[Hurd] Fix deallocation after proc_getprocinfo call

  2014-10-02  Samuel Thibault  <samuel.thibault@ens-lyon.org>

	* gdb/gnu-nat.c (inf_validate_procinfo): Multiply the number of
	elements pi_len by the size of the elements before calling
	vm_deallocate.
        (inf_validate_task_sc): Likewise, and properly deallocate the
        noise array.
  

Hello Samuel,

On Sun, Nov 02, 2014 at 04:25:37PM +0100, Samuel Thibault wrote:
> 2014-10-02  Samuel Thibault  <samuel.thibault@ens-lyon.org>
> 
> 	* gdb/gnu-nat.c (inf_validate_procinfo): Multiply the number of
> 	elements pi_len by the size of the elements before calling
> 	vm_deallocate.
>         (inf_validate_task_sc): Likewise, and properly deallocate the
>         noise array.

Again, sorry about the late review...

I only have a few minor comments, almost trivial in nature. In
the ChangeLog entry above, watch out that the last 2 lines are
indented using spaces intead of tabs.

> diff --git a/gdb/gnu-nat.c b/gdb/gnu-nat.c
> index d17a750..c571190 100644
> --- a/gdb/gnu-nat.c
> +++ b/gdb/gnu-nat.c
> @@ -804,7 +804,7 @@ inf_validate_procinfo (struct inf *inf)
>        inf->nomsg = !!(pi->state & PI_NOMSG);
>        if (inf->nomsg)
>  	inf->traced = !!(pi->state & PI_TRACED);
> -      vm_deallocate (mach_task_self (), (vm_address_t) pi, pi_len);
> +      vm_deallocate (mach_task_self (), (vm_address_t) pi, pi_len * sizeof (*(procinfo_t) 0));

The line is too long (soft limit is 74 characters, hard limit is 80).
Suggest using "sizeof (struct procinfo)", which I think is better
than dereferencing a NULL pointer. This is based on guessing that
type procinfo_t is a pointer to struct procinfo, as suggested by
the code in inf_validate_procinfo.

>        if (noise_len > 0)
>  	vm_deallocate (mach_task_self (), (vm_address_t) noise, noise_len);
>      }
> @@ -844,9 +844,9 @@ inf_validate_task_sc (struct inf *inf)
>  
>    suspend_count = pi->taskinfo.suspend_count;
>  
> -  vm_deallocate (mach_task_self (), (vm_address_t) pi, pi_len);
> +  vm_deallocate (mach_task_self (), (vm_address_t) pi, pi_len * sizeof (*(procinfo_t) 0));

Same as above.

>    if (noise_len > 0)
> -    vm_deallocate (mach_task_self (), (vm_address_t) pi, pi_len);
> +    vm_deallocate (mach_task_self (), (vm_address_t) noise, noise_len);
>  
>    if (inf->task->cur_sc < suspend_count)
>      {

Thank you,
  

Joel Brobecker, le Sun 23 Nov 2014 11:44:52 +0400, a écrit :
> > -      vm_deallocate (mach_task_self (), (vm_address_t) pi, pi_len);
> > +      vm_deallocate (mach_task_self (), (vm_address_t) pi, pi_len * sizeof (*(procinfo_t) 0));
> 
> Suggest using "sizeof (struct procinfo)", which I think is better
> than dereferencing a NULL pointer. This is based on guessing that
> type procinfo_t is a pointer to struct procinfo, as suggested by
> the code in inf_validate_procinfo.

Not, that is not the same: struct procinfo has an open array at its end
(threadinfos[0]), and thus the actually allocated size is variable.

Samuel
  

> > > -      vm_deallocate (mach_task_self (), (vm_address_t) pi, pi_len);
> > > +      vm_deallocate (mach_task_self (), (vm_address_t) pi, pi_len * sizeof (*(procinfo_t) 0));
> > 
> > Suggest using "sizeof (struct procinfo)", which I think is better
> > than dereferencing a NULL pointer. This is based on guessing that
> > type procinfo_t is a pointer to struct procinfo, as suggested by
> > the code in inf_validate_procinfo.
> 
> Not, that is not the same: struct procinfo has an open array at its end
> (threadinfos[0]), and thus the actually allocated size is variable.

OK. I don't know the code well enough to make any better suggestion.
The above does look unusual to me, but if it works and seems to be
the only correct way, let's go with that.

Should I push your patch?
  

Joel Brobecker, le Sun 23 Nov 2014 13:56:29 +0400, a écrit :
> The above does look unusual to me, but if it works and seems to be
> the only correct way, let's go with that.

It is a system interface actually, so we don't really have the choice :)

Samuel