File system testing

discuss the archimedes & risc pc, peripherals and risc os/risc os on pi
Post Reply
User avatar
IanS
Posts: 1407
Joined: Mon Aug 31, 2009 7:02 pm
Contact:

File system testing

Post by IanS » Fri Mar 09, 2018 11:42 pm

As you may have seen I have re-created the ICS IDE podule (it's about as simple as an IDE podule can be). But so far it is working fine with ZIDEFS.
viewtopic.php?f=41&t=5594&start=4440#p196675

However as Dave H mentioned, I don't have a ny series resistor, without which, he cllaims problems can possibly occur.
viewtopic.php?f=41&t=5594&start=4440#p196702

The board was built as a quick-test to see if it could be done easily (I sort-of got a bit fed-up waiting for Baildon recreations, sorry Dave)
So now I have somthing wotking, I can improve on the design if needed.

So far I've only used it with a 1GB compact flash on an A440/1 (with and without backplane). I intend to do some more testing over the weekend on other machines (A3xx, A3000(external, obviously) A7000 and RISC PC).

The question is what sort of tests are useful to be confident that nothing is getting silently corrupted by the hardware.

On a PC my first-pass filesystem confidence test is disk-filltest - https://panthema.net/2013/disk-filltest/ Anyone fancy compiling it for RISC OS?

Anyone got any other suggestions for filesystem confidence tests?

crj
Posts: 846
Joined: Thu May 02, 2013 5:58 pm
Contact:

Re: File system testing

Post by crj » Sat Mar 10, 2018 1:37 am

(I know this is the 32-bit part of the forums, but I'd be interested in filesystem test suites for the Beeb/Master. And possibly somebody has something in BBC Basic that tests both?)

sirbod
Posts: 1133
Joined: Mon Apr 09, 2012 9:44 am
Location: Essex
Contact:

Re: File system testing

Post by sirbod » Sat Mar 10, 2018 9:10 am

Having recently gone through this pain to track down the root cause of disc errors we were seeing on various devices, on the RiscPC and A-series...

You need to test a variety of devices and interfaces, so IDE/SATA/SD/CF, spindle based/SSD. Ensure those that involve adapters work on another IDE interface first, use my patched version of ADFS 2.68 if required to fix all the existing know issues.

Tests you want to perform to be thorough include:
  1. Does the drive format
  2. Identifying the drive after a format
  3. Identifying the drive after a power cycle
  4. Small random reads/writes (<512 bytes)
  5. Large random reads/writes (>32Kb)
  6. All of the above with a slave device attached
  7. All of the above as slave with a master device attached
For read/write tests I'd suggest a combination of OS_File and ADFS_DiscOp, the former to confirm directory updates arent causing corruption and the later to check for random occurances of error 21/23 across the drive. If you see any issues, repeat the test on a known working IDE, to rule out drive/adapter issues.

For your random tests, use a seeded random value so the tests are repeatable.

User avatar
daveejhitchins
Posts: 5910
Joined: Wed Jun 13, 2012 6:23 pm
Location: Newton Aycliffe, County Durham
Contact:

Re: File system testing

Post by daveejhitchins » Sat Mar 10, 2018 12:29 pm

(I sort-of got a bit fed-up waiting for Baildon recreations, sorry Dave)
Tut! :roll: :lol: The problem was due to ‘ringing’. So if you keep the cable short you may be OK. I’m suprised no one has copied this version before. By the way, didn’t the version 5 have a SM IC underneath the PCB?

Dave H :D

User avatar
IanS
Posts: 1407
Joined: Mon Aug 31, 2009 7:02 pm
Contact:

Re: File system testing

Post by IanS » Sat Mar 10, 2018 1:47 pm

daveejhitchins wrote:By the way, didn’t the version 5 have a SM IC underneath the PCB?
Yes, it has a HCT174 as a page latch for the rom. As ZIDEFS is only 8K, it's unnecessary, I only need two bits of page latch (I actually have 3 in the current design so can go up to a 16K image if needed)

User avatar
IanS
Posts: 1407
Joined: Mon Aug 31, 2009 7:02 pm
Contact:

Re: File system testing

Post by IanS » Sat Mar 10, 2018 1:49 pm

sirbod wrote:Having recently gone through this pain to track down the root cause of disc errors we were seeing on various devices, on the RiscPC and A-series...
[...]
Do you have test programs you can share? I could try writing somthing, but I'd then be testing my test programs as well as the hardware.

User avatar
jgharston
Posts: 4119
Joined: Thu Sep 24, 2009 12:22 pm
Location: Whitby/Sheffield
Contact:

Re: File system testing

Post by jgharston » Sun Mar 11, 2018 2:53 am

crj wrote:(I know this is the 32-bit part of the forums, but I'd be interested in filesystem test suites for the Beeb/Master. And possibly somebody has something in BBC Basic that tests both?)
Some filing system test tools here.

Code: Select all

$ bbcbasic
PDP11 BBC BASIC IV Version 0.32
(C) Copyright J.G.Harston 1989,2005-2020
>_

User avatar
IanS
Posts: 1407
Joined: Mon Aug 31, 2009 7:02 pm
Contact:

Re: File system testing

Post by IanS » Sat Apr 07, 2018 4:40 pm

daveejhitchins wrote:(I sort-of got a bit fed-up waiting for Baildon recreations, sorry Dave)
Tut! :roll: :lol: The problem was due to ‘ringing’. So if you keep the cable short you may be OK.
Most of the "cables" I'm using are effectively zero length, I'm mainly using CF to IDE adapters plugged directly onto the PCB. The change that made everything work was the pull-up resistors on the databus going to the IDE connector. They weren't included in my original 16-bit podule layout, as they weren't there on the ICS board. However they can be added fairly easily afterwards, just by bending the common pin of a resistor pack slightly, and soldering them on the back of the board on the bus buffer chips.

I'd also been using HCT chips (again, because that's what the ICS board had), but I tried using 74HC245 buffers, they work without (and with) the pull-up resistors.

Does anyone know what the official spec for the IDE signal levels should be? (I added provision for pull-up resistors on the non-databus signals going to the IDE connector on the 8-bit podule, but with or without makes no difference)

I guess no-one had any file system test tools. I can't be certain that these boards work in all situations, but I have fairly high confidence they are ok, I've now copied many GB's of data between partitions and haven't seen a single error. I found an md5 utility on an old hard disk, and with that I confirm that the data was not corrupted by the copy/move.

Post Reply

Return to “32-bit acorn hardware”