From: Naresh Kamboju naresh.kamboju@linaro.org
I have noticed kernel crash while reading trace file. 'cat /sys/kernel/debug/tracing/trace'
The reason of the kernel crash is few bits in trace.c is missing from Merge tag 'trace-3.10' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace in to linux-linaro. Those are like "struct trace_array *tr " replaced with "struct trace_buffer *buf" and respective changes.
I have applied this patch and tested on linux-linaro branch no more kernel crash reported and trace data is back again.
Signed-off-by: Naresh Kamboju naresh.kamboju@linaro.org --- kernel/trace/trace.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-)
diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c index e4cc806..a68084e 100644 --- a/kernel/trace/trace.c +++ b/kernel/trace/trace.c @@ -2361,9 +2361,9 @@ static void print_func_help_header(struct trace_buffer *buf, struct seq_file *m) seq_puts(m, "# | | | | |\n"); }
-static void print_func_help_header_tgid(struct trace_array *tr, struct seq_file *m) +static void print_func_help_header_tgid(struct trace_buffer *buf, struct seq_file *m) { - print_event_info(tr, m); + print_event_info(buf, m); seq_puts(m, "# TASK-PID TGID CPU# TIMESTAMP FUNCTION\n"); seq_puts(m, "# | | | | | |\n"); } @@ -2380,9 +2380,9 @@ static void print_func_help_header_irq(struct trace_buffer *buf, struct seq_file seq_puts(m, "# | | | |||| | |\n"); }
-static void print_func_help_header_irq_tgid(struct trace_array *tr, struct seq_file *m) +static void print_func_help_header_irq_tgid(struct trace_buffer *buf, struct seq_file *m) { - print_event_info(tr, m); + print_event_info(buf, m); seq_puts(m, "# _-----=> irqs-off\n"); seq_puts(m, "# / _----=> need-resched\n"); seq_puts(m, "# | / _---=> hardirq/softirq\n"); @@ -2693,14 +2693,14 @@ void trace_default_header(struct seq_file *m) if (!(trace_flags & TRACE_ITER_VERBOSE)) { if (trace_flags & TRACE_ITER_IRQ_INFO) if (trace_flags & TRACE_ITER_TGID) - print_func_help_header_irq_tgid(iter->tr, m); + print_func_help_header_irq_tgid(iter->trace_buffer, m); else - print_func_help_header_irq(iter->tr, m); + print_func_help_header_irq(iter->trace_buffer, m); else if (trace_flags & TRACE_ITER_TGID) - print_func_help_header_tgid(iter->tr, m); + print_func_help_header_tgid(iter->trace_buffer, m); else - print_func_help_header(iter->tr, m); + print_func_help_header(iter->trace_buffer, m); } } }
On Fri, 2013-05-24 at 02:03 +0530, naresh.kamboju@linaro.org wrote:
From: Naresh Kamboju naresh.kamboju@linaro.org
I have noticed kernel crash while reading trace file. 'cat /sys/kernel/debug/tracing/trace'
The reason of the kernel crash is few bits in trace.c is missing from Merge tag 'trace-3.10' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace in to linux-linaro. Those are like "struct trace_array *tr " replaced with "struct trace_buffer *buf" and respective changes.
I have applied this patch and tested on linux-linaro branch no more kernel crash reported and trace data is back again.
Signed-off-by: Naresh Kamboju naresh.kamboju@linaro.org
Reviewed-by: Jon Medhurst tixy@linaro.org
Good catch Naresh. I've cc'd John Stultz as it looks like these problems were introduced with the merge of mainline Linux into the linaro-android-3.10-experimental and this patch should be added to that.
Considering the amount of churn in the trace code in 3.10, I wouldn't be 100% confident that we've found all the conflicts with the Android trace patches, but this patch does fix the crashes we were seeing.
On 05/24/2013 12:57 PM, Jon Medhurst (Tixy) wrote:
On Fri, 2013-05-24 at 02:03 +0530, naresh.kamboju@linaro.org wrote:
From: Naresh Kamboju naresh.kamboju@linaro.org
I have noticed kernel crash while reading trace file. 'cat /sys/kernel/debug/tracing/trace'
The reason of the kernel crash is few bits in trace.c is missing from Merge tag 'trace-3.10' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace in to linux-linaro. Those are like "struct trace_array *tr " replaced with "struct trace_buffer *buf" and respective changes.
I have applied this patch and tested on linux-linaro branch no more kernel crash reported and trace data is back again.
Signed-off-by: Naresh Kamboju naresh.kamboju@linaro.org
Reviewed-by: Jon Medhurst tixy@linaro.org
Good catch Naresh. I've cc'd John Stultz as it looks like these problems were introduced with the merge of mainline Linux into the linaro-android-3.10-experimental and this patch should be added to that.
Considering the amount of churn in the trace code in 3.10, I wouldn't be 100% confident that we've found all the conflicts with the Android trace patches, but this patch does fix the crashes we were seeing.
Good. I'll do one more linux-linaro rebuild today to get this fix in then?
Thanks, Andrey
On Fri, 2013-05-24 at 16:28 +0400, Andrey Konovalov wrote:
On 05/24/2013 12:57 PM, Jon Medhurst (Tixy) wrote:
On Fri, 2013-05-24 at 02:03 +0530, naresh.kamboju@linaro.org wrote:
From: Naresh Kamboju naresh.kamboju@linaro.org
I have noticed kernel crash while reading trace file. 'cat /sys/kernel/debug/tracing/trace'
The reason of the kernel crash is few bits in trace.c is missing from Merge tag 'trace-3.10' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace in to linux-linaro. Those are like "struct trace_array *tr " replaced with "struct trace_buffer *buf" and respective changes.
I have applied this patch and tested on linux-linaro branch no more kernel crash reported and trace data is back again.
Signed-off-by: Naresh Kamboju naresh.kamboju@linaro.org
Reviewed-by: Jon Medhurst tixy@linaro.org
Good catch Naresh. I've cc'd John Stultz as it looks like these problems were introduced with the merge of mainline Linux into the linaro-android-3.10-experimental and this patch should be added to that.
Considering the amount of churn in the trace code in 3.10, I wouldn't be 100% confident that we've found all the conflicts with the Android trace patches, but this patch does fix the crashes we were seeing.
Good. I'll do one more linux-linaro rebuild today to get this fix in then?
Well, that is the proper fix for a mis-merge of the Android trace patches which my patch yesterday resolved by simply reverting them (I was trying to get a fix in before the code freeze and didn't get to the bottom of the issue in time.)
It's my opinion that the safest thing for the release would be to stick with the revert patch. Those trace Android patches were new in 3.9 and I'm not aware of anyone actually requiring them (I could be wrong), and as mainline had major trace changes in 3.10 I'm not confident there wouldn't be other bugs lurking. But I wouldn't strongly object either if this new patch was added.
Monday is a public holiday in the UK so I won't be around to boot test the new release candidate candidate that would result from updating the kernel today. And we would have to make sure release candidate build and testing was postponed...
On 05/24/2013 04:44 PM, Jon Medhurst (Tixy) wrote:
On Fri, 2013-05-24 at 16:28 +0400, Andrey Konovalov wrote:
On 05/24/2013 12:57 PM, Jon Medhurst (Tixy) wrote:
On Fri, 2013-05-24 at 02:03 +0530, naresh.kamboju@linaro.org wrote:
From: Naresh Kamboju naresh.kamboju@linaro.org
I have noticed kernel crash while reading trace file. 'cat /sys/kernel/debug/tracing/trace'
The reason of the kernel crash is few bits in trace.c is missing from Merge tag 'trace-3.10' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace in to linux-linaro. Those are like "struct trace_array *tr " replaced with "struct trace_buffer *buf" and respective changes.
I have applied this patch and tested on linux-linaro branch no more kernel crash reported and trace data is back again.
Signed-off-by: Naresh Kamboju naresh.kamboju@linaro.org
Reviewed-by: Jon Medhurst tixy@linaro.org
Good catch Naresh. I've cc'd John Stultz as it looks like these problems were introduced with the merge of mainline Linux into the linaro-android-3.10-experimental and this patch should be added to that.
Considering the amount of churn in the trace code in 3.10, I wouldn't be 100% confident that we've found all the conflicts with the Android trace patches, but this patch does fix the crashes we were seeing.
Good. I'll do one more linux-linaro rebuild today to get this fix in then?
Well, that is the proper fix for a mis-merge of the Android trace patches which my patch yesterday resolved by simply reverting them (I was trying to get a fix in before the code freeze and didn't get to the bottom of the issue in time.)
It's my opinion that the safest thing for the release would be to stick with the revert patch. Those trace Android patches were new in 3.9 and I'm not aware of anyone actually requiring them (I could be wrong), and as mainline had major trace changes in 3.10 I'm not confident there wouldn't be other bugs lurking. But I wouldn't strongly object either if this new patch was added.
Monday is a public holiday in the UK so I won't be around to boot test the new release candidate candidate that would result from updating the kernel today. And we would have to make sure release candidate build and testing was postponed...
OK. Got it. Thanks, Tixy! So ll-20130524.0 is the official release candidate.
Thanks, Andrey
On 05/24/2013 05:44 AM, Jon Medhurst (Tixy) wrote:
On Fri, 2013-05-24 at 16:28 +0400, Andrey Konovalov wrote:
On 05/24/2013 12:57 PM, Jon Medhurst (Tixy) wrote:
On Fri, 2013-05-24 at 02:03 +0530, naresh.kamboju@linaro.org wrote:
From: Naresh Kamboju naresh.kamboju@linaro.org
I have noticed kernel crash while reading trace file. 'cat /sys/kernel/debug/tracing/trace'
The reason of the kernel crash is few bits in trace.c is missing from Merge tag 'trace-3.10' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace in to linux-linaro. Those are like "struct trace_array *tr " replaced with "struct trace_buffer *buf" and respective changes.
I have applied this patch and tested on linux-linaro branch no more kernel crash reported and trace data is back again.
Signed-off-by: Naresh Kamboju naresh.kamboju@linaro.org
Reviewed-by: Jon Medhurst tixy@linaro.org
Good catch Naresh. I've cc'd John Stultz as it looks like these problems were introduced with the merge of mainline Linux into the linaro-android-3.10-experimental and this patch should be added to that.
Considering the amount of churn in the trace code in 3.10, I wouldn't be 100% confident that we've found all the conflicts with the Android trace patches, but this patch does fix the crashes we were seeing.
Good. I'll do one more linux-linaro rebuild today to get this fix in then?
Well, that is the proper fix for a mis-merge of the Android trace patches which my patch yesterday resolved by simply reverting them (I was trying to get a fix in before the code freeze and didn't get to the bottom of the issue in time.)
It's my opinion that the safest thing for the release would be to stick with the revert patch. Those trace Android patches were new in 3.9 and I'm not aware of anyone actually requiring them (I could be wrong), and as mainline had major trace changes in 3.10 I'm not confident there wouldn't be other bugs lurking. But I wouldn't strongly object either if this new patch was added.
Ack. I think Tixy has the right call here. We'll have more time to test Naresh's solution for 13.06. Lets keep the revert for 13.05.
thanks -john
On Fri, 2013-05-24 at 08:58 -0700, John Stultz wrote:
On 05/24/2013 05:44 AM, Jon Medhurst (Tixy) wrote:
On Fri, 2013-05-24 at 16:28 +0400, Andrey Konovalov wrote:
On 05/24/2013 12:57 PM, Jon Medhurst (Tixy) wrote:
On Fri, 2013-05-24 at 02:03 +0530, naresh.kamboju@linaro.org wrote:
From: Naresh Kamboju naresh.kamboju@linaro.org
I have noticed kernel crash while reading trace file. 'cat /sys/kernel/debug/tracing/trace'
The reason of the kernel crash is few bits in trace.c is missing from Merge tag 'trace-3.10' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace in to linux-linaro. Those are like "struct trace_array *tr " replaced with "struct trace_buffer *buf" and respective changes.
I have applied this patch and tested on linux-linaro branch no more kernel crash reported and trace data is back again.
Signed-off-by: Naresh Kamboju naresh.kamboju@linaro.org
Reviewed-by: Jon Medhurst tixy@linaro.org
Good catch Naresh. I've cc'd John Stultz as it looks like these problems were introduced with the merge of mainline Linux into the linaro-android-3.10-experimental and this patch should be added to that.
Considering the amount of churn in the trace code in 3.10, I wouldn't be 100% confident that we've found all the conflicts with the Android trace patches, but this patch does fix the crashes we were seeing.
Good. I'll do one more linux-linaro rebuild today to get this fix in then?
Well, that is the proper fix for a mis-merge of the Android trace patches which my patch yesterday resolved by simply reverting them (I was trying to get a fix in before the code freeze and didn't get to the bottom of the issue in time.)
It's my opinion that the safest thing for the release would be to stick with the revert patch. Those trace Android patches were new in 3.9 and I'm not aware of anyone actually requiring them (I could be wrong), and as mainline had major trace changes in 3.10 I'm not confident there wouldn't be other bugs lurking. But I wouldn't strongly object either if this new patch was added.
Ack. I think Tixy has the right call here. We'll have more time to test Naresh's solution for 13.06.
We did confirm that Naresh's fix not only works for the trivial test case of "cat /sys/kernel/debug/tracing/trace" but also when running the big.LITTLE scheduler tests (which generate and parse trace output) and which originally showed up the problem. So we have no reason to suspect that the patch is not sufficient, I just have a big conservative streak and am somewhat paranoid to boot ;-)
On 05/24/2013 09:26 AM, Jon Medhurst (Tixy) wrote:
On Fri, 2013-05-24 at 08:58 -0700, John Stultz wrote:
On 05/24/2013 05:44 AM, Jon Medhurst (Tixy) wrote:
On Fri, 2013-05-24 at 16:28 +0400, Andrey Konovalov wrote:
On 05/24/2013 12:57 PM, Jon Medhurst (Tixy) wrote:
On Fri, 2013-05-24 at 02:03 +0530, naresh.kamboju@linaro.org wrote:
From: Naresh Kamboju naresh.kamboju@linaro.org
I have noticed kernel crash while reading trace file. 'cat /sys/kernel/debug/tracing/trace'
The reason of the kernel crash is few bits in trace.c is missing from Merge tag 'trace-3.10' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace in to linux-linaro. Those are like "struct trace_array *tr " replaced with "struct trace_buffer *buf" and respective changes.
I have applied this patch and tested on linux-linaro branch no more kernel crash reported and trace data is back again.
Signed-off-by: Naresh Kamboju naresh.kamboju@linaro.org
Reviewed-by: Jon Medhurst tixy@linaro.org
Good catch Naresh. I've cc'd John Stultz as it looks like these problems were introduced with the merge of mainline Linux into the linaro-android-3.10-experimental and this patch should be added to that.
Considering the amount of churn in the trace code in 3.10, I wouldn't be 100% confident that we've found all the conflicts with the Android trace patches, but this patch does fix the crashes we were seeing.
Good. I'll do one more linux-linaro rebuild today to get this fix in then?
Well, that is the proper fix for a mis-merge of the Android trace patches which my patch yesterday resolved by simply reverting them (I was trying to get a fix in before the code freeze and didn't get to the bottom of the issue in time.)
It's my opinion that the safest thing for the release would be to stick with the revert patch. Those trace Android patches were new in 3.9 and I'm not aware of anyone actually requiring them (I could be wrong), and as mainline had major trace changes in 3.10 I'm not confident there wouldn't be other bugs lurking. But I wouldn't strongly object either if this new patch was added.
Ack. I think Tixy has the right call here. We'll have more time to test Naresh's solution for 13.06.
We did confirm that Naresh's fix not only works for the trivial test case of "cat /sys/kernel/debug/tracing/trace" but also when running the big.LITTLE scheduler tests (which generate and parse trace output) and which originally showed up the problem. So we have no reason to suspect that the patch is not sufficient, I just have a big conservative streak and am somewhat paranoid to boot ;-)
I think its all quite reasonable caution. Especially going into a holiday weekend.
thanks -john
linaro-kernel@lists.linaro.org
-
Andrey Konovalov -
John Stultz -
Jon Medhurst (Tixy) -
naresh.kamboju＠linaro.org