kvm: use cpumask_var_t for cpus_hardware_enabled

PATCH In review Rusty Russell 3 participants 6 messages Dec 7, 2008 – Dec 8, 2008 KVM

Share & tools

Rusty Russell

author

2008-12-07 10:56 UTC #0 lore raw

PATCH 2/2 kvm: use cpumask_var_t for cpus_hardware_enabled

This changes cpus_hardware_enabled from a cpumask_t to a cpumask_var_t:
equivalent for CONFIG_CPUMASKS_OFFSTACK=n, otherwise dynamically allocated.

Signed-off-by Rusty Russell <[email protected]>

---
 virt/kvm/kvm_main.c |   20 ++++++++++++++------
 1 file changed, 14 insertions(+), 6 deletions(-)

virt/kvm/kvm_main.c +14 -6

@@ -63,7 +63,7 @@ DEFINE_SPINLOCK(kvm_lock);
 DEFINE_SPINLOCK(kvm_lock);
 LIST_HEAD(vm_list);
 
-static cpumask_t cpus_hardware_enabled;
+static cpumask_var_t cpus_hardware_enabled;
 
 struct kmem_cache *kvm_vcpu_cache;
 EXPORT_SYMBOL_GPL(kvm_vcpu_cache);

@@ -1796,9 +1796,9 @@ static void hardware_enable(void *junk)
 {
 	int cpu = raw_smp_processor_id();
 
-	if (cpu_isset(cpu, cpus_hardware_enabled))
+	if (cpumask_test_cpu(cpu, cpus_hardware_enabled))
 		return;
-	cpu_set(cpu, cpus_hardware_enabled);
+	cpumask_set_cpu(cpu, cpus_hardware_enabled);
 	kvm_arch_hardware_enable(NULL);
 }

@@ -1806,9 +1806,9 @@ static void hardware_disable(void *junk)
 {
 	int cpu = raw_smp_processor_id();
 
-	if (!cpu_isset(cpu, cpus_hardware_enabled))
+	if (!cpumask_test_cpu(cpu, cpus_hardware_enabled))
 		return;
-	cpu_clear(cpu, cpus_hardware_enabled);
+	cpumask_clear_cpu(cpu, cpus_hardware_enabled);
 	kvm_arch_hardware_disable(NULL);
 }

@@ -2042,9 +2042,14 @@ int kvm_init(void *opaque, unsigned int 
 
 	bad_pfn = page_to_pfn(bad_page);
 
+	if (!alloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) {
+		r = -ENOMEM;
+		goto out_free_0;
+	}
+
 	r = kvm_arch_hardware_setup();
 	if (r < 0)
-		goto out_free_0;
+		goto out_free_0a;
 
 	for_each_online_cpu(cpu) {
 		smp_call_function_single(cpu,

@@ -2103,6 +2108,8 @@ out_free_2:
 	on_each_cpu(hardware_disable, NULL, 1);
 out_free_1:
 	kvm_arch_hardware_unsetup();
+out_free_0a:
+	free_cpumask_var(cpus_hardware_enabled);
 out_free_0:
 	__free_page(bad_page);
 out:

@@ -2126,6 +2133,7 @@ void kvm_exit(void)
 	kvm_arch_hardware_unsetup();
 	kvm_arch_exit();
 	kvm_exit_debug();
+	free_cpumask_var(cpus_hardware_enabled);
 	__free_page(bad_page);
 }
 EXPORT_SYMBOL_GPL(kvm_exit);

Avi Kivity

2008-12-07 15:55 UTC #1 lore raw

Review

Quoted 6 lines · Rusty Russell

This changes cpus_hardware_enabled from a cpumask_t to a cpumask_var_t:
equivalent for CONFIG_CPUMASKS_OFFSTACK=n, otherwise dynamically allocated.
(quoted diff) +1 -1
-static cpumask_t cpus_hardware_enabled;
+static cpumask_var_t cpus_hardware_enabled

This isn't on stack, so it isn't buying us anything.

Is the plan to drop cpumask_t? If so, we're penalizing non-stack users by forcing them to go through another pointer (and cacheline).

Signature · 1 line

error compiling committee.c: too many arguments to function

Rusty Russell

author

2008-12-08 06:05 UTC #2 lore raw

Quoted 9 lines · On Monday 08 December 2008 02:25:50 Avi Kivity wrote:

Review
Quoted 6 lines · Rusty Russell
This changes cpus_hardware_enabled from a cpumask_t to a cpumask_var_t:
equivalent for CONFIG_CPUMASKS_OFFSTACK=n, otherwise dynamically allocated.
(quoted diff) +1 -1
-static cpumask_t cpus_hardware_enabled;
+static cpumask_var_t cpus_hardware_enabled
This isn't on stack, so it isn't buying us anything.

It's the CONFIG_NR_CPUS=4096 but nr_cpu_ids=4 case which we win using dynamic allocation. Gotta love distribution kernels.

Quoted 1 line

Is the plan to drop cpumask_t?

Yes. And undefine 'struct cpumask' if CONFIG_CPUMASK_OFFSTACK. That will stop assignment and on-stack declarations for all but the most determined.

Quoted 2 lines

If so, we're penalizing non-stack users
by forcing them to go through another pointer (and cacheline).

Not quite. If !CONFIG_CPUMASK_OFFSTACK, cpumask_var_t == cpumask_t[1]. Blame Linus :)

Cheers,
Rusty.

Avi Kivity

2008-12-08 09:46 UTC #3 lore raw

Quoted 7 lines · Rusty Russell

Quoted 2 lines
This isn't on stack, so it isn't buying us anything.
It's the CONFIG_NR_CPUS=4096 but nr_cpu_ids=4 case which we win using dynamic allocation. Gotta love distribution kernels.

What does it buy? 4096/8 = 512 bytes statically allocated?

I understand passing things as pointers, but allocating everything dynamically is unCish.

Quoted 8 lines

Quoted 2 lines
Is the plan to drop cpumask_t?
Yes. And undefine 'struct cpumask' if CONFIG_CPUMASK_OFFSTACK. That will stop assignment and on-stack declarations for all but the most determined.

Quoted 6 lines

Quoted 2 lines
If so, we're penalizing non-stack users by forcing them to go through another pointer (and cacheline).
Not quite. If !CONFIG_CPUMASK_OFFSTACK, cpumask_var_t == cpumask_t[1]. Blame Linus :)

Hm, is there a C trick which will error out when allocating something on the stack, but work when allocating statically? I can think of something to do the reverse, but that doesn't help.

Maybe a weak or visibility attribute? These don't make sense on function locals.

Signature · 2 lines

I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.

Rusty Russell

author

2008-12-08 11:50 UTC #4 lore raw

Quoted 10 lines · On Monday 08 December 2008 20:16:44 Avi Kivity wrote:

Quoted 7 lines · Rusty Russell
Quoted 2 lines
This isn't on stack, so it isn't buying us anything.
It's the CONFIG_NR_CPUS=4096 but nr_cpu_ids=4 case which we win using dynamic allocation. Gotta love distribution kernels.
What does it buy? 4096/8 = 512 bytes statically allocated?

It adds up, and 4096 seems to be only the start of the insanityH^H^Hfun.

Quoted 7 lines

Quoted 3 lines
Not quite. If !CONFIG_CPUMASK_OFFSTACK, cpumask_var_t == cpumask_t[1]. Blame Linus :)
Hm, is there a C trick which will error out when allocating something on the stack, but work when allocating statically? I can think of something to do the reverse, but that doesn't help.

We also need to prevent assignment, eg:

	*foo = *bar;

Because when we allocate them, we'll cut them to size.

Cheers,
Rusty.

Mike Travis

2008-12-08 14:29 UTC #5 lore raw

Quoted 11 lines · Rusty Russell

Quoted 8 lines · On Monday 08 December 2008 20:16:44 Avi Kivity wrote:
Quoted 6 lines · Rusty Russell
Quoted 2 lines
This isn't on stack, so it isn't buying us anything.
It's the CONFIG_NR_CPUS=4096 but nr_cpu_ids=4 case which we win using dynamic allocation. Gotta love distribution kernels.
What does it buy? 4096/8 = 512 bytes statically allocated?
It adds up, and 4096 seems to be only the start of the insanityH^H^Hfun.

The real win though is when cpumask_size represents the actual size of the cpumask (based on # of possible cpus) instead of the pre-configured size of NR_CPUS. So for 99.9% of the systems (having 64 or fewer cpus), the savings will be 504 bytes not allocated.

Quoted 16 lines

Quoted 7 lines
Quoted 4 lines
Not quite. If !CONFIG_CPUMASK_OFFSTACK, cpumask_var_t == cpumask_t[1]. Blame Linus :)
Hm, is there a C trick which will error out when allocating something on the stack, but work when allocating statically? I can think of something to do the reverse, but that doesn't help.
We also need to prevent assignment, eg:
	*foo = *bar;
Because when we allocate them, we'll cut them to size.
Cheers,
Rusty.