On Wed, 7 May 2014 14:17:21 +0200
<[email protected]> wrote:
> > > > + /* Move child to RT partition */
> > > > + SAFE_ASPRINTF(cleanup, &tid_str, "%ld", (long) tid);
> > > > + SAFE_WRITE(cleanup, 1, cpuset_tasks_fd, tid_str,
> > > > strlen(tid_str)); +
> > > > + free(tid_str);
> > > > + SAFE_CLOSE(cleanup, cpuset_tasks_fd);
> > >
> > > SAFE_FILE_PRINTF(cleanup, CPUSET_RT_TASKS_FILE, "%ld",
> > > (long)tid) ?
> > >
> > > Or is there a good reason to use open(), asprintf(),
> > > write() and close() instead?
> >
> > Well, printf() by itself will do no file updates, but I replaced
> > open(), asprintf(), write(), free() and close() with fopen(),
> > fprintf() and fclose().
>
> You lost me here.
>
> Have you seen the code for the SAFE_FILE_PRINTF()? What it does is
> is fopen(), fprintf() and fclose(). It was designed to simplify
> writing to various kernel pseudo filesystems to one call.
I misread you, sorry. I see now what you mean. I'll be a good boy and
use SAFE_FILE_PRINTF() instead.
>
> > > > + for (nr_matches = fscanf(stream, "%d-%d", &range_first,
> > > > &range_last);
> > > > + nr_matches > 0;
> > > > + nr_matches = fscanf(stream, ",%d-%d",
> > > > &range_first, &range_last)) {
> > > > + if (nr_matches == 1)
> > > > + range_last = range_first;
> > > > +
> > > > + /* Set all bits in range */
> > > > + for (bit = range_first; bit <= range_last;
> > > > bit++)
> > > > + mask |= (1 << bit);
> > >
> > > Are you sure that this would not overflow?
> > >
> > > I guess that it depends on how you have
> > > partitioned your CPUs.
> >
> > As far as I know there is no API stable and architecture portable
> > way of determining how many CPUs a machine has, which makes this
> > whole business a bit tricky. The problem starts with the partrt
> > tool, which should actually support arbitrarily long masks, but
> > this hasn't happened.
>
> There is sysconf(_SC_NPROCESSORS_CONF) which tries to collect this
> information from the system. It first tries /sys/devices/system/cpu/
> then /proc/cpuinfo.
Hmm, didn't know that. Unfortunately this will not help partrt tool,
but it's still interesting information. If I just get the time for it
I'll rewrite partrt as a C program, that would make CPU mask handling
easier.
But I'll make a stab at fixing this in the test case at least.
>
> > This is on the TODO list, so currently the test only guarantees 32
> > CPUs. Should probably be documented somewhere...
>
> Right, at least add this into the documentation.
>
I've added a line about this in the README.
Regards
Mats Liljegren
------------------------------------------------------------------------------
Is your legacy SCM system holding you back? Join Perforce May 7 to find out:
• 3 signs your SCM is hindering your productivity
• Requirements for releasing software faster
• Expert tips and advice for migrating your SCM now
http://p.sf.net/sfu/perforce
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list
