Re: lug-bg: cpuinfo
- Subject: Re: lug-bg: cpuinfo
- From: danchev@email.domain.hidden (George Danchev)
- Date: Sat, 29 Mar 2003 14:42:21 +0200
On 29 03 2003 12:44, sheib wrote:
<em class="quotelev1">> predi godina na lkml imashe takyv thread as i recall.
<em class="quotelev1">> niakoi x86 cpu-ta imat problemi s l2 cache sizing.
<em class="quotelev1">> bootvai s 'cachesize=<actual size>' kato workaround.
<em class="quotelev1">>
<em class="quotelev1">> mojesh da razgledash i arch/i386/kernel/setup.c za da
<em class="quotelev1">> vidish che ne si edinstven s toia problem :)
<em class="quotelev1">> obshto vzeto pri detectvane na tipa cpu (reportvash
<em class="quotelev1">> incorrectly) mu se assignva opredeleno kolichestvo
<em class="quotelev1">> cache, koeto e mn.dyrveno, no drug nachin niama.
naistina interesno... vse pove4e se ubezhdam 4e old nice x86 cpu's sa naj
buggy crap deto mozhe da se sreshne, cheap chips;-) neznam kak stoj vyprosa s
transmeta's crusoe i tehniq code morphing oba4e de, stoto tam mozhe da
korigirat dosta gyvkavo t.e. softuerno harduernite izdunki... ta osven 4e
nqkoj x86 cpu models report-vat wrong cachesize ima i takiva kojto vyobste ne
reportvat nikyf size nemat takyf feature implementedv chip-a, za da stane
oste po-gadno gi pravqt i v nqkolko varianta s razli4en cachesize kojto ne
moze da se razpoznae stoto chip-a e lame, a pyk ne e safe da se setne tvyrdo
kato po-golemiq size value, e.g. tova maj e case-a s PIII Tualatin s 256 i
512 cache variants i se nalaga boottime overwrite za po-golemiq cachesize,
stoto ako se setne 512, pyk boot-ne na cpu s 256 ste stane kofti harduerna
izmama za kernel-a;-)
Az imam edin drug vypros malko v strani: Kak se detect-va (ili se znae
predvaritelno za vseki cpu type) kolko rings na privileges za code execution
ima cpu-to i kak Linux kernel-a upravlqva cpu-to da execute-va (izteglq ot
pametta i execute-va s opredelqni privileges) code na edin ili drug ring.
(kyde se gleda tova v source-a na kernel-a, qvno ste e nqkyde iz asm's ama
vinagi kogato sym se opitval da go razbera maj go podminavam ;-). 4uval sym
4e za x86 cpu's obiknoveno za 4 rings ot kojto kernel-a polzva 2 , e.g. 0 -
kernel code i 3 - user code ? Solaris i OS/2 maj polzvat i 4-te ring-a na
code execution privileges na x86 ili ne vinagi e taka ?
10x. (just interesting)
--
pub 4096R/0E4BD0AB 2003-03-18 (no comments) <danchev_at_spnet.net>
1AE7 7C66 0A26 5BFF DF22 5D55 1C57 0C89 0E4B D0AB <keyserver.bu.edu>
============================================================================
A mail-list of Linux Users Group - Bulgaria (bulgarian linuxers).
http://www.linux-bulgaria.org - Hosted by Internet Group Ltd. - Stara Zagora
To unsubscribe: http://www.linux-bulgaria.org/public/mail_list.html
============================================================================
|