lkml.org 
[lkml]   [1996]   [Nov]   [20]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: Memory Trauam
Date

On Wed, 20 Nov 1996 19:37:07 -0500 (EST), "Richard B. Johnson"
<root@analogic.com> wrote:

>
>On Wed, 20 Nov 1996, Kenneth Albanowski wrote:
>
>>
>> On Tue, 19 Nov 1996, Richard B. Johnson wrote:
>>
>> > One thing is certainly true. Power-on-self-test and check-memory routines
>> > are just for "show". You could lean a lot more by seeing if the system
>> > boots and can then do something useful for several days.
>>
>> Hmm. Odd idea here: how difficult would it be to create a deamon that sits
>> in the background, mmaps in physical pages sequentially, and performs
>> memory tests on them? Obviously you couldn't test any owned or kernel
>> memory this way, but you could run the thing on low priority all the time.
>>
>> --
>> Kenneth Albanowski (kjahds@kjahds.com, CIS: 70705,126)
>>
>>
>I did something sorta like that to test my new Pentium 166 with the Intel
>chip. I made a simple program that declared a local 2048 byte buffer (on
>the stack). It forked a child, then wrote (forever) random numbers to
>the buffer. Of course the child did this too as did its child, etc. When
>the forked failed, I allowed the children to kill each other until everything
>unwound to the original parent. This could continue forever.
>
>Seeing that this worked okay, I had each child write a buffer of data to
>a different file then close. The result was a machine that exercised about
>everything except networking code. After about a day without any crashes,
>I felt that the machine (this machine) was solid.
>
>Cheers,
>Dick Johnson

Based on earlier discussions, it seems like running such a test system
while all caching is disabled in the BIOS would push the memory system
harder... otherwise the primary test is of the caching memory system.
The independent tasks could pause before verifying the written data,
which would expose more memory flaws.

john alvord


\
 
 \ /
  Last update: 2005-03-22 13:38    [W:0.333 / U:0.044 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site