| Message ID | 1446164820-16144-1-git-send-email-icoolidge@google.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Thu, Oct 29, 2015 at 5:27 PM, Ian Coolidge <icoolidge@google.com> wrote: > This allows assert_spin_locked() to be used against > spinlocks that are const qualified. > > For example: > > struct instance_t { > int data; > spinlock_t lock; > }; > > /* > * foo does not mutate instance. > * foo must be called while the lock is held. > */ > int foo(const instance_t *instance) { > assert_spin_locked(&instance->lock); > > /* Code that assumes the lock is held */ > > return 0; > } > > Signed-off-by: Ian Coolidge <icoolidge@google.com> > --- > arch/arm/include/asm/spinlock.h | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/arch/arm/include/asm/spinlock.h b/arch/arm/include/asm/spinlock.h > index 0fa4184..274ceb1 100644 > --- a/arch/arm/include/asm/spinlock.h > +++ b/arch/arm/include/asm/spinlock.h > @@ -118,12 +118,12 @@ static inline int arch_spin_value_unlocked(arch_spinlock_t lock) > return lock.tickets.owner == lock.tickets.next; > } > > -static inline int arch_spin_is_locked(arch_spinlock_t *lock) > +static inline int arch_spin_is_locked(const arch_spinlock_t *lock) > { > return !arch_spin_value_unlocked(READ_ONCE(*lock)); > } > > -static inline int arch_spin_is_contended(arch_spinlock_t *lock) > +static inline int arch_spin_is_contended(const arch_spinlock_t *lock) > { > struct __raw_tickets tickets = READ_ONCE(lock->tickets); > return (tickets.next - tickets.owner) > 1; > -- > 2.6.0.rc2.230.g3dd15c0 > Arnd, Please see this patch for an updated version. I should have noted it is PATCH v2 and replied to the old thread, sorry about that. Anyways, I provided an example of a const spinlock. Here, in that example, the spinlock is not const itself, but it is part of a structure that may be const. I believe the example is a good one. Also, with regards to your feedback, I don't think the length of the function body is a good argument against promising not to mutate it. Thanks for taking a look, Ian
On 31 October 2015 at 18:26, Ian Coolidge <icoolidge@google.com> wrote: > On Thu, Oct 29, 2015 at 5:27 PM, Ian Coolidge <icoolidge@google.com> wrote: >> This allows assert_spin_locked() to be used against >> spinlocks that are const qualified. >> >> For example: >> >> struct instance_t { >> int data; >> spinlock_t lock; >> }; >> >> /* >> * foo does not mutate instance. >> * foo must be called while the lock is held. >> */ >> int foo(const instance_t *instance) { >> assert_spin_locked(&instance->lock); >> >> /* Code that assumes the lock is held */ >> >> return 0; >> } >> >> Signed-off-by: Ian Coolidge <icoolidge@google.com> >> --- >> arch/arm/include/asm/spinlock.h | 4 ++-- >> 1 file changed, 2 insertions(+), 2 deletions(-) >> >> diff --git a/arch/arm/include/asm/spinlock.h b/arch/arm/include/asm/spinlock.h >> index 0fa4184..274ceb1 100644 >> --- a/arch/arm/include/asm/spinlock.h >> +++ b/arch/arm/include/asm/spinlock.h >> @@ -118,12 +118,12 @@ static inline int arch_spin_value_unlocked(arch_spinlock_t lock) >> return lock.tickets.owner == lock.tickets.next; >> } >> >> -static inline int arch_spin_is_locked(arch_spinlock_t *lock) >> +static inline int arch_spin_is_locked(const arch_spinlock_t *lock) >> { >> return !arch_spin_value_unlocked(READ_ONCE(*lock)); >> } >> >> -static inline int arch_spin_is_contended(arch_spinlock_t *lock) >> +static inline int arch_spin_is_contended(const arch_spinlock_t *lock) >> { >> struct __raw_tickets tickets = READ_ONCE(lock->tickets); >> return (tickets.next - tickets.owner) > 1; >> -- >> 2.6.0.rc2.230.g3dd15c0 >> > > Arnd, > > Please see this patch for an updated version. I should have noted it > is PATCH v2 and replied to the old thread, sorry about that. > Anyways, I provided an example of a const spinlock. Here, in that > example, the spinlock is not const itself, but it is part of a > structure that may be const. > I believe the example is a good one. > > Also, with regards to your feedback, I don't think the length of the > function body is a good argument against promising not to mutate it. > Hi, Note that Ard (me) and Arnd are two different persons. I think your patch makes sense in itself, I just wonder if there are any real examples in the current codebase that would require it. Which struct is it that you are looking to constify, and in which context? Regards, Ard.
Hi Ard, :) D'oh, yes, I had assumed you were the same person. It is for a driver that I would eventually like to upstream. I see that there are many cases where assert_spin_lock is used, but perhaps driver developers are underutilizing const, or compiling the kernel with warnings tolerated. On Sat, Oct 31, 2015 at 11:34 AM, Ard Biesheuvel <ard.biesheuvel@linaro.org> wrote: > On 31 October 2015 at 18:26, Ian Coolidge <icoolidge@google.com> wrote: >> On Thu, Oct 29, 2015 at 5:27 PM, Ian Coolidge <icoolidge@google.com> wrote: >>> This allows assert_spin_locked() to be used against >>> spinlocks that are const qualified. >>> >>> For example: >>> >>> struct instance_t { >>> int data; >>> spinlock_t lock; >>> }; >>> >>> /* >>> * foo does not mutate instance. >>> * foo must be called while the lock is held. >>> */ >>> int foo(const instance_t *instance) { >>> assert_spin_locked(&instance->lock); >>> >>> /* Code that assumes the lock is held */ >>> >>> return 0; >>> } >>> >>> Signed-off-by: Ian Coolidge <icoolidge@google.com> >>> --- >>> arch/arm/include/asm/spinlock.h | 4 ++-- >>> 1 file changed, 2 insertions(+), 2 deletions(-) >>> >>> diff --git a/arch/arm/include/asm/spinlock.h b/arch/arm/include/asm/spinlock.h >>> index 0fa4184..274ceb1 100644 >>> --- a/arch/arm/include/asm/spinlock.h >>> +++ b/arch/arm/include/asm/spinlock.h >>> @@ -118,12 +118,12 @@ static inline int arch_spin_value_unlocked(arch_spinlock_t lock) >>> return lock.tickets.owner == lock.tickets.next; >>> } >>> >>> -static inline int arch_spin_is_locked(arch_spinlock_t *lock) >>> +static inline int arch_spin_is_locked(const arch_spinlock_t *lock) >>> { >>> return !arch_spin_value_unlocked(READ_ONCE(*lock)); >>> } >>> >>> -static inline int arch_spin_is_contended(arch_spinlock_t *lock) >>> +static inline int arch_spin_is_contended(const arch_spinlock_t *lock) >>> { >>> struct __raw_tickets tickets = READ_ONCE(lock->tickets); >>> return (tickets.next - tickets.owner) > 1; >>> -- >>> 2.6.0.rc2.230.g3dd15c0 >>> >> >> Arnd, >> >> Please see this patch for an updated version. I should have noted it >> is PATCH v2 and replied to the old thread, sorry about that. >> Anyways, I provided an example of a const spinlock. Here, in that >> example, the spinlock is not const itself, but it is part of a >> structure that may be const. >> I believe the example is a good one. >> >> Also, with regards to your feedback, I don't think the length of the >> function body is a good argument against promising not to mutate it. >> > > Hi, > > Note that Ard (me) and Arnd are two different persons. > > I think your patch makes sense in itself, I just wonder if there are > any real examples in the current codebase that would require it. Which > struct is it that you are looking to constify, and in which context? > > Regards, > Ard.
On Sat, Oct 31, 2015 at 12:46 PM, Ian Coolidge <icoolidge@google.com> wrote: > Hi Ard, > > :) D'oh, yes, I had assumed you were the same person. > > It is for a driver that I would eventually like to upstream. > > I see that there are many cases where assert_spin_lock is used, but > perhaps driver developers are underutilizing const, > or compiling the kernel with warnings tolerated. > >> Note that Ard (me) and Arnd are two different persons. >> >> I think your patch makes sense in itself, I just wonder if there are >> any real examples in the current codebase that would require it. Which >> struct is it that you are looking to constify, and in which context? >> >> Regards, >> Ard. Sorry for the top-post, and I meant 'assert_spin_locked'.
diff --git a/arch/arm/include/asm/spinlock.h b/arch/arm/include/asm/spinlock.h index 0fa4184..274ceb1 100644 --- a/arch/arm/include/asm/spinlock.h +++ b/arch/arm/include/asm/spinlock.h @@ -118,12 +118,12 @@ static inline int arch_spin_value_unlocked(arch_spinlock_t lock) return lock.tickets.owner == lock.tickets.next; } -static inline int arch_spin_is_locked(arch_spinlock_t *lock) +static inline int arch_spin_is_locked(const arch_spinlock_t *lock) { return !arch_spin_value_unlocked(READ_ONCE(*lock)); } -static inline int arch_spin_is_contended(arch_spinlock_t *lock) +static inline int arch_spin_is_contended(const arch_spinlock_t *lock) { struct __raw_tickets tickets = READ_ONCE(lock->tickets); return (tickets.next - tickets.owner) > 1;
This allows assert_spin_locked() to be used against spinlocks that are const qualified. For example: struct instance_t { int data; spinlock_t lock; }; /* * foo does not mutate instance. * foo must be called while the lock is held. */ int foo(const instance_t *instance) { assert_spin_locked(&instance->lock); /* Code that assumes the lock is held */ return 0; } Signed-off-by: Ian Coolidge <icoolidge@google.com> --- arch/arm/include/asm/spinlock.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)