Re: [PATCH] locking: Add __lockfunc to slow path functions

[Waiman Long <longman@redhat.com>]

On 8/9/22 17:13, Namhyung Kim wrote:
> Hello,
>
> On Tue, Aug 9, 2022 at 8:05 AM Waiman Long <longman@redhat.com> wrote:
>> On 8/8/22 13:59, Namhyung Kim wrote:
>>> So that we can skip the functions in the perf lock contention and other
>>> places like /proc/PID/wchan.
>>>
>>> Signed-off-by: Namhyung Kim <namhyung@kernel.org>
>>> ---
>>>    kernel/locking/qrwlock.c   | 4 ++--
>>>    kernel/locking/qspinlock.c | 2 +-
>>>    2 files changed, 3 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/kernel/locking/qrwlock.c b/kernel/locking/qrwlock.c
>>> index 2e1600906c9f..d2ef312a8611 100644
>>> --- a/kernel/locking/qrwlock.c
>>> +++ b/kernel/locking/qrwlock.c
>>> @@ -18,7 +18,7 @@
>>>     * queued_read_lock_slowpath - acquire read lock of a queued rwlock
>>>     * @lock: Pointer to queued rwlock structure
>>>     */
>>> -void queued_read_lock_slowpath(struct qrwlock *lock)
>>> +void __lockfunc queued_read_lock_slowpath(struct qrwlock *lock)
>>>    {
>>>        /*
>>>         * Readers come here when they cannot get the lock without waiting
>>> @@ -63,7 +63,7 @@ EXPORT_SYMBOL(queued_read_lock_slowpath);
>>>     * queued_write_lock_slowpath - acquire write lock of a queued rwlock
>>>     * @lock : Pointer to queued rwlock structure
>>>     */
>>> -void queued_write_lock_slowpath(struct qrwlock *lock)
>>> +void __lockfunc queued_write_lock_slowpath(struct qrwlock *lock)
>>>    {
>>>        int cnts;
>>>
>>> diff --git a/kernel/locking/qspinlock.c b/kernel/locking/qspinlock.c
>>> index 65a9a10caa6f..2b23378775fe 100644
>>> --- a/kernel/locking/qspinlock.c
>>> +++ b/kernel/locking/qspinlock.c
>>> @@ -313,7 +313,7 @@ static __always_inline u32  __pv_wait_head_or_lock(struct qspinlock *lock,
>>>     * contended             :    (*,x,y) +--> (*,0,0) ---> (*,0,1) -'  :
>>>     *   queue               :         ^--'                             :
>>>     */
>>> -void queued_spin_lock_slowpath(struct qspinlock *lock, u32 val)
>>> +void __lockfunc queued_spin_lock_slowpath(struct qspinlock *lock, u32 val)
>>>    {
>>>        struct mcs_spinlock *prev, *next, *node;
>>>        u32 old, tail;
>>
>> For completeness, I think you should also add it to the
>> __pv_queued_spin_unlock() and __pv_queued_spin_unlock_slowpath()
>> function in kernel/locking/qspinlock_paravirt.h. Perhaps even the
>> assembly code in arch/x86/include/asm/qspinlock_paravirt.h.
> Thanks for your comment.  I'm not sure about the asm part, will this be enough?
>
> --- a/arch/x86/include/asm/qspinlock_paravirt.h
> +++ b/arch/x86/include/asm/qspinlock_paravirt.h
> @@ -36,7 +36,7 @@ PV_CALLEE_SAVE_REGS_THUNK(__pv_queued_spin_unlock_slowpath);
>    *   rsi = lockval           (second argument)
>    *   rdx = internal variable (set to 0)
>    */
> -asm    (".pushsection .text;"
> +asm    (".pushsection .spinlock.text;"
>          ".globl " PV_UNLOCK ";"
>          ".type " PV_UNLOCK ", @function;"
>          ".align 4,0x90;"
>
That is what I meant. However, you should also a comment saying that the 
use of .spinlock.text section is equivalent to the use of __lockfunc on 
an equivalent C function to make it clear. Also add the __lockfunc to 
the pseudo code in the comment section.

Cheers,
Longman