* [PATCH 2/4] rt-tests: cyclictest segfault with '-a'
2012-05-02 2:35 [PATCH 1/4] rt-tests: printf format compile warning Frank Rowand
@ 2012-05-02 2:44 ` Frank Rowand
2012-05-02 20:37 ` Darren Hart
2012-05-02 2:45 ` [PATCH 3/4] rt-tests: cyclictest avoid unneeded warning Frank Rowand
` (2 subsequent siblings)
3 siblings, 1 reply; 8+ messages in thread
From: Frank Rowand @ 2012-05-02 2:44 UTC (permalink / raw)
To: linux-rt-users@vger.kernel.org, williams@redhat.com
Cc: jkacur@redhat.com, dvhart@linux.intel.com
This fixes a segfault on ARM when the '-a' option is used.
man sched_setaffinity says to use pthread_setaffinity_np() when using the
POSIX threads API.
Signed-off-by: Frank Rowand <frank.rowand@am.sony.com>
---
src/cyclictest/cyclictest.c | 4 3 + 1 - 0 !
1 file changed, 3 insertions(+), 1 deletion(-)
Index: b/src/cyclictest/cyclictest.c
===================================================================
--- a/src/cyclictest/cyclictest.c
+++ b/src/cyclictest/cyclictest.c
@@ -612,6 +612,7 @@ void *timerthread(void *param)
struct thread_stat *stat = par->stats;
int stopped = 0;
cpu_set_t mask;
+ pthread_t thread;
/* if we're running in numa mode, set our memory node */
if (par->node != -1)
@@ -620,7 +621,8 @@ void *timerthread(void *param)
if (par->cpu != -1) {
CPU_ZERO(&mask);
CPU_SET(par->cpu, &mask);
- if(sched_setaffinity(0, sizeof(mask), &mask) == -1)
+ thread = pthread_self();
+ if(pthread_setaffinity_np(thread, sizeof(mask), &mask) == -1)
warn("Could not set CPU affinity to CPU #%d\n", par->cpu);
}
^ permalink raw reply [flat|nested] 8+ messages in thread* Re: [PATCH 2/4] rt-tests: cyclictest segfault with '-a'
2012-05-02 2:44 ` [PATCH 2/4] rt-tests: cyclictest segfault with '-a' Frank Rowand
@ 2012-05-02 20:37 ` Darren Hart
0 siblings, 0 replies; 8+ messages in thread
From: Darren Hart @ 2012-05-02 20:37 UTC (permalink / raw)
To: frank.rowand
Cc: linux-rt-users@vger.kernel.org, williams@redhat.com,
jkacur@redhat.com
On 05/01/2012 07:44 PM, Frank Rowand wrote:
> This fixes a segfault on ARM when the '-a' option is used.
>
> man sched_setaffinity says to use pthread_setaffinity_np() when using the
> POSIX threads API.
>
> Signed-off-by: Frank Rowand <frank.rowand@am.sony.com>
Compile tested and tried "-a 0", but my ARM kernel/emulation is UP, so
not sure this is of much value.
Tested-by: Darren Hart <dvhart@linux.intel.com>
> ---
> src/cyclictest/cyclictest.c | 4 3 + 1 - 0 !
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> Index: b/src/cyclictest/cyclictest.c
> ===================================================================
> --- a/src/cyclictest/cyclictest.c
> +++ b/src/cyclictest/cyclictest.c
> @@ -612,6 +612,7 @@ void *timerthread(void *param)
> struct thread_stat *stat = par->stats;
> int stopped = 0;
> cpu_set_t mask;
> + pthread_t thread;
>
> /* if we're running in numa mode, set our memory node */
> if (par->node != -1)
> @@ -620,7 +621,8 @@ void *timerthread(void *param)
> if (par->cpu != -1) {
> CPU_ZERO(&mask);
> CPU_SET(par->cpu, &mask);
> - if(sched_setaffinity(0, sizeof(mask), &mask) == -1)
> + thread = pthread_self();
> + if(pthread_setaffinity_np(thread, sizeof(mask), &mask) == -1)
> warn("Could not set CPU affinity to CPU #%d\n", par->cpu);
> }
>
>
--
Darren Hart
Intel Open Source Technology Center
Yocto Project - Linux Kernel
^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH 3/4] rt-tests: cyclictest avoid unneeded warning
2012-05-02 2:35 [PATCH 1/4] rt-tests: printf format compile warning Frank Rowand
2012-05-02 2:44 ` [PATCH 2/4] rt-tests: cyclictest segfault with '-a' Frank Rowand
@ 2012-05-02 2:45 ` Frank Rowand
2012-05-02 20:38 ` Darren Hart
2012-05-02 2:45 ` [PATCH 4/4] rt-tests: cyclictest warn of interaction between '-a', '--smp', and '--numa' Frank Rowand
2012-05-02 20:36 ` [PATCH 1/4] rt-tests: printf format compile warning Darren Hart
3 siblings, 1 reply; 8+ messages in thread
From: Frank Rowand @ 2012-05-02 2:45 UTC (permalink / raw)
To: linux-rt-users@vger.kernel.org, williams@redhat.com
Cc: jkacur@redhat.com, dvhart@linux.intel.com
Avoid annoying warning message when tracing is not requested and the debug
file system is not available.
The same test already protects against calling event_enable_all().
Signed-off-by: Frank Rowand <frank.rowand@am.sony.com>
---
src/cyclictest/cyclictest.c | 5 3 + 2 - 0 !
1 file changed, 3 insertions(+), 2 deletions(-)
Index: b/src/cyclictest/cyclictest.c
===================================================================
--- a/src/cyclictest/cyclictest.c
+++ b/src/cyclictest/cyclictest.c
@@ -1605,8 +1605,9 @@ int main(int argc, char **argv)
if (trace_fd >= 0)
close(trace_fd);
- /* turn off all events */
- event_disable_all();
+ if (enable_events)
+ /* turn off all events */
+ event_disable_all();
/* turn off the function tracer */
fileprefix = procfileprefix;
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH 3/4] rt-tests: cyclictest avoid unneeded warning
2012-05-02 2:45 ` [PATCH 3/4] rt-tests: cyclictest avoid unneeded warning Frank Rowand
@ 2012-05-02 20:38 ` Darren Hart
0 siblings, 0 replies; 8+ messages in thread
From: Darren Hart @ 2012-05-02 20:38 UTC (permalink / raw)
To: frank.rowand
Cc: linux-rt-users@vger.kernel.org, williams@redhat.com,
jkacur@redhat.com
On 05/01/2012 07:45 PM, Frank Rowand wrote:
> Avoid annoying warning message when tracing is not requested and the debug
> file system is not available.
>
> The same test already protects against calling event_enable_all().
>
> Signed-off-by: Frank Rowand <frank.rowand@am.sony.com>
Cross-compile tested for ARM. Visually inspected, looks sane.
Reviewed-by: Darren Hart <dvhart@linux.intel.com>
> ---
> src/cyclictest/cyclictest.c | 5 3 + 2 - 0 !
> 1 file changed, 3 insertions(+), 2 deletions(-)
>
> Index: b/src/cyclictest/cyclictest.c
> ===================================================================
> --- a/src/cyclictest/cyclictest.c
> +++ b/src/cyclictest/cyclictest.c
> @@ -1605,8 +1605,9 @@ int main(int argc, char **argv)
> if (trace_fd >= 0)
> close(trace_fd);
>
> - /* turn off all events */
> - event_disable_all();
> + if (enable_events)
> + /* turn off all events */
> + event_disable_all();
>
> /* turn off the function tracer */
> fileprefix = procfileprefix;
>
--
Darren Hart
Intel Open Source Technology Center
Yocto Project - Linux Kernel
^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH 4/4] rt-tests: cyclictest warn of interaction between '-a', '--smp', and '--numa'
2012-05-02 2:35 [PATCH 1/4] rt-tests: printf format compile warning Frank Rowand
2012-05-02 2:44 ` [PATCH 2/4] rt-tests: cyclictest segfault with '-a' Frank Rowand
2012-05-02 2:45 ` [PATCH 3/4] rt-tests: cyclictest avoid unneeded warning Frank Rowand
@ 2012-05-02 2:45 ` Frank Rowand
2012-05-02 20:41 ` Darren Hart
2012-05-02 20:36 ` [PATCH 1/4] rt-tests: printf format compile warning Darren Hart
3 siblings, 1 reply; 8+ messages in thread
From: Frank Rowand @ 2012-05-02 2:45 UTC (permalink / raw)
To: linux-rt-users@vger.kernel.org, williams@redhat.com
Cc: jkacur@redhat.com, dvhart@linux.intel.com
The '-a' option is always ignored if --smp or --numa is specified. Fix the
warning message to not depend on --smp or --numa occuring first.
Signed-off-by: Frank Rowand <frank.rowand@am.sony.com>
---
src/cyclictest/cyclictest.c | 14 11 + 3 - 0 !
1 file changed, 11 insertions(+), 3 deletions(-)
Index: b/src/cyclictest/cyclictest.c
===================================================================
--- a/src/cyclictest/cyclictest.c
+++ b/src/cyclictest/cyclictest.c
@@ -949,6 +949,7 @@ static char *policyname(int policy)
static void process_options (int argc, char *argv[])
{
int error = 0;
+ int option_affinity = 0;
int max_cpus = sysconf(_SC_NPROCESSORS_CONF);
struct rlimit rlim;
@@ -1000,10 +1001,9 @@ static void process_options (int argc, c
break;
switch (c) {
case 'a':
- if (smp) {
- warn("-a ignored due to --smp\n");
+ option_affinity = 1;
+ if (smp || numa)
break;
- }
if (optarg != NULL) {
affinity = atoi(optarg);
setaffinity = AFFINITY_SPECIFIED;
@@ -1112,6 +1112,14 @@ static void process_options (int argc, c
}
}
+ if (option_affinity) {
+ if (smp) {
+ warn("-a ignored due to --smp\n");
+ } else if (numa) {
+ warn("-a ignored due to --numa\n");
+ }
+ }
+
if (setaffinity == AFFINITY_SPECIFIED) {
if (affinity < 0)
error = 1;
^ permalink raw reply [flat|nested] 8+ messages in thread* Re: [PATCH 4/4] rt-tests: cyclictest warn of interaction between '-a', '--smp', and '--numa'
2012-05-02 2:45 ` [PATCH 4/4] rt-tests: cyclictest warn of interaction between '-a', '--smp', and '--numa' Frank Rowand
@ 2012-05-02 20:41 ` Darren Hart
0 siblings, 0 replies; 8+ messages in thread
From: Darren Hart @ 2012-05-02 20:41 UTC (permalink / raw)
To: frank.rowand
Cc: linux-rt-users@vger.kernel.org, williams@redhat.com,
jkacur@redhat.com
On 05/01/2012 07:45 PM, Frank Rowand wrote:
> The '-a' option is always ignored if --smp or --numa is specified. Fix the
> warning message to not depend on --smp or --numa occuring first.
>
> Signed-off-by: Frank Rowand <frank.rowand@am.sony.com>
Compiled and tested on qemuarm UP. Behaved as expected with --smp --numa
(not built-in) and -a alone.
Tested-by: Darren Hart <dvhart@linux.intel.com>
> ---
> src/cyclictest/cyclictest.c | 14 11 + 3 - 0 !
> 1 file changed, 11 insertions(+), 3 deletions(-)
>
> Index: b/src/cyclictest/cyclictest.c
> ===================================================================
> --- a/src/cyclictest/cyclictest.c
> +++ b/src/cyclictest/cyclictest.c
> @@ -949,6 +949,7 @@ static char *policyname(int policy)
> static void process_options (int argc, char *argv[])
> {
> int error = 0;
> + int option_affinity = 0;
> int max_cpus = sysconf(_SC_NPROCESSORS_CONF);
> struct rlimit rlim;
>
> @@ -1000,10 +1001,9 @@ static void process_options (int argc, c
> break;
> switch (c) {
> case 'a':
> - if (smp) {
> - warn("-a ignored due to --smp\n");
> + option_affinity = 1;
> + if (smp || numa)
> break;
> - }
> if (optarg != NULL) {
> affinity = atoi(optarg);
> setaffinity = AFFINITY_SPECIFIED;
> @@ -1112,6 +1112,14 @@ static void process_options (int argc, c
> }
> }
>
> + if (option_affinity) {
> + if (smp) {
> + warn("-a ignored due to --smp\n");
> + } else if (numa) {
> + warn("-a ignored due to --numa\n");
> + }
> + }
> +
> if (setaffinity == AFFINITY_SPECIFIED) {
> if (affinity < 0)
> error = 1;
>
--
Darren Hart
Intel Open Source Technology Center
Yocto Project - Linux Kernel
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH 1/4] rt-tests: printf format compile warning
2012-05-02 2:35 [PATCH 1/4] rt-tests: printf format compile warning Frank Rowand
` (2 preceding siblings ...)
2012-05-02 2:45 ` [PATCH 4/4] rt-tests: cyclictest warn of interaction between '-a', '--smp', and '--numa' Frank Rowand
@ 2012-05-02 20:36 ` Darren Hart
3 siblings, 0 replies; 8+ messages in thread
From: Darren Hart @ 2012-05-02 20:36 UTC (permalink / raw)
To: frank.rowand; +Cc: linux-rt-users@vger.kernel.org, williams, jkacur
On 05/01/2012 07:35 PM, Frank Rowand wrote:
>
> Fix printf format string to allow for 32 bit and 64 bit word size.
>
> Signed-off-by: Frank Rowand <frank.rowand@am.sony.com>
Tested cross-compilation for ARM, no warnings for this section.
Tested-by: Darren Hart <dvhart@linux.intel.com>
> ---
> src/cyclictest/cyclictest.c | 3 2 + 1 - 0 !
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> Index: b/src/cyclictest/cyclictest.c
> ===================================================================
> --- a/src/cyclictest/cyclictest.c
> +++ b/src/cyclictest/cyclictest.c
> @@ -24,6 +24,7 @@
> #include <time.h>
> #include <errno.h>
> #include <limits.h>
> +#include <inttypes.h>
> #include <linux/unistd.h>
>
> #include <sys/prctl.h>
> @@ -1575,7 +1576,7 @@ int main(int argc, char **argv)
> print_tids(parameters, num_threads);
> if (break_thread_id) {
> printf("# Break thread: %d\n", break_thread_id);
> - printf("# Break value: %lu\n", break_thread_value);
> + printf("# Break value: %" PRIu64 "\n", break_thread_value);
> }
> }
>
>
--
Darren Hart
Intel Open Source Technology Center
Yocto Project - Linux Kernel
^ permalink raw reply [flat|nested] 8+ messages in thread