From: Evan H. <eh...@gm...> - 2006-03-25 18:57:47
|
Tab- Just read this thread, When can we expect the om/ to be read for testing? That will be a huge boon to the userland tools. That was one of the things we had trouble making a good solution for. Evan On 1/28/06, Matt Rechenburg <mos...@t-...> wrote: > Hi Vincent, > > the omctrlfs sounds great :) > > many thanks, > > Matt > > > Vincent Hanquez wrote: > > >What's in openMosix today and tomorrow ? > >=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D > > > > * debug interface moved to debugfs > > ---------------------------------- > > I rewrited the debug code, which was available through /proc/hpc/debug > > to use debugfs. so now to have access to debug files, you need to activ= ate > > CONFIG_OPENMOSIX_DEBUG_FS and then mount debugfs somewhere: > > > > mount -t debugfs none /debug/ > > > > then you'll have an "om" directory there, that contains the needed file= to > > have verbose (migration | syscall | copyuser | files sharing) > > > > This cleanup lead me to more cleanup in the hpc/proc.c and to the next = point. > > > > * openMosix control filesystem > > ------------------------------ > > Instead of hacking into an already messy directory (/proc/) for having > > an openMosix interface (/proc/hpc/ and /proc/$PID/om/), a new filesyste= m > > will be add to provide an efficient clean interface for all openMosix > > purpose: migration and statistics collection. > > > > Here the benefit: > > . remove hack into procfs, which is seriously frown upon these days. > > . provide efficient interface to the userspace tools: > > - not have to read every task /proc/$PID/om/ directory every X ms. > > - give a list of migratable tasks, blocked tasks, .. in a single re= ad. > > - provide various statistics. > > - add scheduler helper in the kernel for userspace. > > > > So, here an example of what will be available there (number are tasks p= ids): > > > > (mount -t omctrlfs none /om) > > /om/ > > > > /om/blocked/ -> list of task that are blocked > > /om/blocked/34 for a reason or another > > /om/blocked/574 (user request, stay reason) > > > > /om/migratable/ -> list of task that are good candidate f= or > > /om/migratable/93 migration, provided by some heuristic > > /om/migratable/7623 in kernel > > > > /om/deputy/ -> list task that are sent to $ip > > /om/deputy/$ip/456 > > /om/deputy/$ip2/345 > > /om/deputy/$ip2/7899 > > > > /om/remote/ -> list task that come from $ip > > /om/remote/$ip/56 > > > > /om/migration -> migration order > > > > > > This is not done, so don't try to mount any omctrlfs yet. > > > > * openMosix dev tree using git > > ------------------------------ > > I've now setup a public git tree, instead of the svn tree. > > > > URI: git://openmosix.snarc.org/kernel/ > > > > here the command that you are likely to use: > > clone it =3D=3D=3D> git clone git://openmosix.snarc.org/kernel/ linux= -2.6-om > > update it =3D=3D=3D> (in the directory) git pull > > > > > > > > > > ------------------------------------------------------- > This SF.net email is sponsored by: Splunk Inc. Do you grep through log fi= les > for problems? Stop! Download the new AJAX search engine that makes > searching your log files as easy as surfing the web. DOWNLOAD SPLUNK! > http://sel.as-us.falkag.net/sel?cmd=3Dlnk&kid=3D103432&bid=3D230486&dat= =3D121642 > _______________________________________________ > openMosix-devel mailing list > ope...@li... > https://lists.sourceforge.net/lists/listinfo/openmosix-devel > |