Re: RCU question

[Andrea Arcangeli]

On Sun, Dec 12, 2004 at 11:22:49AM +0100, Manfred Spraul wrote:
> Andrea Arcangeli wrote:
> 
> >On Sun, Dec 12, 2004 at 09:59:00AM +0100, Manfred Spraul wrote:
> > 
> >
> >>It means that our NMI irq return path should check if it points to a hlt 
> >>instruction and if yes, then increase the saved EIP by one before doing 
> >>the iretd, right?
> >>   
> >>
> >
> >I don't think we'll ever post any event through nmi, so it doesn't
> >matter. We only care to be waken by real irqs, not nmi/smi. Idle loop is
> >fine to ignore the actions of the nmi handlers and to hang into the
> >"hlt".
> > 
> >
> No, You misunderstood the problem:
> 
> sti
> ** NMI handler
> ** normal interrupt arrives, is queued by the cpu
> ** irqd from NMI handler
> ** cpu notices the normal interrupt, handles it.
     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

ok. The above just wasn't obvious to me because iret of an nmi is doing
the same thing that sti does (and the nmi itself is like a cli).
Shouldn't iret wait 1 instruction too or is there a special case about
iret?  The specs only tells sti waits 1 instruction, but they don't tell
anything about iret (nor that it waits nor that it doesn't wait).  I
realized now the link posted here assumes iret isn't going to wait 1
instruction before processing pending irqs which is reasonable given the
specs don't tell anything about iret, but I didn't imagine there was a
difference between sti and iret (I mean only when iret is going to
change the interrupt enable flag from 0 to 1 just like sti does).

Overall this is a very minor issue (unless HZ is 0), it would only
introduce a 1/HZ latency to the irq that get posted while the nmi
handler is running, and the nmi handlers never runs in production.
Forcing idle=poll when the nmi watchdog is enabled is probably a
reasonable fix.

As for the SMI, I wonder how you plan to fix it.  To me it sounds like a
minor mistake that iret isn't equivalent to sti when it toggles the irq
enable bitflag (infact I don't see a way to fix it for smi, though I
know very little about smi).

thanks.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/