On Wed, 2006-12-20 at 11:16 -0800, Tim Nguyen wrote: The SMP problem has been a pain in the butt for a long time with the SF kgdb patch. I've been disabling SMP on 2.6.12 to avoid it.
I'm having problems with the 2.6.12 patch on VmWare were
it looks like a hang in this area but you fix didn't help.
I'm still investigating the problem. Does the patch even
work with SMP un-configured? Might be requirement for LOCAL_APIC.
Seems like it would be simpler for non-SMP to skip the NMI IPI
and the waiting for other processors. Whats the point?
I was wondering with you fix why it's necessary to increase
the premption count by HARDIRQ_OFFSET.
Also I don't know why IRQ_EXIT_OFFSET can be one off of HARDIRQ_OFFSET.
-piet
> Well it turns out that retrying the NMI_IPI is not a good idea. This
> **final** patch presents the minimal changes needed to address the
> issues previously stated and supersedes all previous postings. BTW this
> was derived from the trini's linux-2.6-kgdb-testing tree:
>
> diff --git a/kernel/kgdb.c b/kernel/kgdb.c
> index a60550f..394ab61 100644
> --- a/kernel/kgdb.c
> +++ b/kernel/kgdb.c
> @@ -901,9 +901,11 @@ int kgdb_handle_exception(int ex_vector,
>
> kgdb_disable_hw_debug(linux_regs);
>
> - if (!debugger_step || !kgdb_contthread)
> + if (!debugger_step || !kgdb_contthread) {
> + add_preempt_count(HARDIRQ_OFFSET);
> for (i = 0; i < NR_CPUS; i++)
> spin_lock(&slavecpulocks[i]);
> + }
>
> #ifdef CONFIG_SMP
> /* Make sure we get the other CPUs */
> @@ -1367,6 +1369,7 @@ int kgdb_handle_exception(int ex_vector,
> cpu_relax();
> }
> }
> + sub_preempt_count(IRQ_EXIT_OFFSET);
> }
>
> #ifdef CONFIG_SMP
> diff --git a/kernel/timer.c b/kernel/timer.c
> index 1d7dd62..4c3e7d2 100644
> --- a/kernel/timer.c
> +++ b/kernel/timer.c
> @@ -34,6 +34,7 @@
> #include <linux/cpu.h>
> #include <linux/syscalls.h>
> #include <linux/delay.h>
> +#include <linux/kgdb.h>
>
> #include <asm/uaccess.h>
> #include <asm/unistd.h>
> @@ -1176,6 +1177,10 @@ void update_process_times(int user_tick)
> struct task_struct *p = current;
> int cpu = smp_processor_id();
>
> +#ifdef CONFIG_KGDB
> + if (atomic_read(&cpu_doing_single_step) == cpu)
> + return;
> +#endif
> /* Note: this timer irq context must be accounted for as well.
> */
> if (user_tick)
> account_user_time(p, jiffies_to_cputime(1));
>
> -------------------------------------------------------------------------
> Take Surveys. Earn Cash. Influence the Future of IT
> Join SourceForge.net's Techsay panel and you'll get the chance to share your
> opinions on IT & business topics through brief surveys - and earn cash
> http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
> _______________________________________________
> Kgdb-bugreport mailing list
> Kgdb-bugreport@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/kgdb-bugreport
--
Piet Delaney Phone: (408) 200-5256
Blue Lane Technologies Fax: (408) 200-5299
10450 Bubb Rd.
Cupertino, Ca. 95014 Email: [EMAIL PROTECTED]
signature.asc
Description: This is a digitally signed message part
------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys - and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________ Kgdb-bugreport mailing list Kgdb-bugreport@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/kgdb-bugreport
