Message ID | bb132e9abf4d256e40ad311b106b05d724b290fb.1480607460.git.robin.murphy@arm.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
On 01/12/16 15:55, Robin Murphy wrote: > Under CONFIG_DEBUG_PREEMPT=y, this_cpu_ptr() ends up calling back into > raw_smp_processor_id(), resulting in some hilariously catastrophic > infinite recursion. In the normal case, we have: > > #define this_cpu_ptr(ptr) raw_cpu_ptr(ptr) > > and everything is dandy. However for CONFIG_DEBUG_PREEMPT, this_cpu_ptr() > is defined in terms of my_cpu_offset, wherein the fun begins: > > #define my_cpu_offset per_cpu_offset(smp_processor_id()) > ... > #define smp_processor_id() debug_smp_processor_id() > ... > notrace unsigned int debug_smp_processor_id(void) > { > return check_preemption_disabled("smp_processor_id", ""); > ... > notrace static unsigned int check_preemption_disabled(const char *what1, > const char *what2) > { > int this_cpu = raw_smp_processor_id(); > > and bang. Use raw_cpu_ptr() directly to avoid that. > > Reported-by: Marek Szyprowski <m.szyprowski@samsung.com> > Acked-by: Will Deacon <will.deacon@arm.com> > Signed-off-by: Robin Murphy <robin.murphy@arm.com> I wasn't sure whether commit IDs on for-next/core are stable, but if they are, this could also have: Fixes: 57c82954e77f ("arm64: make cpu number a percpu variable") Robin. > --- > > Since I just reproduced this locally to verify Will's suggestion, it > seemed I might as well just write it up as a patch :) > > arch/arm64/include/asm/smp.h | 4 +++- > 1 file changed, 3 insertions(+), 1 deletion(-) > > diff --git a/arch/arm64/include/asm/smp.h b/arch/arm64/include/asm/smp.h > index a62db952ffcb..d050d720a1b4 100644 > --- a/arch/arm64/include/asm/smp.h > +++ b/arch/arm64/include/asm/smp.h > @@ -41,8 +41,10 @@ DECLARE_PER_CPU_READ_MOSTLY(int, cpu_number); > * We don't use this_cpu_read(cpu_number) as that has implicit writes to > * preempt_count, and associated (compiler) barriers, that we'd like to avoid > * the expense of. If we're preemptible, the value can be stale at use anyway. > + * And we can't use this_cpu_ptr() either, as that winds up recursing back > + * here under CONFIG_DEBUG_PREEMPT=y. > */ > -#define raw_smp_processor_id() (*this_cpu_ptr(&cpu_number)) > +#define raw_smp_processor_id() (*raw_cpu_ptr(&cpu_number)) > > struct seq_file; > >
On Thu, Dec 01, 2016 at 05:16:50PM +0000, Robin Murphy wrote: > On 01/12/16 15:55, Robin Murphy wrote: > > Under CONFIG_DEBUG_PREEMPT=y, this_cpu_ptr() ends up calling back into > > raw_smp_processor_id(), resulting in some hilariously catastrophic > > infinite recursion. In the normal case, we have: > > > > #define this_cpu_ptr(ptr) raw_cpu_ptr(ptr) > > > > and everything is dandy. However for CONFIG_DEBUG_PREEMPT, this_cpu_ptr() > > is defined in terms of my_cpu_offset, wherein the fun begins: > > > > #define my_cpu_offset per_cpu_offset(smp_processor_id()) > > ... > > #define smp_processor_id() debug_smp_processor_id() > > ... > > notrace unsigned int debug_smp_processor_id(void) > > { > > return check_preemption_disabled("smp_processor_id", ""); > > ... > > notrace static unsigned int check_preemption_disabled(const char *what1, > > const char *what2) > > { > > int this_cpu = raw_smp_processor_id(); > > > > and bang. Use raw_cpu_ptr() directly to avoid that. > > > > Reported-by: Marek Szyprowski <m.szyprowski@samsung.com> > > Acked-by: Will Deacon <will.deacon@arm.com> > > Signed-off-by: Robin Murphy <robin.murphy@arm.com> > > I wasn't sure whether commit IDs on for-next/core are stable, but if > they are, this could also have: > > Fixes: 57c82954e77f ("arm64: make cpu number a percpu variable") It depends on which branch is pulled into next. I keep the for-next/core stable at this stage, so I'll include the Fixes like as well. Thanks.
Hi Robin, On 2016-12-01 16:55, Robin Murphy wrote: > Under CONFIG_DEBUG_PREEMPT=y, this_cpu_ptr() ends up calling back into > raw_smp_processor_id(), resulting in some hilariously catastrophic > infinite recursion. In the normal case, we have: > > #define this_cpu_ptr(ptr) raw_cpu_ptr(ptr) > > and everything is dandy. However for CONFIG_DEBUG_PREEMPT, this_cpu_ptr() > is defined in terms of my_cpu_offset, wherein the fun begins: > > #define my_cpu_offset per_cpu_offset(smp_processor_id()) > ... > #define smp_processor_id() debug_smp_processor_id() > ... > notrace unsigned int debug_smp_processor_id(void) > { > return check_preemption_disabled("smp_processor_id", ""); > ... > notrace static unsigned int check_preemption_disabled(const char *what1, > const char *what2) > { > int this_cpu = raw_smp_processor_id(); > > and bang. Use raw_cpu_ptr() directly to avoid that. > > Reported-by: Marek Szyprowski <m.szyprowski@samsung.com> > Acked-by: Will Deacon <will.deacon@arm.com> > Signed-off-by: Robin Murphy <robin.murphy@arm.com> Works fine now. Thanks for the proper fix. Tested-by: Marek Szyprowski <m.szyprowski@samsung.com> > --- > > Since I just reproduced this locally to verify Will's suggestion, it > seemed I might as well just write it up as a patch :) > > arch/arm64/include/asm/smp.h | 4 +++- > 1 file changed, 3 insertions(+), 1 deletion(-) > > diff --git a/arch/arm64/include/asm/smp.h b/arch/arm64/include/asm/smp.h > index a62db952ffcb..d050d720a1b4 100644 > --- a/arch/arm64/include/asm/smp.h > +++ b/arch/arm64/include/asm/smp.h > @@ -41,8 +41,10 @@ DECLARE_PER_CPU_READ_MOSTLY(int, cpu_number); > * We don't use this_cpu_read(cpu_number) as that has implicit writes to > * preempt_count, and associated (compiler) barriers, that we'd like to avoid > * the expense of. If we're preemptible, the value can be stale at use anyway. > + * And we can't use this_cpu_ptr() either, as that winds up recursing back > + * here under CONFIG_DEBUG_PREEMPT=y. > */ > -#define raw_smp_processor_id() (*this_cpu_ptr(&cpu_number)) > +#define raw_smp_processor_id() (*raw_cpu_ptr(&cpu_number)) > > struct seq_file; > Best regards
diff --git a/arch/arm64/include/asm/smp.h b/arch/arm64/include/asm/smp.h index a62db952ffcb..d050d720a1b4 100644 --- a/arch/arm64/include/asm/smp.h +++ b/arch/arm64/include/asm/smp.h @@ -41,8 +41,10 @@ DECLARE_PER_CPU_READ_MOSTLY(int, cpu_number); * We don't use this_cpu_read(cpu_number) as that has implicit writes to * preempt_count, and associated (compiler) barriers, that we'd like to avoid * the expense of. If we're preemptible, the value can be stale at use anyway. + * And we can't use this_cpu_ptr() either, as that winds up recursing back + * here under CONFIG_DEBUG_PREEMPT=y. */ -#define raw_smp_processor_id() (*this_cpu_ptr(&cpu_number)) +#define raw_smp_processor_id() (*raw_cpu_ptr(&cpu_number)) struct seq_file;