This message is from the T13 list server.

> I don't know how anyone could use
> R/W Long on a drive to force
> all the conditions you list
> without having "inside information" on how the drive works.

Eh?  English is a strange thing.  I'm not sure we agree on what "force" means.
 I wonder: do we agree, minimally, that ...

With Read/Write Long we can easily try to create an unreadable Lba: we read
long, we scramble the data, we write long, we read normal.  Maybe that creates
an error, maybe not - after all, ecc has collisions.  But given one isolated
unreadable Lba, it is then a small step to ...

3) We can "read right before and right after the bad sector" to see if the
drive does "return good data and good status".

4) We can "validate" that "read-ahead and read-behind algorithms" don't choke
over the existence of an unreadable Lba.

5) We can "verify" that "Drive Self Test" does or does not complain of
unreadable Lba's.  (The kind I like does not, provided the unreadable Lba
seemingly can be rewritten to become readable.)

2) We can get an idea of whether or not the existence of an unreadable Lba
provokes "Forced reallocation" by looking at how well we can stream data
to/from a range of Lba's that includes that Lba.

1) "Smart values" I know nothing about except the concept related fifth-hand,
but I imagine making an Lba unreadable and then trying to read that Lba ought
to log some Smart data.

What's Sad is that trivial tests like these reveal that some device vendors
don't conduct trivial tests like these themselves.  In summer 2001 it was a
major HDD supplier whose firmware design engineer had the honesty to share
with me that they believe Pio read error protocol is by definition
vendor-specific, because errors don't happen often enough to matter in real
life.  Blech.  Gag.  We help that kind of perspective to grow when we deny
customers standard tools to combat it.

Curiously yours, Pat LaVarre

P.S. Thanks to Jeff W for sharing applications of Read/Write Long.  Anybody
got more?  How about applications of Seek?

>>> Hale Landis 03/12/02 11:20PM >>>
This message is from the T13 list server.


On Tue, 12 Mar 2002 17:39:58 -0600, Wolford, Jeff wrote:
>This message is from the T13 list server.
>We use it force an error on a drive in a specific location.
>Then we can test out a number of drive features to make sure it responds
>appropriately, including (but not limited too):

I can only assume that you have information from the drive
manufacturer (most likely under NDA) that tells you how to force all
these types of errors? I don't know how anyone could use R/W Long on
a drive to force all the conditions you list without having "inside
information" on how the drive works.
...

>>> Wolford, Jeff 03/12/02 04:39PM >>>
...
We use it force an error on a drive in a specific location.

Then we can test out a number of drive features to make sure it responds
appropriately, including (but not limited too):

1) Smart values
2) Forced reallocation
3) Read right before and right after the bad sector 
 (should return good data and good status)
4) Validate read-ahead and read-behind algorithms
5) Drive Self Test verification

I'm sure there are more....

Reply via email to