[Top] [All Lists]

32.96 BogoMIPS on DECstation 5000/133

To: linux-mips
Subject: 32.96 BogoMIPS on DECstation 5000/133
From: "Harald Koerfgen" <>
Date: Wed, 19 Mar 1997 16:19:27 +0100
Comments: Authenticated sender is <>
Organization: Ford Motor Company
Priority: normal
Hello DECstation hackers,

yes, you are reading right. After two weeks of hard work (have i 
mentioned that i am a kernel hacker newbie?) the timer tick code for 
kmin is running. This is what i got so far:

  >>boot 3/tftp

  Linux/MIPS DECStation Boot
  Copyright (C) Paul M. Antoine 1995, 1996 and others, 1994, 1995, 1996

  Linux version 2.1.1 (harry@irgendwo) (gcc version 2.7.2) 
  #1 Mon Mar 17 22:21:03 MET 1997

  Page size is: 1000
  Bitmap size is: 1536 bytes
  Main memory size is: 49152 KB
  System id is: 82030230 
  System has an R3000 CPU
  I-Cache size: 64K, D-Cache size: 128K
  System has firmware type: TCF0
  This DECStation is a: DS5000/1x0
  Added tags
  Setup complete, launching kernel...
  Calibrating delay loop.. ok - 32.96 BogoMIPS
  Memory: 47600k/49148k available (696k kernel code, 852k data) 
  Bad pmd in pte_alloc_kernel: 8003f5bf
  free_area_pte: bad pmd (000365bf)

I commented out console_init(...) in init/main.c and used a 
register_console(pmax_prinf) instead in arch/mips/dec/decstation.c 
(thanks for the hint, Ralf).

If anybody is interested in my patches, please drop me a mail. I've 
tried to put them on the web, but my ISP is having problems with 
uploading, maybe this week... or should i just send them to Paul?

A further investigation revealed:

   - the kernel is dying (silently!) in buffer_init() in fs/buffer.c 
     because the vmalloc(...) for the hash_table fails. 

   - if buffer_init() in init/main.c is commented out, the kernel 
     even prints the (real(tm)) linux banner and dies somewhere in 
     kernel_thread(init, NULL, 0) (CONFIG_PROC_FS not set!).

I'd appreciate, if someone with more insight than me could point me into the 
direction in fixing this vmalloc stuff.

Oh... i nearly forgot to mention: the cache_size() stuff in 
arch/mips/mips1/cache.S also seems to work.

keep hacking,

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