2013-07-29 02:30:11

by Xie XiuQi

[permalink] [raw]
Subject: [PATCH] smp: kill unnecessary variable - csd_flags

We used csd_flags formerly because we allocated csd_data by
kmalloc when "wait == 0". When fail to allocation, we will
fall back to on-stack allocation. "csd_data" might be invalid
after generic_exec_single return.

But now we use per cpu data for single cpu ipi calls, and
csd_data can't fall back to on-stack allocation when "wait == 0".

So csd_flags is unnecessary now. Remove it.

Signed-off-by: Xie XiuQi <[email protected]>
---
kernel/smp.c | 11 +----------
1 files changed, 1 insertions(+), 10 deletions(-)

diff --git a/kernel/smp.c b/kernel/smp.c
index 4dba0f7..cac2b6e 100644
--- a/kernel/smp.c
+++ b/kernel/smp.c
@@ -186,25 +186,16 @@ void generic_smp_call_function_single_interrupt(void)

while (!list_empty(&list)) {
struct call_single_data *csd;
- unsigned int csd_flags;

csd = list_entry(list.next, struct call_single_data, list);
list_del(&csd->list);

- /*
- * 'csd' can be invalid after this call if flags == 0
- * (when called through generic_exec_single()),
- * so save them away before making the call:
- */
- csd_flags = csd->flags;
-
csd->func(csd->info);

/*
* Unlocked CSDs are valid through generic_exec_single():
*/
- if (csd_flags & CSD_FLAG_LOCK)
- csd_unlock(csd);
+ csd_unlock(csd);
}
}

--
1.6.0.2


2013-07-29 12:13:21

by Peter Zijlstra

[permalink] [raw]
Subject: Re: [PATCH] smp: kill unnecessary variable - csd_flags

On Mon, Jul 29, 2013 at 10:29:45AM +0800, Xie XiuQi wrote:
> We used csd_flags formerly because we allocated csd_data by
> kmalloc when "wait == 0". When fail to allocation, we will
> fall back to on-stack allocation. "csd_data" might be invalid
> after generic_exec_single return.
>
> But now we use per cpu data for single cpu ipi calls, and
> csd_data can't fall back to on-stack allocation when "wait == 0".
>
> So csd_flags is unnecessary now. Remove it.

The much simpler argument is that both callsites of
generic_exec_single() do an unconditional csd_lock().

> Signed-off-by: Xie XiuQi <[email protected]>
> ---
> kernel/smp.c | 11 +----------
> 1 files changed, 1 insertions(+), 10 deletions(-)
>
> diff --git a/kernel/smp.c b/kernel/smp.c
> index 4dba0f7..cac2b6e 100644
> --- a/kernel/smp.c
> +++ b/kernel/smp.c
> @@ -186,25 +186,16 @@ void generic_smp_call_function_single_interrupt(void)
>
> while (!list_empty(&list)) {
> struct call_single_data *csd;
> - unsigned int csd_flags;
>
> csd = list_entry(list.next, struct call_single_data, list);
> list_del(&csd->list);
>
> - /*
> - * 'csd' can be invalid after this call if flags == 0
> - * (when called through generic_exec_single()),
> - * so save them away before making the call:
> - */
> - csd_flags = csd->flags;
> -
> csd->func(csd->info);
>
> /*
> * Unlocked CSDs are valid through generic_exec_single():
> */
> - if (csd_flags & CSD_FLAG_LOCK)
> - csd_unlock(csd);
> + csd_unlock(csd);

The comment is completely useless and confusing after this; why do you
leave it in?

2013-07-30 01:47:42

by Xie XiuQi

[permalink] [raw]
Subject: Re: [PATCH] smp: kill unnecessary variable - csd_flags

On 2013/7/29 20:13, Peter Zijlstra wrote:
> On Mon, Jul 29, 2013 at 10:29:45AM +0800, Xie XiuQi wrote:
>> We used csd_flags formerly because we allocated csd_data by
>> kmalloc when "wait == 0". When fail to allocation, we will
>> fall back to on-stack allocation. "csd_data" might be invalid
>> after generic_exec_single return.
>>
>> But now we use per cpu data for single cpu ipi calls, and
>> csd_data can't fall back to on-stack allocation when "wait == 0".
>>
>> So csd_flags is unnecessary now. Remove it.
>
> The much simpler argument is that both callsites of
> generic_exec_single() do an unconditional csd_lock().
>

That's right. All csd_data are protected by csd_lock.

>> Signed-off-by: Xie XiuQi <[email protected]>
>> ---
>> kernel/smp.c | 11 +----------
>> 1 files changed, 1 insertions(+), 10 deletions(-)
>>
>> diff --git a/kernel/smp.c b/kernel/smp.c
>> index 4dba0f7..cac2b6e 100644
>> --- a/kernel/smp.c
>> +++ b/kernel/smp.c
>> @@ -186,25 +186,16 @@ void generic_smp_call_function_single_interrupt(void)
>>
>> while (!list_empty(&list)) {
>> struct call_single_data *csd;
>> - unsigned int csd_flags;
>>
>> csd = list_entry(list.next, struct call_single_data, list);
>> list_del(&csd->list);
>>
>> - /*
>> - * 'csd' can be invalid after this call if flags == 0
>> - * (when called through generic_exec_single()),
>> - * so save them away before making the call:
>> - */
>> - csd_flags = csd->flags;
>> -
>> csd->func(csd->info);
>>
>> /*
>> * Unlocked CSDs are valid through generic_exec_single():
>> */
>> - if (csd_flags & CSD_FLAG_LOCK)
>> - csd_unlock(csd);
>> + csd_unlock(csd);
>
> The comment is completely useless and confusing after this; why do you
> leave it in?
>

Thanks for your comment, I'll remove the comment and send a new version soon.