the state of the shell
So, I spent all evening trying to figure out _exactly_ which differences between Chris Beard's tree and cvs-head were preventing us from getting to a shell. Here's what I found:
- The drivers/gecko/lasi.c changes made between 1999/10/19 and 1999/10/25 made the console unbearably slow. At first glance I suspect we're not handling interrupts properly, but I haven't looked at it in any sort of detail. I'll get to it soon, unless someone else does first.
- I reverted someone else's one-line change to drivers/char/serial.c that was causing the console to barf when sash tried to display something. This was ultimately the only hard problem to track down.
- We need ``argv_init[1] = NULL;'' in init/main.c, at least for the moment, so that we don't get sash's usage screen.
- There are problems with the IODC console. If I set console=tty and boot with IODC, I get:
Attempting to execute '/sbin/init'
It's for the PA-RISC!
It's executable!
It has an appropriate version!
The checksum checks out!
Well, it *smells* like a valid SOM executable
(start_brk) 40017000
(end_code) 0003c000
(start_code) 00001000
(end_data) 40008000
(start_stack) 20000028
(brk) 40017000
Kernel panic: Attempted to kill the idle task!
In swapper task - not syncing
I've not even begun to debug this. If you want to just get to a prompt, turn off the IODC console and set ``console=ttyS0'' in arch/parisc/boot/boot_code/ipl_c.c
Aside from that, using the ramdisk that Chris made (on puffin.external.hp.com), you should be able to boot to a sash prompt, at least on the A180. I've not tried it on any other machine; I'll probably try it first thing tomorrow.
Good luck--
-Phil
[parisc-linux] Wuhoo! A shell!
First off, many thanks to everyone who worked hard to get us these base parts completed, especially Philipp, Helge and Grant.
Next, many thanks to Thomas, Chris and especially Phil for not just getting seperate trees working, but one in CVS that's easily reproducable and which we can move forward with.
Obviously there's a lot of stuff to be cleared up, but these problems don't see too difficult.
- lasi.c needs to have some of the interrupt management cleared up so that we have a reasonable serial console speed. From what Phil says it sounds pretty easy. I'm not exaggerating when I say it takes 15 minutes to get to a shell on my 712/80.
- serial console code (as mentioned by Phil)
- overal console management needs to be handled properly. This includes finishing up the STI interface, the keyboard controller, and making it so that the system chooses a console based on the firmware console path settings. This seems pretty reasonable.
- it sounds like we might be at a point now where getting networking isn't so unreasonable. We have the base of the tulip and i82596 controllers going.
In any case, we can now safely say that 'Yes, PA-RISC Linux now boots to a shell'. Congrats, everyone.
- Alex
[parisc-linux] Configuration patches
Ok, as of this instant, if you checkout a fresh kernel from CVS, and hit [enter] all the way through the ``make config'' and use ramdisk.bin-19991019-workingsash from the FTP site, you should get to sash prompt when you boot the kernel image.
You will also feel the pain of the serial interrupts.
This has been verified to /work/ on an A180C.
Chris