Re: [PATCH] sched/debug: initialize sd_sysctl_cpus if !CONFIG_CPUMASK_OFFSTACK

From: Joe Lawrence
Date: Fri Feb 01 2019 - 17:33:24 EST


On 1/31/19 8:44 AM, Masayoshi Mizuma wrote:
On Wed, Jan 30, 2019 at 09:14:00PM +0100, Peter Zijlstra wrote:
On Tue, Jan 29, 2019 at 10:12:45AM -0500, Masayoshi Mizuma wrote:
From: Hidetoshi Seto <seto.hidetoshi@xxxxxxxxxxxxxx>

register_sched_domain_sysctl() copies the cpu_possible_mask into
sd_sysctl_cpus, but only if sd_sysctl_cpus hasn't already been
allocated (ie, CONFIG_CPUMASK_OFFSTACK is set). However, when
CONFIG_CPUMASK_OFFSTACK is not set, sd_sysctl_cpus is left uninitialized
(all zeroes) and the kernel may fail to initialize sched_domain sysctl
entries for all possible cpus.

This is visible to the user if the kernel is booted with maxcpus=n, or
if ACPI tables have been modified to leave cpus offline, and then
checking for missing /proc/sys/kernel/sched_domain/cpu* entries.

Fix this by separating the allocataion and initialization, and adding
a flag to initialize the possible cpu entries while system booting only.

Signed-off-by: Hidetoshi Seto <seto.hidetoshi@xxxxxxxxxxxxxx>
Reviewed-by: Masayoshi Mizuma <m.mizuma@xxxxxxxxxxxxxx>
Tested-by: Syuuichirou Ishii <ishii.shuuichir@xxxxxxxxxxxxxx>
Tested-by: Tarumizu, Kohei <tarumizu.kohei@xxxxxxxxxxxxxx>
---
kernel/sched/debug.c | 4 ++++
1 file changed, 4 insertions(+)

diff --git a/kernel/sched/debug.c b/kernel/sched/debug.c
index de3de997e245..9c6637f3e21d 100644
--- a/kernel/sched/debug.c
+++ b/kernel/sched/debug.c
@@ -310,6 +310,7 @@ static struct ctl_table *sd_alloc_ctl_cpu_table(int cpu)
static cpumask_var_t sd_sysctl_cpus;
static struct ctl_table_header *sd_sysctl_header;
+static int register_sched_domain_sysctl_on_boot = 1;
void register_sched_domain_sysctl(void)
{
@@ -344,9 +345,12 @@ void register_sched_domain_sysctl(void)
if (!cpumask_available(sd_sysctl_cpus)) {
if (!alloc_cpumask_var(&sd_sysctl_cpus, GFP_KERNEL))
return;
+ }
+ if (register_sched_domain_sysctl_on_boot) {
/* init to possible to not have holes in @cpu_entries */
cpumask_copy(sd_sysctl_cpus, cpu_possible_mask);
+ register_sched_domain_sysctl_on_boot = 0;
}
for_each_cpu(i, sd_sysctl_cpus) {

I change it like the below. By keeping the initial value 0 it can go
into .bss instead of .data.

Great, thanks!
Should I re-post the patch as v2?

- Masa


Peter's revision tests fine on my aarch64 box and would get my ACK, however it ends up getting re-posted :)

Thanks,

-- Joe


--- a/kernel/sched/debug.c
+++ b/kernel/sched/debug.c
@@ -315,6 +315,7 @@ void register_sched_domain_sysctl(void)
{
static struct ctl_table *cpu_entries;
static struct ctl_table **cpu_idx;
+ static bool init_done = false;
char buf[32];
int i;
@@ -344,7 +345,10 @@ void register_sched_domain_sysctl(void)
if (!cpumask_available(sd_sysctl_cpus)) {
if (!alloc_cpumask_var(&sd_sysctl_cpus, GFP_KERNEL))
return;
+ }
+ if (!init_done) {
+ init_done = true;
/* init to possible to not have holes in @cpu_entries */
cpumask_copy(sd_sysctl_cpus, cpu_possible_mask);
}


-- Joe