Re: [9fans] some installation problems (was "panic: vmap")

[Russ Cox <rsc@swtch.com>]

> The wiki gives the following instructions to modify /lib/vgadb:
> ramfs
> ed /lib/vgadb

> If you select Boot at startup, you aren't given write permission to
> /lib/vgadb, and as a result, these instructions can't be followed.  If
> you select Install at startup, you can do this just fine.  However,
> executing "ramfs" produces the following output: '/bin/boot' file does
> not exist.  Luckily you don't even need the temporary filesystem,
> because a write to /lib/vgadb works just fine.  If this behavior is
> how it should be, the wiki should be changed, but I will wait on that
> until it's decided what the desired behavior is.

The ramfs is for ed's /tmp files, not /lib/vgadb.
I think on the boot floppy /tmp is not writable.

> The following is just a suggestion.  When Install is selected, a few
> files are missing from /bin.  One of them is p.  If a user is forced
> to exist in a pre-install text screen, this would be a big help in
> reading files.

There are a lot of files missing from bin.  It's hard to get them all
onto a floppy image, which is what you're booting off of when you
choose install.

> The last is in regards to my video card.  When I boot and select the
> vga monitor, I am taken to a vga screen running an extremely slowly
> printing and responding terminal, and the mouse cursor is a distorted
> image.  When started by answering vga as the monitor, the vga terminal
> (slowly) prints messages resembling "idle stat 172 put 177 scr
> f02f18a4 F0162156".  I can instead pick none for the monitor and start
> the vga screen with aux/vga, the result being the same slow terminal,
> but without the idle messages.

This means that the video acceleration in the kernel isn't
driving your card quite right and is timing out.
If you echo hwaccel off >/dev/vgactl this will go away.
Doing it before running aux/vga is easiest.

Another solution should be to run with monitor=vesa.

I doubt your cursor shows up either, but that's not related.
I fixed the cursor code in the native nvidia driver just
now (I think).  Running with monitor=vesa will avoid this
bug too.

Russ