On Fridayen den 14 December 2001 12.44, Borsenkow Andrej wrote:
> > > open("/dev/sg0", O_RDWR) = -1 ENOENT (No such file or
> > > directory)
> > >
> > > Please, immediately after reboot:
> > >
> > > ls -l /dev/sg*
> >
> > [root@oden /]# ls -l /dev/sg*
> > lr-xr-xr-x 1 root root 36 Dec 14 2001 /dev/sg0 ->
> > scsi/host0/bus0/target0/lun0/generic
> > lr-xr-xr-x 1 root root 36 Dec 14 2001 /dev/sg1 ->
> > scsi/host0/bus0/target1/lun0/generic
>
> Sigh ... what is supposed to happen is
>
> - program tries to open /dev/sg0
> - if it is missing (because sg is not yet loaded) devfs generates LOOKUP
> event and sends it to devfsd
> - devfsd then loads sg module
> - sg driver registers /dev/sg0 and devfs returns handle to it to a
> program
>
> What happens is
>
> - /dev/sg0 and /dev/sg1 do exist so no LOOOKUP event is ever generated
> - LOOKUP is first generated on access to /dev/sg2
> - but sg driver registers sg0 and sg1! *Not* sg2 So lookup for /dev/sg2
> (and all others) still fails.
> - the second time you access it it works.
This is very advanced for me, also I have a slight cold today affecting my
synopsis activities it seems...
DevFs is _VERY_ new to me, I have never ever used it because a.) I'm to used
to plain /dev/ structure, b.) devfs is supposed to be buggy, unsafe, etc.
So, I don't really know much else than that it's supposed to create the
devices, transparent, on-the-fly, right?
> > These lines?
> >
> > if [ -d /lib/dev-state ]; then
> > cp -af /lib/dev-state/* /dev > /dev/null 2>&1
> > fi
>
> Yes, these. THESE LINES ARE EVIL I will continue to repeat it over and
> over. Since I removed them I never have any problem. You will need to
> configure devfsd to create any "persistent" nodes (I have a single
> /dev/modem link) - or please bug Juan and Thierry to release news
> devfs/devfsd with much better persistent nodes management.
could you explain just exactly why this is bad... (shit I can't think
straight today...)
> To Fabrice: Fabrice, we are on cooker not in mandrake NG. We need to fix
> problems not provide workaround. If everybody will preload all modules
> (or will turn devfs off) it will never be debugged and fixed.
>
> -andrej
Agree... You can't just ignore the devfs stuff. Either totally remove it from
the distro or make it work. I for one do not have the expertice needed to
make it work, but with help I might be able to help out.
--
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
| Oden Eriksson, Deserve-IT Networks, Jokkmokk, Sweden.
| Mandrake Linux release 8.2 (Cooker) for i586
| Current uptime with kernel 2.4.16-7mdksmp: 2 hours
| cpu0 @ 814.28 bm, fan 4530 rpm, temp +32�C
| cpu1 @ 815.92 bm, fan 4440 rpm, temp +31.5�C
