Re: 1.5.3: clock stalls - confirmation

In message <[email protected]>, Eyal Lebedinsky writes:
>I checked the situation tonight again. If I turn off the console
>vt100 terminal (it is flaky and will crash is left on for too long)
>then a message to the console will block until the terminal is
>turned on (no problem here) but all other serial ports seem to
>block (ppp stalls, bad) and the clock stops too (bad).

Done this my self a few times.  Powered up the pc532, started up
the ppp link, but forgot to start cu on the console.  It works for
a short time then everything starts locking up.  I was under the
impression that it was syslog that was blocking, and everything
would work until it had a message to log.  Then it would block on
syslog.  Pretty soon, you can't get anything done at all.

>Are we waiting with interrupts off? I am looking at arch/dev/scn.c
>and cannot see anything more than blocking on one serial port, so
>clock ints should be honoured - right? I am still unsure about the
>preemptability of the kernel.

I guess it could be this, but it seems to me that many more things
would be broken then.  I used to use 2 ppp lines at the same time,
to different machines.  I would think that this would have some
problems if it was something as bad as waiting with interrupts off.
As I remember, heavy use could make the machine lose up to a minute
a day, so there is (or was) something causing lost clock interrupts,
but it didn't have much other noticable effect.

Jon