Re: [PATCH v2] Disable get_ptrace_pid for NetBSD

[Simon Marchi <simark@simark.ca>]

On 2020-03-18 5:55 p.m., Kamil Rytarowski wrote:
> @@ -313,8 +325,12 @@ inf_ptrace_target::kill ()
>    target_mourn_inferior (inferior_ptid);
>  }
> 
> +#ifndef __NetBSD__
>  /* Return which PID to pass to ptrace in order to observe/control the
> -   tracee identified by PTID.  */
> +   tracee identified by PTID.
> +
> +   Unlike most other Operating Systems, NetBSD tracks both pid and lwp
> +   and avoids this function.  */

To align this function with our current way of documenting functions, please move
this comment to the declaration of get_ptrace_pid, in inf-ptrace.h, and put this here:

  /* See inf-ptrace.h.  */

> 
>  pid_t
>  get_ptrace_pid (ptid_t ptid)
> @@ -328,6 +344,7 @@ get_ptrace_pid (ptid_t ptid)
>      pid = ptid.pid ();
>    return pid;
>  }
> +#endif
> 
>  /* Resume execution of thread PTID, or all threads if PTID is -1.  If
>     STEP is nonzero, single-step it.  If SIGNAL is nonzero, give it
> @@ -336,15 +353,12 @@ get_ptrace_pid (ptid_t ptid)
>  void
>  inf_ptrace_target::resume (ptid_t ptid, int step, enum gdb_signal signal)
>  {
> -  pid_t pid;
>    int request;
> 
>    if (minus_one_ptid == ptid)
>      /* Resume all threads.  Traditionally ptrace() only supports
>         single-threaded processes, so simply resume the inferior.  */
> -    pid = inferior_ptid.pid ();
> -  else
> -    pid = get_ptrace_pid (ptid);
> +    ptid = inferior_ptid;

I'm not certain about this part (which I suggested).  With the existing code,
if inferior_ptid is {pid=10, lwp=20}, we'll call ptrace with pid=10.  With the
patch, we'll call ptrace with pid=20.

I'm not entirely sure if/when resume can be called with ptid == minus_one_ptid.
And if it's called with minus_one_ptid, is it possible that ptid.pid != ptid.lwp?

In any case, it's probably safer to do:

  ptid = ptid_t (inferior_ptid.pid ());

to retain the existing behavior.  WDYT?

> 
>    if (catch_syscall_enabled () > 0)
>      request = PT_SYSCALL;
> @@ -365,7 +379,7 @@ inf_ptrace_target::resume (ptid_t ptid, int step, enum gdb_signal signal)
>       where it was.  If GDB wanted it to start some other way, we have
>       already written a new program counter value to the child.  */
>    errno = 0;
> -  ptrace (request, pid, (PTRACE_TYPE_ARG3)1, gdb_signal_to_host (signal));
> +  gdb_ptrace (request, ptid, (PTRACE_TYPE_ARG3)1, gdb_signal_to_host (signal));
>    if (errno != 0)
>      perror_with_name (("ptrace"));
>  }
> @@ -528,7 +542,7 @@ inf_ptrace_target::xfer_partial (enum target_object object,
>  				 const gdb_byte *writebuf,
>  				 ULONGEST offset, ULONGEST len, ULONGEST *xfered_len)
>  {
> -  pid_t pid = get_ptrace_pid (inferior_ptid);
> +  ptid_t ptid = inferior_ptid;
> 
>    switch (object)
>      {
> @@ -552,7 +566,7 @@ inf_ptrace_target::xfer_partial (enum target_object object,
>  	piod.piod_len = len;
> 
>  	errno = 0;
> -	if (ptrace (PT_IO, pid, (caddr_t)&piod, 0) == 0)
> +	if (gdb_ptrace (PT_IO, ptid, (caddr_t)&piod, 0) == 0)
>  	  {
>  	    /* Return the actual number of bytes read or written.  */
>  	    *xfered_len = piod.piod_len;
> @@ -565,7 +579,7 @@ inf_ptrace_target::xfer_partial (enum target_object object,
>  	  return TARGET_XFER_EOF;
>        }
>  #endif
> -      *xfered_len = inf_ptrace_peek_poke (pid, readbuf, writebuf,
> +      *xfered_len = inf_ptrace_peek_poke (ptid.pid (), readbuf, writebuf,
>  					  offset, len);
>        return *xfered_len != 0 ? TARGET_XFER_OK : TARGET_XFER_EOF;

I'm pretty sure that this change is not right.  For non-NetBSD targets, this
used to pass `inferior_ptid.lwp` (the result of get_ptrace_pid).  Now, we pass
`inferior_ptid.pid`.

I think you'll need to pass the ptid to inf_ptrace_peek_poke and use gdb_ptrace
inside that function too.

Simon