Tape-caused hang (was IPChains question (SOLVED))

[jc at trillian.mit.edu (John Chambers)]

Mike writes:
	 On 2000-05-16 at 09:53 -0400, Jerry Callen wrote:
	 
	 > Ron Peterson wrote:
	 > > [...]
	 > > So I sat down at the server
	 > > console, and I couldn't even log in.  Couldn't use ssh to get in
	 > > either.  I was contemplating hitting the power switch (ouch), but tried
	 > > turning off the tape drive first.  Presto.  Everything came back.
	 > > 
	 > > Yikes.
	 > 
	 > Yikes is right! This sounds like lameness in a device driver. 
	 > 
	 > This is the sort of thing MS advocates love to point to as proof
	 > that Linux is inferior to NT. (Never mind whether NT does it better
	 > or not...)
	 
	 There's no solution to this.  If the hardware hangs, any OS hangs.

On the contrary, it's only fatal if the hardware failure locks up the
cpu.   If  the  cpu is still running, it could very well time out the
operation and return an ETIMEDOUT to the process.  Not doing this has
long  been  one of the things I've thought that the original Unix got
wrong, and which has not been fixed.   But  if  the  cpu  (and  clock
interrupts)  are  still  alive, it would be very easy to recover from
the failure and eliminate all the zombies that bad drivers produce.

-
Subcription/unsubscription/info requests: send e-mail with
"subscribe", "unsubscribe", or "info" on the first line of the
message body to discuss-request at blu.org (Subject line is ignored).