On 2011-02-15 13:32, Marcelo Tosatti wrote:
> On Tue, Feb 08, 2011 at 12:55:39PM +0100, Jan Kiszka wrote:
>> Only for walking the list of VMs, we do not need to hold the preemption
>> disabling kvm_lock. Convert stat services, the cpufreq callback and
>> mmu_shrink to RCU. For the latter, special c
On Tue, Feb 08, 2011 at 12:55:39PM +0100, Jan Kiszka wrote:
> Only for walking the list of VMs, we do not need to hold the preemption
> disabling kvm_lock. Convert stat services, the cpufreq callback and
> mmu_shrink to RCU. For the latter, special care is required to
> synchronize its list_move_ta
On 2011-02-10 15:47, Avi Kivity wrote:
> On 02/10/2011 04:34 PM, Jan Kiszka wrote:
>> On 2011-02-10 15:26, Avi Kivity wrote:
>>> On 02/10/2011 03:47 PM, Jan Kiszka wrote:
>>
>> Accept for mmu_shrink, which is write but not delete, thus works
>> without
>> that slow synchronize
On 02/10/2011 04:34 PM, Jan Kiszka wrote:
On 2011-02-10 15:26, Avi Kivity wrote:
> On 02/10/2011 03:47 PM, Jan Kiszka wrote:
Accept for mmu_shrink, which is write but not delete, thus works without
that slow synchronize_rcu.
>>>
>>> I don't really see how you can implement l
On 2011-02-10 15:26, Avi Kivity wrote:
> On 02/10/2011 03:47 PM, Jan Kiszka wrote:
Accept for mmu_shrink, which is write but not delete, thus works without
that slow synchronize_rcu.
>>>
>>> I don't really see how you can implement list_move_rcu(), it has to be
>>> atomic or othe
On 02/10/2011 03:47 PM, Jan Kiszka wrote:
>>
>> Accept for mmu_shrink, which is write but not delete, thus works without
>> that slow synchronize_rcu.
>
> I don't really see how you can implement list_move_rcu(), it has to be
> atomic or other users will see a partial vm_list.
Right, even if
On 2011-02-10 14:19, Avi Kivity wrote:
> On 02/10/2011 03:14 PM, Jan Kiszka wrote:
>> On 2011-02-10 13:57, Avi Kivity wrote:
>>> On 02/10/2011 02:56 PM, Avi Kivity wrote:
> What's the benefit? The downside is a bit more complexity as you need an
> additional callback handler.
>
On 02/10/2011 03:14 PM, Jan Kiszka wrote:
On 2011-02-10 13:57, Avi Kivity wrote:
> On 02/10/2011 02:56 PM, Avi Kivity wrote:
>>> What's the benefit? The downside is a bit more complexity as you need an
>>> additional callback handler.
>>
>>
>> synchronize_rcu() can be very slow (its a systemw
On 2011-02-10 13:57, Avi Kivity wrote:
> On 02/10/2011 02:56 PM, Avi Kivity wrote:
>>> What's the benefit? The downside is a bit more complexity as you need an
>>> additional callback handler.
>>
>>
>> synchronize_rcu() can be very slow (its a systemwide operation), and
>> mmu_shrink() can be call
On 02/10/2011 02:56 PM, Avi Kivity wrote:
What's the benefit? The downside is a bit more complexity as you need an
additional callback handler.
synchronize_rcu() can be very slow (its a systemwide operation), and
mmu_shrink() can be called often on a loaded system.
In fact this just shows
On 02/10/2011 02:45 PM, Jan Kiszka wrote:
>>>
>>> There is no list_move_tail_rcu().
>>
>> ...specifically not for this one.
>
> Well, we can add one if needed (and if possible).
I can have a look, at least at the lower hanging fruits.
Please keep rcu->parent in the loop.
>
>>>
>>> Why
On 2011-02-10 13:34, Avi Kivity wrote:
> On 02/10/2011 01:31 PM, Jan Kiszka wrote:
>>>
@@ -3607,10 +3607,14 @@ static int mmu_shrink(struct shrinker *shrink,
int nr_to_scan, gfp_t gfp_mask)
spin_unlock(&kvm->mmu_lock);
srcu_read_unloc
On 02/10/2011 01:31 PM, Jan Kiszka wrote:
>
>>
>> @@ -3607,10 +3607,14 @@ static int mmu_shrink(struct shrinker *shrink, int
nr_to_scan, gfp_t gfp_mask)
>>spin_unlock(&kvm->mmu_lock);
>>srcu_read_unlock(&kvm->srcu, idx);
>>}
>> - if (kvm_freed)
>> -
On 2011-02-10 11:16, Avi Kivity wrote:
> On 02/08/2011 01:55 PM, Jan Kiszka wrote:
>> Only for walking the list of VMs, we do not need to hold the preemption
>> disabling kvm_lock. Convert stat services, the cpufreq callback and
>> mmu_shrink to RCU. For the latter, special care is required to
>> s
On 02/08/2011 01:55 PM, Jan Kiszka wrote:
Only for walking the list of VMs, we do not need to hold the preemption
disabling kvm_lock. Convert stat services, the cpufreq callback and
mmu_shrink to RCU. For the latter, special care is required to
synchronize its list_move_tail with kvm_destroy_vm.
15 matches
Mail list logo
