[PATCH]: PTRACE_PEEKUSER redux...

[PATCH]: PTRACE_PEEKUSER redux...

[David S. Miller]

Daniel, I've been thinking about that PTRACE_PEEKUSER dependency issue
in linux-nat.c some more.  And it's kind of unreasonable to expect
people to get the latest and greatest Sparc Linux kernel in order to
get a working gdb setup.

Especially when we can handle this gracefully, and without much pain.

In all cases, no matter what ptrace operation is specified, if the
child does not exist the error return we will get is -ESRCH (task not
found for pid) or -EPERM (trying to trace init or similar).

Specifically checking for those two error codes allows us to cleanly
handle Linux targets that do not need to implement PTRACE_PEEKUSER.

Therefore, what do you think about the following patch?

2006-04-06  David S. Miller  <davem@sunset.davemloft.net>

	* linux-nat.c (linux_nat_thread_alive): Thread is alive
	as long as errno is neither -ESRCH nor -EPERM.  This allows
	to handle cleanly the case where PTRACE_PEEKUSER is not
	a supported ptrace operation for a given Linux target.

--- linux-nat.c.~1~	2006-04-05 18:08:11.000000000 -0700
+++ linux-nat.c	2006-04-06 21:55:49.000000000 -0700
@@ -2382,7 +2383,7 @@ linux_nat_thread_alive (ptid_t ptid)
 			"LLTA: PTRACE_PEEKUSER %s, 0, 0 (%s)\n",
 			target_pid_to_str (ptid),
 			errno ? safe_strerror (errno) : "OK");
-  if (errno)
+  if (errno == -ESRCH || errno == -EPERM)
     return 0;
 
   return 1;

Re: [PATCH]: PTRACE_PEEKUSER redux...

[Daniel Jacobowitz]

On Thu, Apr 06, 2006 at 10:03:57PM -0700, David S. Miller wrote:
> In all cases, no matter what ptrace operation is specified, if the
> child does not exist the error return we will get is -ESRCH (task not
> found for pid) or -EPERM (trying to trace init or similar).

This relies on all platforms doing the permission check before the
operation check.  Well, I guess it's still better than what we had
before, right?

> 2006-04-06  David S. Miller  <davem@sunset.davemloft.net>
> 
> 	* linux-nat.c (linux_nat_thread_alive): Thread is alive
> 	as long as errno is neither -ESRCH nor -EPERM.  This allows
> 	to handle cleanly the case where PTRACE_PEEKUSER is not
> 	a supported ptrace operation for a given Linux target.

OK, except, please reverse the changelog and the patch :-)  This
comment should be in the source code, or we'll break it again later.

-- 
Daniel Jacobowitz
CodeSourcery

Re: [PATCH]: PTRACE_PEEKUSER redux...

[David S. Miller]

From: Daniel Jacobowitz <drow@false.org>
Date: Fri, 7 Apr 2006 09:16:23 -0400

> On Thu, Apr 06, 2006 at 10:03:57PM -0700, David S. Miller wrote:
> > In all cases, no matter what ptrace operation is specified, if the
> > child does not exist the error return we will get is -ESRCH (task not
> > found for pid) or -EPERM (trying to trace init or similar).
> 
> This relies on all platforms doing the permission check before the
> operation check.  Well, I guess it's still better than what we had
> before, right?

I definitely think so.

> > 2006-04-06  David S. Miller  <davem@sunset.davemloft.net>
> > 
> > 	* linux-nat.c (linux_nat_thread_alive): Thread is alive
> > 	as long as errno is neither -ESRCH nor -EPERM.  This allows
> > 	to handle cleanly the case where PTRACE_PEEKUSER is not
> > 	a supported ptrace operation for a given Linux target.
> 
> OK, except, please reverse the changelog and the patch :-)  This
> comment should be in the source code, or we'll break it again later.

Ok, will do.  Thanks a lot for reviewing.