hal, usbstick and IBM R40

[Thomas Gufler]

Hi,

thanks for the answer and your help.

On Tuesday 07 September 2004 22:15, David Zeuthen wrote:
> Hey,
>
> Sorry for the delay,
>
> On Sun, 2004-09-05 at 17:20 +0200, Thomas Gufler wrote:
> > Hi,
> >
> > I have a strange problem using a IBM usb-stick on a IBM R40 Laptop (with
> > hal-cvs+fstab-sync).
> >
> > There are three different cases:
> > 1) The usbstick is not recognized at all. hal does recognize many (all?)
> > other devices, but no usb devices though all needed usb modules are
> > loaded. This often happens when the computer was rebooted before. A
> > /dev/sda is created by udev (lshal shows 30 devices)
> > 2) The usbstick is recognized but the partition on the usbstick is not
> > detected, so I get a /dev/sda but not a /dev/sda1. (lshal shows 45
> > devices) 3) The usbstick and the partition are recognized :-) (lshal
> > shows 46 devices)
> >
> > The output of 'HALD_VERBOSE=1 ./hald --daemon=no' for case 2 and 3 is
> > attached below. If the lshal output would also be interesting please let
> > me know.
> >
> > I use this usb stick also on a desktop computer (Athlon XP 2200+, SIS 735
> > chipset) without any problem (also with hal+fstab-sync). Both computers
> > use a gentoo linux system with kernel 2.6.7, udev-030, dbus-cvs, hal-cvs
> > (both from yesterday), on the laptop usb needs the uhci_hcd and on the
> > desktop the ohci_hcd module.
> >
> > Any suggestions?
>
> Yes, your system is obviously sufficiently misconfigured to trigger some
> bugs in hal - which is a good thing insofar we can make hal more
> resistant :-).
>
> So, I'm troubled by seeing this
>
>         14:54:20.372 [I] linux/osspec.c:1421: Queing up seqnum=165,
> sysfspath=/class/net/eth0, subsys=pci 14:54:30.475 [I] linux/osspec.c:1421:
> Queing up seqnum=165, sysfspath=/class/net/eth0, subsys=pci 14:54:40.599
> [I] linux/osspec.c:1421: Queing up seqnum=165, sysfspath=/class/net/eth0,
> subsys=pci 14:54:50.714 [I] linux/osspec.c:1421: Queing up seqnum=165,
> sysfspath=/class/net/eth0, subsys=pci 14:55:00.835 [I] linux/osspec.c:1421:
> Queing up seqnum=165, sysfspath=/class/net/eth0, subsys=pci
>
> that occurs roughly every ten seconds. The first other "real" hotplug event
> is 324:
>
>         14:54:41.529 [I] linux/osspec.c:1421: Queing up seqnum=324,
> sysfspath=/devices/pci0000:00/0000:00:1d.1/usb2/2-1, subsys=usb
> 14:54:41.529 [I] linux/osspec.c:1421: Queing up seqnum=324,
> sysfspath=/devices/pci0000:00/0000:00:1d.1/usb2/2-1, subsys=usb
>
> and it seems you have two hal.hotplug scripts installed :-). Btw, we now
> print a big annoying warning saying that duplicates are ignored. I
> suppose you should bug the hal-cvs ebuild maintainer (if applicable).

Yes, there were two hal.hotplug scripts (hal.hotplug, 20-hal.hotplug). I 
removed hal.hotplug after some testing (see below).
I have switched away from the hal-cvs ebuild, building now by hand. Probably 
the hal-cvs.ebuild didn't cleanly remove, so hal.hotplug persisted. 

>
> The reason hal locks up is simply that we're waiting for the hotplug
> event with sequence number 166; you know it may be delayed for any
> amount of time (from hal's point of view). We can fix this issue only
> when the kernel exports the initial sequence number, and thanks to Kay
> this may be sooner than later. Until then I suppose you need to fix your
> system manually.
>
> Care to find out what is triggering the SEQNUM 165? I'm kind of curious;
> it's probably something that execute the hotplug scripts.

I don't know how to find out "what is triggering the SEQNUM 165". 
However I played with my /etc/modules.autoload.d/kernel-2.6 and when I removed 
the prism54 module from the autoload file hal seems to work fine, even with 
two hal.hotplug files (hal-cvs still from Saturday). The prism54 module is 
still loaded (maybe by hotplug?). 
So putting the prism54 wireless lan module into the autoload file seems to be 
a bad idea ;-) 

>
> Good luck,
> David

Thanks,
Thomas 
_______________________________________________
hal mailing list
hal at freedesktop.org
http://freedesktop.org/mailman/listinfo/hal