Announce: kdb v4.4 x86-64 updates for for kernel 2.6.6

Announce: kdb v4.4 x86-64 updates for for kernel 2.6.6

[Keith Owens]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

KDB (Linux Kernel Debugger) has been updated.

ftp://oss.sgi.com/projects/kdb/download/v4.4/

kdb-v4.4-2.6.6-x86-64-2.bz2 is available.  The x86-64 patch is still a
work in progress, use with care.  Changelog extract.

2004-05-15 Jack F. Vogel <jfv@bluesong.net>
	* port to 2.6.6 for x86_64

2003-12-15 Cliff Neighbors <cliff@fabric7.com>
	* initial port from i386 to x86_64


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)
Comment: Exmh version 2.1.1 10/15/1999

iD8DBQFAxmgii4UHNye0ZOoRAhUyAJkBDBkWlmPmLHPfbMykdzd7u11/NACgylLR
62/ewGczr+kFxRYq1X9ATUI=
=8Jrf
-----END PGP SIGNATURE-----

Announce: kdb v4.3 x86-64 updates for for kernel 2.4.25

[Keith Owens]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

KDB (Linux Kernel Debugger) has been updated.

ftp://oss.sgi.com/projects/kdb/download/v4.3/

kdb-v4.3-2.4.25-x86-64-1.bz2 is available.  The x86-64 patch is still a
work in progress, use with care.  Changelog extract.

2004-06-23 Jack Vogel <jfv@bluesong.net>
	* Port patch to 2.4.25 and sync with Keith's
	  4.3 common and i386 code.
	* By internal (IBM) request add cr8 to register dump

2003-12-15 Cliff Neighbors <cliff@fabric7.com>
	* initial port from i386 to x86_64

x86-64 for 2.4.25 requires kdb-v4.3-2.4.25-common-3 or later.


When Marcelo releases 2.4.27, I will do a backport of kdb v4.4 from 2.6
to 2.4.27, this is probably the last update for kdb v4.3.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)
Comment: Exmh version 2.1.1 10/15/1999

iD8DBQFA47pli4UHNye0ZOoRAq1MAKC7AAgihexkB5cmB689cdybKh6EDACfcTZw
5v/umYBg6Cn9lc4xZOnsq9I=
=RVDU
-----END PGP SIGNATURE-----

Re: Announce: kdb v4.4 x86-64 updates for for kernel 2.6.6

[Andi Kleen]

Keith Owens <kaos@sgi.com> writes:

> KDB (Linux Kernel Debugger) has been updated.
>
> ftp://oss.sgi.com/projects/kdb/download/v4.4/
>
> kdb-v4.4-2.6.6-x86-64-2.bz2 is available.  The x86-64 patch is still a
> work in progress, use with care.  Changelog extract.
>
> 2004-05-15 Jack F. Vogel <jfv@bluesong.net>
> 	* port to 2.6.6 for x86_64

It uses the wrong interfaces. The x86-64 die notifier was exactly 
designed to avoid putting KDB hooks all over the kernel, but you
added them anyways. Please fix that.

In theory with the die hooks interface kdb could be a loadable (although
not unloadable) module btw.

Some comments:

+#if defined(CONFIG_SMP) && defined(CONFIG_KDB)
+static void do_ack_apic_irq(void)
+{
+       ack_APIC_irq();
+}
+#endif

NMIs don't need to be ACKed in the x86 APIC. The only reason to ack
something is that a higher priority interrupt can run, but there is no
higher priority interrupt than an NMI.

+#if defined(CONFIG_SMP) && defined(CONFIG_KDB)
+       /*
+        * Call the debugger to see if this NMI is due
+        * to a KDB requested IPI. If so, it will handle
+        */
+       if (kdb_ipi(regs, do_ack_apic_irq)){
+               nmi_exit();
+               return;
+       }
+#endif

Please just grab DIE_NMI_IPI in the die chain, which you're already
using. The NMI handler can veto the NMI then using NOTIFY_BAD.
The ack is also not needed.


@@ -367,6 +376,10 @@
        handle_BUG(regs);
        __die(str, regs, err);
        oops_end();
+#ifdef  CONFIG_KDB
+        kdb_diemsg = str;
+       kdb(KDB_REASON_OOPS, err, regs);
+#endif  /* CONFIG_KDB */

__die already has a die chain hook for that. Use that instead.

+#if !defined(CONFIG_KDB)
 DO_ERROR( 3, SIGTRAP, "int3", int3);
+#endif

and 

+#ifdef CONFIG_KDB
+/*
+ * KDB Breakpoint vector
+ */
+asmlinkage int do_int3(struct pt_regs * regs, long error_code)
+{
+       if (kdb(KDB_REASON_BREAK, error_code, regs))
+               return 0;
+       do_trap(3, SIGTRAP, "int3", regs, error_code, NULL);
+       return 0;
+}

Unnecessary when you have a die chain handler (which you have
already). Take a look at how DO_ERROR is defined.

-       set_intr_gate_ist(18,&machine_check, MCE_STACK); 
+#ifdef  CONFIG_KDB
+       {
+               set_intr_gate(18, &machine_check);
+       }
+       kdb_enablehwfault();

and

+void
+kdba_enable_mce(void)
+{
+       /* No longer required, arch/x86_64/kernel/bluesmoke.c does the job now *
/
+}


kdb_enablehwfault only calls kdba_enable_mce and is a complete 
nop. Can be just removed.

Overall about 90% of your changes to arch/x86_64/kernel/traps.c are
unnecessary and the rest could be easily moved into arch/x86_64/kdb,
giving an completely independent and low maintainance kdb.

-Andi