All of lore.kernel.org
 help / color / mirror / Atom feed
From: Tzvetomir Stoyanov <tz.stoyanov@gmail.com>
To: Sameeruddin Shaik <sameeruddin.shaik8@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>,
	Linux Trace Devel <linux-trace-devel@vger.kernel.org>
Subject: Re: [PATCH] libtracefs: An API to set the filtering of functions
Date: Wed, 10 Mar 2021 07:28:19 +0200	[thread overview]
Message-ID: <CAPpZLN7xgt2jsUycoNksdXOtOXiCZP5ZTP3F5_KUw30qapo=rw@mail.gmail.com> (raw)
In-Reply-To: <CAK7tX=aU41LzhjmDZHWUL2qLQ0M8hRJKfg7QcY9FYDpTHfUq3g@mail.gmail.com>

On Tue, Mar 9, 2021 at 6:51 PM Sameeruddin Shaik
<sameeruddin.shaik8@gmail.com> wrote:
>
> This is my test code, I tried handling the triple pointers in the library,
> but when it comes to user program, at first "errs" has NULL at the
> end(checked it using the gdb), but when we try to dereference and
> print, gradually position holding NULL in "errs" is getting assinged
> to some other location, and i am getting the segmentation fault.
>
> >Since a triple pointer is difficult to manage in the code, you could have:
> >
> >        const char **e = NULL;
> >
> >
> >               if (errs) {
> >                        e = realloc(sizeof(*e), j + 1);
> >                        e[j++] = filters[i];
> >                }
> >
> >Then at the end:
> >
> >        if (errs)
> >                *errs = e;
>
> i didn't got the clear picture of what, the above code snippet is doing.
> particularly the below line
>
> >e = realloc(sizeof(*e), j + 1);
 It should be just
   e = realloc(e, (j + 1) * sizeof(char *));
we need to store one more pointer to char,
 On the first call, e is NULL and j is 0 and this behaves
 as a regular malloc:
   e = realloc(NULL, 1 * sizeof(char *));
on each subsequent call it allocates memory for
one additional pointer to char, at the end of the existing
array e.
Note that the allocated memory is not set to 0, it contains
random data.

>
> From man page i got
>
> void *realloc(void *ptr, size_t size);
>
> could you please brief me with an example?
>
> Subject: [PATCH] test: Test code
> diff --git a/test.c b/test.c
> index d38fc92..3e90711 100644
> --- a/test.c
> +++ b/test.c
> @@ -1,7 +1,39 @@
>  #include <tracefs.h>
> +#include <stdio.h>
> +#include <fcntl.h>
> +#include <string.h>
> +#include <unistd.h>
> +#include <stdlib.h>
> +
> +
> +
> +#define INST "dummy"
> +
> +const char *filters[] = {"kvm_pmu_reset", "kvm_pmu_init",
> "dir_item_err", "tzvetomir", "sameer", "steve", NULL};
>
>  int main()
>  {
> -    tracefs_tracing_dir();
> +    struct tracefs_instance *instance = NULL;
> +    const char ***errs;

For the user of the API, errs is just an array of strings,
regular double pointer. It is not allocated by the caller,
so it is important to be initialised with NULL;

 const char **errs = NULL;

> +    int ret;
> +    int fd;
> +    int n;
> +    int i = 0;
> +
> +    errs = malloc(sizeof(char *));

No need to allocate it here, as the size is not known -
it depends on the number of failed filters. The API
will do the job.

> +    printf("%s\n",tracefs_tracing_dir());
> +    instance = tracefs_instance_create(INST);
> +    ret = tracefs_function_filter(instance, filters, "kvm", 1, errs);

As the errs will be allocated by the API, you should pass the address
of it:
   ret = tracefs_function_filter(instance, filters, "kvm", 1, &errs);
That makes the pointer triple, and it is important to be initialised
with NULL.

If the caller is not interested of the errors, it could pass NULL:
   ret = tracefs_function_filter(instance, filters, "kvm", 1, NULL);

> +
> +    if(ret < -1)
> +        if (errs)
> +            while (errs[i])
> +                printf("%s\n", *errs[i++]);
Yes, that is exactly how it should be used, except
that additional "*":
if (ret < 0 && errs) {
    while (errs[i])
          printf("%s\n", errs[i++]);
}

> +
> +    getchar();
> +    tracefs_instance_destroy(instance);
> +    tracefs_instance_free(instance);
> +    free(errs);
>      return 0;
>  }
>
> On Tue, Mar 9, 2021 at 9:51 PM Sameeruddin shaik
> <sameeruddin.shaik8@gmail.com> wrote:
> >
> > This new API will write the function filters into the
> > set_ftrace_filter file.
> >
> > tracefs_function_filter()
> >
> > https://bugzilla.kernel.org/show_bug.cgi?id=210643
> >
> > Signed-off-by: Sameeruddin shaik <sameeruddin.shaik8@gmail.com>
> >
> > diff --git a/include/tracefs.h b/include/tracefs.h
> > index f3eec62..a2249d0 100644
> > --- a/include/tracefs.h
> > +++ b/include/tracefs.h
> > @@ -50,6 +50,7 @@ int tracefs_trace_on(struct tracefs_instance *instance);
> >  int tracefs_trace_off(struct tracefs_instance *instance);
> >  int tracefs_trace_on_fd(int fd);
> >  int tracefs_trace_off_fd(int fd);
> > +int tracefs_function_filter(struct tracefs_instance *instance, const char **filters, const char *module, bool reset, const char ***errs);
> >
> >  /**
> >   * tracefs_trace_on_get_fd - Get a file descriptor of "tracing_on" in given instance
> > diff --git a/src/tracefs-tools.c b/src/tracefs-tools.c
> > index e2dfc7b..4e168df 100644
> > --- a/src/tracefs-tools.c
> > +++ b/src/tracefs-tools.c
> > @@ -18,6 +18,7 @@
> >  #include "tracefs-local.h"
> >
> >  #define TRACE_CTRL     "tracing_on"
> > +#define TRACE_FILTER      "set_ftrace_filter"
> >
> >  static const char * const options_map[] = {
> >         "unknown",
> > @@ -387,3 +388,96 @@ void tracefs_option_clear(struct tracefs_options_mask *options, enum tracefs_opt
> >         if (options && id > TRACEFS_OPTION_INVALID)
> >                 options->mask &= ~(1ULL << (id - 1));
> >  }
> > +
> > +static int controlled_write(const char *filter_path, const char **filters, const char *module, bool reset, const char ***errs)
> > +{
> > +       int flags = reset ? O_TRUNC : O_APPEND;
> > +       int write_size;
> > +       char *each_str;
> > +       int ret = 0;
> > +       int j = 0;
> > +       int size;
> > +       int slen;
> > +       int fd;
> > +       int i;
> > +
> > +       fd = open(filter_path, O_WRONLY | flags);
> > +       if (fd < 0)
> > +               return 1;
> > +
> > +       for (i = 0; filters[i]; i++) {
> > +               slen = strlen(filters[i]);
> > +               if (!slen)
> > +                       continue;
> > +
> > +               if (module)
> > +                       /* Adding 5 extra bytes for ":mod:"*/
> > +                       slen += strlen(module) + 5;
> > +
> > +               /* Adding 2 extra byte for the space and '\0' at the end*/
> > +               slen += 2;
> > +               each_str = calloc(1, slen);
> > +               if (!each_str)
> > +                       return 1;
> > +               if (module)
> > +                       write_size = snprintf(each_str, slen, "%s:mod:%s ", filters[i], module);
> > +               else
> > +                       write_size = snprintf(each_str, slen, "%s ", filters[i]);
> > +
> > +               size = write(fd, each_str, write_size);
> > +               /* compare written bytes*/
> > +               if (size < write_size) {
> > +                       if (errs) {
> > +                               errs[j++] = &filters[i];
> > +                               ret -= 1;
> > +                       }
> > +               }
> > +               free(each_str);
> > +       }
> > +       errs[j] = NULL;
> > +       close(fd);
> > +       return ret;
> > +}
> > +
> > +/**
> > + * tracefs_function_filter - write to set_ftrace_filter file to trace particular functions
> > + * @instance: ftrace instance, can be NULL for top tracing instance
> > + * @filters: An array of function names ending with a NULL pointer
> > + * @module: Module to be traced
> > + * @reset: set to true to reset the file before applying the filter
> > + * @errs: A pointer to array of constant strings that will be allocated
> > + * on negative return of this function, pointing to the filters that failed.
> > + * May be NULL, in which case this field will be ignored.
> > + *
> > + * The @filters is an array of strings, where each string will be used to set
> > + * a function or functions to be traced.
> > + *
> > + * If @reset is true, then all functions in the filter are cleared before
> > + * adding functions from @filters. Otherwise, the functions set by @filters
> > + * will be appended to the filter file
> > + *
> > + * returns -x on filter errors (where x is number of failed filter srtings)
> > + * and if @errs is not NULL will be an allocated string array pointing to the
> > + * strings in @filters that failed and must be freed with free().
> > + *
> > + * returns 1 on general errors not realted to setting the filter.
> > + * @errs is not set even if supplied.
> > + *
> > + * return 0 on success and @errs is not set.
> > + */
> > +int tracefs_function_filter(struct tracefs_instance *instance, const char **filters, const char *module, bool reset, const char ***errs)
> > +{
> > +       char *ftrace_filter_path;
> > +       int ret = 0;
> > +
> > +       if (!filters)
> > +               return 1;
> > +
> > +       ftrace_filter_path = tracefs_instance_get_file(instance, TRACE_FILTER);
> > +       if (!ftrace_filter_path)
> > +               return 1;
> > +
> > +       ret = controlled_write(ftrace_filter_path, filters, module, reset, errs);
> > +       tracefs_put_tracing_file(ftrace_filter_path);
> > +       return ret;
> > +}
> > --
> > 2.7.4
> >



-- 
Tzvetomir (Ceco) Stoyanov
VMware Open Source Technology Center

  reply	other threads:[~2021-03-10  5:29 UTC|newest]

Thread overview: 23+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-03-10 16:21 [PATCH] libtracefs: An API to set the filtering of functions Sameeruddin shaik
2021-03-10  5:28 ` Tzvetomir Stoyanov
2021-03-10 16:51 ` Sameeruddin Shaik
2021-03-10  5:28   ` Tzvetomir Stoyanov [this message]
  -- strict thread matches above, loose matches on Subject: below --
2021-03-06 11:20 Sameeruddin shaik
2021-03-05 12:20 ` Tzvetomir Stoyanov
2021-03-05 14:39   ` Steven Rostedt
2021-03-05 14:54     ` Steven Rostedt
2021-03-06  1:55       ` Sameeruddin Shaik
2021-03-06  3:39         ` Steven Rostedt
2021-03-06  4:29           ` Sameeruddin Shaik
2021-03-06  5:19             ` Steven Rostedt
2021-03-06 15:05         ` Steven Rostedt
2021-03-08 23:53           ` Sameeruddin Shaik
2021-03-02 17:15 Sameeruddin shaik
2021-03-01 18:17 ` Steven Rostedt
2021-03-02  4:21   ` Tzvetomir Stoyanov
2021-03-02  5:14     ` Sameeruddin Shaik
2021-03-02 13:15       ` Steven Rostedt
2021-03-03  1:16   ` Sameeruddin Shaik
2021-03-02  1:28     ` Steven Rostedt
2021-03-04  8:59       ` Tzvetomir Stoyanov
2021-03-04  9:43         ` Sameeruddin Shaik

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to='CAPpZLN7xgt2jsUycoNksdXOtOXiCZP5ZTP3F5_KUw30qapo=rw@mail.gmail.com' \
    --to=tz.stoyanov@gmail.com \
    --cc=linux-trace-devel@vger.kernel.org \
    --cc=rostedt@goodmis.org \
    --cc=sameeruddin.shaik8@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.