Re: kdb4.3 for linux 2.6 on i386

[Prashanth T <prasht@xxxxxxxxxx>]

Jim,   I had tried this earlier, taking COMMON code
kdb-4.3-2.6.0-test11-common1.patch  (by Keith) and i386 code from
your code on linux 2.6.0.   I had encountered the same two problems
which you have mentioned and the fixes rightly said by you,are in
COMMON code itself.  Attached below is the patch for those two
issues to be made to the  COMMON code: 
Correct me if I am wrong.

1.  local_bh_enable( ) is to be replaced with __local_bh_enable( ) since
    in an SMP environment, all CPUs other than the intial CPU would
    have their IRQs disabled.   Because of this, they would dump
    their stack when they encounter the WARN_ON check in
    local_bh_enable( ) code. 

2.  I think tasklist_lock comes into picture only in case of SMP and
    to check if this lock is acquired by any other process,
    rwlock_is_locked( ) routine is to be used.


Thanks
Prashanth

Jim Houston wrote:

> On Mon, 2004-01-19 at 08:14, Prashanth T wrote:
>  
>
> > Jim,
> >    I came accross your patches on kdb-4.3 for linux2.6 on i386 at
> >                                                                  
> > http://www.ccur.com/isd_oss.asp
> >    I believe that your COMMON code patch (+ some other changes superceded)
> >    are included in the common patch put by Keith
> > ftp://oss.sgi.com/projects/kdb/download/v4.3/kdb-v4.3-2.6.0-test11-common-1.bz2
> >
> > I was wondering why your i386 patches are NOT seen there in the list?
> > Or is it under progress?  
> >    
>
> Hi Prashanth, Keith,
>
> I had assumed that Keith Owen's release included the i386
> support.  I guess I was premature in declaring my patch obsolete.
>
> I spent this most of the day doing a "quick" merge to see if Keith's
> common code would work with my i386 changes.  It does.
> I'm working with a 2.6.1 kernel and have a couple problems which
> are in the common code.
>
>     -   This warning when I do a go command:
>         "Badness in local_bh_enable at kernel/softirq.c:121"
>
>     -   The kdb kill command has compile errors.
>
> I hope to have a new patch on the web tomorrow.
>
> Jim Houston - Concurrent Computer Corp.
>
>
>
>
>
>  
--- linux-2.6.0-kdb-COMMON/kdb/kdbmain.c        2004-01-20 10:41:15.575577432 
+0530
+++ linux-2.6.0-kdb-COMMON-mod/kdb/kdbmain.c    2004-01-19 13:40:53.000000000 
+0530
@@ -1967,7 +1967,7 @@
                KDB_DEBUG_STATE("kdb 15", result);
                kdb_bp_install_local(regs);
                kdba_enable_lbr();
-               local_bh_enable();
+               __local_bh_enable();
                KDB_STATE_CLEAR(KDB_CONTROL);
        }
 
@@ -3168,11 +3168,13 @@
        }
        spin_unlock(&p->sighand->siglock);
        /* Where's read_trylock? */
-       if (tasklist_lock.write_lock) {
+#if defined(CONFIG_SMP)
+       if (rwlock_is_locked(&tasklist_lock)) {
                kdb_printf("Can't do kill command now.\n"
                        "The runqueue lock is held somewhere else in kernel, 
try again later\n");
                return 0;
        }
+#endif
        info.si_signo = sig;
        info.si_errno = 0;
        info.si_code = SI_USER;