lkml.org 
[lkml]   [2000]   [Feb]   [10]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
SubjectRe: 2.3.43-pre[25] do not work with HIGHMEM=64GB

On 10 Feb 2000, Christoph Rohland wrote:

> With Stevens patch 2.3.43-5 works again for me.
>
> But during shm tests I get:
>
> hm #1: c113a2a0, c113a0c0.

> Trace; c022b2cc <call_spurious_interrupt+253b/91b7>

it's not necesserily a bug (in fact it's a valid scenario) - are these
messages also accompanied by bad application behavior or kernel crashes?

if you still see problems then you might want to try Manfred's debugging
patchset he sent yesterday.

If no real problems apart from these messages, then could you apply
smp-2.3.43-E1 (attached) to pre8-2.3.43, it removes some paranoia and
debugging stuff. (also fixes the small window for a deadlock scenario
Manfred noticed, plus does the tlb_state initialization at compile time)
(pre8 already includes Stephen's fix as well)

-- mingo
--- linux/arch/i386/kernel/smp.c.orig2 Thu Feb 10 07:13:25 2000
+++ linux/arch/i386/kernel/smp.c Thu Feb 10 07:13:48 2000
@@ -348,12 +348,7 @@
__flush_tlb_one(flush_va);
} else
leave_mm(cpu);
- } else {
- extern void show_stack (void *);
- printk("hm #1: %p, %p.\n", flush_mm, cpu_tlbstate[cpu].active_mm);
- show_stack(NULL);
}
- __flush_tlb();
ack_APIC_irq();
clear_bit(cpu, &flush_cpumask);
}
@@ -383,7 +378,7 @@
* Temporarily this turns IRQs off, so that lockups are
* detected by the NMI watchdog.
*/
- spin_lock_irq(&tlbstate_lock);
+ spin_lock(&tlbstate_lock);

flush_mm = mm;
flush_va = va;
@@ -399,7 +394,7 @@

flush_mm = NULL;
flush_va = 0;
- spin_unlock_irq(&tlbstate_lock);
+ spin_unlock(&tlbstate_lock);
}

void flush_tlb_current_task(void)
\
 
 \ /
  Last update: 2005-03-22 13:56    [W:0.034 / U:0.052 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site