[indiana-discuss] 1st batch of Release Notes posted

[Jan Damborsky]

Hi Barbara,


Barbara.Lundquist at Sun.COM wrote:
> Jan, Can you resend RN submissions for these 2? I'm not finding them.

sure - please see below for attached draft for release notes relating
to that two issues mentioned.

Please let me know, if you have any questions or you would need
more information.

Thank you,
Jan


[A] 6690824 fdisk(1M) corrupts Solaris VTOC, if there is Linux swap 
primary partition defined

* Description of the problem
When installing on Solaris2 partition which precedes Linux swap one
(which should be supported configuration) after installation finishes
fdisk(1M) corrupts Solaris2 VTOC when marking Solaris2 partition "active"
and thus makes installed Solaris unusable.

* workarounds
[1] Avoid defining Linux swap as primary partition on the same disk
where Solaris2 primary partition exists

[2] Change type of Linux swap partition to something else (neither 0xbf 
not 0x82)
during installation. After installation finishes, it is possible to 
change it back
to 0x82.

[3] Create Solaris2 partition outside the installer and mark it active.
Then install into this preexisting partition. Calling fdisk(1M) for
activating the partition after installation finishes will be then
null operation and corruption will not occur.

[B] 6690194 fdisk(1M) doesn't clear VTOC when partition starts after or 
exceeds 4GB limit

* Description of the problem

When new Solaris2 partition is created or when existing one is 
resized/moved, fdisk(1M)
will clear primary as well as backup VTOCs, so that default one is 
picked by kernel.
The problem is that appropriate operations which take care of clearing 
the VTOC
(lseek(2), write(2)) are 32bit limited. This causes following issues:

* Solaris2 partition, which starts after 4GB limit can't be resized. 
Even deleting
and subsequent creation doesn't help as long as starting sector is not 
changed.

* When fdisk(1M) applies operation of clearing the VTOC to Solaris2 
partition which
 crosses 4GB boundary, it actually clears couple of sectors within 4GB 
due to the
 32bit overflow. If there are valid data defined there, they are corrupted.

 For example, if there is NTFS partition occupying first 4GB, during 
creation of Solaris2
 partition fdisk(1M) actually clears about 6 random sectors somewhere in 
NTFS or
 boot block (depending on start&size of Solaris2 partition to be created).

* workaround - too restrictive from my point of view - I will let Barry 
comment on this
[1] Solaris2 partition must always fit into first 4GB, otherwise data 
corruption might occur.