>>> On 2/18/2009 at  1:58 PM, Ivan Warren <[email protected]> wrote: 
-snip-
> I was wondering...
> 
> Why on earth does one need to mkinitrd/zipl after adding a DASD volume
> to z/Linux ?

Basically, some historical, performance, and data integrity reasons.

When the DASD driver initializes, particularly in an LPAR, which devices do you 
want it to pay attention to?  Unlike most distributed systems, an LPAR can be 
set up to "see" all the devices that are attached to the box.  Those devices 
can come and go rather dynamically.  For Linux, this presents a problem, since 
device names until just recently, were assigned based on the order in which 
they were detected, or in the order specified in the DASD driver parameters.

Additionally, you don't _want_ Linux to try to look at all the devices out 
there.  In some shops, it can take forever for Linux to come up because it's 
trying to interrogate thousands of devices to figure out what they are.  After 
that, a simple mistake by someone running as root could wipe out a DASD volume 
that doesn't really have anything to do with Linux.

You can overcome some of this by telling the kernel to ignore ranges of device 
numbers.  That value gets stored in the boot parms, which are written out by 
zipl.  (There's no equivalent to grub that will read a parm file at boot time.) 
 So, if your I/O config changes, you still need to re-run zipl.

Which DASD volumes get activated at boot time is controlled by what is put into 
the initrd.  If you want that list to change, you need to rebuild the initrd.  
For SLES, at least, if you add or remove DASD volumes via YaST, that gets done 
for you under the covers.  If you do it manually, of course you're on your own.


Mark Post

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390

Reply via email to