[PATCH v1] Extend -e flag to provide the user possibility of listing all system calls in a class.

[Dmitry V. Levin]

On Thu, Jul 07, 2016 at 04:17:16PM +0200, aminkhorsandi at gmail.com wrote:
> From: Amin Khorsandi <amin.khorsandi at infinera.com>
> 
> ---
>  strace.1  |  5 +++++
>  syscall.c | 26 +++++++++++++++++++++++++-
>  2 files changed, 30 insertions(+), 1 deletion(-)
> 
> diff --git a/strace.1 b/strace.1
> index b7dddc0..c2088fa 100644
> --- a/strace.1
> +++ b/strace.1
> @@ -403,6 +403,11 @@ Furthermore, using the abbreviation will ensure that you don't
>  accidentally forget to include a call like
>  .B lstat
>  in the list.  Betchya woulda forgot that one.
> +To see a list of all system calls in a class you can use
> +.BR "-e\ trace?" = classname .
> +For example you can see what are the system calls in the network
> +class running strace using the option
> +.BR "-e\ trace?" = network .

This extension of "-e trace=" syntax might appear to be confusing,
especially if it isn't consistent with traditional "-e trace=" syntax,
see below.

>  .TP
>  .BR "\-e\ trace" = process
>  Trace all system calls which involve process management.  This
> diff --git a/syscall.c b/syscall.c
> index c61f827..fb20f65 100644
> --- a/syscall.c
> +++ b/syscall.c
> @@ -503,6 +503,25 @@ lookup_class(const char *s)
>  	return -1;
>  }
>  
> +static void
> +print_trace_class(const char* trace_class)
> +{
> +	unsigned int i;
> +	unsigned pers = 0; /* only print system calls in personality 0 */

Why?  Traditional "-e trace=" syntax is not limited to personality 0.

> +	int class = lookup_class(trace_class);

Traditional "-e trace=" syntax accepts lists.

> +
> +	if (class != -1) {
> +		printf("List of system calls in %s class\n", trace_class);
> +		printf("%s\n", "-------------------------------------");

Why printf instead of traditional fprintf(shared_log)?

> +		for (i = 0; i < nsyscall_vec[pers]; i++)
> +			if (sysent_vec[pers][i].sys_flags & class) {
> +			printf("\t%s \n", sysent_vec[pers][i].sys_name);
> +		}

This indentation is misleading.

> +	} else {
> +		error_msg_and_die("invalid class name '%s'", trace_class);
> +	}
> +}
> +
>  void
>  qualify(const char *s)
>  {
> @@ -518,7 +537,12 @@ qualify(const char *s)
>  	opt = &qual_options[0];
>  	for (i = 0; (p = qual_options[i].option_name); i++) {
>  		unsigned int len = strlen(p);
> -		if (strncmp(s, p, len) == 0 && s[len] == '=') {
> +		if (strncmp(s, p, len) == 0 && s[len] == '?' && s[len + 1] == '=') {
> +			s += len + 2;
> +			print_trace_class(s);
> +			exit(0);

Traditional "-e trace=" syntax doesn't stop after the first option.

> +		}
> +		else if (strncmp(s, p, len) == 0 && s[len] == '=') {
>  			opt = &qual_options[i];
>  			s += len + 1;
>  			break;

In short, either this new extension should match traditional "-e trace="
syntax, or it shouldn't be called "-e trace?=".


-- 
ldv
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://lists.strace.io/pipermail/strace-devel/attachments/20160711/a057bb5d/attachment.bin>