False positives in /sbin/badblocks ?

[moliver-fC0AHe2n+mcIvw5+aKnW+Pd9D2ou9A/h at public.gmane.org]

I just got a new WD hard drive, 80GB, which I'm running
in an Adaptec enclosure from my laptop (running Debian Sarge) via
a USB 2.0 connection through an adapter in the PCMCIA slot (the
machine natively has only USB 1.1).

I decided to put the HDD through its paces using
badblocks -w (the version that writes 0xaa, reads
it, and repeats with 0x55, 0xff, 0x00).  It found
19 allegedly bad blocks, out of 78157044, which I
thought was a little disappointing.  But when I
checked them out by running "badblocks -w" on
a small interval containing the supposedly bad
block, none of them turned up bad the second time.
(Also the tests with 0xaa, 0x55, 0xff, 0x00 yielded
four disjoint sets of "bad" blocks, which was already
suspicious.)

So my current working hypothesis is that the hard
drive is actually fine, no bad blocks at all, but
every now and then there's a random error somewhere
in the flow of sending information from CPU to PCMCIA
to USB to adapter to hard disk and back.  Does that
sound right?  If so, how concerned should I be about
19 errors in 40 billion chances (figuring it's a 32-bit
bus, and there's a chance for error at read time and
another at write time)?



----- Fine messaggio inoltrato. -----

--
The Toronto Linux Users Group.      Meetings: http://tlug.ss.org
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://tlug.ss.org/subscribe.shtml