On 2003-03-03, Arik Baratz wrote: > > From: Beni Cherniavsky [mailto:[EMAIL PROTECTED] > > > Yesterday my linux hang so that I had to reset it, then fsck failed > > and hang and then the partiotion table got broken :-(. It seems that > > there is a physical problem with the drive, it's giving "short read > > error"s. Couldn't find any better area to fail than around the boot > > record of the second partion?!? Now the rest of the locigal > > partitions linked list is inaccessible... > > I highly sympathize. > > > I've STFWed, I'm now using gpart to recover the other partions and > > I'll try to copy it to a new disk and proceed with recovery there... > > I just wonder what does "short read" mean - what kind of hardware > > failure gives that? In most errors it says that read returned > > 4[56]??? bytes out of 66??? requested (I don't remember the exact > > numbers). > > From what I gather, the read() function for the device usually > returns the number of bytes read. This number is then compared with > the number of bytes that were requested. If the numbers mismatch - > i.e. read() returns less bytes than it requested - it's a "short > read". > > I haven't actually searched the sources. > That's about what I guessed too for the API meaning of these errors but I wondered what actually happens in the hardware that results in such results. I would assume that the last part of the requested range is damaged but this result repeats over at least several thousands consecutive sectors, so this makes little sense...
-- Beni Cherniavsky <[EMAIL PROTECTED]> True elegance has negative overhead so it isn't subject to tradeoffs. ================================================================= To unsubscribe, send mail to [EMAIL PROTECTED] with the word "unsubscribe" in the message body, e.g., run the command echo unsubscribe | mail [EMAIL PROTECTED]
