procfs issue.

[Alfred Perlstein]

First off, nice job fixing up sys_process.c it's a lot cleaner now and
the races seem to be gone, however there may still be a problem.

Please see: kern/29741:
  ptrace(pid);ptrace(ppid) makes pid and ppid unkillable
http://www.freebsd.org/cgi/query-pr.cgi?pr=kern%2F29741

It looks like the following delta (submitted by Tim J. Robbins) may fix it:

Index: sys_process.c
===
RCS file: /home/ncvs/src/sys/kern/sys_process.c,v
retrieving revision 1.89
diff -u -r1.89 sys_process.c
--- sys_process.c   12 Apr 2002 21:17:37 -  1.89
+++ sys_process.c   13 Apr 2002 02:16:14 -
@@ -332,11 +332,13 @@
struct fpreg fpreg;
struct reg reg;
} r;
-   struct proc *p;
+   struct proc *curp, *p, *pp;
struct thread *td2;
int error, write;
int proctree_locked = 0;
 
+   curp = td-td_proc;
+
/*
 * Do copyin() early before getting locks and lock proctree before
 * locking the process.
@@ -421,6 +423,17 @@
error = EBUSY;
goto fail;
}
+
+   /* Can't trace an ancestor if you're being traced. */
+   if (curp-p_flag  P_TRACED) {
+   for (pp = curp-p_pptr; pp != NULL; pp = pp-p_pptr) {
+   if (pp == p) {
+   error = EINVAL;
+   goto fail;
+   }
+   }
+   }
+
 
/* OK */
break;




thanks,
-- 
-Alfred Perlstein [[EMAIL PROTECTED]]
'Instead of asking why a piece of software is using 1970s technology,
 start asking why software is ignoring 30 years of accumulated wisdom.'
Tax deductible donations for FreeBSD: http://www.freebsdfoundation.org/

To Unsubscribe: send mail to [EMAIL PROTECTED]
with unsubscribe freebsd-hackers in the body of the message

Re: procfs issue.

[Dag-Erling Smorgrav]

Alfred Perlstein [EMAIL PROTECTED] writes:
 It looks like the following delta (submitted by Tim J. Robbins) may
 fix it:

It looks correct to me, please commit (unless John has any
objections?)

DES
-- 
Dag-Erling Smorgrav - [EMAIL PROTECTED]

To Unsubscribe: send mail to [EMAIL PROTECTED]
with unsubscribe freebsd-hackers in the body of the message

RE: procfs issue.

[John Baldwin]

On 13-Apr-2002 Alfred Perlstein wrote:
 First off, nice job fixing up sys_process.c it's a lot cleaner now and
 the races seem to be gone, however there may still be a problem.
 
 Please see: kern/29741:
   ptrace(pid);ptrace(ppid) makes pid and ppid unkillable
 http://www.freebsd.org/cgi/query-pr.cgi?pr=kern%2F29741
 
 It looks like the following delta (submitted by Tim J. Robbins) may fix it:
 
 Index: sys_process.c
 ===
 RCS file: /home/ncvs/src/sys/kern/sys_process.c,v
 retrieving revision 1.89
 diff -u -r1.89 sys_process.c
 --- sys_process.c 12 Apr 2002 21:17:37 -  1.89
 +++ sys_process.c 13 Apr 2002 02:16:14 -
 @@ -332,11 +332,13 @@
   struct fpreg fpreg;
   struct reg reg;
   } r;
 - struct proc *p;
 + struct proc *curp, *p, *pp;
   struct thread *td2;
   int error, write;
   int proctree_locked = 0;
  
 + curp = td-td_proc;
 +
   /*
* Do copyin() early before getting locks and lock proctree before
* locking the process.
 @@ -421,6 +423,17 @@
   error = EBUSY;
   goto fail;
   }
 +
 + /* Can't trace an ancestor if you're being traced. */
 + if (curp-p_flag  P_TRACED) {
 + for (pp = curp-p_pptr; pp != NULL; pp = pp-p_pptr) {
 + if (pp == p) {
 + error = EINVAL;
 + goto fail;
 + }
 + }
 + }
 +
  
   /* OK */
   break;

There's an inferior() function which you might be able to use for that check. 
Or not.  Anyways, I think that the proctree_lock is sufficient to protect
reading of P_TRACED since we always hold that lock while setting it, so we
don't need to lock curp (thank goodness since that could easily lead to
deadlocks) and proctree_lock will protect the walking of p_pptr so that should
be safe from a locking perspective.

I'll leave the actual bug in the PR up to someone else though as I'm not but so
familiar with ptrace(2).

-- 

John Baldwin [EMAIL PROTECTED]http://www.FreeBSD.org/~jhb/
Power Users Use the Power to Serve!  -  http://www.FreeBSD.org/

To Unsubscribe: send mail to [EMAIL PROTECTED]
with unsubscribe freebsd-hackers in the body of the message