[PATCH] x86/hap: Defer P2M TLB flushes

[PATCH] x86/hap: Defer P2M TLB flushes

[Ross Lagerwall]

Like the EPT code, defer TLB flushes to reduce the number of flushes and
avoid holding the P2M lock while flushing. This can substantially
improve performance in some scenarios.

The cases where the TLB needs to be flushed without deferring are
already handled by the call to p2m_tlb_flush_sync() in p2m_free_ptp().

Suggested-by: Roger Pau Monne <roger.pau@citrix.com>
Signed-off-by: Ross Lagerwall <ross.lagerwall@citrix.com>
---
 xen/arch/x86/mm/hap/hap.c | 14 ++++++++++++--
 1 file changed, 12 insertions(+), 2 deletions(-)

diff --git a/xen/arch/x86/mm/hap/hap.c b/xen/arch/x86/mm/hap/hap.c
index a337752bf488..1eba995c7b39 100644
--- a/xen/arch/x86/mm/hap/hap.c
+++ b/xen/arch/x86/mm/hap/hap.c
@@ -813,16 +813,26 @@ static void cf_check hap_update_paging_modes(struct vcpu *v)
 
 static void cf_check
 hap_write_p2m_entry_post(struct p2m_domain *p2m, unsigned int oflags)
+{
+    if ( oflags & _PAGE_PRESENT )
+    {
+        ASSERT(p2m->defer_flush);
+        p2m->need_flush = true;
+    }
+}
+
+static void cf_check
+hap_p2m_tlb_flush(struct p2m_domain *p2m)
 {
     struct domain *d = p2m->domain;
 
-    if ( oflags & _PAGE_PRESENT )
-        guest_flush_tlb_mask(d, d->dirty_cpumask);
+    guest_flush_tlb_mask(d, d->dirty_cpumask);
 }
 
 void hap_p2m_init(struct p2m_domain *p2m)
 {
     p2m->write_p2m_entry_post = hap_write_p2m_entry_post;
+    p2m->tlb_flush = hap_p2m_tlb_flush;
 }
 
 static unsigned long cf_check hap_gva_to_gfn_real_mode(
-- 
2.53.0

Re: [PATCH] x86/hap: Defer P2M TLB flushes

[Roger Pau Monné]

The subject should possibly mention NPT, as reading HAP one would
assume it applies to both EPT and NPT (and the EPT side is already
done here):

x86/hap: Defer NPT P2M TLB flushes

On Wed, Mar 18, 2026 at 05:13:24PM +0000, Ross Lagerwall wrote:
> Like the EPT code, defer TLB flushes to reduce the number of flushes and
> avoid holding the P2M lock while flushing. This can substantially
> improve performance in some scenarios.
> 
> The cases where the TLB needs to be flushed without deferring are
> already handled by the call to p2m_tlb_flush_sync() in p2m_free_ptp().
> 
> Suggested-by: Roger Pau Monne <roger.pau@citrix.com>
> Signed-off-by: Ross Lagerwall <ross.lagerwall@citrix.com>
> ---
>  xen/arch/x86/mm/hap/hap.c | 14 ++++++++++++--
>  1 file changed, 12 insertions(+), 2 deletions(-)
> 
> diff --git a/xen/arch/x86/mm/hap/hap.c b/xen/arch/x86/mm/hap/hap.c
> index a337752bf488..1eba995c7b39 100644
> --- a/xen/arch/x86/mm/hap/hap.c
> +++ b/xen/arch/x86/mm/hap/hap.c
> @@ -813,16 +813,26 @@ static void cf_check hap_update_paging_modes(struct vcpu *v)
>  
>  static void cf_check
>  hap_write_p2m_entry_post(struct p2m_domain *p2m, unsigned int oflags)
> +{
> +    if ( oflags & _PAGE_PRESENT )
> +    {
> +        ASSERT(p2m->defer_flush);

Here I'm unsure whether we might want to do:

if ( !p2m->defer_flush )
{
    ASSERT_UNREACHABLE();
    guest_flush_tlb_mask(d, d->dirty_cpumask);
}

Instead of just asserting that the defer_flush field is set, to be on
the safe(r) side.

Thanks, Roger.