![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
So while trying to debug my kernel, I found a bug in another peice of software... it turns out ksymoops doesn't behavior correctly on cases where nm/objdump decide a file format is ambiguous. Ksymoops patch: http://dev.gentoo.org/~robbat2/xxs1500/ksymoops-2.4.9-target-fix.patch.gz
Latest kernel patch: http://dev.gentoo.org/~robbat2/xxs1500/linux-xxs1500-20040825.patch.gz
Had markedly more success with the XXS1500 today. After digging in the old config file, I found out that I'd totally missed HW_HAS_PCI, so the kernel was acting as if PCI didn't exist. Sorted it out in the Kconfigs, and a bunch more things fell into place. PCMCIA and PCI _nearly_ work. See my kernel patch above, and look for the two XXX markers. I need to see about the actual wiring schematic of the system to solve those two.
Cleaned up the include tree a bit, properly creating the asm/mach-xxs1500 instead of having the header directly in asm.
Cleaned up the au1000 handling of the boot cmdline, so that it actually uses CONFIG_CMDLINE, and makes life easier (editing your params straight into the file is nasty).
I suspect there is a bug somewhere in the NoOp I/O scheduler, as if I use it, I get a lot of panics. Switched back to the deadline scheduler, works much better :-).
Got my initial NFSRoot mostly up and working, a few slight glitches left, as the initial NFSRoot I got my hands on is for 2.4 only, I'll have to ask `Kumba if he has 2.6 mipsel stage available, or could build be a few packages I could use to convert the stage as needed. Packages needed: pcmcia-cs*, module-init-tools, nfs-utils, portmap
Once I've got the initial NFSRoot up (it uses glibc), I can build some uclibc stages, and nfsroot into them, ready to create some embedded disk images. I need to see about adding mtd-utils into the Gentoo tree first, as I'll need it for doing this.
Console logs in the usual place.
Latest kernel patch: http://dev.gentoo.org/~robbat2/xxs1500/linux-xxs1500-20040825.patch.gz
Had markedly more success with the XXS1500 today. After digging in the old config file, I found out that I'd totally missed HW_HAS_PCI, so the kernel was acting as if PCI didn't exist. Sorted it out in the Kconfigs, and a bunch more things fell into place. PCMCIA and PCI _nearly_ work. See my kernel patch above, and look for the two XXX markers. I need to see about the actual wiring schematic of the system to solve those two.
Cleaned up the include tree a bit, properly creating the asm/mach-xxs1500 instead of having the header directly in asm.
Cleaned up the au1000 handling of the boot cmdline, so that it actually uses CONFIG_CMDLINE, and makes life easier (editing your params straight into the file is nasty).
I suspect there is a bug somewhere in the NoOp I/O scheduler, as if I use it, I get a lot of panics. Switched back to the deadline scheduler, works much better :-).
Got my initial NFSRoot mostly up and working, a few slight glitches left, as the initial NFSRoot I got my hands on is for 2.4 only, I'll have to ask `Kumba if he has 2.6 mipsel stage available, or could build be a few packages I could use to convert the stage as needed. Packages needed: pcmcia-cs*, module-init-tools, nfs-utils, portmap
Once I've got the initial NFSRoot up (it uses glibc), I can build some uclibc stages, and nfsroot into them, ready to create some embedded disk images. I need to see about adding mtd-utils into the Gentoo tree first, as I'll need it for doing this.
Console logs in the usual place.