[v1] srcu: Use suitable gfp_flags for the init_srcu_struct_nodes()

[PATCH] srcu: Use suitable gfp_flags for the init_srcu_struct_nodes()

Posted by Zqiang 1 month, 3 weeks ago

In some kernels which is set convert_to_big to SRCU_SIZING_INIT,
for use the init_srcu_struct*() to initialized srcu structure,
the is_static parameters is always false, the memory allocation
for srcu_sup structure's->node can use GFP_KERNEL flags.

Signed-off-by: Zqiang <qiang.zhang@linux.dev>
---
 kernel/rcu/srcutree.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
index ea3f128de06f..e4571b569752 100644
--- a/kernel/rcu/srcutree.c
+++ b/kernel/rcu/srcutree.c
@@ -262,7 +262,7 @@ static int init_srcu_struct_fields(struct srcu_struct *ssp, bool is_static)
 	ssp->srcu_sup->srcu_gp_seq_needed_exp = SRCU_GP_SEQ_INITIAL_VAL;
 	ssp->srcu_sup->srcu_last_gp_end = ktime_get_mono_fast_ns();
 	if (READ_ONCE(ssp->srcu_sup->srcu_size_state) == SRCU_SIZE_SMALL && SRCU_SIZING_IS_INIT()) {
-		if (!init_srcu_struct_nodes(ssp, GFP_ATOMIC))
+		if (!init_srcu_struct_nodes(ssp, !is_static ? GFP_KERNEL : GFP_ATOMIC))
 			goto err_free_sda;
 		WRITE_ONCE(ssp->srcu_sup->srcu_size_state, SRCU_SIZE_BIG);
 	}
-- 
2.48.1

Re: [PATCH] srcu: Use suitable gfp_flags for the init_srcu_struct_nodes()

Posted by Joel Fernandes 1 month, 3 weeks ago


> On Dec 13, 2025, at 10:56 PM, Zqiang <qiang.zhang@linux.dev> wrote:
> 
> In some kernels which is set convert_to_big to SRCU_SIZING_INIT,
> for use the init_srcu_struct*() to initialized srcu structure,
> the is_static parameters is always false, the memory allocation
> for srcu_sup structure's->node can use GFP_KERNEL flags.
> 
> Signed-off-by: Zqiang <qiang.zhang@linux.dev>
> ---
> kernel/rcu/srcutree.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
> index ea3f128de06f..e4571b569752 100644
> --- a/kernel/rcu/srcutree.c
> +++ b/kernel/rcu/srcutree.c
> @@ -262,7 +262,7 @@ static int init_srcu_struct_fields(struct srcu_struct *ssp, bool is_static)
>    ssp->srcu_sup->srcu_gp_seq_needed_exp = SRCU_GP_SEQ_INITIAL_VAL;
>    ssp->srcu_sup->srcu_last_gp_end = ktime_get_mono_fast_ns();
>    if (READ_ONCE(ssp->srcu_sup->srcu_size_state) == SRCU_SIZE_SMALL && SRCU_SIZING_IS_INIT()) {
> -        if (!init_srcu_struct_nodes(ssp, GFP_ATOMIC))
> +        if (!init_srcu_struct_nodes(ssp, !is_static ? GFP_KERNEL : GFP_ATOMIC))

Nit: please avoid double negatives, becomes a bit harder to read:

Instead,
is_static ? GFP_ATOMIC : GFP_KERNEL

Is it also worthwhile adding a might_sleep() here for additional robustness?

Thanks.



>            goto err_free_sda;
>        WRITE_ONCE(ssp->srcu_sup->srcu_size_state, SRCU_SIZE_BIG);
>    }
> --
> 2.48.1
> 
>

Re: [PATCH] srcu: Use suitable gfp_flags for the init_srcu_struct_nodes()

Posted by Zqiang 1 month, 3 weeks ago

> 
> > 
> > On Dec 13, 2025, at 10:56 PM, Zqiang <qiang.zhang@linux.dev> wrote:
> >  
> >  In some kernels which is set convert_to_big to SRCU_SIZING_INIT,
> >  for use the init_srcu_struct*() to initialized srcu structure,
> >  the is_static parameters is always false, the memory allocation
> >  for srcu_sup structure's->node can use GFP_KERNEL flags.
> >  
> >  Signed-off-by: Zqiang <qiang.zhang@linux.dev>
> >  ---
> >  kernel/rcu/srcutree.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >  
> >  diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
> >  index ea3f128de06f..e4571b569752 100644
> >  --- a/kernel/rcu/srcutree.c
> >  +++ b/kernel/rcu/srcutree.c
> >  @@ -262,7 +262,7 @@ static int init_srcu_struct_fields(struct srcu_struct *ssp, bool is_static)
> >  ssp->srcu_sup->srcu_gp_seq_needed_exp = SRCU_GP_SEQ_INITIAL_VAL;
> >  ssp->srcu_sup->srcu_last_gp_end = ktime_get_mono_fast_ns();
> >  if (READ_ONCE(ssp->srcu_sup->srcu_size_state) == SRCU_SIZE_SMALL && SRCU_SIZING_IS_INIT()) {
> >  - if (!init_srcu_struct_nodes(ssp, GFP_ATOMIC))
> >  + if (!init_srcu_struct_nodes(ssp, !is_static ? GFP_KERNEL : GFP_ATOMIC))
> > 
> Nit: please avoid double negatives, becomes a bit harder to read:
> 
> Instead,
> is_static ? GFP_ATOMIC : GFP_KERNEL

Ok, will do that.

> 
> Is it also worthwhile adding a might_sleep() here for additional robustness?

Would it be more appropriate to add might_sleep() before
allocating ssp->srcu_sup ?

Thanks
Zqiang


> 
> Thanks.
> 
> > 
> > goto err_free_sda;
> >  WRITE_ONCE(ssp->srcu_sup->srcu_size_state, SRCU_SIZE_BIG);
> >  }
> >  --
> >  2.48.1
> >
>

Re: [PATCH] srcu: Use suitable gfp_flags for the init_srcu_struct_nodes()

Posted by Joel Fernandes 1 month, 3 weeks ago


> On Dec 14, 2025, at 12:28 PM, Zqiang <qiang.zhang@linux.dev> wrote:
> 
> 
>> 
>> 
>>> 
>>>> On Dec 13, 2025, at 10:56 PM, Zqiang <qiang.zhang@linux.dev> wrote:
>>> 
>>> In some kernels which is set convert_to_big to SRCU_SIZING_INIT,
>>> for use the init_srcu_struct*() to initialized srcu structure,
>>> the is_static parameters is always false, the memory allocation
>>> for srcu_sup structure's->node can use GFP_KERNEL flags.
>>> 
>>> Signed-off-by: Zqiang <qiang.zhang@linux.dev>
>>> ---
>>> kernel/rcu/srcutree.c | 2 +-
>>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>> 
>>> diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
>>> index ea3f128de06f..e4571b569752 100644
>>> --- a/kernel/rcu/srcutree.c
>>> +++ b/kernel/rcu/srcutree.c
>>> @@ -262,7 +262,7 @@ static int init_srcu_struct_fields(struct srcu_struct *ssp, bool is_static)
>>> ssp->srcu_sup->srcu_gp_seq_needed_exp = SRCU_GP_SEQ_INITIAL_VAL;
>>> ssp->srcu_sup->srcu_last_gp_end = ktime_get_mono_fast_ns();
>>> if (READ_ONCE(ssp->srcu_sup->srcu_size_state) == SRCU_SIZE_SMALL && SRCU_SIZING_IS_INIT()) {
>>> - if (!init_srcu_struct_nodes(ssp, GFP_ATOMIC))
>>> + if (!init_srcu_struct_nodes(ssp, !is_static ? GFP_KERNEL : GFP_ATOMIC))
>>> 
>> Nit: please avoid double negatives, becomes a bit harder to read:
>> 
>> Instead,
>> is_static ? GFP_ATOMIC : GFP_KERNEL
> 
> Ok, will do that.
> 
>> 
>> Is it also worthwhile adding a might_sleep() here for additional robustness?
> 
> Would it be more appropriate to add might_sleep() before
> allocating ssp->srcu_sup ?

Actually this is probably not needed because the slab allocator already does that for sleepable allocations.

So feel free to ignore the suggestion.;-)

The one reason to do it might just be for documentation.

Thanks.

Re: [PATCH] srcu: Use suitable gfp_flags for the init_srcu_struct_nodes()

Posted by Zqiang 1 month, 3 weeks ago

> 
> > 
> > On Dec 14, 2025, at 12:28 PM, Zqiang <qiang.zhang@linux.dev> wrote:
> >  
> >  
> > 
> > > 
> > > 
> > > 
> >  
> >  On Dec 13, 2025, at 10:56 PM, Zqiang <qiang.zhang@linux.dev> wrote:
> >  
> >  In some kernels which is set convert_to_big to SRCU_SIZING_INIT,
> >  for use the init_srcu_struct*() to initialized srcu structure,
> >  the is_static parameters is always false, the memory allocation
> >  for srcu_sup structure's->node can use GFP_KERNEL flags.
> >  
> >  Signed-off-by: Zqiang <qiang.zhang@linux.dev>
> >  ---
> >  kernel/rcu/srcutree.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >  
> >  diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
> >  index ea3f128de06f..e4571b569752 100644
> >  --- a/kernel/rcu/srcutree.c
> >  +++ b/kernel/rcu/srcutree.c
> >  @@ -262,7 +262,7 @@ static int init_srcu_struct_fields(struct srcu_struct *ssp, bool is_static)
> >  ssp->srcu_sup->srcu_gp_seq_needed_exp = SRCU_GP_SEQ_INITIAL_VAL;
> >  ssp->srcu_sup->srcu_last_gp_end = ktime_get_mono_fast_ns();
> >  if (READ_ONCE(ssp->srcu_sup->srcu_size_state) == SRCU_SIZE_SMALL && SRCU_SIZING_IS_INIT()) {
> >  - if (!init_srcu_struct_nodes(ssp, GFP_ATOMIC))
> >  + if (!init_srcu_struct_nodes(ssp, !is_static ? GFP_KERNEL : GFP_ATOMIC))
> >  
> > 
> > > 
> > > Nit: please avoid double negatives, becomes a bit harder to read:
> > >  
> > >  Instead,
> > >  is_static ? GFP_ATOMIC : GFP_KERNEL
> > > 
> >  
> >  Ok, will do that.
> >  
> > 
> > > 
> > > Is it also worthwhile adding a might_sleep() here for additional robustness?
> > > 
> >  
> >  Would it be more appropriate to add might_sleep() before
> >  allocating ssp->srcu_sup ?
> > 
> Actually this is probably not needed because the slab allocator already does that for sleepable allocations.

Yes, I also find the might_alloc() already exists in the slub allocator. 

Thanks
Zqiang

> 
> So feel free to ignore the suggestion.;-)
> 
> The one reason to do it might just be for documentation.
> 
> Thanks.
>