On Tue, Jan 25, 2005 at 05:10:44PM +0000, Jean-Luc Coulon (f5ibh) wrote:
> Le 25.01.2005 17:26:19, Sjoerd Simons a écrit :
>
> [ destructive compression ]
>
> >
> >Oh that's a good hint.. First i saw fstab-sync running in your trace,
> >which is
> >not necessary so it's probably best to remove it.
>
> How to remove it?
rm /etc/hal/device.d/50-fstab-sync.hal
You had to use fstab-sync for automounting with gvm in the time before we had
the current pmount solution.
>
> >
> >Second if you strace it, hal runs as root, if you don't it runs as
> >user
> >haldaemon...
>
> Well, you told me to run: hald... run hald --daemon=no --verbose=yes
> If I do that with a normal user, I get the following error
Yeah, i wasn't expecting a permission problem..
>
> [EMAIL PROTECTED] % /usr/sbin/hald --daemon=no --verbose=yes
> ~
> 17:42:36.712 [I] hald.c:394: hal 0.4.6
> 17:42:36.713 [I] hald.c:398: Will not daemonize
> 17:42:36.740 [E] hald_dbus.c:1945: dbus_bus_acquire_service():
> Connection ":1.54" is not allowed to own the service
> "org.freedesktop.Hal" due to security policies in the configuration
> file
That's correct, hal can only run as either root or as usr hal. You can just add
--drop-privileges to the options, so it behave the same way as it does when
running normally..
> What are the permissions of your cd drives ?
>
> brw-rw---- 1 root cdrom 3, 0 2005-01-25 13:27 /dev/hda
> brw-rw---- 1 root cdrom 22, 0 2005-01-25 13:27 /dev/hdc
Looks okay. Is user hal member of group cdrom ?
> >Also you can use ``strace -e ioctl,open,close -p $(pidof hald)'' to
> >strace your
> >running hald.. Probably it will show permission denied or somthing
> >like
> >that.
>
> Yes, right
>
> [EMAIL PROTECTED] /home/jean-luc # strace -e ioctl,open,close -p
> 19238
> Process 19238 attached - interrupt to quit
> open("/dev/hdc", O_RDONLY|O_NONBLOCK|O_EXCL|O_LARGEFILE) = 3
> ioctl(3, CDROM_DRIVE_STATUS, 0x7fffffff) = 1
> close(3) = 0
hdc seems to work fine here? Does inserting volumes in that driver work ?
> open("/dev/hda", O_RDONLY|O_NONBLOCK|O_EXCL|O_LARGEFILE) = -1 EACCES
> (Permission denied)
> But it is not the case with the previous version of hal...
>
> open("/dev/hda", O_RDONLY|O_NONBLOCK|O_LARGEFILE) = 3
> ioctl(3, CDROM_DRIVE_STATUS, 0x7fffffff) = 4
Okay, compare the two opens.. The ``new'' hal adds O_EXCL, which is correct,
although i have no clue what triggered the difference..
But this shouldn't cause E_ACCESS, if access is denied to the E_EXCL you get
E_BUSY..
> If this can helps, I remarked while doing my tests that fam is
> something doing bad things. The drive is marked busy so I cannot eject
> it nor unmount it without killing fam. But I don't know the exact
> conditions this occured.
Could you disable fam while testing hal, just to be sure it doesn't gets in way
in some weird way..
I'm also still guessing what the exact problem is here, but i think we're
slowing getting there :)
Sjoerd
--
Everything should be made as simple as possible, but not simpler.
-- Albert Einstein
