Partial update of bootscripts

Bruce Dubbs bruce.dubbs at gmail.com
Mon Jul 11 08:55:20 PDT 2011


Bryan Kadzban wrote:
> Bruce Dubbs wrote:
>> 5. Updated udev_retry to handle the /run directory for events that
>> occur between starting udevd and mounting / read/write.
> 
> Good, it's using the same thing the rule_generator.functions file is
> using (see choose_rules_file): "udevadm info --run".
> 
> Though it looks a little strange to be searching $PATH for udevadm
> there, yet be specifying /sbin/udevadm through the rest of the script.
> Consistency might be good.

Good point.  I added the path.

> It actually probably makes sense to pull this out into its own change?
> We'll need it with newer udev versions no matter what happens to the
> rest (and maybe current udev versions; I'm not sure).  Up to you though.

I'm not sure what you mean.

>> Actually this script may need to be reviewed in more detail, because
>> /run is always mounted r/w before udevd is started.
> 
> Hmm, not sure this why this matters?  :-)
> 
> /run is mounted read-write, yes -- that's the only reason these rules
> files would have been successfully written there.  This whole setup is
> here to catch the case where the rules are generated before the *root*
> FS is mounted read-write (so /etc/udev/rules.d is not writable).  And
> AFAIK the root FS can't be mounted read-write before udevadm trigger
> happens, unless I'm missing something.  (Because the root FS's device
> node is not present.)

One purpose of the /run directory is to allow (eventually) / to be able 
to remain read only.  Things like mtab, blkid.tab, etc will migrate to 
/run.   I'd think dynamic udev rules would move there too.  I don't know 
why any rules would have to be delayed and that would mean that 
udev_retry would be unnecessary.

   -- Bruce



More information about the lfs-dev mailing list