etch/s390 TODO

[Bastian Blank]

Hi folks

As s390 is currently not in a good shape, I tried to produce a list of
things which needs to be done for a possible etch release of this
architecture. Please provide some feedback.

Hardware configuration
======================
In linux 2.6, the complete hardware configuration on s390 is done via
sysfs. This needs some program which reads the needed informations from
a config and puts them into sysfs.

There are two call modes:
- via udev,
- by an init script.

Via udev
--------
udev gets events when the channels got available and can call our
program which sets up this one channel and the assigned driver.

There is one possible problem with that: udev paralizes the events,
which makes it impossible to predict the order of the devices.

Another problem: we can't enable the devices in config order, only in
channel number order. This can be fixed with a mix between both modes.

Both problems needs some discussion with udev upstream.

By an init script
-----------------
The tool will just crawl through all available channels and configure
those, which have a config entry or it will crawl through the config
entries and configure any available channels.

Hardware configuration utility for d-i
======================================
There needs to be some changed in s390-dasd and s390-netdevice to read
the informations from sysfs and generate configs usable by the real
configuration tools.

One problem: We need to make sure that the order of the devices in the
installer is the same than in the installed system.

s390 support in partman
=======================
There is some type of s390/dasd support available for parted, it just
completely disables the check for the sector size, as parted hardcodes
this in too many places and dasds don't follow this. Maybe this can be
fixed in a proper way.

partman needs to be fixed to use a correct block size for the devices
(1024 byte blocks on a 4096 bytes sectors device just don't work).

Bastian

[Sven Luther]

Post by Bastian Blank
s390 support in partman
=======================
There is some type of s390/dasd support available for parted, it just
completely disables the check for the sector size, as parted hardcodes
this in too many places and dasds don't follow this. Maybe this can be
fixed in a proper way.
partman needs to be fixed to use a correct block size for the devices
(1024 byte blocks on a 4096 bytes sectors device just don't work).

I disabled the dasd patch in parted 1.6.23, because it didn't apply cleanly,
and there where some changes which made it non-trivial for me to fix it, i
asked you on irc to look at it, and think i also mailed here, but nothing
happened since then. Otavio has now taken a more active role than me in parted
maintenance, as has K.G., so it would be best to contact them about this
maybe.

One issue with the DASD patch, is that it was from some guy who now seems MIA
or something, as i wrote him too about this, and it is a rather consequent
patch, and thus needs official FSF copyright assignement to be merged
upstream, which is really what we want to do.

Someone with DASD knowledge and hardware is probably best placed to look at
it.

Friendly,

Sven Luther

[Bastian Blank]

Post by Sven Luther
I disabled the dasd patch in parted 1.6.23, because it didn't apply cleanly,
and there where some changes which made it non-trivial for me to fix it, i
asked you on irc to look at it, and think i also mailed here, but nothing
happened since then. Otavio has now taken a more active role than me in parted
maintenance, as has K.G., so it would be best to contact them about this
maybe.

I fixed it at least one time and it worked after that.

Bastian

[Sven Luther]

Post by Bastian Blank

Post by Sven Luther
I disabled the dasd patch in parted 1.6.23, because it didn't apply cleanly,
and there where some changes which made it non-trivial for me to fix it, i
asked you on irc to look at it, and think i also mailed here, but nothing
happened since then. Otavio has now taken a more active role than me in parted
maintenance, as has K.G., so it would be best to contact them about this
maybe.

I fixed it at least one time and it worked after that.

Yes, but then the code it is based on has changed again, and it needs
refixing, which is why it is not included. I asked you back then.

Now if you have fixed it again since 1.6.23, then it is another matter, can
you please try enabling the patch and confirm it builds and works indeed ?

Friendly,

Sven Luther