Re: [PATCH v4 14/20] kernel-shark: Provide merging of multiple data streams

["Yordan Karadzhov (VMware)" <y.karadz@gmail.com>]

On 19.11.20 г. 4:00 ч., Steven Rostedt wrote:
> On Wed, 18 Nov 2020 16:49:57 +0200
> "Yordan Karadzhov (VMware)" <y.karadz@gmail.com> wrote:
> 
>> The C API provides loading of the trace data in two different forms.
>> The firs one is an array of kshark_entries and is being used by the
> 
>    "first one"
> 
>> KernelShark GUI. The second is a matrix-like structure that has all
>> the fields of the kshark_entry stored in separate arrays, forming the
>> columns of the matrix. The second form of the data is used by
>> trace-cruncher. In this patch we add methods for merging of several
>> data streams into a single data set. Both kshark_entries and matrix
>> forms of the data are supported. This patch includes a simple example
>> that demonstrate how to open a file that contains multiple buffers.
>> Each buffers is loaded into a separate Data stream and those streams
>> are merged together.
>>
>> Signed-off-by: Yordan Karadzhov (VMware) <y.karadz@gmail.com>
>> ---
>>   examples/CMakeLists.txt    |   4 +
>>   examples/multibufferload.c |  60 +++++++++
>>   src/libkshark.c            | 255 +++++++++++++++++++++++++++++++++++++
>>   src/libkshark.h            |  47 +++++++
>>   4 files changed, 366 insertions(+)
>>   create mode 100644 examples/multibufferload.c
>>
>> diff --git a/examples/CMakeLists.txt b/examples/CMakeLists.txt
>> index 8d40e42c..831eee24 100644
>> --- a/examples/CMakeLists.txt
>> +++ b/examples/CMakeLists.txt
>> @@ -8,6 +8,10 @@ message(STATUS "datafilter")
>>   add_executable(dfilter          datafilter.c)
>>   target_link_libraries(dfilter   kshark)
>>   
>> +message(STATUS "multibufferload")
>> +add_executable(mbload          multibufferload.c)
>> +target_link_libraries(mbload   kshark)
>> +
>>   # message(STATUS "datahisto")
>>   # add_executable(dhisto          datahisto.c)
>>   # target_link_libraries(dhisto   kshark)
>> diff --git a/examples/multibufferload.c b/examples/multibufferload.c
>> new file mode 100644
>> index 00000000..70b2733a
>> --- /dev/null
>> +++ b/examples/multibufferload.c
>> @@ -0,0 +1,60 @@
>> +#include <stdio.h>
>> +#include <stdlib.h>
>> +
>> +#include "libkshark.h"
>> +#include "libkshark-tepdata.h"
>> +
>> +const char *default_file = "trace.dat";
>> +
>> +void put_entry(struct kshark_entry *e)
>> +{
>> +	char *entry_str = kshark_dump_entry(e);
>> +	puts(entry_str);
>> +	free(entry_str);
>> +}
>> +
>> +int main(int argc, char **argv)
>> +{
>> +	struct kshark_context *kshark_ctx;
>> +	struct kshark_entry **data = NULL;
>> +	ssize_t r, n_rows;
>> +	int sd;
>> +
>> +	/* Create a new kshark session. */
>> +	kshark_ctx = NULL;
>> +	if (!kshark_instance(&kshark_ctx))
>> +		return 1;
>> +
>> +	/* Open a trace data file produced by trace-cmd. */
>> +	if (argc > 1)
>> +		sd = kshark_open(kshark_ctx, argv[1]);
>> +	else
>> +		sd = kshark_open(kshark_ctx, default_file);
> 
> I'm confused. It doesn't look like this merges more than one stream.


Hi Steven,

Here it only opens the "top" buffer in the file.

> 
> -- Steve
> 
>> +
>> +	if (sd < 0) {
>> +		kshark_free(kshark_ctx);
>> +		return 1;
>> +	}
>> +
>> +	/* Initialize data streams for all buffers in this file. */
>> +	kshark_tep_init_all_buffers(kshark_ctx, sd);

And here is the place where it opens in separate streams the other 
buffers that are in the same file.


>> +
>> +	/* Load all buffers. */
>> +	n_rows = kshark_load_all_entries(kshark_ctx, &data);
>> +

Here it loads and merges the entries from all buffers.

Thanks!
Yordan

>> +	/* Print to the screen the first 20 entries. */
>> +	for (r = 0; r < 20; ++r)
>> +		put_entry(data[r]);
>> +
>> +	/* Free the memory. */
>> +	for (r = 0; r < n_rows; ++r)
>> +		free(data[r]);
>> +	free(data);
>> +
>> +	kshark_close_all(kshark_ctx);
>> +
>> +	/* Close the session. */
>> +	kshark_free(kshark_ctx);
>> +
>> +	return 0;
>> +}