Wessel, Jason wrote:
>> It is not limited to KGDBoE.
> You are correct. This is a generic problem with kgdbwait on x86_64.
>>>Perhaps you can experiment a bit further, and report back as I do not
>>>have your exact setup with SMP on x86_64 at the moment?
>> Well, this would explain why this has gone unnoticed so far. :-)
> I setup a an SMP x86_64 system and yes indeed there is a significant
> problem there.
I have verified that in non -rt kernel it's even worse -- kernel just
locks up instead of giving BUG trace -- yet somehow I was sure it worked for
me before. :-/
> I further tested the patches I created, and they fix both cases for
> kgdbwait. On x86_64 the rs232 automatically becomes a late init device
Good to know -- I haven't gotten around to testing them myself... :-/
> because it requires the exception stack be available to issue a
> breakpoint(). Another way to make a change that is KGDB specific for
> this arch if early init is still truly needed for the rs232 is to add
> kgdb specific code right after the place that the exception stack is
> initialized.
>>> kernel/kgdb.c | 23 +++++++++++------------
>>> 1 file changed, 11 insertions(+), 12 deletions(-)
>>>Index: linux-2.6.21-standard/kernel/kgdb.c
>>>===================================================================
>>>--- linux-2.6.21-standard.orig/kernel/kgdb.c
>>>+++ linux-2.6.21-standard/kernel/kgdb.c
>>>@@ -1788,17 +1788,6 @@ late_initcall(kgdb_late_entry);
>>> */
>>> void breakpoint(void)
>>> {
>>>- if (kgdb_initialized != 1) {
>>>- kgdb_early_entry();
>>>- if (kgdb_initialized == 1)
>>>- printk(KERN_CRIT "Waiting for connection from remote "
>>>- "gdb...\n");
>>>- else {
>>>- printk(KERN_CRIT "KGDB cannot initialize I/O yet.\n");
>>>- return;
>>>- }
>>>- }
>>>-
>>> atomic_set(&kgdb_setting_breakpoint, 1);
>>> wmb();
>>> BREAKPOINT();
>>>@@ -1883,7 +1872,17 @@ static int __init opt_kgdb_enter(char *s
>>> if (kgdb_initialized)
>>> return 0;
>>>
>>>- /* Call breakpoint() which will take care of init. */
>>>+ if (kgdb_initialized != 1) {
>>>+ kgdb_early_entry();
>>>+ if (kgdb_initialized == 1)
>>>+ printk(KERN_CRIT "Waiting for connection from remote "
>>>+ "gdb...\n");
>>>+ else {
>>>+ printk(KERN_CRIT "KGDB cannot initialize I/O yet.\n");
>>>+ return;
>>>+ }
>>>+ }
>>>+
>>> breakpoint();
>>
>> I don't see how this might help.
> This patch does directly have to do with the problem however it does fix
Did you mean "doesn't have to do"?
> a problem where breakpoint() references a function which is in the
> .init_text and this generates a warning with the latest rc kernels.
> Also moving the function into the early init makes for the situation
> where the low level initialization can only get called one time as a
> part of the init sequence.
>>> kernel/kgdb.c | 11 ++++++-----
>>> 1 file changed, 6 insertions(+), 5 deletions(-)
>>>Index: linux-2.6.21-standard/kernel/kgdb.c
>>>===================================================================
>>>--- linux-2.6.21-standard.orig/kernel/kgdb.c
>>>+++ linux-2.6.21-standard/kernel/kgdb.c
>>>@@ -1682,20 +1682,21 @@ DECLARE_TASKLET(kgdb_tasklet_breakpoint,
>>> */
>>> static void __init kgdb_early_entry(void) {
>>>+ /* Let the architecture do any setup that it needs to. */
>>>+ kgdb_arch_init();
>>>+
>>> /*
>>> * Don't try and do anything until the architecture is able to
>>> * setup the exception stack. In this case, it is up to the
>>> * architecture to hook in and look at us when they are ready.
>>> */
>>>+
>>> if(!CHECK_EXCEPTION_STACK()) {
Worth fixing if stmt style while at it. :-)
>>> kgdb_initialized = -1;
>>>- tasklet_schedule(&kgdb_tasklet_breakpoint);
>>>+ /* any kind of break point is deferred to late_init */
>>> return;
> Note I removed the tasklet and it is put into a late breakpoint() call.
> I had tried out both kgdb over rs232 and kgdboe on the x86_64 SMP with
> kgdbwait and it works now with no hang at boot.
Oops, I have indeed managed to overlook this particular change. :-<
> Jason.
Thanks,
Sergei
-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
_______________________________________________
Kgdb-bugreport mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/kgdb-bugreport
