usb drive not recognized, maybe timing?

[Kay Sievers]

On Tue, Sep 13, 2005 at 11:24:55AM +0400, Eugene Crosser wrote:
> Kay Sievers wrote:
> 
> >>I have hal 5.0.4 running (with --retain-privileges) on kernel 2.6.13,
> >>with udev 0.64. When a usb drive is inserted, hald recognizes is as
> >>"usb_noserial", not as a block device with a filesystem.  I suspect that
> >>it has something to do with "Cannot open" messages in the HAL debug
> >>output (attached).  If I start hald when the USB drive is already
> >>unserted, it *is* recognized and appears in `lshal' output.
> >>
> >>Now, the interesting part.  When I manually cat the sys files that hald
> >>could not open, a few seconds later, most of them are in place and
> >>perfectly readable.  On the other hand, if I watch dmesg, I can see that
> >>there is a several seconds' pause between "usb-storage: device found at
> >>20" and final "usb-storage: device scan complete" message.  Maybe hald
> >>is trying to open sysfs files while the kernel did not create them yet?
> > 
> > 
> > That should be covered by udev managing the whole hotplug process.
> > Does:
> >   cat /proc/sys/kernel/hotplug
> > print /sbin/udevsend?
> 
> Nope, not really.  If I put /sbin/udevsend into
> /proc/sys/kernel/hotplug, USB insertion information never gets to
> messagebus at all, even as "usb_noserial".  This is hald output that I get:

HAL can't work if /proc/sys/kernel/hotplug is not set to /sbin/udevsend
And the hotplug multiplexing or the udev rules need to plug HAL into the event
processing. HAL also depends on a specific udev version to be able to "coldplug"
on startup.

The things are getting more complex and the interdependencies are
growing with every release. I think the effort for a single person to get it
running from scratch is much too high, you may better start of with a
working setup with the packages your distro provides and try to replace
it part by part if you want to hack on it.

Good luck,
Kay