Re: How do I get udev to replay coldplug events for devices that did not load?

[David Goodenough <david.goodenough@btconnect.com>]

On Monday 22 November 2010, Kay Sievers wrote:
> On Mon, Nov 22, 2010 at 12:57, David Goodenough
> 
> <david.goodenough@btconnect.com> wrote:
> > Is there any way to get udev to replay coldplug events for device that it
> > could not run when it first tried to?
> > 
> > My situation is that I have my kernel modules not in the rootfs (for a
> > good reason, not just for the sake of it) and they have not been 
mounted
> > when udev fires off its coldplug events.  Does udev notice failed device
> > creation events and keep them anywhere that can be reused?
> > 
> > I could simply fix all these modules into the kernel, but some wireless
> > drivers (it happens to be these I am having problems with) seem to
> > work better when loaded as modules (or is that just an old memory that
> > is no longer true).
> 
> Udev will not anything like this on its own. And I don't know any
> distro which allows/supports /lib/* to be on a non-root filesystem.
> 
> You can add some custom code though that triggers all events again
> after all filesystems are mounted. That should work in usual setups.
> 
> Kay
My problem is that I have a mixture of distros.  The kernel (and the modules
on a NAND file system) come from OpenWRT, and the root file system
comes from Debian.  The NAND is large enough to hold the kernel, but not
an initramfs, and certainly not the root file system.  But the board does 
have a USB port, so I put the root file system on a USB stick and link to the 
modules on the NAND so that I am always using the modules that the
kernel needs.  

When you say "add some custom code", any clues as to where I should 
be looking to do this (I am not very familiar with the udev kernel code)
or is this simply a matter of asking for the list twice?  It is only the PCI
devices (and the only things on the PCI bus are the wireless cards)
that I need to re-scan.

David