Messages in this thread | | | Subject | Re: [RFC GIT PULL] softirq: Consolidation and stack overrun fix | From | Benjamin Herrenschmidt <> | Date | Wed, 25 Sep 2013 06:55:47 +1000 |
| |
On Tue, 2013-09-24 at 15:56 +0200, Frederic Weisbecker wrote: > On Tue, Sep 24, 2013 at 02:42:57PM +1000, Benjamin Herrenschmidt wrote: > > On Tue, 2013-09-24 at 04:44 +0200, Frederic Weisbecker wrote: > > > So the safest way to fix this is to unconditionally call do_softirq() > > > from irq_exit(). > > > A performance penalty may come along but safety primes. > > > > > > We should probably do that and work on longer term solutions (Kconfig > > > based arch switch, etc...) > > > for the next merge window? > > > > As you prefer, though I'm keen on getting the "fast" version in RHEL7 if > > RH will take it :-) > > So what is the fast version? Converting __do_softirq() to do_softirq() > unconditionally. > > RH will accept any fix that goes upstream.
No, me fixing powerpc do_IRQ to do irq_exit run on the irq stack, and your fix for everybody else with an ifdef such that x86_64 and powerpc get to skip the additional stack switch.
> > > > From the generic code POV, it's a one-liner #ifdef to select between > > do_softirq and __do_softirq() right ? Then it's up to the arch to > > #define I_CAN_DO_FAST ! > > I'd rather say #define I_CAN_DO_SAFE :) > > But I guess the kind of symbol we want is some ARCH_HAS_IRQ_STACK_LOW_HANDLER
ARCH_IRQ_EXIT_ON_IRQ_STACK
Cheers, Ben.
> > > > > I'll respin the series plus the regression fix, unless somebody has a > > > better solution. > > > > Cheers, > > Ben. > > > >
| |