* RE: [PATCH]_stp_free_percpu function definition responding to _stp_alloc_percpu
@ 2006-10-18 0:05 Keshavamurthy, Anil S
2006-10-18 4:09 ` bibo,mao
0 siblings, 1 reply; 3+ messages in thread
From: Keshavamurthy, Anil S @ 2006-10-18 0:05 UTC (permalink / raw)
To: Mao, Bibo, systemtap
Will this not break the older kernels?
How can systemtap makes sure that it is
backward compatible while supporting newer kernels?
-Anil Keshavamurthy
Open Source Technology Center/SSG
Intel Corp.
-----Original Message-----
From: systemtap-owner@sourceware.org
[mailto:systemtap-owner@sourceware.org] On Behalf Of bibo,mao
Sent: Monday, October 16, 2006 4:08 AM
To: systemtap@sourceware.org
Subject: [PATCH]_stp_free_percpu function definition responding to
_stp_alloc_percpu
Hi,
On the newest 2.6.19-rc2 smp machines, stap fails to run. It
is because _stp_free_percpu is defined as free_percpu macro and
one function at the same time. This patch modifies this, and
replaces for_each_cpu with for_each_possible_cpu macro.
thanks
bibo,mao
--- src.org/runtime/alloc.c 2006-01-25 18:08:47.000000000 +0800
+++ src/runtime/alloc.c 2006-10-16 18:17:40.000000000 +0800
@@ -18,14 +18,9 @@
#define kmalloc_node(size,flags,node) kmalloc(size,flags)
#endif /* LINUX_VERSION_CODE */
-#ifdef CONFIG_SMP
-#define _stp_free_percpu(ptr) free_percpu(ptr)
-#else
-#define _stp_free_percpu(ptr) kfree(ptr)
-#endif
-
#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,15)
#define _stp_alloc_percpu(size) __alloc_percpu(size, 8)
+#define _stp_free_percpu(ptr) free_percpu(ptr)
#else
#ifdef CONFIG_SMP
/* This is like alloc_percpu() except it simply takes a size, instead
of a type. */
@@ -37,7 +32,7 @@ void *_stp_alloc_percpu(size_t size)
if (!pdata)
return NULL;
- for_each_cpu(i) {
+ for_each_possible_cpu(i) {
int node = cpu_to_node(i);
if (node_online(node))
@@ -68,7 +63,7 @@ void _stp_free_percpu(const void *objp)
int i;
struct percpu_data *p = (struct percpu_data *) (~(unsigned long)
objp);
- for_each_cpu(i)
+ for_each_possible_cpu(i)
kfree(p->ptrs[i]);
kfree(p);
}
@@ -80,6 +75,7 @@ void *_stp_alloc_percpu(size_t size)
memset(ret, 0, size);
return ret;
}
+#define _stp_free_percpu(ptr) kfree(ptr)
#endif /* CONFIG_SMP */
#endif /* LINUX_VERSION_CODE */
#endif /* _ALLOC_C_ */
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH]_stp_free_percpu function definition responding to _stp_alloc_percpu
2006-10-18 0:05 [PATCH]_stp_free_percpu function definition responding to _stp_alloc_percpu Keshavamurthy, Anil S
@ 2006-10-18 4:09 ` bibo,mao
0 siblings, 0 replies; 3+ messages in thread
From: bibo,mao @ 2006-10-18 4:09 UTC (permalink / raw)
To: Keshavamurthy, Anil S; +Cc: systemtap
ooooh, this patch will break the older kernels,
for_each_possible_cpu macro is not defined before 2.6.16
version, but modification about _stp_free_percpu does not
break the older kernels. I verified it on 2.6.9 smp kernel
version.
thanks
bibo,mao
Keshavamurthy, Anil S wrote:
> Will this not break the older kernels?
>
> How can systemtap makes sure that it is
> backward compatible while supporting newer kernels?
>
>
> -Anil Keshavamurthy
> Open Source Technology Center/SSG
> Intel Corp.
>
> -----Original Message-----
> From: systemtap-owner@sourceware.org
> [mailto:systemtap-owner@sourceware.org] On Behalf Of bibo,mao
> Sent: Monday, October 16, 2006 4:08 AM
> To: systemtap@sourceware.org
> Subject: [PATCH]_stp_free_percpu function definition responding to
> _stp_alloc_percpu
>
> Hi,
> On the newest 2.6.19-rc2 smp machines, stap fails to run. It
> is because _stp_free_percpu is defined as free_percpu macro and
> one function at the same time. This patch modifies this, and
> replaces for_each_cpu with for_each_possible_cpu macro.
>
> thanks
> bibo,mao
>
> --- src.org/runtime/alloc.c 2006-01-25 18:08:47.000000000 +0800
> +++ src/runtime/alloc.c 2006-10-16 18:17:40.000000000 +0800
> @@ -18,14 +18,9 @@
> #define kmalloc_node(size,flags,node) kmalloc(size,flags)
> #endif /* LINUX_VERSION_CODE */
>
> -#ifdef CONFIG_SMP
> -#define _stp_free_percpu(ptr) free_percpu(ptr)
> -#else
> -#define _stp_free_percpu(ptr) kfree(ptr)
> -#endif
> -
> #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,15)
> #define _stp_alloc_percpu(size) __alloc_percpu(size, 8)
> +#define _stp_free_percpu(ptr) free_percpu(ptr)
> #else
> #ifdef CONFIG_SMP
> /* This is like alloc_percpu() except it simply takes a size, instead
> of a type. */
> @@ -37,7 +32,7 @@ void *_stp_alloc_percpu(size_t size)
> if (!pdata)
> return NULL;
>
> - for_each_cpu(i) {
> + for_each_possible_cpu(i) {
> int node = cpu_to_node(i);
>
> if (node_online(node))
> @@ -68,7 +63,7 @@ void _stp_free_percpu(const void *objp)
> int i;
> struct percpu_data *p = (struct percpu_data *) (~(unsigned long)
> objp);
>
> - for_each_cpu(i)
> + for_each_possible_cpu(i)
> kfree(p->ptrs[i]);
> kfree(p);
> }
> @@ -80,6 +75,7 @@ void *_stp_alloc_percpu(size_t size)
> memset(ret, 0, size);
> return ret;
> }
> +#define _stp_free_percpu(ptr) kfree(ptr)
> #endif /* CONFIG_SMP */
> #endif /* LINUX_VERSION_CODE */
> #endif /* _ALLOC_C_ */
^ permalink raw reply [flat|nested] 3+ messages in thread
* [PATCH]_stp_free_percpu function definition responding to _stp_alloc_percpu
@ 2006-10-16 11:08 bibo,mao
0 siblings, 0 replies; 3+ messages in thread
From: bibo,mao @ 2006-10-16 11:08 UTC (permalink / raw)
To: systemtap
Hi,
On the newest 2.6.19-rc2 smp machines, stap fails to run. It
is because _stp_free_percpu is defined as free_percpu macro and
one function at the same time. This patch modifies this, and
replaces for_each_cpu with for_each_possible_cpu macro.
thanks
bibo,mao
--- src.org/runtime/alloc.c 2006-01-25 18:08:47.000000000 +0800
+++ src/runtime/alloc.c 2006-10-16 18:17:40.000000000 +0800
@@ -18,14 +18,9 @@
#define kmalloc_node(size,flags,node) kmalloc(size,flags)
#endif /* LINUX_VERSION_CODE */
-#ifdef CONFIG_SMP
-#define _stp_free_percpu(ptr) free_percpu(ptr)
-#else
-#define _stp_free_percpu(ptr) kfree(ptr)
-#endif
-
#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,15)
#define _stp_alloc_percpu(size) __alloc_percpu(size, 8)
+#define _stp_free_percpu(ptr) free_percpu(ptr)
#else
#ifdef CONFIG_SMP
/* This is like alloc_percpu() except it simply takes a size, instead
of a type. */
@@ -37,7 +32,7 @@ void *_stp_alloc_percpu(size_t size)
if (!pdata)
return NULL;
- for_each_cpu(i) {
+ for_each_possible_cpu(i) {
int node = cpu_to_node(i);
if (node_online(node))
@@ -68,7 +63,7 @@ void _stp_free_percpu(const void *objp)
int i;
struct percpu_data *p = (struct percpu_data *) (~(unsigned long) objp);
- for_each_cpu(i)
+ for_each_possible_cpu(i)
kfree(p->ptrs[i]);
kfree(p);
}
@@ -80,6 +75,7 @@ void *_stp_alloc_percpu(size_t size)
memset(ret, 0, size);
return ret;
}
+#define _stp_free_percpu(ptr) kfree(ptr)
#endif /* CONFIG_SMP */
#endif /* LINUX_VERSION_CODE */
#endif /* _ALLOC_C_ */
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2006-10-18 4:09 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2006-10-18 0:05 [PATCH]_stp_free_percpu function definition responding to _stp_alloc_percpu Keshavamurthy, Anil S
2006-10-18 4:09 ` bibo,mao
-- strict thread matches above, loose matches on Subject: below --
2006-10-16 11:08 bibo,mao
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).