Skip to contentSkip to conversation

sched/uclamp: Fix overzealous type replacement

PATCH Applied Valentin Schneider 4 participants 8 messages Nov 13, 2019 – Nov 15, 2019 Scheduler

v1 of 2 Applied Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 sched/uclamp: Fix overzealous type replacement

Some uclamp helpers had their return type changed from 'unsigned int' to 'enum uclamp_id' by commit

0413d7f33e60 ("sched/uclamp: Always use 'enum uclamp_id' for clamp_id values")

but it happens that some *actually* do return an unsigned int value. Those are the helpers that return a utilization value: uclamp_rq_max_value() and uclamp_eff_value(). Fix those up.

Note that this doesn't lead to any obj diff using a relatively recent aarch64 compiler (8.3-2019.03). The current code of e.g. uclamp_eff_value() already figures out that the return value is 11 bits (bits_per(1024)) and doesn't seem to do anything funny. I'm still marking this as fixing the above commit to be on the safe side.

Fixes 0413d7f33e60 ("sched/uclamp: Always use 'enum uclamp_id' for clamp_id values")
---
 kernel/sched/core.c  | 4 ++--
 kernel/sched/sched.h | 2 +-
 2 files changed, 3 insertions(+), 3 deletions(-)
kernel/sched/core.c +2 -2
index 513a4794ff36..71a45025cd2e 100644
@@ -853,7 +853,7 @@ static inline void uclamp_idle_reset(struct rq *rq, enum uclamp_id clamp_id,
 }
 
 static inline
-enum uclamp_id uclamp_rq_max_value(struct rq *rq, enum uclamp_id clamp_id,
+unsigned int uclamp_rq_max_value(struct rq *rq, enum uclamp_id clamp_id,
 				   unsigned int clamp_value)
 {
 	struct uclamp_bucket *bucket = rq->uclamp[clamp_id].bucket;
@@ -918,7 +918,7 @@ uclamp_eff_get(struct task_struct *p, enum uclamp_id clamp_id)
 	return uc_req;
 }
 
-enum uclamp_id uclamp_eff_value(struct task_struct *p, enum uclamp_id clamp_id)
+unsigned int uclamp_eff_value(struct task_struct *p, enum uclamp_id clamp_id)
 {
 	struct uclamp_se uc_eff;
 
kernel/sched/sched.h +1 -1
index 05c282775f21..280a3c735935 100644
@@ -2300,7 +2300,7 @@ static inline void cpufreq_update_util(struct rq *rq, unsigned int flags) {}
 #endif /* CONFIG_CPU_FREQ */
 
 #ifdef CONFIG_UCLAMP_TASK
-enum uclamp_id uclamp_eff_value(struct task_struct *p, enum uclamp_id clamp_id);
+unsigned int uclamp_eff_value(struct task_struct *p, enum uclamp_id clamp_id);
 
 static __always_inline
 unsigned int uclamp_util_with(struct rq *rq, unsigned int util,
Signature · 1 line
2.22.0
Reviewed-by on v1 1/1
#1 lore raw
Quoted 15 lines · Valentin Schneider

Some uclamp helpers had their return type changed from 'unsigned int' to 'enum uclamp_id' by commit

0413d7f33e60 ("sched/uclamp: Always use 'enum uclamp_id' for clamp_id values")

but it happens that some *actually* do return an unsigned int value. Those are the helpers that return a utilization value: uclamp_rq_max_value() and uclamp_eff_value(). Fix those up.

Note that this doesn't lead to any obj diff using a relatively recent aarch64 compiler (8.3-2019.03). The current code of e.g. uclamp_eff_value() already figures out that the return value is 11 bits (bits_per(1024)) and doesn't seem to do anything funny. I'm still marking this as fixing the above commit to be on the safe side.

Quoted 3 lines
Fixes 0413d7f33e60 ("sched/uclamp: Always use 'enum uclamp_id' for clamp_id values")

---

The changelog could be a bit simpler and more explicitly say 0413d7f33e60 wrongly changed the return type of some functions. For a second I thought something weird is going inside these functions.

But that could be just me :-)

Thanks!

--
Qais Yousef

#2 lore raw
Quoted 16 lines · Qais Yousef
Quoted 15 lines · Valentin Schneider

Some uclamp helpers had their return type changed from 'unsigned int' to 'enum uclamp_id' by commit

0413d7f33e60 ("sched/uclamp: Always use 'enum uclamp_id' for clamp_id values")

but it happens that some *actually* do return an unsigned int value. Those are the helpers that return a utilization value: uclamp_rq_max_value() and uclamp_eff_value(). Fix those up.

Note that this doesn't lead to any obj diff using a relatively recent aarch64 compiler (8.3-2019.03). The current code of e.g. uclamp_eff_value() already figures out that the return value is 11 bits (bits_per(1024)) and doesn't seem to do anything funny. I'm still marking this as fixing the above commit to be on the safe side.

Quoted 8 lines
Quoted 3 lines
Fixes 0413d7f33e60 ("sched/uclamp: Always use 'enum uclamp_id' for clamp_id values")

---

The changelog could be a bit simpler and more explicitly say 0413d7f33e60 wrongly changed the return type of some functions. For a second I thought something weird is going inside these functions.

The first paragraph is exactly that, no? The rest (that starts with "Note that") is just optional stuff I look into because I was curious.

Quoted 9 lines

But that could be just me :-)

Thanks!

--
Qais Yousef

#3 lore raw
Quoted 21 lines · Valentin Schneider
Quoted 20 lines · Qais Yousef
Quoted 18 lines · Valentin Schneider

Some uclamp helpers had their return type changed from 'unsigned int' to 'enum uclamp_id' by commit

0413d7f33e60 ("sched/uclamp: Always use 'enum uclamp_id' for clamp_id values")

but it happens that some *actually* do return an unsigned int value. Those are the helpers that return a utilization value: uclamp_rq_max_value() and uclamp_eff_value(). Fix those up.

Note that this doesn't lead to any obj diff using a relatively recent aarch64 compiler (8.3-2019.03). The current code of e.g. uclamp_eff_value() already figures out that the return value is 11 bits (bits_per(1024)) and doesn't seem to do anything funny. I'm still marking this as fixing the above commit to be on the safe side.

Fixes 0413d7f33e60 ("sched/uclamp: Always use 'enum uclamp_id' for clamp_id values")

---

Quoted 7 lines
Quoted 4 lines

The changelog could be a bit simpler and more explicitly say 0413d7f33e60 wrongly changed the return type of some functions. For a second I thought something weird is going inside these functions.

The first paragraph is exactly that, no? The rest (that starts with "Note that") is just optional stuff I look into because I was curious.

The way it read to me is that the function was returning uclamp_id as unsigned int, hence my confusion/comment. Anyway, it's not a big deal. It's not really a problem.

Thanks

--
Qais Yousef

maintainer
#4 lore raw
In reply to Valentin Schneider
Quoted 37 lines · Valentin Schneider

Some uclamp helpers had their return type changed from 'unsigned int' to 'enum uclamp_id' by commit

0413d7f33e60 ("sched/uclamp: Always use 'enum uclamp_id' for clamp_id values")

but it happens that some *actually* do return an unsigned int value. Those are the helpers that return a utilization value: uclamp_rq_max_value() and uclamp_eff_value(). Fix those up.

Note that this doesn't lead to any obj diff using a relatively recent aarch64 compiler (8.3-2019.03). The current code of e.g. uclamp_eff_value() already figures out that the return value is 11 bits (bits_per(1024)) and doesn't seem to do anything funny. I'm still marking this as fixing the above commit to be on the safe side.

Fixes 0413d7f33e60 ("sched/uclamp: Always use 'enum uclamp_id' for clamp_id values")
---
 kernel/sched/core.c  | 4 ++--
 kernel/sched/sched.h | 2 +-
 2 files changed, 3 insertions(+), 3 deletions(-)
kernel/sched/core.c +1 -1
index 513a4794ff36..71a45025cd2e 100644
@@ -853,7 +853,7 @@ static inline void uclamp_idle_reset(struct rq *rq, enum uclamp_id clamp_id,
 }
 
 static inline
-enum uclamp_id uclamp_rq_max_value(struct rq *rq, enum uclamp_id clamp_id,
+unsigned int uclamp_rq_max_value(struct rq *rq, enum uclamp_id clamp_id,
                                   unsigned int clamp_value)
 {
        struct uclamp_bucket *bucket = rq->uclamp[clamp_id].bucket;

kernel/sched/core.c uclamp_eff_get(struct task_struct *p, enum uclamp_id clamp_id) · line 924

Quoted 8 lines
kernel/sched/core.c +1 -1
@@ -918,7 +918,7 @@ uclamp_eff_get(struct task_struct *p, enum uclamp_id clamp_id)
        return uc_req;
 }
 
-enum uclamp_id uclamp_eff_value(struct task_struct *p, enum uclamp_id clamp_id)
+unsigned int uclamp_eff_value(struct task_struct *p, enum uclamp_id clamp_id)
 {
        struct uclamp_se uc_eff;
 

And static inline enum uclamp_id uclamp_none(enum uclamp_id clamp_id) ?

Should it be fixed as well as it can return SCHED_CAPACITY_SCALE ?

Quoted 16 lines
kernel/sched/sched.h +1 -2
index 05c282775f21..280a3c735935 100644
@@ -2300,7 +2300,7 @@ static inline void cpufreq_update_util(struct rq *rq, unsigned int flags) {}
 #endif /* CONFIG_CPU_FREQ */
 
 #ifdef CONFIG_UCLAMP_TASK
-enum uclamp_id uclamp_eff_value(struct task_struct *p, enum uclamp_id clamp_id);
+unsigned int uclamp_eff_value(struct task_struct *p, enum uclamp_id clamp_id);
 
 static __always_inline
 unsigned int uclamp_util_with(struct rq *rq, unsigned int util,
--

2.22.0

#5 lore raw
Quoted 50 lines · Vincent Guittot
Quoted 37 lines · Valentin Schneider

Some uclamp helpers had their return type changed from 'unsigned int' to 'enum uclamp_id' by commit

0413d7f33e60 ("sched/uclamp: Always use 'enum uclamp_id' for clamp_id values")

but it happens that some *actually* do return an unsigned int value. Those are the helpers that return a utilization value: uclamp_rq_max_value() and uclamp_eff_value(). Fix those up.

Note that this doesn't lead to any obj diff using a relatively recent aarch64 compiler (8.3-2019.03). The current code of e.g. uclamp_eff_value() already figures out that the return value is 11 bits (bits_per(1024)) and doesn't seem to do anything funny. I'm still marking this as fixing the above commit to be on the safe side.

Fixes 0413d7f33e60 ("sched/uclamp: Always use 'enum uclamp_id' for clamp_id values")
---
 kernel/sched/core.c  | 4 ++--
 kernel/sched/sched.h | 2 +-
 2 files changed, 3 insertions(+), 3 deletions(-)
kernel/sched/core.c +1 -1
index 513a4794ff36..71a45025cd2e 100644
@@ -853,7 +853,7 @@ static inline void uclamp_idle_reset(struct rq *rq, enum uclamp_id clamp_id,
 }
 
 static inline
-enum uclamp_id uclamp_rq_max_value(struct rq *rq, enum uclamp_id clamp_id,
+unsigned int uclamp_rq_max_value(struct rq *rq, enum uclamp_id clamp_id,
                                   unsigned int clamp_value)
 {
        struct uclamp_bucket *bucket = rq->uclamp[clamp_id].bucket;

kernel/sched/core.c uclamp_eff_get(struct task_struct *p, enum uclamp_id clamp_id) · line 924

Quoted 8 lines
kernel/sched/core.c +1 -1
@@ -918,7 +918,7 @@ uclamp_eff_get(struct task_struct *p, enum uclamp_id clamp_id)
        return uc_req;
 }
 
-enum uclamp_id uclamp_eff_value(struct task_struct *p, enum uclamp_id clamp_id)
+unsigned int uclamp_eff_value(struct task_struct *p, enum uclamp_id clamp_id)
 {
        struct uclamp_se uc_eff;
 

And static inline enum uclamp_id uclamp_none(enum uclamp_id clamp_id) ?

Quoted 1 line

Should it be fixed as well as it can return SCHED_CAPACITY_SCALE ?

Indeed. That should return unsigned int too as it's returning the uclamp value.

Thanks

--
Qais Yousef

#6 lore raw
In reply to Vincent Guittot
Quoted 5 lines · Vincent Guittot

And static inline enum uclamp_id uclamp_none(enum uclamp_id clamp_id) ?

Should it be fixed as well as it can return SCHED_CAPACITY_SCALE ?

Right, thanks for staring at this. I'll go double check the diff and fix up any strays.

#7 lore raw
In reply to Valentin Schneider
The following commit has been merged into the sched/urgent branch of tip:

The following commit has been merged into the sched/urgent branch of tip:

Commit-ID:     16d6bbe7982e99f7f326ec6f088c0aaf85efa69d
Gitweb:        https://git.kernel.org/tip/16d6bbe7982e99f7f326ec6f088c0aaf85efa69d
Author:        Valentin Schneider <[email protected]>
AuthorDate:    Wed, 13 Nov 2019 16:53:34 
Committer:     Ingo Molnar <[email protected]>
CommitterDate: Fri, 15 Nov 2019 11:02:19 +01:00

sched/uclamp: Fix overzealous type replacement

Some uclamp helpers had their return type changed from 'unsigned int' to 'enum uclamp_id' by commit:

0413d7f33e60 ("sched/uclamp: Always use 'enum uclamp_id' for clamp_id values")

but it happens that some *actually* do return an unsigned int value. Those are the helpers that return a utilization value: uclamp_rq_max_value() and uclamp_eff_value(). Fix those up.

Note that this doesn't lead to any obj diff using a relatively recent aarch64 compiler (8.3-2019.03). The current code of e.g. uclamp_eff_value() already figures out that the return value is 11 bits (bits_per(1024)) and doesn't seem to do anything funny. I'm still marking this as fixing the above commit to be on the safe side.

---
 kernel/sched/core.c  | 4 ++--
 kernel/sched/sched.h | 2 +-
 2 files changed, 3 insertions(+), 3 deletions(-)
kernel/sched/core.c +2 -2
index 44123b4..a4f76d3 100644
@@ -853,7 +853,7 @@ static inline void uclamp_idle_reset(struct rq *rq, enum uclamp_id clamp_id,
 }
 
 static inline
-enum uclamp_id uclamp_rq_max_value(struct rq *rq, enum uclamp_id clamp_id,
+unsigned int uclamp_rq_max_value(struct rq *rq, enum uclamp_id clamp_id,
 				   unsigned int clamp_value)
 {
 	struct uclamp_bucket *bucket = rq->uclamp[clamp_id].bucket;
@@ -918,7 +918,7 @@ uclamp_eff_get(struct task_struct *p, enum uclamp_id clamp_id)
 	return uc_req;
 }
 
-enum uclamp_id uclamp_eff_value(struct task_struct *p, enum uclamp_id clamp_id)
+unsigned int uclamp_eff_value(struct task_struct *p, enum uclamp_id clamp_id)
 {
 	struct uclamp_se uc_eff;
 
kernel/sched/sched.h +1 -1
index c8870c5..49ed949 100644
@@ -2309,7 +2309,7 @@ static inline void cpufreq_update_util(struct rq *rq, unsigned int flags) {}
 #endif /* CONFIG_CPU_FREQ */
 
 #ifdef CONFIG_UCLAMP_TASK
-enum uclamp_id uclamp_eff_value(struct task_struct *p, enum uclamp_id clamp_id);
+unsigned int uclamp_eff_value(struct task_struct *p, enum uclamp_id clamp_id);
 
 static __always_inline
 unsigned int uclamp_util_with(struct rq *rq, unsigned int util,