[PATCH] fix maxcpus=1 oops in show_stat()

!MAILaRCHIVE_VOTE_RePLACE
Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]
To: Linus Torvalds <torvalds@...>
Cc: Alexey Dobriyan <adobriyan@...>, Alexey Dobriyan <adobriyan@...>, Len Brown <lenb@...>, Michal Piotrowski <michal.k.k.piotrowski@...>, Andrew Morton <akpm@...>, Andi Kleen <ak@...>, Jan Beulich <jbeulich@...>, <linux-kernel@...>
Date: Thursday, August 30, 2007 - 11:26 pm

Alexey Dobriyan reports that maxcpus=1 is still broken in 2.6.23-rc4:
if CONFIG_HOTPLUG_CPU is not set, x86_64 bootup oopses in show_stat() -
for_each_possible_cpu accesses a per-cpu area which was never set up.

Alexey identified commit 61ec7567db103d537329b0db9a887db570431ff4
(ACPI: boot correctly with "nosmp" or "maxcpus=0") as the origin;
but it's not really to blame, just exposes a bug in 2.6.23-rc1's commit
8b3b295502444340dd0701855ac422fbf32e161d (Especially when !CONFIG_HOTPLUG_CPU,
avoid needlessy allocating resources for CPUs that can never become available).

rc1's test for max_cpus < 2 in start_kernel() wasn't working because
max_cpus was still NR_CPUS at that point: until rc4 moved the maxcpus
parsing earlier.  Now it sets cpu_possible_map to 1 before allocating
all possible per-cpu areas; then smp_init() expands cpu_possible_map
to cpu_present_map (0xf in my case) later on.

rc1's commit has good intentions, but expects cpu_present_map to be
limited by maxcpus, which is only the case on i386.  cpus_and(possible,
possible,present) might be good, but needs an audit of cpu_present_map
uses - there may well be assumptions that any cpu present is possible.
So stay safe for now and just revert those #ifndef CONFIG_HOTPLUG_CPU
optimizations in rc1's commit.

Signed-off-by: Hugh Dickins <hugh@veritas.com>

--- 2.6.23-rc4/init/main.c	2007-08-28 04:42:58.000000000 +0100
+++ linux/init/main.c	2007-08-30 20:45:26.000000000 +0100
@@ -397,10 +397,6 @@ static void __init smp_init(void)
 {
 	unsigned int cpu;
 
-#ifndef CONFIG_HOTPLUG_CPU
-	cpu_possible_map = cpu_present_map;
-#endif
-
 	/* FIXME: This should be done in userspace --RR */
 	for_each_present_cpu(cpu) {
 		if (num_online_cpus() >= max_cpus)
@@ -545,10 +541,6 @@ asmlinkage void __init start_kernel(void
 	setup_arch(&command_line);
 	setup_command_line(command_line);
 	unwind_setup();
-#ifndef CONFIG_HOTPLUG_CPU
-	if (max_cpus < 2)
-		cpu_possible_map = cpu_online_map;
-#endif
 	setup_per_cpu_areas();
 	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
 
-
Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]

Messages in current thread:
2.6.23-rc4: maxcpus still broken, Alexey Dobriyan, (Tue Aug 28, 7:14 am)
Re: 2.6.23-rc4: maxcpus still broken, Michal Piotrowski, (Tue Aug 28, 7:35 pm)
Re: 2.6.23-rc4: maxcpus still broken, Alexey Dobriyan, (Wed Aug 29, 12:39 am)
Re: 2.6.23-rc4: maxcpus still broken, Hugh Dickins, (Wed Aug 29, 1:03 am)
Re: 2.6.23-rc4: maxcpus still broken, Hugh Dickins, (Wed Aug 29, 4:20 pm)
Re: 2.6.23-rc4: maxcpus still broken, Alexey Dobriyan, (Wed Aug 29, 1:23 am)
Re: 2.6.23-rc4: maxcpus still broken, Len Brown, (Wed Aug 29, 2:41 am)
Re: 2.6.23-rc4: maxcpus still broken, Alexey Dobriyan, (Wed Aug 29, 3:25 am)
[PATCH] fix maxcpus=1 oops in show_stat(), Hugh Dickins, (Thu Aug 30, 11:26 pm)
Re: [PATCH] fix maxcpus=1 oops in show_stat(), Alexey Dobriyan, (Fri Aug 31, 3:37 am)
Re: [PATCH] fix maxcpus=1 oops in show_stat(), Linus Torvalds, (Fri Aug 31, 12:33 am)