Developers Club geek daily blog

1 year, 3 months ago
Recent habrastatya about distinctions in byte by byte identical files caused from memory depths (and a mailbox) a small piece of my correspondence with one of the engineers who were responsible at that time for the line of the disks MPG in the Fujitsu company. For convenience of English nonspeaking readers, I give transfer from English under a cat.

Venerable sir,
In 2001 I already communicated with you concerning problems with my disk Fujitsu MPG3409AH. Now I faced one more problem — I am afraid, much more the worst. Whether still I can address you? If is not present, prompt the person who is responsible for it whom I can address.

Monday, July 29, 2002, 8:57:37 AM AM, you wrote:
gffc> What problem you have with your disk?

Let's begin with bureaucratic parts:

Model: MPG3409AH
Serial number: VLxxxxxxxxCF (August, 2001)
Audit of a firmware: A9

The problem is as follows: occasionally the only bit in the read data changes with 1 on 0 — but is exclusive if there is a data exchange between HDD on primary the channel and CD ROM on secondary the channel.

The bit arrangement always same — xxxx1xxx turns into xxxx0xxx on XXXXX02E shift approximately everyone 50 read or written megabytes, but is absolutely accidental.

For example:

Shift in the file — the expected value — the read value
26002E 5A 52
C2D02E 8C 84
28002E 99 91

For the first time I noticed a problem, having copied the zip-file from a kompat-disk on the winchester: the file from a disk opened normally, and from the winchester — no; file comparison showed that in a specified way — from 1 in 0 — it was reset one bits. Then, comparing the 130-megabyte file on other svezhezapisanny compact disk to its initial copy on the winchester, I found out that copies sometimes match, and sometimes — no (!!!). Having requested bit-by-bit printout of discrepancies, I received similar result: information, read from the winchester, was spoiled from time to time. The byte damaged in the previous attempt of reading in the next attempt was correct and vice versa.

At first I sinned on memory levels in my computer. Put memory with support of ECC and to a heap added to it a cooler — fruitlessly. Suspected CD ROM and began to compare the file on a compact disk to the file on the winchester, and the copy of the file on other computer. Comparison on a network always was successful, comparison with the winchester did not. Suspected the hard disk controller (i845D chipset). Transferred the winchester to the computer with older materinka (DELL of biennial prescription — so a chipset there garanitrovanno another and CD ROM too) — and comparison errors "the winchester with CD" were reproduced.

For me it initially looks as a beaten cell in an internal cache of the winchester. However I cannot understand one — why I cannot repeat a problem when copying from Master-vinchecter on the Slave-winchester on the same controller — or from the problem winchester on him? Perhaps, because the data stream in that case flows twice more slowly, than when copying between Primary and Secondary IDE controllers?

Suspicions are cast also by my first problem with which I addressed you few years ago — it was that the disk Fujitsu of earlier series, MPD, in a random way tightly hung up — so that the RESET button but only power supply switching off — during appeals to CD ROM on Secondary IDE the controller did not help.

If not this additional, but necessary condition (an exchange between the winchester and CD ROM), I would not be so discouraged. Whether you met similar behavior?

I will repeat a set of sufficient and necessary conditions for emergence of a problem:
  1. The experimental winchester has to be exposed as Master on primary IDE the channel;
  2. CD ROM has to be exposed as Master on secondary IDE the channel;
  3. The experimental winchester and CD ROM have to transfer data at the same time actively.

The factors which are not influencing reproduction of a problem (the problem is reproduced):
  • after replacement of CD ROM of the drive
  • after replacement of RAM
  • after replacement of the power supply unit
  • after replacement of the motherboard
  • on other computer

At the same time, if to change at least one parameter from the list necessary, the bug ceases to be reproduced:
  • if to switch a vichester to secondary the controller, and CD ROM on primary;
  • if to put the winchester (or CD ROM) as Slave;
  • if during data reading from the winchester CD ROM stands idle.

Thank you in advance for your attention.
In two months I received from Fujitsu the new winchester with the updated audit of a firmware. It worked normally …
And morals of this history is as follows: if I was not accustomed beaten to compare the files copied somewhere, nobody would learn anything for a long time...

This article is a translation of the original post at habrahabr.ru/post/274235/
If you have any questions regarding the material covered in the article above, please, contact the original author of the post.
If you have any complaints about this article or you want this article to be deleted, please, drop an email here: sysmagazine.com@gmail.com.

We believe that the knowledge, which is available at the most popular Russian IT blog habrahabr.ru, should be accessed by everyone, even though it is poorly translated.
Shared knowledge makes the world better.
Best wishes.

comments powered by Disqus