Re: [NTLK] Experiencing Serious ATA 1.0b8-D Data Corruption (I can reproduce it).

From: Paul Guyot (pguyot_at_kallisys.net)
Date: Thu Oct 18 2001 - 13:52:31 EDT


>I've got a MP2000 w/latest BIOS update (717260).

BIOS Update? You probably mean system update (and the latest is 710031 :)

>TURN OFF AUTOMOUNTING OF THE STORE (this will save you some headaches
>later. You don't want to experience a lockup every time you insert the
>card).

There is a workaround for that, see below.

>Upload this to the internal store of your Newton:
>
>http://www.unna.org/unna/books/Misc_Non-Fiction/Hacker_Crackdown.pkg
>
>(2.6MB)
>
>It will take up 1348kB of space (from overview button).
>
>Before upload I have 1958kB free (1565kB used). After upload I have
>585kB free (used 2938kB). That's a difference of 1353B free (or 1373kB
>used).
>
>Now transfer (file) it to the new Card store. OH DEAR! 2,427kB has been
>used up on the card!
>
>Before transfer I have 4082kB free (13kB used). After transfer I have
>1668kB free (2427kB used). The package overview still says the package
>is using 1348kB os space.
>
>OK we know something serious has gone wrong.

Er, this actually doesn't mean anything.
Linear stores and ATA stores arithmetic are very different (because
their structure is different). This difference you experience can
come from a problem on the ATA store or not.

Indeed, on ATA stores, when it says X KB free, it means 2 * X 512
sectors entirely empty. It doesn't count partially free sectors,
although the store can use it. The reason is that ATA Support doesn't
reserve sectors for transaction purposes but uses any free sector
(well, actually, the closest free sector) to save time with spinning
devices (not sure this is pertinent) but also to not use always the
same sectors which, with flash technology, would be bad for them
(nevertheless, the map is always at the same place, but it can be
moved if the sectors it is on appear to be dead).
So if I just indicate the sum of all available byte in sectors, the
result is much larger, but you can have a store full error (because
there is no free sector left) while memory infos tends to prove the
contrary.
Of course, ATA Support can use the free bytes here and there.

Nevertheless, there is of course a problem and, as you think, it's
very probably there (and it also explains the abnormally high value
of used memory).

>Let's now try and tick (select) and delete the package.
>
>It looks like it worked.
>
>But try and click on the memory info ("i", "Memory Info").
>
>The "i" stays highlighted and it looks like the Newton is locked up.

Here, I got without any problem the infos on my 16 MB CF card, but it
still said 2.4 MB used.

>BTW I think there should be a way to disable all automouting in the
>settings (perhaps there is and I've missed it).

Well, it's not already written in the documentation.
Having the system prefs open (only) -> erases the stores which are
mounted. This applies actually via the system mechanism and it works
when you mount a store (either automounted stores when inserting a
card or manually mounted stores).

Having the ATA Support prefs open -> do not auto mount stores.

>I've worked far into the next morning isolating and then reproducing
>this. So don't expect a reply for a while ;-)

Thanks for the full report. This helps. And since I have a problem
when doing the same steps (although not exactly yours, but it's very
probably linked), I think I'll be able to fix it. If I'm less tired
than I have been these last days (yes, that's why the project doesn't
advance), it should be fixed for next release.

Thanks again.

Regards,

Paul

-- 
Home page: http://www.kallisys.com/
Newton-powered WebServer: http://newt.dyndns.org:8080/

-- This is the Newtontalk mailinglist - http://www.newtontalk.net To unsubscribe or manage: visit the above link or mailto:newtontalk-request_at_newtontalk.net?Subject=unsubscribe



This archive was generated by hypermail 2.1.2 : Thu Nov 01 2001 - 10:02:18 EST