x86/iommu: avoid MSI address and data writes if IRT index hasn't changed

Attempt to reduce the MSI entry writes, and the associated checking whether
memory decoding and MSI-X is enabled for the PCI device, when the MSI data
hasn't changed.

When using Interrupt Remapping the MSI entry will contain an index into
the remapping table, and it's in such remapping table where the MSI vector
and destination CPU is stored.  As such, when using interrupt remapping,
changes to the interrupt affinity shouldn't result in changes to the MSI
entry, and the MSI entry update can be avoided.

Signal from the IOMMU update_ire_from_msi hook whether the MSI data or
address fields have changed, and thus need writing to the device registers.
Such signaling is done by returning 1 from the function.  Otherwise
returning 0 means no update of the MSI fields, and thus no write
required.

Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>

diff --git a/xen/arch/x86/hpet.c b/xen/arch/x86/hpet.c
index 51ff7f12f5c0af759401e1939fd3065d971cc5c4..1bca8c8b670d0e1bc930b2b892e54e254bda103d 100644 (file)
--- a/xen/arch/x86/hpet.c
+++ b/xen/arch/x86/hpet.c
@@ -283,8 +283,12 @@ static int hpet_msi_write(struct hpet_event_channel *ch, struct msi_msg *msg)
     {
         int rc = iommu_update_ire_from_msi(&ch->msi, msg);
 
-        if ( rc )
+        if ( rc < 0 )
             return rc;
+        /*
+         * Always propagate writes, to avoid having to pass a flag for handling
+         * a forceful write in the resume from suspension case.
+         */
     }
 
     hpet_write32(msg->data, HPET_Tn_ROUTE(ch->idx));

diff --git a/xen/arch/x86/hvm/vmx/vmx.c b/xen/arch/x86/hvm/vmx/vmx.c
index 23b7ecd77f85b6b2c90d9c7d057868846c14a63f..4883bd823da12ed8ef1bf3991f80576ececf3bcd 100644 (file)
--- a/xen/arch/x86/hvm/vmx/vmx.c
+++ b/xen/arch/x86/hvm/vmx/vmx.c
@@ -417,7 +417,18 @@ static int cf_check vmx_pi_update_irte(const struct vcpu *v,
 
     ASSERT_PDEV_LIST_IS_READ_LOCKED(msi_desc->dev->domain);
 
-    return iommu_update_ire_from_msi(msi_desc, &msg);
+    rc = iommu_update_ire_from_msi(msi_desc, &msg);
+    if ( rc > 0 )
+    {
+        /*
+         * Callers of vmx_pi_update_irte() won't propagate the updated MSI
+         * fields to the hardware, must assert there are no changes.
+         */
+        ASSERT_UNREACHABLE();
+        rc = -EILSEQ;
+    }
+
+    return rc;
 
  unlock_out:
     spin_unlock_irq(&desc->lock);

diff --git a/xen/arch/x86/msi.c b/xen/arch/x86/msi.c
index 6c11d76015fb32072dccc78b345272583b9195b9..163ccf8747206dbbf995b2c2be91b54362415e29 100644 (file)
--- a/xen/arch/x86/msi.c
+++ b/xen/arch/x86/msi.c
@@ -184,7 +184,8 @@ void msi_compose_msg(unsigned vector, const cpumask_t *cpu_mask, struct msi_msg
                 MSI_DATA_VECTOR(vector);
 }
 
-static int write_msi_msg(struct msi_desc *entry, struct msi_msg *msg)
+static int write_msi_msg(struct msi_desc *entry, struct msi_msg *msg,
+                         bool force)
 {
     entry->msg = *msg;
 
@@ -194,7 +195,7 @@ static int write_msi_msg(struct msi_desc *entry, struct msi_msg *msg)
 
         ASSERT(msg != &entry->msg);
         rc = iommu_update_ire_from_msi(entry, msg);
-        if ( rc )
+        if ( rc < 0 || (rc == 0 && !force) )
             return rc;
     }
 
@@ -259,7 +260,7 @@ void cf_check set_msi_affinity(struct irq_desc *desc, const cpumask_t *mask)
     msg.address_lo |= MSI_ADDR_DEST_ID(dest);
     msg.dest32 = dest;
 
-    write_msi_msg(msi_desc, &msg);
+    write_msi_msg(msi_desc, &msg, false);
 }
 
 void __msi_set_enable(pci_sbdf_t sbdf, int pos, int enable)
@@ -522,7 +523,7 @@ int __setup_msi_irq(struct irq_desc *desc, struct msi_desc *msidesc,
     desc->msi_desc = msidesc;
     desc->handler = handler;
     msi_compose_msg(desc->arch.vector, desc->arch.cpu_mask, &msg);
-    ret = write_msi_msg(msidesc, &msg);
+    ret = write_msi_msg(msidesc, &msg, false);
     if ( unlikely(ret) )
     {
         desc->handler = &no_irq_type;
@@ -1403,7 +1404,7 @@ int pci_restore_msi_state(struct pci_dev *pdev)
         type = entry->msi_attrib.type;
 
         msg = entry->msg;
-        write_msi_msg(entry, &msg);
+        write_msi_msg(entry, &msg, true);
 
         for ( i = 0; ; )
         {

diff --git a/xen/drivers/passthrough/amd/iommu_intr.c b/xen/drivers/passthrough/amd/iommu_intr.c
index c0273059cb1de772f98f99a2dae36964c5a0f26c..9abdc38053d70fc184a2584930e9cc1ee2530d45 100644 (file)
--- a/xen/drivers/passthrough/amd/iommu_intr.c
+++ b/xen/drivers/passthrough/amd/iommu_intr.c
@@ -492,7 +492,7 @@ static int update_intremap_entry_from_msi_msg(
                get_ivrs_mappings(iommu->seg)[alias_id].intremap_table);
     }
 
-    return 0;
+    return fresh;
 }
 
 static struct amd_iommu *_find_iommu_for_device(int seg, int bdf)
@@ -546,7 +546,7 @@ int cf_check amd_iommu_msi_msg_update_ire(
     rc = update_intremap_entry_from_msi_msg(iommu, bdf, nr,
                                             &msi_desc->remap_index,
                                             msg, &data);
-    if ( !rc )
+    if ( rc > 0 )
     {
         for ( i = 1; i < nr; ++i )
             msi_desc[i].remap_index = msi_desc->remap_index + i;

diff --git a/xen/drivers/passthrough/vtd/intremap.c b/xen/drivers/passthrough/vtd/intremap.c
index 1aeaeb5ec595a8e74baf2062511b8c23ddb3e568..b3b53518e294405b03f165a7a4bdab7a42ad09e6 100644 (file)
--- a/xen/drivers/passthrough/vtd/intremap.c
+++ b/xen/drivers/passthrough/vtd/intremap.c
@@ -506,6 +506,7 @@ static int msi_msg_to_remap_entry(
     unsigned int index, i, nr = 1;
     unsigned long flags;
     const struct pi_desc *pi_desc = msi_desc->pi_desc;
+    bool alloc = false;
 
     if ( msi_desc->msi_attrib.type == PCI_CAP_ID_MSI )
         nr = msi_desc->msi.nvec;
@@ -529,6 +530,7 @@ static int msi_msg_to_remap_entry(
         index = alloc_remap_entry(iommu, nr);
         for ( i = 0; i < nr; ++i )
             msi_desc[i].remap_index = index + i;
+        alloc = true;
     }
     else
         index = msi_desc->remap_index;
@@ -601,7 +603,7 @@ static int msi_msg_to_remap_entry(
     unmap_vtd_domain_page(iremap_entries);
     spin_unlock_irqrestore(&iommu->intremap.lock, flags);
 
-    return 0;
+    return alloc;
 }
 
 int cf_check msi_msg_write_remap_rte(

diff --git a/xen/include/xen/iommu.h b/xen/include/xen/iommu.h
index 77a514019cc6169594b0167ddb4d30a73d643125..984f0735d4a9bb1a4f8b343ad6e5f5b9b430530a 100644 (file)
--- a/xen/include/xen/iommu.h
+++ b/xen/include/xen/iommu.h
@@ -435,6 +435,12 @@ extern struct page_list_head iommu_pt_cleanup_list;
 bool arch_iommu_use_permitted(const struct domain *d);
 
 #ifdef CONFIG_X86
+/*
+ * Return values:
+ *  - < 0 on error.
+ *  - 0 on success and no need to write msi_msg to the hardware.
+ *  - 1 on success and msi_msg must be propagated to the hardware.
+ */
 static inline int iommu_update_ire_from_msi(
     struct msi_desc *msi_desc, struct msi_msg *msg)
 {