Skip to contentSkip to conversation

kprobes: get rid of distinct type warning

PATCH Heiko Carstens 4 participants 10 messages Dec 21, 2009 – Jan 9, 2010

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

From: Heiko Carstens <[email protected]>

Every time I see this:

kernel/kprobes.c: In function 'register_kretprobe':
kernel/kprobes.c:1038: warning: comparison of distinct pointer types lacks a cast

I'm wondering if something changed in common code and we need to do something for s390. Apparently that's not the case. Let's get rid of this annoying warning.

Cc Ananth N Mavinakayanahalli <[email protected]>
---
 kernel/kprobes.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/kprobes.c +1 -1
@@ -1035,7 +1035,7 @@ int __kprobes register_kretprobe(struct 
 	/* Pre-allocate memory for max kretprobe instances */
 	if (rp->maxactive <= 0) {
 #ifdef CONFIG_PREEMPT
-		rp->maxactive = max(10, 2 * num_possible_cpus());
+		rp->maxactive = max(10U, 2 * num_possible_cpus());
 #else
 		rp->maxactive = num_possible_cpus();
 #endif
#1 lore raw
Quoted 7 lines · Heiko Carstens

From: Heiko Carstens <[email protected]>

Every time I see this:

kernel/kprobes.c: In function 'register_kretprobe':
kernel/kprobes.c:1038: warning: comparison of distinct pointer types lacks a cast

Quoted 6 lines

I'm wondering if something changed in common code and we need to do something for s390. Apparently that's not the case. Let's get rid of this annoying warning.

Cc Ananth N Mavinakayanahalli <[email protected]>

Thanks Heiko. I'd tested this on x86 and powerpc and neither throws the warning. No wonder I missed it.

author Acked-by Ananth N Mavinakayanahalli
#2 lore raw
Quoted 12 lines · Ananth N Mavinakayanahalli
Quoted 11 lines · Heiko Carstens

From: Heiko Carstens <[email protected]>

Every time I see this:

kernel/kprobes.c: In function 'register_kretprobe':
kernel/kprobes.c:1038: warning: comparison of distinct pointer types lacks a cast

I'm wondering if something changed in common code and we need to do something for s390. Apparently that's not the case. Let's get rid of this annoying warning.

Quoted 7 lines
Quoted 2 lines
Cc Ananth N Mavinakayanahalli <[email protected]>

Thanks Heiko. I'd tested this on x86 and powerpc and neither throws the warning. No wonder I missed it.

Of course the patch wouldn't help for CONFIG_PREEMPT and !CONFIG_SMP since we would have a comparison of a signed and and unsigned value again *sigh*.

So here's an updated version that should always work:

Subject: [PATCH] kprobes: get rid of distinct type warning

From: Heiko Carstens <[email protected]>

Every time I see this:

kernel/kprobes.c: In function 'register_kretprobe':
kernel/kprobes.c:1038: warning: comparison of distinct pointer types lacks a cast

I'm wondering if something changed in common code and we need to do something for s390. Apparently that's not the case. Let's get rid of this annoying warning.

---
 kernel/kprobes.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/kprobes.c +1 -1
@@ -1035,7 +1035,7 @@ int __kprobes register_kretprobe(struct 
 	/* Pre-allocate memory for max kretprobe instances */
 	if (rp->maxactive <= 0) {
 #ifdef CONFIG_PREEMPT
-		rp->maxactive = max(10, 2 * num_possible_cpus());
+		rp->maxactive = max_t(unsigned int, 10, 2*num_possible_cpus());
 #else
 		rp->maxactive = num_possible_cpus();
 #endif
#3 lore raw

[tip:perf/urgent] kprobes: Fix distinct type warning

Commit-ID:  c2ef6661ce62e26a8c0978e521fab646128a144b
Gitweb:     http://git.kernel.org/tip/c2ef6661ce62e26a8c0978e521fab646128a144b
Author:     Heiko Carstens <[email protected]>
AuthorDate: Mon, 21 Dec 2009 13:02:24 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Mon, 28 Dec 2009 10:25:31 +0100

kprobes: Fix distinct type warning

Every time I see this:

kernel/kprobes.c: In function 'register_kretprobe':
kernel/kprobes.c:1038: warning: comparison of distinct pointer types lacks a cast

I'm wondering if something changed in common code and we need to do something for s390. Apparently that's not the case. Let's get rid of this annoying warning.

LKML-Reference: <[email protected]>

---
 kernel/kprobes.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
kernel/kprobes.c +1 -1
index e5342a3..b7df302 100644
@@ -1035,7 +1035,7 @@ int __kprobes register_kretprobe(struct kretprobe *rp)
 	/* Pre-allocate memory for max kretprobe instances */
 	if (rp->maxactive <= 0) {
 #ifdef CONFIG_PREEMPT
-		rp->maxactive = max(10, 2 * num_possible_cpus());
+		rp->maxactive = max_t(unsigned int, 10, 2*num_possible_cpus());
 #else
 		rp->maxactive = num_possible_cpus();
 #endif
#4 lore raw
In reply to Heiko Carstens
Quoted 2 lines · On Mon, 21 Dec 2009 13:02:24 +0100 Heiko Carstens <[email protected]>…

Of course the patch wouldn't help for CONFIG_PREEMPT and !CONFIG_SMP since we would have a comparison of a signed and and unsigned value again *sigh*.

We should fix that, shouldn't we? Rather than working around it at one caller site.

: #if NR_CPUS > 1
: #define num_online_cpus() cpumask_weight(cpu_online_mask)
: #define num_possible_cpus() cpumask_weight(cpu_possible_mask)
: #define num_present_cpus() cpumask_weight(cpu_present_mask)
: #define num_active_cpus() cpumask_weight(cpu_active_mask)
: #define cpu_online(cpu) cpumask_test_cpu((cpu), cpu_online_mask)
: #define cpu_possible(cpu) cpumask_test_cpu((cpu), cpu_possible_mask)
: #define cpu_present(cpu) cpumask_test_cpu((cpu), cpu_present_mask)
: #define cpu_active(cpu) cpumask_test_cpu((cpu), cpu_active_mask)
: #else
: #define num_online_cpus() 1
: #define num_possible_cpus() 1
: #define num_present_cpus() 1
: #define num_active_cpus() 1
: #define cpu_online(cpu) ((cpu) == 0)
: #define cpu_possible(cpu) ((cpu) == 0)
: #define cpu_present(cpu) ((cpu) == 0)
: #define cpu_active(cpu) ((cpu) == 0)
: #endif

The num_*() "functions" return unsigned on SMP and int on UP. This is wrong.

The cpu_*() "functions" got lucky and return int in both cases.

Personally I think it's neatest if a quantity which can never be negative is held in an unsigned type. Than includes anything starting with "num". But for expediency's sake we could live with making these things consistently return "int".

Alas, changing those four num_*() "functions" to return int on SMP is a pretty wide-reaching change and will probably expose warts.

#5 lore raw
Quoted 34 lines · Andrew Morton
Quoted 2 lines · On Mon, 21 Dec 2009 13:02:24 +0100 Heiko Carstens <[email protected]>…

Of course the patch wouldn't help for CONFIG_PREEMPT and !CONFIG_SMP since we would have a comparison of a signed and and unsigned value again *sigh*.

We should fix that, shouldn't we? Rather than working around it at one caller site.

: #if NR_CPUS > 1
: #define num_online_cpus() cpumask_weight(cpu_online_mask)
: #define num_possible_cpus() cpumask_weight(cpu_possible_mask)
: #define num_present_cpus() cpumask_weight(cpu_present_mask)
: #define num_active_cpus() cpumask_weight(cpu_active_mask)
: #define cpu_online(cpu) cpumask_test_cpu((cpu), cpu_online_mask)
: #define cpu_possible(cpu) cpumask_test_cpu((cpu), cpu_possible_mask)
: #define cpu_present(cpu) cpumask_test_cpu((cpu), cpu_present_mask)
: #define cpu_active(cpu) cpumask_test_cpu((cpu), cpu_active_mask)
: #else
: #define num_online_cpus() 1
: #define num_possible_cpus() 1
: #define num_present_cpus() 1
: #define num_active_cpus() 1
: #define cpu_online(cpu) ((cpu) == 0)
: #define cpu_possible(cpu) ((cpu) == 0)
: #define cpu_present(cpu) ((cpu) == 0)
: #define cpu_active(cpu) ((cpu) == 0)
: #endif

The num_*() "functions" return unsigned on SMP and int on UP. This is wrong.

The cpu_*() "functions" got lucky and return int in both cases.

Quoted 7 lines

Personally I think it's neatest if a quantity which can never be negative is held in an unsigned type. Than includes anything starting with "num". But for expediency's sake we could live with making these things consistently return "int".

Alas, changing those four num_*() "functions" to return int on SMP is a pretty wide-reaching change and will probably expose warts.

Looks like there are quite a lot of num_* function usages in the kernel. Some seem to assume they return an int some assume an unsigned int. Don't know if it's worth changing anything here.
Maybe Rusty has an opinion.

#6 lore raw
Quoted 40 lines · Heiko Carstens
Quoted 39 lines · Andrew Morton
Quoted 2 lines · On Mon, 21 Dec 2009 13:02:24 +0100 Heiko Carstens <[email protected]>…

Of course the patch wouldn't help for CONFIG_PREEMPT and !CONFIG_SMP since we would have a comparison of a signed and and unsigned value again *sigh*.

We should fix that, shouldn't we? Rather than working around it at one caller site.

: #if NR_CPUS > 1
: #define num_online_cpus() cpumask_weight(cpu_online_mask)
: #define num_possible_cpus() cpumask_weight(cpu_possible_mask)
: #define num_present_cpus() cpumask_weight(cpu_present_mask)
: #define num_active_cpus() cpumask_weight(cpu_active_mask)
: #define cpu_online(cpu) cpumask_test_cpu((cpu), cpu_online_mask)
: #define cpu_possible(cpu) cpumask_test_cpu((cpu), cpu_possible_mask)
: #define cpu_present(cpu) cpumask_test_cpu((cpu), cpu_present_mask)
: #define cpu_active(cpu) cpumask_test_cpu((cpu), cpu_active_mask)
: #else
: #define num_online_cpus() 1
: #define num_possible_cpus() 1
: #define num_present_cpus() 1
: #define num_active_cpus() 1
: #define cpu_online(cpu) ((cpu) == 0)
: #define cpu_possible(cpu) ((cpu) == 0)
: #define cpu_present(cpu) ((cpu) == 0)
: #define cpu_active(cpu) ((cpu) == 0)
: #endif

The num_*() "functions" return unsigned on SMP and int on UP. This is wrong.

The cpu_*() "functions" got lucky and return int in both cases.

Personally I think it's neatest if a quantity which can never be negative is held in an unsigned type. Than includes anything starting with "num". But for expediency's sake we could live with making these things consistently return "int".

Quoted 7 lines
Quoted 2 lines

Alas, changing those four num_*() "functions" to return int on SMP is a pretty wide-reaching change and will probably expose warts.

Looks like there are quite a lot of num_* function usages in the kernel. Some seem to assume they return an int some assume an unsigned int. Don't know if it's worth changing anything here.
Maybe Rusty has an opinion.

If we have to go one way or the other, go with unsigned.

What does such a patch look like?

Thanks,
Rusty.

#7 lore raw
Quoted 15 lines · Rusty Russell
Quoted 14 lines · Heiko Carstens
Quoted 12 lines · Andrew Morton

The num_*() "functions" return unsigned on SMP and int on UP. This is wrong.

The cpu_*() "functions" got lucky and return int in both cases.

Personally I think it's neatest if a quantity which can never be negative is held in an unsigned type. Than includes anything starting with "num". But for expediency's sake we could live with making these things consistently return "int".

Alas, changing those four num_*() "functions" to return int on SMP is a pretty wide-reaching change and will probably expose warts.

Quoted 8 lines
Quoted 4 lines

Looks like there are quite a lot of num_* function usages in the kernel. Some seem to assume they return an int some assume an unsigned int. Don't know if it's worth changing anything here.
Maybe Rusty has an opinion.

If we have to go one way or the other, go with unsigned.

What does such a patch look like?

Something like this, doesn't even trigger new warnings on an !SMP defconfig build:

Subject: [PATCH] cpumask: let num_*_cpus() function always return unsigned values

From: Heiko Carstens <[email protected]>

Dependent on CONFIG_SMP the num_*_cpus() functions return unsigned or signed values.
Let them always return unsigned values to avoid strange casts.

---
 include/linux/cpumask.h |    8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
include/linux/cpumask.h +4 -4
@@ -90,10 +90,10 @@ extern const struct cpumask *const cpu_a
 #define cpu_present(cpu)	cpumask_test_cpu((cpu), cpu_present_mask)
 #define cpu_active(cpu)		cpumask_test_cpu((cpu), cpu_active_mask)
 #else
-#define num_online_cpus()	1
-#define num_possible_cpus()	1
-#define num_present_cpus()	1
-#define num_active_cpus()	1
+#define num_online_cpus()	1U
+#define num_possible_cpus()	1U
+#define num_present_cpus()	1U
+#define num_active_cpus()	1U
 #define cpu_online(cpu)		((cpu) == 0)
 #define cpu_possible(cpu)	((cpu) == 0)
 #define cpu_present(cpu)	((cpu) == 0)
#8 lore raw
Quoted 48 lines · On Tue, 5 Jan 2010 09:40:27 +0100 Heiko Carstens <[email protected]>…
Quoted 15 lines · Rusty Russell
Quoted 14 lines · Heiko Carstens
Quoted 12 lines · Andrew Morton

The num_*() "functions" return unsigned on SMP and int on UP. This is wrong.

The cpu_*() "functions" got lucky and return int in both cases.

Personally I think it's neatest if a quantity which can never be negative is held in an unsigned type. Than includes anything starting with "num". But for expediency's sake we could live with making these things consistently return "int".

Alas, changing those four num_*() "functions" to return int on SMP is a pretty wide-reaching change and will probably expose warts.

Quoted 8 lines
Quoted 4 lines

Looks like there are quite a lot of num_* function usages in the kernel. Some seem to assume they return an int some assume an unsigned int. Don't know if it's worth changing anything here.
Maybe Rusty has an opinion.

If we have to go one way or the other, go with unsigned.

What does such a patch look like?

Something like this, doesn't even trigger new warnings on an !SMP defconfig build:

Subject: [PATCH] cpumask: let num_*_cpus() function always return unsigned values

From: Heiko Carstens <[email protected]>

Dependent on CONFIG_SMP the num_*_cpus() functions return unsigned or signed values.
Let them always return unsigned values to avoid strange casts.

---
 include/linux/cpumask.h |    8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
include/linux/cpumask.h +0 -1
@@ -90,10 +90,10 @@ extern const struct cpumask *const cpu_a
 #define cpu_present(cpu)	cpumask_test_cpu((cpu), cpu_present_mask)
 #define cpu_active(cpu)		cpumask_test_cpu((cpu), cpu_active_mask)
 #else
-#define num_online_cpus()	1

include/linux/cpumask.h extern const struct cpumask *const cpu_a · line 99

Quoted 10 lines
include/linux/cpumask.h +4 -3
-#define num_possible_cpus()	1
-#define num_present_cpus()	1
-#define num_active_cpus()	1
+#define num_online_cpus()	1U
+#define num_possible_cpus()	1U
+#define num_present_cpus()	1U
+#define num_active_cpus()	1U
 #define cpu_online(cpu)		((cpu) == 0)
 #define cpu_possible(cpu)	((cpu) == 0)
 #define cpu_present(cpu)	((cpu) == 0)

I assume that this fixes the kprobes warning, so

Commit-ID:  c2ef6661ce62e26a8c0978e521fab646128a144b
Gitweb:     http://git.kernel.org/tip/c2ef6661ce62e26a8c0978e521fab646128a144b
Author:     Heiko Carstens <[email protected]>
AuthorDate: Mon, 21 Dec 2009 13:02:24 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Mon, 28 Dec 2009 10:25:31 +0100

becomes unneeded?

#9 lore raw
Quoted 22 lines · Andrew Morton
Quoted 7 lines · On Tue, 5 Jan 2010 09:40:27 +0100 Heiko Carstens <[email protected]>…
include/linux/cpumask.h +0 -1
@@ -90,10 +90,10 @@ extern const struct cpumask *const cpu_a
 #define cpu_present(cpu)	cpumask_test_cpu((cpu), cpu_present_mask)
 #define cpu_active(cpu)		cpumask_test_cpu((cpu), cpu_active_mask)
 #else
-#define num_online_cpus()	1

include/linux/cpumask.h extern const struct cpumask *const cpu_a · line 99

Quoted 10 lines
include/linux/cpumask.h +4 -3
-#define num_possible_cpus()	1
-#define num_present_cpus()	1
-#define num_active_cpus()	1
+#define num_online_cpus()	1U
+#define num_possible_cpus()	1U
+#define num_present_cpus()	1U
+#define num_active_cpus()	1U
 #define cpu_online(cpu)		((cpu) == 0)
 #define cpu_possible(cpu)	((cpu) == 0)
 #define cpu_present(cpu)	((cpu) == 0)

I assume that this fixes the kprobes warning, so

Quoted 8 lines
Commit-ID:  c2ef6661ce62e26a8c0978e521fab646128a144b
Gitweb:     http://git.kernel.org/tip/c2ef6661ce62e26a8c0978e521fab646128a144b
Author:     Heiko Carstens <[email protected]>
AuthorDate: Mon, 21 Dec 2009 13:02:24 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Mon, 28 Dec 2009 10:25:31 +0100

becomes unneeded?

Nah, that would be too easy. The code still compares an int (10) and an unsigned int. The only thing that could be changed with this patch is that max_t isn't needed anymore, so that the patch below would work as well (10U).
But I don't think it's worth it.

kernel/kprobes.c +1 -1
index b7df302..85222a9 100644
@@ -1035,7 +1035,7 @@ int __kprobes register_kretprobe(struct kretprobe *rp)
 	/* Pre-allocate memory for max kretprobe instances */
 	if (rp->maxactive <= 0) {
 #ifdef CONFIG_PREEMPT
-		rp->maxactive = max_t(unsigned int, 10, 2*num_possible_cpus());
+		rp->maxactive = max(10U, 2*num_possible_cpus());
 #else
 		rp->maxactive = num_possible_cpus();
 #endif