Jukka,

As promised, I have updated chap-rf.xml based on your RAIDframe
advice.  Please note that I have also corrected a small problem with
an example in the document.  The correction is that the kernel dump
flag was shown to be 'dw' in the documentation, but it should read
'dp'.  'dp' is also what is stated to be the correct flag elsewhere in
the documentation and it corresponds with fstab(5).

Finally, there is a problem with booting from the bootblock on
/dev/sd0a; however, not from /dev/sd1a.  I'm not quite sure how to
remedy this and wonder if it is something that should be documented as
a sparc64-specific problem.  I found this URL which partially
addresses the problem:

  http://www.netbsd.org/Ports/sparc64/faq.html#fast-dmmu-miss

My screen looks like this when I execute 'boot' or 'boot disk0' or
'boot disk0 netbsd', which all do the same thing:

  Rebooting with command: boot
  Boot device: disk  File and args:
  NetBSD IEEE 1275 Bootblock
  .Inode not directory
  Fast Data Access MMU Miss
  ok _

But it really isn't ok.  ;^)  I don't quite understand how this could
happen.  I was successfully booting disk0 without any problems prior
to this RAID-1 configuration.  Any thoughts?

Ryan


*** chap-rf.xml	2005/08/16 16:53:07	1.1
--- chap-rf.xml	2005/08/16 17:10:34	1.2
***************
*** 983,988 ****
--- 983,1039 ----
     d:  19541088         0     unused      0     0
    </screen>
  
+       <para>On sparc64:</para>
+ 
+   <screen>
+   &rprompt; <command>disklabel raid0</command>
+ 
+   8 partitions:
+   #        size    offset     fstype  [fsize bsize cpg/sgs]
+    a:  19015680         0     4.2BSD   1024  8192    64
+    b:    525184  19015680     swap
+    c:  19540864         0     unused      0     0     0
+ 
+ 
+   &rprompt; <command>disklabel wd1</command>
+ 
+   8 partitions:
+   #        size    offset     fstype  [fsize bsize cpg/sgs]
+    a:  19541088         0     RAID
+    c:  19541088         0     unused      0     0
+    </screen>
+ 
+       <para>
+         Each RAID set has a 64 block reserved area (see RF_PROTECTED_SECTORS in
+         &lt;dev/raidframe/raidframevar.h&gt;) in the beginning of the set
+         to store the internal RAID structures.
+       </para>
+ 
+   <screen>
+   &rprompt; <command>dc</command>
+   64              # RF_PROTECTED_SECTORS
+   19015680        # offset of raid0b
+   +p
+   19015744        # offset of swap within wd1
+   q
+   </screen>
+ 
+       <para>We know now that real offset of the still-nonexisting wd1b 
+         is 19015744 and size is 525184. Next we need to add wd1b to 
+ 	wd1's disklabel.</para>
+ 
+   <screen>
+   &rprompt; <command>disklabel wd1 &gt; disklabel.wd1</command>
+   &rprompt; <command>vi disklabel.wd1</command>
+ 
+   8 partitions:
+   #        size    offset     fstype  [fsize bsize cpg/sgs]
+    a:  19541025        63       RAID
+    b:   2097152  19015807       swap
+    c:  19541025        63     unused      0     0
+    d:  19541088         0     unused      0     0
+   </screen>
+ 
        <para>Next we install the new disklabel.</para>
  
    <screen>
***************
*** 1047,1053 ****
  
  /dev/raid0a  /  ffs  rw  1  1
  /dev/raid0b  none  swap  sw  0  0
! /dev/wd0b  none  swap  dw  0  0
  kernfs    /kern  kernfs  rw
  procfs    /proc  procfs  rw</screen>
  
--- 1098,1104 ----
  
  /dev/raid0a  /  ffs  rw  1  1
  /dev/raid0b  none  swap  sw  0  0
! /dev/wd0b  none  swap  dp  0  0
  kernfs    /kern  kernfs  rw
  procfs    /proc  procfs  rw</screen>
  



On Mon, Aug 15, 2005 at 08:22:51PM +0200, Jukka Salmi wrote:
> Hi,
> 
> Ryan Cresawn --> netbsd-docs (2005-08-15 12:09:24 -0400):
> > http://www.netbsd.org/guide/en/chap-rf.html#chap-rf-setup-kerneldump
> > 
> > The trouble I'm having is that the sparc64 documentation for this
> > section is absent.
> 
> Until about two weeks ago, the whole chapter 16.3.7. ("Setting up
> kernel dumps") was missing. It was commented out some time ago while
> applying a documentation update. For details see PR [1]misc/30427.
> 
> I recalculated the numbers for i386 only because I don't have a sparc64
> system to test with. This also means: what follows in this message is
> what I think is correct, but I can't test it, so use it with care.
> 
> 
> > The use of the 'dc' command gets a bit confusing due to the offset of
> > 'sd1a' on sparc64 being 0 as opposed to 63 on i386.
> 
> Just substitute 0 for 63 (see below).
> 
> 
> > What is equally
> > confusing to me is that since 'sd1a' and 'sd1c' have the same size it
> > appears I have left no room for a swap partition at 'sd1b'.
> 
> That's the same situation as described for i386 in the documentation.
> In your case, the RAID partition (sd1a) uses the whole disk (sd1c);
> thus these two partitions have the same offset size. Your swap partition
> (raid0b) resides _within_ the RAID partition (sd1a). To calculate the
> offset of your dump partition, add RF_PROTECTED_SECTORS to raid0b's
> offset; the size is the same as raid0b's size. In your case, this is:
> 
> # disklabel raid0
> [...]
> #        size    offset     fstype [fsize bsize cpg/sgs]
>  a:  33804288         0     4.2BSD   2048 16384 28368
>  b:   1573888  33804288       swap
>  c:  35378432         0     unused      0     0
> 
> # disklabel sd1
> [...]
> #        size    offset     fstype [fsize bsize cpg/sgs]
>  a:  35378533         0       RAID
>  b:   1573888  33804352       swap
>  c:  35378533         0     unused      0     0
> 
> 
> > I am happy to document the sparc64-specific section of the
> > RAIDframe documentation and send a unified diff file if that would be
> > desirable.
> 
> That would be great. The file to patch is [2]htdocs/guide/en/chap-rf.xml.
> 
> 
> Cheers, Jukka
> 
> [1] http://www.netbsd.org/cgi-bin/query-pr-single.pl?number=30427
> [2] http://cvsweb.netbsd.org/bsdweb.cgi/htdocs/guide/en/chap-rf.xml
> 
> -- 
> bashian roulette:
> $ ((RANDOM%6)) || rm -rf ~