for errors or omissions, or for damages resulting from
> .\" the use of the information contained herein.  The author(s) may not
> .\" have taken the same level of care in the production of this manual,
> .\" which is licensed free of charge, as they might when working
> .\" professionally.
> .\"
> .\" Formatted or processed versions of this manual, if unaccompanied by
> .\" the source, must acknowledge the copyright and authors of this work.
> .\"
> .TH PTHREAD_SETSCHEDPARAM 3 2008-11-07 "Linux" "Linux Programmer's Manual"
> .SH NAME
> pthread_setschedparam, pthread_setschedparam \- set/get
> scheduling policy and parameters of a thread
> .SH SYNOPSIS
> .nf
> .B #include <pthread.h>
> 
> .BI "pthread_setschedparam(pthread_t " thread ", int " policy ,
> .BI "                      const struct sched_param *" param );
> .BI "pthread_getschedparam(pthread_t " thread ", int *" policy ,
> .BI "                      struct sched_param *" param );
> .sp
> Compile and link with \fI\-pthread\fP.

One question regarding the synopsis. Is there a reason with the 
"restrict" keyword is not used for pthread_getschedparam(). Accordingly 
to /usr/include/pthread.h:

extern int pthread_getschedparam (pthread_t __target_thread,
                                   int *__restrict __policy,
                                   struct sched_param *__restrict __param);


> .SH DESCRIPTION
> The
> .BR pthread_setschedparam ()
> function sets the scheduling policy and parameters of the thread
> .IR thread .
> 
> .I policy
> specifies the new scheduling policy for
> .IR thread .
> The supported values for
> .IR policy ,
> and their semantics, are described in
> .BR sched_setscheduler (2).
> .\" FIXME . pthread_setschedparam() places no restriction on the policy,
> .\" but pthread_attr_setschedpolicy() restricts policy to RR/FIFO/OTHER
> .\" http://sourceware.org/bugzilla/show_bug.cgi?id=7013

Perhaps it's my poor english... But why "new scheduling policy"? I may 
just want to change the priority, within the same scheduling policy...

> The structure pointed to by
> .I param
> specifies the new scheduling parameters for
> .IR thread .

Similar comment here. I may just want to change the scheduling policy, 
without changing the priority. ( I admit, this one is seldom compared to 
the previous one ).

> Scheduling parameters are maintained in the following structure:
> 
> .in +4n
> .nf
> struct sched_param {
>     int sched_priority;     /* Scheduling priority */
> };
> .fi
> .in
> 
> As can be seen, only one scheduling parameter is supported
> (this is the only parameter specified by POSIX.1-2001.)

No, it think this is not correct? If a system support SCHED_SPORADIC, 
then the sched_param structure has additional fields.


> For details of the permitted ranges for scheduling priorities
> in each scheduling policy, see
> .BR sched_setscheduler (2).

> The
> .BR pthread_getschedparam ()
> function returns the scheduling policy and parameters of the thread
> .IR thread ,
> in the buffers pointed to by
> .I policy
> and
> .IR param ,
> respectively.
> The returned priority value is that set by the most recent
> .BR pthread_setschedparam (),
> .BR pthread_setschedprio (3),
> or
> .BR pthread_create (3)
> call that affected
> .IR thread .

Hmm, that's perfectly right from a POSIX point of view. Knowing how 
Linux implements threads, I have been interested about the effect of 
sched_setscheduler() on a MT-process (since NPTL uses 1:1 model, this 
should be a NOP).

I tested the following program against the stable glibc-2.7... 
Apparently, it seems that sched_setscheduler() might affect the main 
thread priority as well.

--
#include <stdio.h>
#include <pthread.h>
#include <sched.h>
#include <unistd.h>

void
print_schedinfo(const char* thread)
{
    struct sched_param param;
    int policy;
    int rc;

    rc = pthread_getschedparam(pthread_self(), &policy, &param);
    if (rc!=0) printf("##%d\n", rc);
    printf("%s > Policy=%s, prio=%d\n",
           thread,
           (policy==SCHED_FIFO) ? "FIFO" : "*NOT* FIFO",
	  param.sched_priority);
}

// dummy thread...
//
void*
thread(void* ignore)
{
    sleep(3);
    print_schedinfo("dummy thread");
    pthread_exit(NULL);
}

int
main()
{
    struct sched_param param;
    int    policy;
    int    rc;
    pthread_t tid;

    // create dummy thread
    //
    pthread_create(&tid, NULL, thread, NULL);
    param.sched_priority=1;

    // now we shall change the process policy/prio using
    // sched_setscheduler().
    // Normally: this should be a NOP. But due to the way Linux
    // implements threads, I am suspecting that this shall affect
    // the main thread
    //
    rc=sched_setscheduler(0, SCHED_FIFO, &param);
    if (rc==-1) printf("sched_setscheduler FAILED\n");

    // print my scheduling info
    //
    print_schedinfo("main");

    // join dummy thread and terminate
    //
    pthread_join(tid, NULL);
    return 0;
}

--

> The returned priority does not reflect any temporary priority adjustments
> as a result of calls to any priority inheritance or
> priority ceiling functions (see, for example,
> .BR pthread_mutexattr_setprioceiling (3)
> and
> .BR pthread_mutexattr_setprotocol (3)).
> .\" FIXME . nptl/pthread_setschedparam.c has the following
> .\"   /* If the thread should have higher priority because of some
> .\"      PTHREAD_PRIO_PROTECT mutexes it holds, adjust the priority. */
> .\" Eventually (perhaps after writing the mutexattr pages), we
> .\" may want to add something on the topic to this page.

If I understood correctly, the thread can't lower the priority that he 
has resulting from priority inheritance/ceiling. That makes perfectly 
sense to me, otherwise this could defeat the purpose of such schemes, 
among others avoiding priority inversion.

> .SH RETURN VALUE
> On success, these functions return 0;
> on error, they return a non-zero error number.
> If
> .BR pthread_setschedparam ()
> fails, the scheduling policy and parameters of
> .I thread
> are not changed.
> .SH ERRORS
> Both of these functions may fail with the following error:
> .TP
> .B ESRCH
> No thread with the ID
> .I thread
> could be found.

Should the ERSCH error description be consistent across pthread_* linux
man-pages? See for instance pthread_attr_setaffinity_np().

> .PP
> .BR pthread_setschedparam ()
> may additionally fail with the following errors:
> .TP
> .B EINVAL
> .I policy
> is not a recognized policy, or
> .I param
> does not make sense for the
> .IR policy .

I got troubled by the "may additionally", as "may" has a particular 
meaning in POSIX.1...

But I guess, you just want to express that pthread_setschedparam() shall 
fail if the policy or the param is invalid, right?

> .TP
> .B EPERM
> The caller does not have appropriate privileges
> .RB ( CAP_SYS_NICE )
> to set the specified scheduling policy and parameters.
> See
> .BR sched_setscheduler (2)
> for more details on what scheduling changes require privileges.
> .PP
> POSIX.1-2001 also documents an
> .B ENOTSUP
> ("attempt was made to set the policy or scheduling parameters
> to an unsupported value") error for
> .BR pthread_setschedparam ().

... but it doesn't seem to be used by Linux/Glibc...

> .\" .SH VERSIONS
> .\" Available since glibc 2.0
> .SH CONFORMING TO
> POSIX.1-2001.
> .SH NOTES
> See
> .BR sched_setscheduler (2)
> for a description of the effect of changing a thread's
> scheduling policy and priority.
> .SH EXAMPLE
> The program below demonstrates the use of
> .BR pthread_setschedparam ()
> and
> .BR pthread_getschedparam (),
> as well as the use of a number of other scheduling-related
> pthreads functions.
> 
> In the following run, the main thread sets its scheduling policy to
> .BR SCHED_FIFO
> with a priority of 10,
> and initializes a thread attributes object with
> a scheduling policy attribute of
> .BR SCHED_RR
> and a scheduling priority attribute of 20.
> The program then sets (using
> .BR pthread_attr_setinheritsched (3))
> the inherit scheduler attribute of the thread attributes object to
> .BR PTHREAD_EXPLICIT_SCHED ,
> meaning that threads created using this attributes object should
> take their scheduling attributes from the thread attributes object.
> The program then creates a thread using the thread attributes object,
> and that thread displays its scheduling policy and priority.
> .in +4n
> .nf
> 
> $ \fBsu\fP      # Need privilege to set real-time scheduling policies
> Password:
> # \fB./a.out \-mf10 \-ar20 \-i e\fP
> Scheduler settings of main thread
>     policy=SCHED_FIFO, priority=10
> 
> Scheduler settings in \(aqattr\(aq
>     policy=SCHED_RR, priority=20
>     inheritsched is EXPLICIT
> 
> Scheduler attributes of new thread
>     policy=SCHED_RR, priority=20
> .fi
> .in
> 
> In the above output, one can see that the scheduling policy and priority
> were taken from the values specified in the thread attributes object.
> 
> The next run is the same as the previous,
> except that the inherit scheduler attribute is set to
> .BR PTHREAD_INHERIT_SCHED ,
> meaning that threads created using the thread attributes object should
> ignore the scheduling attributes specified in the attributes object
> and instead take their scheduling attributes from the creating thread.
> 
> .in +4n
> .nf
> # \fB./a.out \-mf10 \-ar20 \-i i\fP
> Scheduler settings of main thread
>     policy=SCHED_FIFO, priority=10
> 
> Scheduler settings in \(aqattr\(aq
>     policy=SCHED_RR, priority=20
>     inheritsched is INHERIT
> 
> Scheduler attributes of new thread
>     policy=SCHED_FIFO, priority=10
> .fi
> .in
> 
> In the above output, one can see that the scheduling policy and priority
> were taken from the creating thread,
> rather than the thread attributes object.

A classical trap is that people don't set inheritsched to explicit, and 
by default it is inherit... You could illustrate this by the following 
example: ./a.out -mf10 -ar20.

( I don't know however if it is the right place to speak about such 
things ).


> .SS Program source
> \&
> .nf
> /* pthreads_sched_test.c */
> 
> #include <pthread.h>
> #include <stdio.h>
> #include <stdlib.h>
> #include <unistd.h>
> #include <errno.h>
> 
> /* Simple error handling functions */
> 
> #define errExit(msg)            { perror(msg); exit(EXIT_FAILURE); }
> 
> #define errExitEN(en, msg)      { errno = en; perror(msg); \\
>                                   exit(EXIT_FAILURE); }
> 
> static void
> usage(char *prog_name, char *msg)
> {
>     if (msg != NULL)
>         fputs(msg, stderr);
> 
>     fprintf(stderr, "Usage: %s [options]\\n", prog_name);
>     fprintf(stderr, "Options are:\\n");
> #define fpe(msg) fprintf(stderr, "\\t%s", msg);          /* Shorter */
>     fpe("\-a<policy><prio> Set scheduling policy and priority in\\n");
>     fpe("                 thread attributes object\\n");
>     fpe("                 <policy> can be\\n");
>     fpe("                     f  SCHED_FIFO\\n");
>     fpe("                     r  SCHED_RR\\n");
>     fpe("                     o  SCHED_OTHER\\n");
>     fpe("\-A               Use default thread attributes object\\n");
>     fpe("\-i {e|s}         Set inherit scheduler attribute to\\n");
>     fpe("                 \(aqexplicit\(aq or \(aqinherit\(aq\\n");
>     fpe("\-m<policy><prio> Set scheduling policy and priority on\\n");
>     fpe("                 main thread before pthread_create() call\\n");
>     exit(EXIT_FAILURE);
> } /* usage */
> 
> static int
> get_policy(char p, int *policy)
> {
>     switch (p) {
>     case \(aqf\(aq: *policy = SCHED_FIFO;     return 1;
>     case \(aqr\(aq: *policy = SCHED_RR;       return 1;
>     case \(aqo\(aq: *policy = SCHED_OTHER;    return 1;
>     default:  return 0;
>     }
> } /* get_policy */
> 
> static void
> display_sched_attr(int policy, struct sched_param *param)
> {
>     printf("    policy=%s, priority=%d\\n",
>             (policy == SCHED_FIFO)  ? "SCHED_FIFO" :
>             (policy == SCHED_RR)    ? "SCHED_RR" :
>             (policy == SCHED_OTHER) ? "SCHED_OTHER" :
>             "???",
>             param\->sched_priority);
> } /* display_sched_attr */
> 
> static void
> display_thread_sched_attr(char *msg)
> {
>     int policy, s;
>     struct sched_param param;
> 
>     s = pthread_getschedparam(pthread_self(), &policy, &param);
>     if (s != 0)
>         errExitEN(s, "pthread_getschedparam");
> 
>     printf("%s\\n", msg);
>     display_sched_attr(policy, &param);
> } /* display_thread_sched_attr */
> 
> static void *
> thread_start(void *arg)
> {
>     display_thread_sched_attr("Scheduler attributes of new thread");
> 
>     return NULL;
> } /* thread_start */
> 
> int
> main(int argc, char *argv[])
> {
>     int s, opt, inheritsched, use_null_attrib, policy;
>     pthread_t thread;
>     pthread_attr_t attr;
>     pthread_attr_t *attrp;
>     char *attr_sched_str, *main_sched_str, *inheritsched_str;
>     struct sched_param param;
> 
>     /* Process command\-line options */
> 
>     use_null_attrib = 0;
>     attr_sched_str = NULL;
>     main_sched_str = NULL;
>     inheritsched_str = NULL;
> 
>     while ((opt = getopt(argc, argv, "a:Ai:m:")) != \-1) {
>         switch (opt) {
>         case \(aqa\(aq: attr_sched_str = optarg;      break;
>         case \(aqA\(aq: use_null_attrib = 1;          break;
>         case \(aqi\(aq: inheritsched_str = optarg;    break;
>         case \(aqm\(aq: main_sched_str = optarg;      break;
>         default:  usage(argv[0], "Unrecognized option\\n");
>         }
>     }
> 
>     if (use_null_attrib &&
>             (inheritsched_str != NULL || attr_sched_str != NULL))
>         usage(argv[0], "Can\(aqt specify \-A with \-i or \-a\\n");
> 
>     /* Optionally set scheduling attributes of main thread,
>        and display the attributes */
> 
>     if (main_sched_str != NULL) {
>         if (!get_policy(main_sched_str[0], &policy))
>             usage(argv[0], "Bad policy for main thread (\-s)\\n");
>         param.sched_priority = strtol(&main_sched_str[1], NULL, 0);
> 
>         s = pthread_setschedparam(pthread_self(), policy, &param);
>         if (s != 0)
>             errExitEN(s, "pthread_setschedparam");
>     }
> 
>     display_thread_sched_attr("Scheduler settings of main thread");
>     printf("\\n");
> 
>     /* Initialize thread attributes object according to options */
> 
>     attrp = NULL;
> 
>     if (!use_null_attrib) {
>         s = pthread_attr_init(&attr);
>         if (s != 0)
>             errExitEN(s, "pthread_attr_init");
>         attrp = &attr;
>     }
> 
>     if (inheritsched_str != NULL) {
>         if (inheritsched_str[0] == \(aqe\(aq)
>             inheritsched = PTHREAD_EXPLICIT_SCHED;
>         else if (inheritsched_str[0] == \(aqi\(aq)
>             inheritsched = PTHREAD_INHERIT_SCHED;
>         else
>             usage(argv[0], "Value for \-i must be \(aqe\(aq or \(aqi\(aq\\n");
> 
>         s = pthread_attr_setinheritsched(&attr, inheritsched);
>         if (s != 0)
>             errExitEN(s, "pthread_attr_setinheritsched");
>     }
> 
>     if (attr_sched_str != NULL) {
>         if (!get_policy(attr_sched_str[0], &policy))
>             usage(argv[0],
>                     "Bad policy for \(aqattr\(aq (\-a)\\n");
>         param.sched_priority = strtol(&attr_sched_str[1], NULL, 0);
> 
>         s = pthread_attr_setschedpolicy(&attr, policy);
>         if (s != 0)
>             errExitEN(s, "pthread_attr_setschedpolicy");
>         s = pthread_attr_setschedparam(&attr, &param);
>         if (s != 0)
>             errExitEN(s, "pthread_attr_setschedparam");
>     }
> 
>     /* If we initialized a thread attributes object, display
>        the scheduling attributes that were set in the object */
> 
>     if (attrp != NULL) {
>         s = pthread_attr_getschedparam(&attr, &param);
>         if (s != 0)
>             errExitEN(s, "pthread_attr_getschedparam");
>         s = pthread_attr_getschedpolicy(&attr, &policy);
>         if (s != 0)
>             errExitEN(s, "pthread_attr_getschedpolicy");
> 
>         printf("Scheduler settings in \(aqattr\(aq\\n");
>         display_sched_attr(policy, &param);
> 
>         s = pthread_attr_getinheritsched(&attr, &inheritsched);
>         printf("    inheritsched is %s\\n",
>                 (inheritsched == PTHREAD_INHERIT_SCHED)  ? "INHERIT" :
>                 (inheritsched == PTHREAD_EXPLICIT_SCHED) ? "EXPLICIT" :
>                 "???");
>         printf("\\n");
>     }
> 
>     /* Create a thread that will display its scheduling attributes */
> 
>     s = pthread_create(&thread, attrp, &thread_start, NULL);
>     if (s != 0)
>         errExitEN(s, "pthread_create");
> 
>     /* Destroy unneeded thread attributes object */
> 
>     s = pthread_attr_destroy(&attr);
>     if (s != 0)
>         errExitEN(s, "pthread_attr_destroy");
> 
>     s = pthread_join(thread, NULL);
>     if (s != 0)
>         errExitEN(s, "pthread_join");
> 
>     exit(EXIT_SUCCESS);
> } /* main */
> .fi
> .SH SEE ALSO
> .BR sched_get_priority_min (2),
> .BR sched_setscheduler (2),
> .BR pthread_attr_init (3),
> .BR pthread_attr_setinheritsched (3),
> .BR pthread_attr_setschedparam (3),
> .BR pthread_attr_setschedpolicy (3),
> .BR pthread_setschedprio (3),
> .BR pthread_create (3),
> .BR pthread_self (3),
> .BR pthreads (7)
> 
--
To unsubscribe from this list: send the line "unsubscribe linux-man" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: For review: pthread_setschedparam.3

[Michael Kerrisk]

Hi Loïc,

On Sat, Nov 15, 2008 at 4:09 PM, Loic Domaigne <tech-Z4JMKDdsf89Wk0Htik3J/w@public.gmane.org> wrote:
> Hallo Michael,
>
> enclosed my review comments.


Thanks for yet another review!

[...]

>> .TH PTHREAD_SETSCHEDPARAM 3 2008-11-07 "Linux" "Linux Programmer's Manual"
>> .SH NAME
>> pthread_setschedparam, pthread_setschedparam \- set/get
>> scheduling policy and parameters of a thread
>> .SH SYNOPSIS
>> .nf
>> .B #include <pthread.h>
>>
>> .BI "pthread_setschedparam(pthread_t " thread ", int " policy ,
>> .BI "                      const struct sched_param *" param );
>> .BI "pthread_getschedparam(pthread_t " thread ", int *" policy ,
>> .BI "                      struct sched_param *" param );
>> .sp
>> Compile and link with \fI\-pthread\fP.
>
> One question regarding the synopsis. Is there a reason with the "restrict"
> keyword is not used for pthread_getschedparam(). Accordingly to
> /usr/include/pthread.h:

The only real reason is that to date *no* man pages include it.

> extern int pthread_getschedparam (pthread_t __target_thread,
>                                  int *__restrict __policy,
>                                  struct sched_param *__restrict __param);
>
>
[...]

>> .I policy
>> specifies the new scheduling policy for
>> .IR thread .
>> The supported values for
>> .IR policy ,
>> and their semantics, are described in
>> .BR sched_setscheduler (2).
>> .\" FIXME . pthread_setschedparam() places no restriction on the policy,
>> .\" but pthread_attr_setschedpolicy() restricts policy to RR/FIFO/OTHER
>> .\" http://sourceware.org/bugzilla/show_bug.cgi?id=7013
>
> Perhaps it's my poor english... But why "new scheduling policy"? I may just
> want to change the priority, within the same scheduling policy...

I see what you are saying, but I'm not sure this matters too much --
yes, the new policy can be the same policy as the old.

>> The structure pointed to by
>> .I param
>> specifies the new scheduling parameters for
>> .IR thread .
>
> Similar comment here. I may just want to change the scheduling policy,
> without changing the priority. ( I admit, this one is seldom compared to the
> previous one ).

Same comment as above ;-).

>> Scheduling parameters are maintained in the following structure:
>>
>> .in +4n
>> .nf
>> struct sched_param {
>>    int sched_priority;     /* Scheduling priority */
>> };
>> .fi
>> .in
>>
>> As can be seen, only one scheduling parameter is supported
>> (this is the only parameter specified by POSIX.1-2001.)
>
> No, it think this is not correct? If a system support SCHED_SPORADIC, then
> the sched_param structure has additional fields.

Yes.  Strictly speaking you are right.  Of course, Linux doesn't
support SCHED_SPORADIC.  I think the simplest solution is just to
remove the part "(this is the only parameter specified by
POSIX.1-2001.)", and I've also done that in
pthread_attr_setschedparam.3.

[...]

>> The
>> .BR pthread_getschedparam ()
>> function returns the scheduling policy and parameters of the thread
>> .IR thread ,
>> in the buffers pointed to by
>> .I policy
>> and
>> .IR param ,
>> respectively.
>> The returned priority value is that set by the most recent
>> .BR pthread_setschedparam (),
>> .BR pthread_setschedprio (3),
>> or
>> .BR pthread_create (3)
>> call that affected
>> .IR thread .
>
> Hmm, that's perfectly right from a POSIX point of view. Knowing how Linux
> implements threads, I have been interested about the effect of
> sched_setscheduler() on a MT-process (since NPTL uses 1:1 model, this should
> be a NOP).
>
> I tested the following program against the stable glibc-2.7... Apparently,
> it seems that sched_setscheduler() might affect the main thread priority as
> well.
>
> --
> #include <stdio.h>
> #include <pthread.h>
> #include <sched.h>
> #include <unistd.h>
>
> void
> print_schedinfo(const char* thread)
> {
>   struct sched_param param;
>   int policy;
>   int rc;
>
>   rc = pthread_getschedparam(pthread_self(), &policy, &param);
>   if (rc!=0) printf("##%d\n", rc);
>   printf("%s > Policy=%s, prio=%d\n",
>          thread,
>          (policy==SCHED_FIFO) ? "FIFO" : "*NOT* FIFO",
>          param.sched_priority);
> }
>
> // dummy thread...
> //
> void*
> thread(void* ignore)
> {
>   sleep(3);
>   print_schedinfo("dummy thread");
>   pthread_exit(NULL);
> }
>
> int
> main()
> {
>   struct sched_param param;
>   int    policy;
>   int    rc;
>   pthread_t tid;
>
>   // create dummy thread
>   //
>   pthread_create(&tid, NULL, thread, NULL);
>   param.sched_priority=1;
>
>   // now we shall change the process policy/prio using
>   // sched_setscheduler().
>   // Normally: this should be a NOP. But due to the way Linux
>   // implements threads, I am suspecting that this shall affect
>   // the main thread
>   //
>   rc=sched_setscheduler(0, SCHED_FIFO, &param);
>   if (rc==-1) printf("sched_setscheduler FAILED\n");
>
>   // print my scheduling info
>   //
>   print_schedinfo("main");
>
>   // join dummy thread and terminate
>   //
>   pthread_join(tid, NULL);
>   return 0;
> }

I have to look into this further.

>> The returned priority does not reflect any temporary priority adjustments
>> as a result of calls to any priority inheritance or
>> priority ceiling functions (see, for example,
>> .BR pthread_mutexattr_setprioceiling (3)
>> and
>> .BR pthread_mutexattr_setprotocol (3)).
>> .\" FIXME . nptl/pthread_setschedparam.c has the following
>> .\"   /* If the thread should have higher priority because of some
>> .\"      PTHREAD_PRIO_PROTECT mutexes it holds, adjust the priority. */
>> .\" Eventually (perhaps after writing the mutexattr pages), we
>> .\" may want to add something on the topic to this page.
>
> If I understood correctly, the thread can't lower the priority that he has
> resulting from priority inheritance/ceiling. That makes perfectly sense to
> me, otherwise this could defeat the purpose of such schemes, among others
> avoiding priority inversion.

Okay.  I may revise the text later (once I write the other pages).

[...]

>> .B ESRCH
>> No thread with the ID
>> .I thread
>> could be found.
>
> Should the ERSCH error description be consistent across pthread_* linux
> man-pages? See for instance pthread_attr_setaffinity_np().

Yes, I haven't been too consistent on that point.  I changed the ESRCH
text in all the pages to be just:

"No thread with the ID _thread_ could be found."

>> .PP
>> .BR pthread_setschedparam ()
>> may additionally fail with the following errors:
>> .TP
>> .B EINVAL
>> .I policy
>> is not a recognized policy, or
>> .I param
>> does not make sense for the
>> .IR policy .
>
> I got troubled by the "may additionally", as "may" has a particular meaning
> in POSIX.1...
>
> But I guess, you just want to express that pthread_setschedparam() shall
> fail if the policy or the param is invalid, right?

Yes.

[...]

>> .PP
>> POSIX.1-2001 also documents an
>> .B ENOTSUP
>> ("attempt was made to set the policy or scheduling parameters
>> to an unsupported value") error for
>> .BR pthread_setschedparam ().
>
> ... but it doesn't seem to be used by Linux/Glibc...

Exactly.  That's why I don't list it in the ERRORS proper, just as an
add-on sentence.

[...]

>> .SH EXAMPLE

[...]

>> The next run is the same as the previous,
>> except that the inherit scheduler attribute is set to
>> .BR PTHREAD_INHERIT_SCHED ,
>> meaning that threads created using the thread attributes object should
>> ignore the scheduling attributes specified in the attributes object
>> and instead take their scheduling attributes from the creating thread.
>>
>> .in +4n
>> .nf
>> # \fB./a.out \-mf10 \-ar20 \-i i\fP
>> Scheduler settings of main thread
>>    policy=SCHED_FIFO, priority=10
>>
>> Scheduler settings in \(aqattr\(aq
>>    policy=SCHED_RR, priority=20
>>    inheritsched is INHERIT
>>
>> Scheduler attributes of new thread
>>    policy=SCHED_FIFO, priority=10
>> .fi
>> .in
>>
>> In the above output, one can see that the scheduling policy and priority
>> were taken from the creating thread,
>> rather than the thread attributes object.
>
> A classical trap is that people don't set inheritsched to explicit, and by
> default it is inherit... You could illustrate this by the following example:
> ./a.out -mf10 -ar20.
>
> ( I don't know however if it is the right place to speak about such things
> ).

I added a sentence noting that if we omit the "-i i" option in the
second run, then the result would be the same, since the default is
INHERIT.

[...]

Thanks for the review Loic!

Cheers,

Michael

-- 
Michael Kerrisk
Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/
git://git.kernel.org/pub/scm/docs/man-pages/man-pages.git
man-pages online: http://www.kernel.org/doc/man-pages/online_pages.html
Found a bug? http://www.kernel.org/doc/man-pages/reporting_bugs.html
--
To unsubscribe from this list: send the line "unsubscribe linux-man" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: For review: pthread_setschedparam.3

[Loic Domaigne]

Hi Michael,

I've only a minor comment left.

I've set the following reminder:

<reminder>

- effect of sched_setscheduler() on MT-process.
- priority adjustment => later, when dealing with the topic on prio 
inheritance/ceiling.

</reminder>

Cheers,
Loïc.
--

>>> .PP
>>> .BR pthread_setschedparam ()
>>> may additionally fail with the following errors:
>>> .TP
>>> .B EINVAL
>>> .I policy
>>> is not a recognized policy, or
>>> .I param
>>> does not make sense for the
>>> .IR policy .
>> I got troubled by the "may additionally", as "may" has a particular meaning
>> in POSIX.1...
>>
>> But I guess, you just want to express that pthread_setschedparam() shall
>> fail if the policy or the param is invalid, right?
> 
> Yes.

What about using "pthread_setschedparam() can additionally [...]" in a 
similar fashion to pthread_setcancelstate(3)" ?
--
To unsubscribe from this list: send the line "unsubscribe linux-man" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: For review: pthread_setschedparam.3

[Michael Kerrisk]

Hi Loic,

On Mon, Nov 17, 2008 at 2:22 PM, Loic Domaigne <tech-Z4JMKDdsf89Wk0Htik3J/w@public.gmane.org> wrote:
> Hi Michael,
>
> I've only a minor comment left.
>
> I've set the following reminder:
>
> <reminder>
>
> - effect of sched_setscheduler() on MT-process.

Will follow up shortly.

> - priority adjustment => later, when dealing with the topic on prio
> inheritance/ceiling.

There remains the FIXME in the page source.

> </reminder>
>
> Cheers,
> Loïc.
> --
>
>>>> .PP
>>>> .BR pthread_setschedparam ()
>>>> may additionally fail with the following errors:
>>>> .TP
>>>> .B EINVAL
>>>> .I policy
>>>> is not a recognized policy, or
>>>> .I param
>>>> does not make sense for the
>>>> .IR policy .
>>>
>>> I got troubled by the "may additionally", as "may" has a particular
>>> meaning
>>> in POSIX.1...
>>>
>>> But I guess, you just want to express that pthread_setschedparam() shall
>>> fail if the policy or the param is invalid, right?
>>
>> Yes.
>
> What about using "pthread_setschedparam() can additionally [...]" in a
> similar fashion to pthread_setcancelstate(3)" ?

Done.

Thanks,

Michael


-- 
Michael Kerrisk
Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/
git://git.kernel.org/pub/scm/docs/man-pages/man-pages.git
man-pages online: http://www.kernel.org/doc/man-pages/online_pages.html
Found a bug? http://www.kernel.org/doc/man-pages/reporting_bugs.html
--
To unsubscribe from this list: send the line "unsubscribe linux-man" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: For review: pthread_setschedparam.3

[Michael Kerrisk]

Hi Loic,

Okay, here I'll revisit this piece.

>> The returned priority value is that set by the most recent
>> .BR pthread_setschedparam (),
>> .BR pthread_setschedprio (3),
>> or
>> .BR pthread_create (3)
>> call that affected
>> .IR thread .
> 
> Hmm, that's perfectly right from a POSIX point of view. Knowing how 
> Linux implements threads, I have been interested about the effect of 
> sched_setscheduler() on a MT-process (since NPTL uses 1:1 model, this 
> should be a NOP).

Why should it be a NOP?

Thanks,

Michael

> I tested the following program against the stable glibc-2.7... 
> Apparently, it seems that sched_setscheduler() might affect the main 
> thread priority as well.
> 
> -- 
> #include <stdio.h>
> #include <pthread.h>
> #include <sched.h>
> #include <unistd.h>
> 
> void
> print_schedinfo(const char* thread)
> {
>    struct sched_param param;
>    int policy;
>    int rc;
> 
>    rc = pthread_getschedparam(pthread_self(), &policy, &param);
>    if (rc!=0) printf("##%d\n", rc);
>    printf("%s > Policy=%s, prio=%d\n",
>           thread,
>           (policy==SCHED_FIFO) ? "FIFO" : "*NOT* FIFO",
>       param.sched_priority);
> }
> 
> // dummy thread...
> //
> void*
> thread(void* ignore)
> {
>    sleep(3);
>    print_schedinfo("dummy thread");
>    pthread_exit(NULL);
> }
> 
> int
> main()
> {
>    struct sched_param param;
>    int    policy;
>    int    rc;
>    pthread_t tid;
> 
>    // create dummy thread
>    //
>    pthread_create(&tid, NULL, thread, NULL);
>    param.sched_priority=1;
> 
>    // now we shall change the process policy/prio using
>    // sched_setscheduler().
>    // Normally: this should be a NOP. But due to the way Linux
>    // implements threads, I am suspecting that this shall affect
>    // the main thread
>    //
>    rc=sched_setscheduler(0, SCHED_FIFO, &param);
>    if (rc==-1) printf("sched_setscheduler FAILED\n");
> 
>    // print my scheduling info
>    //
>    print_schedinfo("main");
> 
>    // join dummy thread and terminate
>    //
>    pthread_join(tid, NULL);
>    return 0;
> }


-- 
Michael Kerrisk
Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/
git://git.kernel.org/pub/scm/docs/man-pages/man-pages.git
man-pages online: http://www.kernel.org/doc/man-pages/online_pages.html
Found a bug? http://www.kernel.org/doc/man-pages/reporting_bugs.html

--
To unsubscribe from this list: send the line "unsubscribe linux-man" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: For review: pthread_setschedparam.3

[Loic Domaigne]

Gidday Michael,

> Okay, here I'll revisit this piece.
> 
>>> The returned priority value is that set by the most recent
>>> .BR pthread_setschedparam (),
>>> .BR pthread_setschedprio (3),
>>> or
>>> .BR pthread_create (3)
>>> call that affected
>>> .IR thread .
>>
>> Hmm, that's perfectly right from a POSIX point of view. Knowing how 
>> Linux implements threads, I have been interested about the effect of 
>> sched_setscheduler() on a MT-process (since NPTL uses 1:1 model, this 
>> should be a NOP).
> 
> Why should it be a NOP?

Because it is stated in SuSv3 TC2:

<copy>
Each process shall be controlled by an associated scheduling policy and
priority. These parameters may be specified by explicit application
execution of the sched_setscheduler() or sched_setparam() functions.

The effect of the process scheduling attributes on individual threads
in the process is dependent on the scheduling contention scope of the
threads, see <ref>Thread Scheduling</ref>:

(*) For threads with system scheduling contention scope, the process
scheduling attributes shall have no effect on the scheduling attributes
or behavior either of the visible POSIX thread or an underlying kernel
scheduling entity dedicated to that thread.

[...]

</copy>

Since NPTL uses 1:1 model, i.e. all threads have system scheduling 
contention scope. Hence, sched_setscheduler() should not affect the 
current threads scheduling.

It happens that I know this part of the standard quite well, because I 
reworked it together with David Butenhof 4 years ago (XSH ERN 52)... We 
attempt to describe in a better way what 1003.1-2004 meant in this regard.

By the way, on GNU/Linux sched_setscheduler() might possibly affect the 
scheduling of the calling thread (and not only the main thread, as 
stated in my previous email). This was the true some Glibc version ago...

See also:
http://groups.google.de/group/comp.unix.programmer/browse_thread/thread/2531f90c0b05ef95/e66e69ad57349f7f
https://www.opengroup.org/sophocles/show_mail.tpl?CALLER=show_archive.tpl&source=L&listname=austin-group-l&id=8143
http://groups.google.de/group/comp.unix.programmer/browse_thread/thread/5c305e82eef3f975/3cbb9722f1f49c15
http://www.qnx.com/developers/docs/6.3.2/neutrino/lib_ref/s/sched_setscheduler.html


HTH,
Loïc.
--
To unsubscribe from this list: send the line "unsubscribe linux-man" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

sched_setscheduler.3 and threads (was: For review: pthread_setschedparam.3)

[Loic Domaigne]

Hello Michael,

I've investigated the effects of sched_setscheduler() on threads. Things 
seem worse than I expected.

AFAICS, you can change scheduling attribute of a thread using 
sched_setscheduler(3) without the changes being reflected in the 
thread's scheduling parameters as retrieved with pthread_getschedparam(3).

Output from the schedtest program below:

<copy>

main   > call sched_setscheduler(0,FIFO,1)
main   > pthread_getschedparam() : Policy=FIFO, prio=1
main   > sched_getscheduler()    : Policy=FIFO, prio=1
-----
Thread > pthread_getschedparam() : Policy=*NOT* FIFO, prio=0
Thread > sched_getscheduler()    : Policy=*NOT* FIFO, prio=0
-----
Thread > call sched_setscheduler(0,FIFO,2)
Thread > pthread_getschedparam() : Policy=*NOT* FIFO, prio=0
Thread > sched_getscheduler()    : Policy=FIFO, prio=2
-----
main   > pthread_getschedparam() : Policy=FIFO, prio=1
main   > sched_getscheduler()    : Policy=FIFO, prio=1
-----

</copy>

Let's hope that my program is broken!

A+,
Loïc
--


/***************************************************************************/
/* schedtest.c- test interaction between sched_setscheduler and threads
/***************************************************************************/
/*
  * compile with: cc -pthread schedtest.c -o schedtest
  */
#include <errno.h>
#include <string.h>
#include <stdio.h>
#include <stdlib.h>
#include <pthread.h>
#include <unistd.h>

/**********************************************************************/
/* our macro for errors checking                                      */
/**********************************************************************/
#define COND_CHECK(func, cond, errv) \
if ( (cond) ) \
{ \
    fprintf(stderr, "\n[CHECK FAILED at %s:%d]\n| %s(...)=%d (%s)\n\n",\
                  __FILE__,__LINE__,func,errv,strerror(errv)); \
                     exit(EXIT_FAILURE); \
}

#define UnixCheck(func,rc)    COND_CHECK(func, (rc==-1), errno)
#define PthreadCheck(func,rc) COND_CHECK(func,(rc!=0), rc)

void
sched_fifo(const char* thr, int prio)
{
    struct sched_param param;
    int rc;

    printf("%s > call sched_setscheduler(0,FIFO,%d)\n", thr, prio);
    param.sched_priority = prio;
    rc = sched_setscheduler(0, SCHED_FIFO, &param);
    UnixCheck("sched_setscheduler", rc);
}

void
sched_info(const char* thr)
{
    struct sched_param param;
    int    policy;
    int    rc;

    rc = pthread_getschedparam(pthread_self(), &policy, &param);
    PthreadCheck("pthread_getschedparam", rc);
    printf("%s > pthread_getschedparam() : Policy=%s, prio=%d\n",
           thr,
           (policy==SCHED_FIFO) ? "FIFO" : "*NOT* FIFO",
	  param.sched_priority);
    policy = sched_getscheduler(0);
    UnixCheck("sched_getscheduler", policy);
    rc = sched_getparam(0, &param);
    UnixCheck("sched_getparam", rc);
    printf("%s > sched_getscheduler()    : Policy=%s, prio=%d\n",
           thr,
           (policy==SCHED_FIFO) ? "FIFO" : "*NOT* FIFO",
	  param.sched_priority);
    printf("-----\n");
}

void*
thread(void* ignore)
{
    sleep(2);
    sched_info("Thread");
    sched_fifo("Thread", 2);
    sched_info("Thread");
    return NULL;
}

int
main()
{
    pthread_t tid;
    int       rc;

    rc = pthread_create(&tid, NULL, thread, NULL);
    PthreadCheck("pthread_create", rc);
    sched_fifo("main  ", 1);
    sched_info("main  ");
    rc = pthread_join(tid, NULL);
    PthreadCheck("pthread_join", rc);
    sched_info("main  ");
    return EXIT_SUCCESS;
}
--
To unsubscribe from this list: send the line "unsubscribe linux-man" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html