[Soc-coordination] Progress: week 2, d-i MTD
avtobiff at gmail.com
Mon Jun 8 20:33:55 UTC 2009
Before GSoC started I investigated if QEMU or nandsim,
a linux kernel module, could emulate MTD somehow. I found
that getting QEMU to play with the -mtdblock argument was
really hard (I did not get this to work). Instead, I rolled with
the nandsim module which looks like it can at least be used
for testing the first steps of the project, i.e. recognize MTD,
creating the MTD device files for partman.
* Installed emdebian armel cross-toolchain.
* Did a git svn checkout of d-i.
* Compiled a lot of kernels, trying to get the realview or
integratorcp boards to work. (They have onboard MTD so
QEMU argument -mtdblock should work with them, it did
* Got d-i/kernel to show MTD (/proc/mtd and /dev/mtdX)
* Installed debian armel on a QEMU disk. (This is needed
because d-i does not cross compile. Hardware target is
a balloon board, which is an ARM machine.)
* Read up on d-i build system, and built d-i.
* Generally took it quite easy because I just returned to my
new apartment after a finished master thesis in London. :-)
* Modified partman-base to not ignore MTD, built the udeb
and tried it out. I had trouble with either including it (on
i386) or running it with d-i (on armel) satisfactory.
I managed to get the udebs included in the armel netboot
images. There was some dependency problem though
which I haven't solved yet.
Finally, I also got the udeb included in the i386 netboot
Next 2 weeks
* Solve the remaining issues with the build and test
* Look into kernel-wedge to produce required kernel udebs
(first of all nandsim so I can test MTD identification code,
and then ubi and ubifs).
* Synchronize my proposed strategy of changes and
additions to real world d-i code. (So far my strategy is to
dig inte partman and hw-detect code more.)
* Take first steps towards recognizing MTD and integrating
this with d-i.
More information about the Soc-coordination