On Mon, 2010-09-27 at 15:11 -0700, Kyle Williams wrote:
> I got the pid files too, but didn't do the log files. Doh, that's
> another thing to fix.
> What other XML databases does it create?
That should be it, I don't know if it creates any other transient files
-- the code base is way too complex. A way would be to monitor the
execution with
strace -o trace=open
and see which files get opened that are not elfs for writing :)
> On Mon, Sep 27, 2010 at 2:56 PM, Inaky Perez-Gonzalez
> <inaky at linux.intel.com> wrote:
> On Mon, 2010-09-27 at 14:52 -0700, Kyle Williams wrote:
> > I was also working on something simular with the 1.4.0
> version of the
> > Driver.
> > Your dirty hack was the same one that came to my mind.
> >
> >
> > So I have a wimaxd-wmx0, wimaxd-wmx1, wimaxd-wmx2, etc...
> > Same for wimaxcu-wmx0, wimaxcu-wmx1, etc...
> >
> >
> > I changed the code to look for config.wmx0.xml,
> config.wmx1.xml, etc.
> > Also had to change the listening port in each XML file for
> each card,
> > 1234, 1235, etc..
>
>
> Good to hear it works!
>
> I recommend you go with the different prefix approach --
> otherwise when
> the code modifies the xml databases, they are going to
> conflict. Unless
> of course (and this is way cleaner) you also add the
> 'diferentiation'
> string to all the paths that the code accesses (databases, log
> files and
> config file).
>
> I think a rough list wouldbe:
>
> /var/run/wimaxd.pid
> /var/log/wimax/*log
> /var/lib/WiMAX_{DB,Def}.bin
>
> If you do that in 1.5, I'd love a patch to add it to the main
> tree!!
>
> maybe we could forfait the config.xml one if we did something
> like
> adding the interface's index number to a base port number
> (again, this
> is super dirty but a quick hack):
>
> $ cat /sys/class/net/wmx0/ifindex
> 5
>
> from config.xml, SdkPort is 49001, so wmx0 uses 49001+5, wmx1
>
> $ cat /sys/class/net/wmx1/ifindex
> 8
>
> would use 49001+8
>
> and then another patch to wimaxcu to take a '-i' argument that
> defaults
> to wmx0 would round the whole thing.
>
>
> >
> > It's really a rough work around, but I successfully got 3
> different
> > cards working on 1 box last night, so I'm content.
> >
> >
> > Thank you for the advice with the 1.5 version! :)
> >
> >
> > - Kyle
> >
> >
> >
> > On Mon, Sep 27, 2010 at 1:24 PM, Inaky Perez-Gonzalez
> > <inaky at linux.intel.com> wrote:
> > Hi Kyle
> >
> >
> > On Sat, 2010-09-25 at 03:50 -0700, Kyle Williams
> wrote:
> > > Hello,
> > >
> > >
> > > I'm trying to use three Intel EchoPeak 5150 cards
> working on
> > a single
> > > PC. I noticed with the latest wimaxd has the "-i"
> option
> > which allows
> > > the specification of the interface for wimaxd to
> control,
> > but the
> > > wimaxcu program does not allow the same type of
> > specification. Is
> > > there some way to specify which wimaxd instance
> the wimaxcu
> > program
> > > controls?
> > >
> > >
> > > The objective is to get three Echopeek 5150 cards
> to all
> > work on the
> > > same system.
> > > Is this possible at this time?
> > > Will the wimaxcu program support this down the
> road?
> >
> >
> > It is possible but very dirty. Unfortunately the
> daemon
> > doesn't have
> > provisions to identify different interfaces at the
> RPC level.
> >
> > The ugly trick I use is configure three builds with
> three
> > different
> > prefixes (to avoid the databases getting corrupted)
> and then
> > modify
> > each's configuration file (prefix/wimax/config.xml)
> so that
> > <sdkportnum>
> > are different.
> >
> > somehting like:
> >
> > tar xf wimax-1.5.tar.gz
> > for v in 0 1 2
> > do
> > mkdir build-wmx$v
> > (cd build-wmx$v && ../wimax/configure
> --prefix=/opt/wimax/wmx
> > $v/
> > --program-suffix=-wmx$v && make install)
> > done
> >
> > [change the port numbers
> > in /opt/wimax/wmx*/etc/wimax/config.xml]
> >
> > I then add the /opt/wimax/wmx*/bin paths to my PATH
> and just
> > have to
> > start
> >
> > $ for v in 0 1 2; do wimaxd-wmx$v -i wmx$v -b; done
> >
> > and in the command line
> >
> > $ wimaxcu-wmx0 scan ..
> > etc
> >
> > As I said, it is dirty as heck, but it does the
> trick.
> >
> >
> >
> > Then start once instance of each daemon with -i
> IFACENAME and
> > call the
> >
> >
> >
>
>
>
>
>
>
