linux-mips-fnet
[Top] [All Lists]

Re: Slower BogoMIPS rating on 1.3.62 than on 1.2.10 :-)

To: linux-mips@fnet.fr
Subject: Re: Slower BogoMIPS rating on 1.3.62 than on 1.2.10 :-)
From: Warner Losh <imp@village.org>
Date: Sat, 17 Feb 1996 18:27:46 -0700
In-reply-to: Your message of Fri, 16 Feb 1996 23:40:17 MST
: Anyway, sounds like I have some hacking to do to the keyboard driver.

Trivia for the keyboard hackers out there (Ralf likely knew this
already, since I stole his code):

PC keyboards have at least two scan modes.  Mode 1 you get a scan code
when the key goes down, <n> repeats of it per the autorepeat stuff,
and then an 0xf0 followed by the key that is coming up.  In Mode 2 you
when key k goes down, you get scan k, <n> repeats, and then a k + 0x80
when it comes back up.  Mode 1 and Mode 2 scan codes are completely
different.  Linux and X (I think) presume Mode 2.

The Deskstation starts out in Mode 1.  Putting it in Mode 2 either by
using the FreeBSD source (which sends 0x4d to to keyboard port) or the
Linux stuff on the #if 0 of the pica (which sends 41) or by the method
used in initalize_kbd (which sends a 65) gets things bascially working
for me.  For reasons unknown, initialize_kbd really hurts my
keyboard's brains, so I've disabled it when you are running on a
Deskstation.  With this disable, I get a single bogus scan code, but I
might be able to "eat" it before interrupts are enabled.

I now get to the VFS: Insert root floppy and press ENTER stage.
However, the read never seems to complete.  Sound familiar to anybody?
If not, then I'll take a look into this and see what is up.  Actually,
now that I'm this far, it is time to drop back 10 and do some of the
cleanup work that I need to do:
        o  Get rid of ifdefs in slots.h and mipsconfig.h through
           proper TLB initialization.
        o  Load the kernel at 0x80100000 (KSEG0 + 1M) on the
           deskstation to get around the 640K limitations of the
           motherboard (0x800a0000-0x800fffff is where the ISA bus is
           memory mapped).  Make sure that the low memory isn't
           wasted, especially on ACER machines.  Alternatively, figure
           out how to have a hole in the image...
        o  Hack MILO to:
           +    More minimal in verbage on boot (old patch integration)
           +    Compressed kernels
           +    Get the commands to actually work, rather than always
                boot linux
           +    Get the command args from the ARC BIOS to work on the
                Deskstation.
           +    Patches to allow a minimalistic MILO
           +    Proper memory stuff on the Deskstation: SIMM
                detection, etc.
        o  Look for all the XXX fixme places in the kernel that I put
           in and fix them.
        o  Take a look at the Kernel Debugging via Serial port stuff
           that Andy did and see if I can get them working.
        o  Get rid of the single spurious bogus scan code
        o  Figure out the FDC driver problems
        o  Get the Bustech SCSI driver working
        o  Investigate removing some of the if (mips_machtype == xxx)
           code by better use of "virtual" functions (aka function
           pointers in structs) to allow for easier machine
           integration and more unified kernels (long term goal).
        o  continue work on various howtos that I have out there,
           especially the word smithing of the linux-mips howto.

More progress.  Life is good.  Must drink microbrews in honor of
progress and Drew's birthday...

Warner

<Prev in Thread] Current Thread [Next in Thread>