Re: [PATCH 7/7] x86: Tell about irq stack coverage

2013-09-25 Thread Linus Torvalds
On Wed, Sep 25, 2013 at 5:40 PM, Joe Perches wrote: > > Huh? That matches all the ARCH_HAS_ patterns. Right. And they are all crap. lib/string.c is a prime example of something that should never have happened. The ARCH_HAS_xyz pattern is totally retarded. It's wrong. For big conceptual

Re: [PATCH 7/7] x86: Tell about irq stack coverage

2013-09-25 Thread Joe Perches
On Wed, 2013-09-25 at 17:21 -0700, Andrew Morton wrote: > On Wed, 25 Sep 2013 16:08:20 -0700 Linus Torvalds > wrote: > > > On Wed, Sep 25, 2013 at 9:18 AM, Frederic Weisbecker > > wrote: > > > +#ifdef CONFIG_X86_64 > > > +# define __ARCH_IRQ_EXIT_ON_IRQ_STACK > > > +#endif > > > > Ok, see

Re: [PATCH 7/7] x86: Tell about irq stack coverage

2013-09-25 Thread Andrew Morton
On Wed, 25 Sep 2013 16:08:20 -0700 Linus Torvalds wrote: > On Wed, Sep 25, 2013 at 9:18 AM, Frederic Weisbecker > wrote: > > +#ifdef CONFIG_X86_64 > > +# define __ARCH_IRQ_EXIT_ON_IRQ_STACK > > +#endif > > Ok, see the previous email, this should just be a single > > select

Re: [PATCH 7/7] x86: Tell about irq stack coverage

2013-09-25 Thread Linus Torvalds
On Wed, Sep 25, 2013 at 9:18 AM, Frederic Weisbecker wrote: > +#ifdef CONFIG_X86_64 > +# define __ARCH_IRQ_EXIT_ON_IRQ_STACK > +#endif Ok, see the previous email, this should just be a single select IRQ_EXIT_ON_IRQ_STACK in under the "config X86_64" header in arch/x86/Kconfig. And as of

[PATCH 7/7] x86: Tell about irq stack coverage

2013-09-25 Thread Frederic Weisbecker
x86-64 runs irq_exit() under the irq stack. So it can afford to run softirqs in hardirq exit without the need to switch the stacks. The hardirq stack is good enough for that. Now x86-64 runs softirqs in the hardirq stack anyway, so what we mostly skip is some needless per cpu refcounting updates

[PATCH 7/7] x86: Tell about irq stack coverage

2013-09-25 Thread Frederic Weisbecker
x86-64 runs irq_exit() under the irq stack. So it can afford to run softirqs in hardirq exit without the need to switch the stacks. The hardirq stack is good enough for that. Now x86-64 runs softirqs in the hardirq stack anyway, so what we mostly skip is some needless per cpu refcounting updates

Re: [PATCH 7/7] x86: Tell about irq stack coverage

2013-09-25 Thread Linus Torvalds
On Wed, Sep 25, 2013 at 9:18 AM, Frederic Weisbecker fweis...@gmail.com wrote: +#ifdef CONFIG_X86_64 +# define __ARCH_IRQ_EXIT_ON_IRQ_STACK +#endif Ok, see the previous email, this should just be a single select IRQ_EXIT_ON_IRQ_STACK in under the config X86_64 header in arch/x86/Kconfig.

Re: [PATCH 7/7] x86: Tell about irq stack coverage

2013-09-25 Thread Andrew Morton
On Wed, 25 Sep 2013 16:08:20 -0700 Linus Torvalds torva...@linux-foundation.org wrote: On Wed, Sep 25, 2013 at 9:18 AM, Frederic Weisbecker fweis...@gmail.com wrote: +#ifdef CONFIG_X86_64 +# define __ARCH_IRQ_EXIT_ON_IRQ_STACK +#endif Ok, see the previous email, this should just be a

Re: [PATCH 7/7] x86: Tell about irq stack coverage

2013-09-25 Thread Joe Perches
On Wed, 2013-09-25 at 17:21 -0700, Andrew Morton wrote: On Wed, 25 Sep 2013 16:08:20 -0700 Linus Torvalds torva...@linux-foundation.org wrote: On Wed, Sep 25, 2013 at 9:18 AM, Frederic Weisbecker fweis...@gmail.com wrote: +#ifdef CONFIG_X86_64 +# define __ARCH_IRQ_EXIT_ON_IRQ_STACK

Re: [PATCH 7/7] x86: Tell about irq stack coverage

2013-09-25 Thread Linus Torvalds
On Wed, Sep 25, 2013 at 5:40 PM, Joe Perches j...@perches.com wrote: Huh? That matches all the ARCH_HAS_foo patterns. Right. And they are all crap. lib/string.c is a prime example of something that should never have happened. The ARCH_HAS_xyz pattern is totally retarded. It's wrong. For big