hvm: passthrough MSI-X mask bit acceleration

Add a new parameter to DOMCTL_bind_pt_irq to allow Xen to know the
guest physical address of MSI-X table. Also add a new MMIO intercept
handler to intercept that gpa in order to handle MSI-X vector mask
bit operation in the hypervisor. This reduces the load of device model
considerably if the guest does mask and unmask frequently

Signed-off-by: Qing He <qing.he@intel.com>

11 files changed, 331 insertions, 5 deletions

diff --git a/tools/libxc/xc_domain.c b/tools/libxc/xc_domain.c
index 5e977307cf..6ed4f52ec3 100644
--- a/tools/libxc/xc_domain.c
+++ b/tools/libxc/xc_domain.c
@@ -920,7 +920,8 @@ int xc_domain_update_msi_irq(
     uint32_t domid,
     uint32_t gvec,
     uint32_t pirq,
-    uint32_t gflags)
+    uint32_t gflags,
+    uint64_t gtable)
 {
     int rc;
     xen_domctl_bind_pt_irq_t *bind;
@@ -936,6 +937,7 @@ int xc_domain_update_msi_irq(
     bind->machine_irq = pirq;
     bind->u.msi.gvec = gvec;
     bind->u.msi.gflags = gflags;
+    bind->u.msi.gtable = gtable;
 
     rc = do_domctl(xc_handle, &domctl);
     return rc;
diff --git a/tools/libxc/xenctrl.h b/tools/libxc/xenctrl.h
index dedaa0dd69..93fbd8be67 100644
--- a/tools/libxc/xenctrl.h
+++ b/tools/libxc/xenctrl.h
@@ -1092,7 +1092,8 @@ int xc_domain_update_msi_irq(
     uint32_t domid,
     uint32_t gvec,
     uint32_t pirq,
-    uint32_t gflags);
+    uint32_t gflags,
+    uint64_t gtable);
 
 int xc_domain_unbind_msi_irq(int xc_handle,
                              uint32_t domid,
diff --git a/xen/arch/x86/hvm/hvm.c b/xen/arch/x86/hvm/hvm.c
index ae4acb3787..c19d36a9d3 100644
--- a/xen/arch/x86/hvm/hvm.c
+++ b/xen/arch/x86/hvm/hvm.c
@@ -308,6 +308,9 @@ int hvm_domain_initialise(struct domain *d)
     spin_lock_init(&d->arch.hvm_domain.irq_lock);
     spin_lock_init(&d->arch.hvm_domain.uc_lock);
 
+    INIT_LIST_HEAD(&d->arch.hvm_domain.msixtbl_list);
+    spin_lock_init(&d->arch.hvm_domain.msixtbl_list_lock);
+
     hvm_init_guest_time(d);
 
     d->arch.hvm_domain.params[HVM_PARAM_HPET_ENABLED] = 1;
@@ -348,11 +351,15 @@ int hvm_domain_initialise(struct domain *d)
     return rc;
 }
 
+extern void msixtbl_pt_cleanup(struct domain *d);
+
 void hvm_domain_relinquish_resources(struct domain *d)
 {
     hvm_destroy_ioreq_page(d, &d->arch.hvm_domain.ioreq);
     hvm_destroy_ioreq_page(d, &d->arch.hvm_domain.buf_ioreq);
 
+    msixtbl_pt_cleanup(d);
+
     /* Stop all asynchronous timer actions. */
     rtc_deinit(d);
     if ( d->vcpu[0] != NULL )
diff --git a/xen/arch/x86/hvm/intercept.c b/xen/arch/x86/hvm/intercept.c
index 0e110e00dc..107b87d2f7 100644
--- a/xen/arch/x86/hvm/intercept.c
+++ b/xen/arch/x86/hvm/intercept.c
@@ -35,14 +35,16 @@
 extern struct hvm_mmio_handler hpet_mmio_handler;
 extern struct hvm_mmio_handler vlapic_mmio_handler;
 extern struct hvm_mmio_handler vioapic_mmio_handler;
+extern struct hvm_mmio_handler msixtbl_mmio_handler;
 
-#define HVM_MMIO_HANDLER_NR 3
+#define HVM_MMIO_HANDLER_NR 4
 
 static struct hvm_mmio_handler *hvm_mmio_handlers[HVM_MMIO_HANDLER_NR] =
 {
     &hpet_mmio_handler,
     &vlapic_mmio_handler,
-    &vioapic_mmio_handler
+    &vioapic_mmio_handler,
+    &msixtbl_mmio_handler
 };
 
 static int hvm_mmio_access(struct vcpu *v,
diff --git a/xen/arch/x86/hvm/vmsi.c b/xen/arch/x86/hvm/vmsi.c
index 65682894e9..538637b452 100644
--- a/xen/arch/x86/hvm/vmsi.c
+++ b/xen/arch/x86/hvm/vmsi.c
@@ -193,3 +193,283 @@ int vmsi_deliver(struct domain *d, int pirq)
     return 1;
 }
 
+/* MSI-X mask bit hypervisor interception */
+struct msixtbl_entry
+{
+    struct list_head list;
+    atomic_t refcnt;    /* how many bind_pt_irq called for the device */
+
+    /* TODO: resolve the potential race by destruction of pdev */
+    struct pci_dev *pdev;
+    unsigned long gtable;       /* gpa of msix table */
+    unsigned long table_len;
+    unsigned long table_flags[MAX_MSIX_TABLE_ENTRIES / BITS_PER_LONG + 1];
+
+    struct rcu_head rcu;
+};
+
+static struct msixtbl_entry *msixtbl_find_entry(
+    struct vcpu *v, unsigned long addr)
+{
+    struct msixtbl_entry *entry;
+    struct domain *d = v->domain;
+
+    list_for_each_entry( entry, &d->arch.hvm_domain.msixtbl_list, list )
+        if ( addr >= entry->gtable &&
+             addr < entry->gtable + entry->table_len )
+            return entry;
+
+    return NULL;
+}
+
+static void __iomem *msixtbl_addr_to_virt(
+    struct msixtbl_entry *entry, unsigned long addr)
+{
+    int idx, nr_page;
+
+    if ( !entry )
+        return NULL;
+
+    nr_page = (addr >> PAGE_SHIFT) -
+              (entry->gtable >> PAGE_SHIFT);
+
+    if ( !entry->pdev )
+        return NULL;
+
+    idx = entry->pdev->msix_table_idx[nr_page];
+    if ( !idx )
+        return NULL;
+
+    return (void *)(fix_to_virt(idx) +
+                    (addr & ((1UL << PAGE_SHIFT) - 1)));
+}
+
+static int msixtbl_read(
+    struct vcpu *v, unsigned long address,
+    unsigned long len, unsigned long *pval)
+{
+    unsigned long offset;
+    struct msixtbl_entry *entry;
+    void *virt;
+    int r = X86EMUL_UNHANDLEABLE;
+
+    rcu_read_lock();
+
+    if ( len != 4 )
+        goto out;
+
+    offset = address & (PCI_MSIX_ENTRY_SIZE - 1);
+    if ( offset != PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET)
+        goto out;
+
+    entry = msixtbl_find_entry(v, address);
+    virt = msixtbl_addr_to_virt(entry, address);
+    if ( !virt )
+        goto out;
+
+    *pval = readl(virt);
+    r = X86EMUL_OKAY;
+
+out:
+    rcu_read_unlock();
+    return r;
+}
+
+static int msixtbl_write(struct vcpu *v, unsigned long address,
+                        unsigned long len, unsigned long val)
+{
+    unsigned long offset;
+    struct msixtbl_entry *entry;
+    void *virt;
+    int nr_entry;
+    int r = X86EMUL_UNHANDLEABLE;
+
+    rcu_read_lock();
+
+    if ( len != 4 )
+        goto out;
+
+    entry = msixtbl_find_entry(v, address);
+    nr_entry = (address - entry->gtable) % PCI_MSIX_ENTRY_SIZE;
+
+    offset = address & (PCI_MSIX_ENTRY_SIZE - 1);
+    if ( offset != PCI_MSIX_ENTRY_VECTOR_CTRL_OFFSET)
+    {
+        set_bit(nr_entry, &entry->table_flags);
+        goto out;
+    }
+
+    /* exit to device model if address/data has been modified */
+    if ( test_and_clear_bit(nr_entry, &entry->table_flags) )
+        goto out;
+
+    virt = msixtbl_addr_to_virt(entry, address);
+    if ( !virt )
+        goto out;
+
+    writel(val, virt);
+    r = X86EMUL_OKAY;
+
+out:
+    rcu_read_unlock();
+    return r;
+}
+
+static int msixtbl_range(struct vcpu *v, unsigned long addr)
+{
+    struct msixtbl_entry *entry;
+    void *virt;
+
+    rcu_read_lock();
+
+    entry = msixtbl_find_entry(v, addr);
+    virt = msixtbl_addr_to_virt(entry, addr);
+
+    rcu_read_unlock();
+
+    return !!virt;
+}
+
+struct hvm_mmio_handler msixtbl_mmio_handler = {
+    .check_handler = msixtbl_range,
+    .read_handler = msixtbl_read,
+    .write_handler = msixtbl_write
+};
+
+static struct msixtbl_entry *add_msixtbl_entry(struct domain *d,
+                                               struct pci_dev *pdev,
+                                               uint64_t gtable)
+{
+    struct msixtbl_entry *entry;
+    u32 len;
+
+    entry = xmalloc(struct msixtbl_entry);
+    if ( !entry )
+        return NULL;
+
+    memset(entry, 0, sizeof(struct msixtbl_entry));
+        
+    INIT_LIST_HEAD(&entry->list);
+    INIT_RCU_HEAD(&entry->rcu);
+    atomic_set(&entry->refcnt, 0);
+
+    len = pci_msix_get_table_len(pdev);
+    entry->table_len = len;
+    entry->pdev = pdev;
+    entry->gtable = (unsigned long) gtable;
+
+    list_add_rcu(&entry->list, &d->arch.hvm_domain.msixtbl_list);
+
+    return entry;
+}
+
+static void free_msixtbl_entry(struct rcu_head *rcu)
+{
+    struct msixtbl_entry *entry;
+
+    entry = container_of (rcu, struct msixtbl_entry, rcu);
+
+    xfree(entry);
+}
+
+static void del_msixtbl_entry(struct msixtbl_entry *entry)
+{
+    list_del_rcu(&entry->list);
+    call_rcu(&entry->rcu, free_msixtbl_entry);
+}
+
+int msixtbl_pt_register(struct domain *d, int pirq, uint64_t gtable)
+{
+    irq_desc_t *irq_desc;
+    struct msi_desc *msi_desc;
+    struct pci_dev *pdev;
+    struct msixtbl_entry *entry;
+    int r = -EINVAL;
+
+    /* pcidevs_lock already held */
+    irq_desc = domain_spin_lock_irq_desc(d, pirq, NULL);
+
+    if ( irq_desc->handler != &pci_msi_type )
+        goto out;
+
+    msi_desc = irq_desc->msi_desc;
+    if ( !msi_desc )
+        goto out;
+
+    pdev = msi_desc->dev;
+
+    spin_lock(&d->arch.hvm_domain.msixtbl_list_lock);
+
+    list_for_each_entry( entry, &d->arch.hvm_domain.msixtbl_list, list )
+        if ( pdev == entry->pdev )
+            goto found;
+
+    entry = add_msixtbl_entry(d, pdev, gtable);
+    if ( !entry )
+    {
+        spin_unlock(&d->arch.hvm_domain.msixtbl_list_lock);
+        goto out;
+    }
+
+found:
+    atomic_inc(&entry->refcnt);
+
+    spin_unlock(&d->arch.hvm_domain.msixtbl_list_lock);
+
+out:
+    spin_unlock_irq(&irq_desc->lock);
+    return r;
+
+}
+
+void msixtbl_pt_unregister(struct domain *d, int pirq)
+{
+    irq_desc_t *irq_desc;
+    struct msi_desc *msi_desc;
+    struct pci_dev *pdev;
+    struct msixtbl_entry *entry;
+
+    /* pcidevs_lock already held */
+    irq_desc = domain_spin_lock_irq_desc(d, pirq, NULL);
+
+    if ( irq_desc->handler != &pci_msi_type )
+        goto out;
+
+    msi_desc = irq_desc->msi_desc;
+    if ( !msi_desc )
+        goto out;
+
+    pdev = msi_desc->dev;
+
+    spin_lock(&d->arch.hvm_domain.msixtbl_list_lock);
+
+    list_for_each_entry( entry, &d->arch.hvm_domain.msixtbl_list, list )
+        if ( pdev == entry->pdev )
+            goto found;
+
+    spin_unlock(&d->arch.hvm_domain.msixtbl_list_lock);
+
+
+out:
+    spin_unlock(&irq_desc->lock);
+    return;
+
+found:
+    if ( !atomic_dec_and_test(&entry->refcnt) )
+        del_msixtbl_entry(entry);
+
+    spin_unlock(&d->arch.hvm_domain.msixtbl_list_lock);
+    spin_unlock(&irq_desc->lock);
+}
+void msixtbl_pt_cleanup(struct domain *d, int pirq)
+{
+    struct msixtbl_entry *entry, *temp;
+
+    spin_lock(&d->arch.hvm_domain.msixtbl_list_lock);
+
+    list_for_each_entry_safe( entry, temp,
+                              &d->arch.hvm_domain.msixtbl_list, list )
+        del_msixtbl_entry(entry);
+
+    spin_unlock(&d->arch.hvm_domain.msixtbl_list_lock);
+}
diff --git a/xen/arch/x86/msi.c b/xen/arch/x86/msi.c
index dc136f7f62..0c753f22aa 100644
--- a/xen/arch/x86/msi.c
+++ b/xen/arch/x86/msi.c
@@ -839,3 +839,23 @@ int pci_restore_msi_state(struct pci_dev *pdev)
     return 0;
 }
 
+unsigned int pci_msix_get_table_len(struct pci_dev *pdev)
+{
+    int pos;
+    u16 control;
+    u8 bus, slot, func;
+    unsigned int len;
+
+    bus = pdev->bus;
+    slot = PCI_SLOT(pdev->devfn);
+    func = PCI_FUNC(pdev->devfn);
+
+    pos = pci_find_cap_offset(bus, slot, func, PCI_CAP_ID_MSIX);
+    if ( !pos )
+        return 0;
+
+    control = pci_conf_read16(bus, slot, func, msix_control_reg(pos));
+    len = msix_table_size(control) * PCI_MSIX_ENTRY_SIZE;
+
+    return len;
+}
diff --git a/xen/drivers/passthrough/io.c b/xen/drivers/passthrough/io.c
index dce6814a25..40ed8ed12c 100644
--- a/xen/drivers/passthrough/io.c
+++ b/xen/drivers/passthrough/io.c
@@ -58,6 +58,9 @@ static void pt_irq_time_out(void *data)
     pirq_guest_eoi(irq_map->dom, machine_gsi);
 }
 
+extern int msixtbl_pt_register(struct domain *d, int pirq, uint64_t gtable);
+extern int msixtbl_pt_unregister(struct domain *d, int pirq);
+
 int pt_irq_create_bind_vtd(
     struct domain *d, xen_domctl_bind_pt_irq_t *pt_irq_bind)
 {
@@ -115,6 +118,8 @@ int pt_irq_create_bind_vtd(
                 spin_unlock(&d->event_lock);
                 return rc;
             }
+            if ( pt_irq_bind->u.msi.gtable )
+                msixtbl_pt_register(d, pirq, pt_irq_bind->u.msi.gtable);
         }
         else if (hvm_irq_dpci->mirq[pirq].gmsi.gvec != pt_irq_bind->u.msi.gvec
                 ||hvm_irq_dpci->msi_gvec_pirq[pt_irq_bind->u.msi.gvec] != pirq)
@@ -259,6 +264,7 @@ int pt_irq_destroy_bind_vtd(
         if ( list_empty(&hvm_irq_dpci->mirq[machine_gsi].digl_list) )
         {
             pirq_guest_unbind(d, machine_gsi);
+            msixtbl_pt_unregister(d, machine_gsi);
             if ( pt_irq_need_timer(hvm_irq_dpci->mirq[machine_gsi].flags) )
                 kill_timer(&hvm_irq_dpci->hvm_timer[domain_irq_to_vector(d, machine_gsi)]);
             hvm_irq_dpci->mirq[machine_gsi].dom   = NULL;
diff --git a/xen/include/asm-x86/hvm/domain.h b/xen/include/asm-x86/hvm/domain.h
index 475c031df0..70e38bff80 100644
--- a/xen/include/asm-x86/hvm/domain.h
+++ b/xen/include/asm-x86/hvm/domain.h
@@ -75,6 +75,10 @@ struct hvm_domain {
     /* Pass-through */
     struct hvm_iommu       hvm_iommu;
 
+    /* hypervisor intercepted msix table */
+    struct list_head       msixtbl_list;
+    spinlock_t             msixtbl_list_lock;
+
     struct viridian_domain viridian;
 
     bool_t                 hap_enabled;
diff --git a/xen/include/asm-x86/msi.h b/xen/include/asm-x86/msi.h
index a5f67e1655..56387c6004 100644
--- a/xen/include/asm-x86/msi.h
+++ b/xen/include/asm-x86/msi.h
@@ -81,6 +81,8 @@ extern void teardown_msi_vector(int vector);
 extern int msi_free_vector(struct msi_desc *entry);
 extern int pci_restore_msi_state(struct pci_dev *pdev);
 
+extern unsigned int pci_msix_get_table_len(struct pci_dev *pdev);
+
 struct msi_desc {
 	struct {
 		__u8	type	: 5; 	/* {0: unused, 5h:MSI, 11h:MSI-X} */
diff --git a/xen/include/public/domctl.h b/xen/include/public/domctl.h
index 042ccf185c..c2845e12f3 100644
--- a/xen/include/public/domctl.h
+++ b/xen/include/public/domctl.h
@@ -485,6 +485,7 @@ struct xen_domctl_bind_pt_irq {
         struct {
             uint8_t gvec;
             uint32_t gflags;
+            uint64_t gtable;
         } msi;
     } u;
 };
diff --git a/xen/include/xen/pci.h b/xen/include/xen/pci.h
index 8ed5058e99..c60b8e3691 100644
--- a/xen/include/xen/pci.h
+++ b/xen/include/xen/pci.h
@@ -29,7 +29,8 @@
 #define PCI_BDF(b,d,f)  ((((b) & 0xff) << 8) | PCI_DEVFN(d,f))
 #define PCI_BDF2(b,df)  ((((b) & 0xff) << 8) | ((df) & 0xff))
 
-#define MAX_MSIX_TABLE_PAGES    8    /* 2048 entries */
+#define MAX_MSIX_TABLE_ENTRIES  2048
+#define MAX_MSIX_TABLE_PAGES    8
 struct pci_dev {
     struct list_head alldevs_list;
     struct list_head domain_list;