Messages in this thread | | | Date | Mon, 15 Mar 1999 22:39:24 +0100 | From | Kurt Garloff <> | Subject | Re: [patch] recover lost ticks |
| |
On Mon, Mar 15, 1999 at 01:28:27PM +0100, Ingo Molnar wrote: > > I am not fixing the symptom, I am recovering the bug at runtime kprinting > > the useful information as in your patch but with the difference that then > > I am allowing the machine to run fine until the problem will get fixed > > (and to get it fixed you may need to wait some time like in the scsi > > code). > > the SCSI case is pure paranoia and an unmaintained subsystem. Fixing (one > of ) the effects of overlong cli() just delays the real fix. (why should > we bother fixing them when system time runs just fine) Some other > side-effects of overlong clis, apart from lost timer ticks: lost > characters on the serial line, lost keyboard events. If someone doing > strange things can live with these other effects, he sure can live with > lost timer ticks as well.
Well, almost any SCSI driver does things like for (msec=1000; msec--; ) udelay (1000); after a bus RESET was issued or detected. And the IRQ-safe io_request spinlock is held.
I am not happy with this, and posted something to l-scsi, but nobody was proposing a good solution other than major SCSI code rework in 2.3.
Maybe you can come up with a nice proposal. Since then, I prefer Arca's patch which does both detect the problem and still tries to the user from trouble.
Regards, -- Dipl.Phys. Kurt Garloff <kurt@garloff.de> [Dortmund, FRG] Plasma physics, high perf. computing [Linux-ix86,-axp, DUX] PGP key: see mailheader / key servers [Linux SCSI driver: DC390] [unhandled content-type:application/pgp-signature] | |