Christof Schulze
2009-Sep-26 13:25 UTC
various kde programs spin in a poll/read loop against their respective config files
Hello everyone,
the discussion in -current and the behavior of my hard disk caused me to
investigate. Whenever kde programs are run on this system, the hard disk
will not spin down (even when excessive timeouts for syncs are in effect)
So I ran truss against kwallet and plasma-desktop to find out they do the
same thing:
stat("/home/erika/.kde4/share/apps/kwallet",{ mode=drwx------
,inode=56314,size=3,blksize=4096 }) = 0 (0x0)
clock_gettime(4,{10454.430272035 }) = 0 (0x0)
clock_gettime(4,{10454.430363671 }) = 0 (0x0)
clock_gettime(4,{10454.430432397 }) = 0 (0x0)
clock_gettime(4,{10454.430516490 }) = 0 (0x0)
read(8,0x808f9102c,4096) ERR#35 'Resource temporarily
unavailable'
poll({3/POLLIN 9/POLLIN 8/POLLIN 5/POLLIN},4,0) = 0 (0x0)
clock_gettime(4,{10454.430734403 }) = 0 (0x0)
read(8,0x808f9102c,4096) ERR#35 'Resource temporarily
unavailable'
clock_gettime(4,{10454.430882473 }) = 0 (0x0)
read(8,0x808f9102c,4096) ERR#35 'Resource temporarily
unavailable'
poll({3/POLLIN 9/POLLIN 8/POLLIN 5/POLLIN},4,499) = 0 (0x0)
clock_gettime(4,{10454.939752100 }) = 0 (0x0)
read(8,0x808f9102c,4096) ERR#35 'Resource temporarily
unavailable'
clock_gettime(4,{10454.939905198 }) = 0 (0x0)
stat("/home/erika/.kde4/share/apps/kwallet",{ mode=drwx------
,inode=56314,size=3,blksize=4096 }) = 0 (0x0)
clock_gettime(4,{10454.940151049 }) = 0 (0x0)
clock_gettime(4,{10454.940242126 }) = 0 (0x0)
clock_gettime(4,{10454.940309456 }) = 0 (0x0)
clock_gettime(4,{10454.940392710 }) = 0 (0x0)
read(8,0x808f9102c,4096) ERR#35 'Resource temporarily
unavailable'
poll({3/POLLIN 9/POLLIN 8/POLLIN 5/POLLIN},4,0) = 0 (0x0)
clock_gettime(4,{10454.940612580 }) = 0 (0x0)
read(8,0x808f9102c,4096) ERR#35 'Resource temporarily
unavailable'
clock_gettime(4,{10454.940760929 }) = 0 (0x0)
read(8,0x808f9102c,4096) ERR#35 'Resource temporarily
unavailable'
Shouldn't this be handled by fam/gamin in order to avoid this overhead?
While it does not produce notable load it prevents my processor from
saving power as well as spinning down the disks. I will gladly provide
more information on this if necessary.
This is a FreeBSD 7.2-STABLE FreeBSD 7.2-STABLE #0: Tue Jul 7 04:56:00
CEST 2009
root@ccschu935:/usr/obj/usr/src/sys/GENERIC amd64
system
Regards
Christof
Patrick Lamaiziere
2009-Sep-27 07:07 UTC
various kde programs spin in a poll/read loop against their respective config files
Le Sat, 26 Sep 2009 15:25:17 +0200, Christof Schulze <christof.schulze@gmx.com> a ?crit :> Hello everyone,Hi,> the discussion in -current and the behavior of my hard disk caused me > to investigate. Whenever kde programs are run on this system, the > hard disk will not spin down (even when excessive timeouts for syncs > are in effect) So I ran truss against kwallet and plasma-desktop to > find out they do the same thing: > > stat("/home/erika/.kde4/share/apps/kwallet",{ mode=drwx------ > ,inode=56314,size=3,blksize=4096 }) = 0 (0x0)> Shouldn't this be handled by fam/gamin in order to avoid this > overhead? While it does not produce notable load it prevents my > processor from saving power as well as spinning down the disks.KDE with fam is broken, according to http://wiki.freebsd.org/KDE4 A work-around is to change the PollInterval in kdedrc baby-jane:~$ less .kde4/share/config/kdedrc [$Version] update_info=kded.upd:kde3.0 [DirWatch] PollInterval=60000 (IMHO, for KDE issues you should ask on the kde-freebsd@kde.org mailing list). See also this thread : http://mail.kde.org/pipermail/kde-freebsd/2009-September/006447.html Regards.