[PATCH v4 0/2] x86/hvm/viridian: APIC assist

[PATCH v4 0/2] x86/hvm/viridian: APIC assist

[Paul Durrant]

This patch series enables use of the 'APIC assist' enlightenment in Xen.

See section 13.3.4.1 of the Microsoft Hypervisor Top Level Function
Specification v4.0b at:

https://msdn.microsoft.com/en-us/virtualization/hyperv_on_windows/develop/tlfs

for more information.

Patch #1 modifies the viridian code to keep the guest APIC assist pages
mapped for the lifetime of the domain (and incorporates a straightforward
data-structure change).

Patch #2 adds a new viridian code and a configuration option to the
toolstack that, when enabled, allows the local APIC emulation to make use
the APIC assist pages.


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

[PATCH v5 1/2] x86/hvm/viridian: keep APIC assist page mapped...

[Paul Durrant]

... for the lifetime of the domain.

If Xen is to make use of the APIC assist enlightenment then a persistent
mapping needs to be kept, rather than the temporary one which is currently
used only to initialize the page content.

This patch also adds a comment block at the top of the source with
information on the latest version of the spec. from Microsoft and the
current URL where it may be found.

Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
Cc: Keir Fraser <keir@xen.org>
Cc: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
---

v5:
 - Get rid of the stored page_info pointer and instead derive it from
   the va on teardown.

v4:
 - Added missing teardown required if the MSR is re-written.

v2:
 - Re-instated warning if the initialization of an APIC assist page fails.
 - Added up-to-date information about the viridian specification to the
   comment block at the top of the source.
---
 xen/arch/x86/hvm/hvm.c             |  2 +
 xen/arch/x86/hvm/viridian.c        | 77 ++++++++++++++++++++++++++++----------
 xen/include/asm-x86/hvm/viridian.h |  7 +++-
 3 files changed, 65 insertions(+), 21 deletions(-)

diff --git a/xen/arch/x86/hvm/hvm.c b/xen/arch/x86/hvm/hvm.c
index 4ea51d7..f446ee4 100644
--- a/xen/arch/x86/hvm/hvm.c
+++ b/xen/arch/x86/hvm/hvm.c
@@ -2606,6 +2606,8 @@ int hvm_vcpu_initialise(struct vcpu *v)
 
 void hvm_vcpu_destroy(struct vcpu *v)
 {
+    viridian_vcpu_deinit(v);
+
     hvm_all_ioreq_servers_remove_vcpu(v->domain, v);
 
     if ( hvm_altp2m_supported() )
diff --git a/xen/arch/x86/hvm/viridian.c b/xen/arch/x86/hvm/viridian.c
index 1ee22aa..abbca3c 100644
--- a/xen/arch/x86/hvm/viridian.c
+++ b/xen/arch/x86/hvm/viridian.c
@@ -1,7 +1,12 @@
 /******************************************************************************
  * viridian.c
  *
- * An implementation of the Viridian hypercall interface.
+ * An implementation of some Viridian enlightenments. See Microsoft's
+ * Hypervisor Top Level Functional Specification (v4.0b) at:
+ *
+ * https://msdn.microsoft.com/en-us/virtualization/hyperv_on_windows/develop/tlfs
+ *
+ * for more information.
  */
 
 #include <xen/sched.h>
@@ -163,7 +168,7 @@ static void dump_apic_assist(const struct vcpu *v)
 {
     const union viridian_apic_assist *aa;
 
-    aa = &v->arch.hvm_vcpu.viridian.apic_assist;
+    aa = &v->arch.hvm_vcpu.viridian.apic_assist.msr;
 
     printk(XENLOG_G_INFO "%pv: VIRIDIAN APIC_ASSIST: enabled: %x pfn: %lx\n",
            v, aa->fields.enabled, (unsigned long)aa->fields.pfn);
@@ -217,9 +222,9 @@ static void enable_hypercall_page(struct domain *d)
 static void initialize_apic_assist(struct vcpu *v)
 {
     struct domain *d = v->domain;
-    unsigned long gmfn = v->arch.hvm_vcpu.viridian.apic_assist.fields.pfn;
+    unsigned long gmfn = v->arch.hvm_vcpu.viridian.apic_assist.msr.fields.pfn;
     struct page_info *page = get_page_from_gfn(d, gmfn, NULL, P2M_ALLOC);
-    uint8_t *p;
+    void *va;
 
     /*
      * We don't yet make use of the APIC assist page but by setting
@@ -227,25 +232,49 @@ static void initialize_apic_assist(struct vcpu *v)
      * bound to support the MSR. We therefore do just enough to keep windows
      * happy.
      *
-     * See http://msdn.microsoft.com/en-us/library/ff538657%28VS.85%29.aspx for
-     * details of how Windows uses the page.
+     * See section 13.3.4.1 of the specification for details of this
+     * enlightenment.
      */
 
-    if ( !page || !get_page_type(page, PGT_writable_page) )
+    if ( !page )
+        goto fail;
+
+    if ( !get_page_type(page, PGT_writable_page) )
     {
-        if ( page )
-            put_page(page);
-        gdprintk(XENLOG_WARNING, "Bad GMFN %lx (MFN %lx)\n", gmfn,
-                 page ? page_to_mfn(page) : INVALID_MFN);
-        return;
+        put_page(page);
+        goto fail;
     }
 
-    p = __map_domain_page(page);
+    va = __map_domain_page_global(page);
+    if ( !va )
+    {
+        put_page_and_type(page);
+        goto fail;
+    }
 
-    *(u32 *)p = 0;
+    *(uint32_t *)va = 0;
 
-    unmap_domain_page(p);
+    v->arch.hvm_vcpu.viridian.apic_assist.va = va;
+    return;
+
+ fail:
+    gdprintk(XENLOG_WARNING, "Bad GMFN %lx (MFN %lx)\n", gmfn,
+             page ? page_to_mfn(page) : INVALID_MFN);
+}
+
+static void teardown_apic_assist(struct vcpu *v)
+{
+    void *va = v->arch.hvm_vcpu.viridian.apic_assist.va;
+    struct page_info *page;
+
+    if ( !va )
+        return;
+
+    v->arch.hvm_vcpu.viridian.apic_assist.va = NULL;
+
+    page = mfn_to_page(domain_page_map_to_mfn(va));
 
+    unmap_domain_page_global(va);
     put_page_and_type(page);
 }
 
@@ -374,9 +403,10 @@ int wrmsr_viridian_regs(uint32_t idx, uint64_t val)
 
     case VIRIDIAN_MSR_APIC_ASSIST:
         perfc_incr(mshv_wrmsr_apic_msr);
-        v->arch.hvm_vcpu.viridian.apic_assist.raw = val;
+        teardown_apic_assist(v); /* release any previous mapping */
+        v->arch.hvm_vcpu.viridian.apic_assist.msr.raw = val;
         dump_apic_assist(v);
-        if (v->arch.hvm_vcpu.viridian.apic_assist.fields.enabled)
+        if ( v->arch.hvm_vcpu.viridian.apic_assist.msr.fields.enabled )
             initialize_apic_assist(v);
         break;
 
@@ -485,7 +515,7 @@ int rdmsr_viridian_regs(uint32_t idx, uint64_t *val)
 
     case VIRIDIAN_MSR_APIC_ASSIST:
         perfc_incr(mshv_rdmsr_apic_msr);
-        *val = v->arch.hvm_vcpu.viridian.apic_assist.raw;
+        *val = v->arch.hvm_vcpu.viridian.apic_assist.msr.raw;
         break;
 
     case VIRIDIAN_MSR_REFERENCE_TSC:
@@ -521,6 +551,11 @@ int rdmsr_viridian_regs(uint32_t idx, uint64_t *val)
     return 1;
 }
 
+void viridian_vcpu_deinit(struct vcpu *v)
+{
+    teardown_apic_assist(v);
+}
+
 static DEFINE_PER_CPU(cpumask_t, ipi_cpumask);
 
 int viridian_hypercall(struct cpu_user_regs *regs)
@@ -721,7 +756,7 @@ static int viridian_save_vcpu_ctxt(struct domain *d, hvm_domain_context_t *h)
     for_each_vcpu( d, v ) {
         struct hvm_viridian_vcpu_context ctxt;
 
-        ctxt.apic_assist = v->arch.hvm_vcpu.viridian.apic_assist.raw;
+        ctxt.apic_assist = v->arch.hvm_vcpu.viridian.apic_assist.msr.raw;
 
         if ( hvm_save_entry(VIRIDIAN_VCPU, v->vcpu_id, h, &ctxt) != 0 )
             return 1;
@@ -747,7 +782,9 @@ static int viridian_load_vcpu_ctxt(struct domain *d, hvm_domain_context_t *h)
     if ( hvm_load_entry(VIRIDIAN_VCPU, h, &ctxt) != 0 )
         return -EINVAL;
 
-    v->arch.hvm_vcpu.viridian.apic_assist.raw = ctxt.apic_assist;
+    v->arch.hvm_vcpu.viridian.apic_assist.msr.raw = ctxt.apic_assist;
+    if ( v->arch.hvm_vcpu.viridian.apic_assist.msr.fields.enabled )
+        initialize_apic_assist(v);
 
     return 0;
 }
diff --git a/xen/include/asm-x86/hvm/viridian.h b/xen/include/asm-x86/hvm/viridian.h
index c4319d7..19ecc06 100644
--- a/xen/include/asm-x86/hvm/viridian.h
+++ b/xen/include/asm-x86/hvm/viridian.h
@@ -21,7 +21,10 @@ union viridian_apic_assist
 
 struct viridian_vcpu
 {
-    union viridian_apic_assist apic_assist;
+    struct {
+        union viridian_apic_assist msr;
+        void *va;
+    } apic_assist;
 };
 
 union viridian_guest_os_id
@@ -117,6 +120,8 @@ viridian_hypercall(struct cpu_user_regs *regs);
 void viridian_time_ref_count_freeze(struct domain *d);
 void viridian_time_ref_count_thaw(struct domain *d);
 
+void viridian_vcpu_deinit(struct vcpu *v);
+
 #endif /* __ASM_X86_HVM_VIRIDIAN_H__ */
 
 /*
-- 
2.1.4


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

[PATCH v5 2/2] x86/hvm/viridian: Enable APIC assist enlightenment

[Paul Durrant]

This patch adds code to enable the APIC assist enlightenment which,
under certain conditions, means that the guest can avoid an EOI of
the local APIC and thereby avoid a VMEXIT.

Use of the enlightenment by the hypervisor is under control of the
toolstack, and is added to the default set.

Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
Cc: Ian Jackson <ian.jackson@eu.citrix.com>
Cc: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Cc: Wei Liu <wei.liu2@citrix.com>
Cc: Keir Fraser <keir@xen.org>
Cc: Jan Beulich <jbeulich@suse.com>
Cc: Andrew Cooper <andrew.cooper3@citrix.com>
---

v4:
 - Re-worded xl.cfg text and added missing LIBXL_HAVE_ definition

v3:
 - Re-instated read-modify-write for forwards compatibility
 - Fix a coding style issue

v2:
 - Removed some code duplication and unnecessary read-modify-write
   operations on the APIC assist word.
 - Stated in the xl.cfg text that the enlightenment has no effect if
   posted interrupts are in use.
 - Added the enlightenment to the default set.
---
 docs/man/xl.cfg.pod.5              | 12 +++++++-
 tools/libxl/libxl.h                |  6 ++++
 tools/libxl/libxl_dom.c            |  4 +++
 tools/libxl/libxl_types.idl        |  1 +
 xen/arch/x86/hvm/viridian.c        | 59 ++++++++++++++++++++++++++++++++++----
 xen/arch/x86/hvm/vlapic.c          | 58 +++++++++++++++++++++++++++++++++----
 xen/include/asm-x86/hvm/viridian.h |  5 ++++
 xen/include/public/hvm/params.h    |  7 ++++-
 8 files changed, 139 insertions(+), 13 deletions(-)

diff --git a/docs/man/xl.cfg.pod.5 b/docs/man/xl.cfg.pod.5
index 56b1117..abf638c 100644
--- a/docs/man/xl.cfg.pod.5
+++ b/docs/man/xl.cfg.pod.5
@@ -1484,10 +1484,20 @@ This set incorporates use of hypercalls for remote TLB flushing.
 This enlightenment may improve performance of Windows guests running
 on hosts with higher levels of (physical) CPU contention.
 
+=item B<apic_assist>
+
+This set incorporates use of the APIC assist page to avoid EOI of
+the local APIC.
+This enlightenment may improve performance of guests that make use of
+per-vcpu event channel upcall vectors.
+Note that this enlightenment will have no effect if the guest is
+using APICv posted interrupts.
+
 =item B<defaults>
 
 This is a special value that enables the default set of groups, which
-is currently the B<base>, B<freq> and B<time_ref_count> groups.
+is currently the B<base>, B<freq>, B<time_ref_count> and B<apic_assist>
+groups.
 
 =item B<all>
 
diff --git a/tools/libxl/libxl.h b/tools/libxl/libxl.h
index f9e3ef5..055a7b4 100644
--- a/tools/libxl/libxl.h
+++ b/tools/libxl/libxl.h
@@ -218,6 +218,12 @@
 #define LIBXL_HAVE_SOFT_RESET 1
 
 /*
+ * LIBXL_HAVE_APIC_ASSIST indicates that the 'apic_assist' value
+ * is present in the viridian enlightenment enumeration.
+ */
+#define LIBXL_HAVE_APIC_ASSIST 1
+
+/*
  * libxl ABI compatibility
  *
  * The only guarantee which libxl makes regarding ABI compatibility
diff --git a/tools/libxl/libxl_dom.c b/tools/libxl/libxl_dom.c
index b825b98..09d3bca 100644
--- a/tools/libxl/libxl_dom.c
+++ b/tools/libxl/libxl_dom.c
@@ -211,6 +211,7 @@ static int hvm_set_viridian_features(libxl__gc *gc, uint32_t domid,
         libxl_bitmap_set(&enlightenments, LIBXL_VIRIDIAN_ENLIGHTENMENT_BASE);
         libxl_bitmap_set(&enlightenments, LIBXL_VIRIDIAN_ENLIGHTENMENT_FREQ);
         libxl_bitmap_set(&enlightenments, LIBXL_VIRIDIAN_ENLIGHTENMENT_TIME_REF_COUNT);
+        libxl_bitmap_set(&enlightenments, LIBXL_VIRIDIAN_ENLIGHTENMENT_APIC_ASSIST);
     }
 
     libxl_for_each_set_bit(v, info->u.hvm.viridian_enable) {
@@ -253,6 +254,9 @@ static int hvm_set_viridian_features(libxl__gc *gc, uint32_t domid,
     if (libxl_bitmap_test(&enlightenments, LIBXL_VIRIDIAN_ENLIGHTENMENT_HCALL_REMOTE_TLB_FLUSH))
         mask |= HVMPV_hcall_remote_tlb_flush;
 
+    if (libxl_bitmap_test(&enlightenments, LIBXL_VIRIDIAN_ENLIGHTENMENT_APIC_ASSIST))
+        mask |= HVMPV_apic_assist;
+
     if (mask != 0 &&
         xc_hvm_param_set(CTX->xch,
                          domid,
diff --git a/tools/libxl/libxl_types.idl b/tools/libxl/libxl_types.idl
index 632c009..e3be957 100644
--- a/tools/libxl/libxl_types.idl
+++ b/tools/libxl/libxl_types.idl
@@ -221,6 +221,7 @@ libxl_viridian_enlightenment = Enumeration("viridian_enlightenment", [
     (2, "time_ref_count"),
     (3, "reference_tsc"),
     (4, "hcall_remote_tlb_flush"),
+    (5, "apic_assist"),
     ])
 
 libxl_hdtype = Enumeration("hdtype", [
diff --git a/xen/arch/x86/hvm/viridian.c b/xen/arch/x86/hvm/viridian.c
index abbca3c..0f548ce 100644
--- a/xen/arch/x86/hvm/viridian.c
+++ b/xen/arch/x86/hvm/viridian.c
@@ -227,11 +227,6 @@ static void initialize_apic_assist(struct vcpu *v)
     void *va;
 
     /*
-     * We don't yet make use of the APIC assist page but by setting
-     * the CPUID3A_MSR_APIC_ACCESS bit in CPUID leaf 40000003 we are duty
-     * bound to support the MSR. We therefore do just enough to keep windows
-     * happy.
-     *
      * See section 13.3.4.1 of the specification for details of this
      * enlightenment.
      */
@@ -255,6 +250,7 @@ static void initialize_apic_assist(struct vcpu *v)
     *(uint32_t *)va = 0;
 
     v->arch.hvm_vcpu.viridian.apic_assist.va = va;
+    v->arch.hvm_vcpu.viridian.apic_assist.vector = -1;
     return;
 
  fail:
@@ -262,6 +258,59 @@ static void initialize_apic_assist(struct vcpu *v)
              page ? page_to_mfn(page) : INVALID_MFN);
 }
 
+static uint32_t *get_apic_assist_word(struct vcpu *v)
+{
+    if ( !(viridian_feature_mask(v->domain) & HVMPV_apic_assist) )
+        return NULL;
+
+    return v->arch.hvm_vcpu.viridian.apic_assist.va;
+}
+
+void viridian_start_apic_assist(struct vcpu *v, int vector)
+{
+    uint32_t *va = get_apic_assist_word(v);
+
+    if ( !va )
+        return;
+
+    /*
+     * If there is already an assist pending then something has gone
+     * wrong and the VM will most likely hang so force a crash now
+     * to make the problem clear.
+     */
+    if ( v->arch.hvm_vcpu.viridian.apic_assist.vector >= 0 )
+        domain_crash(v->domain);
+
+    v->arch.hvm_vcpu.viridian.apic_assist.vector = vector;
+    *va |= 1u;
+}
+
+bool_t viridian_complete_apic_assist(struct vcpu *v, int *vector)
+{
+    uint32_t *va = get_apic_assist_word(v);
+
+    if ( !va )
+        return 0;
+
+    if ( *va & 1u )
+        return 0; /* Interrupt not yet processed by the guest */
+
+    *vector = v->arch.hvm_vcpu.viridian.apic_assist.vector;
+    v->arch.hvm_vcpu.viridian.apic_assist.vector = -1;
+    return 1;
+}
+
+void viridian_abort_apic_assist(struct vcpu *v)
+{
+    uint32_t *va = get_apic_assist_word(v);
+
+    if ( !va )
+        return;
+
+    *va &= ~1u;
+    v->arch.hvm_vcpu.viridian.apic_assist.vector = -1;
+}
+
 static void teardown_apic_assist(struct vcpu *v)
 {
     void *va = v->arch.hvm_vcpu.viridian.apic_assist.va;
diff --git a/xen/arch/x86/hvm/vlapic.c b/xen/arch/x86/hvm/vlapic.c
index 01a8430..aac4263 100644
--- a/xen/arch/x86/hvm/vlapic.c
+++ b/xen/arch/x86/hvm/vlapic.c
@@ -38,6 +38,7 @@
 #include <asm/hvm/support.h>
 #include <asm/hvm/vmx/vmx.h>
 #include <asm/hvm/nestedhvm.h>
+#include <asm/hvm/viridian.h>
 #include <public/hvm/ioreq.h>
 #include <public/hvm/params.h>
 
@@ -95,6 +96,18 @@ static int vlapic_find_highest_vector(const void *bitmap)
     return (fls(word[word_offset*4]) - 1) + (word_offset * 32);
 }
 
+static int vlapic_find_lowest_vector(const void *bitmap)
+{
+    const uint32_t *word = bitmap;
+    unsigned int word_offset;
+
+    /* Work forwards through the bitmap (first 32-bit word in every four). */
+    for ( word_offset = 0; word_offset < NR_VECTORS / 32; word_offset++)
+        if ( word[word_offset * 4] )
+            return (ffs(word[word_offset * 4]) - 1) + (word_offset * 32);
+
+    return -1;
+}
 
 /*
  * IRR-specific bitmap update & search routines.
@@ -1157,7 +1170,7 @@ int vlapic_virtual_intr_delivery_enabled(void)
 int vlapic_has_pending_irq(struct vcpu *v)
 {
     struct vlapic *vlapic = vcpu_vlapic(v);
-    int irr, isr;
+    int irr, vector, isr;
 
     if ( !vlapic_enabled(vlapic) )
         return -1;
@@ -1170,10 +1183,27 @@ int vlapic_has_pending_irq(struct vcpu *v)
          !nestedhvm_vcpu_in_guestmode(v) )
         return irr;
 
+    /*
+     * If APIC assist was used then there may have been no EOI so
+     * we need to clear the requisite bit from the ISR here, before
+     * comparing with the IRR.
+     */
+    if ( viridian_complete_apic_assist(v, &vector) &&
+         vector != -1 )
+        vlapic_clear_vector(vector, &vlapic->regs->data[APIC_ISR]);
+
     isr = vlapic_find_highest_isr(vlapic);
     isr = (isr != -1) ? isr : 0;
     if ( (isr & 0xf0) >= (irr & 0xf0) )
+    {
+        /*
+         * There's already a higher priority vector pending so
+         * we need to abort any previous APIC assist to ensure there
+         * is an EOI.
+         */
+        viridian_abort_apic_assist(v);
         return -1;
+    }
 
     return irr;
 }
@@ -1181,13 +1211,29 @@ int vlapic_has_pending_irq(struct vcpu *v)
 int vlapic_ack_pending_irq(struct vcpu *v, int vector, bool_t force_ack)
 {
     struct vlapic *vlapic = vcpu_vlapic(v);
+    int isr;
 
-    if ( force_ack || !vlapic_virtual_intr_delivery_enabled() )
-    {
-        vlapic_set_vector(vector, &vlapic->regs->data[APIC_ISR]);
-        vlapic_clear_irr(vector, vlapic);
-    }
+    if ( !force_ack &&
+         vlapic_virtual_intr_delivery_enabled() )
+        return 1;
+
+    if ( vlapic_test_vector(vector, &vlapic->regs->data[APIC_TMR]) )
+        goto done;
+
+    isr = vlapic_find_lowest_vector(&vlapic->regs->data[APIC_ISR]);
+    if ( isr >= 0 && isr < vector )
+        goto done;
+
+    /*
+     * This vector is edge triggered and there are no lower priority
+     * vectors pending, so we can use APIC assist to avoid exiting
+     * for EOI.
+     */
+    viridian_start_apic_assist(v, vector);
 
+done:
+    vlapic_set_vector(vector, &vlapic->regs->data[APIC_ISR]);
+    vlapic_clear_irr(vector, vlapic);
     return 1;
 }
 
diff --git a/xen/include/asm-x86/hvm/viridian.h b/xen/include/asm-x86/hvm/viridian.h
index 19ecc06..336e547 100644
--- a/xen/include/asm-x86/hvm/viridian.h
+++ b/xen/include/asm-x86/hvm/viridian.h
@@ -24,6 +24,7 @@ struct viridian_vcpu
     struct {
         union viridian_apic_assist msr;
         void *va;
+        int vector;
     } apic_assist;
 };
 
@@ -122,6 +123,10 @@ void viridian_time_ref_count_thaw(struct domain *d);
 
 void viridian_vcpu_deinit(struct vcpu *v);
 
+void viridian_start_apic_assist(struct vcpu *v, int vector);
+bool_t viridian_complete_apic_assist(struct vcpu *v, int *vector);
+void viridian_abort_apic_assist(struct vcpu *v);
+
 #endif /* __ASM_X86_HVM_VIRIDIAN_H__ */
 
 /*
diff --git a/xen/include/public/hvm/params.h b/xen/include/public/hvm/params.h
index 73d4718..e69c72c 100644
--- a/xen/include/public/hvm/params.h
+++ b/xen/include/public/hvm/params.h
@@ -115,12 +115,17 @@
 #define _HVMPV_hcall_remote_tlb_flush 4
 #define HVMPV_hcall_remote_tlb_flush (1 << _HVMPV_hcall_remote_tlb_flush)
 
+/* Use APIC assist */
+#define _HVMPV_apic_assist 5
+#define HVMPV_apic_assist (1 << _HVMPV_apic_assist)
+
 #define HVMPV_feature_mask \
         (HVMPV_base_freq | \
          HVMPV_no_freq | \
          HVMPV_time_ref_count | \
          HVMPV_reference_tsc | \
-         HVMPV_hcall_remote_tlb_flush)
+         HVMPV_hcall_remote_tlb_flush | \
+         HVMPV_apic_assist)
 
 #endif
 
-- 
2.1.4


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: [PATCH v5 2/2] x86/hvm/viridian: Enable APIC assist enlightenment

[Jan Beulich]

>>> On 17.03.16 at 13:13, <paul.durrant@citrix.com> wrote:
> @@ -262,6 +258,59 @@ static void initialize_apic_assist(struct vcpu *v)
>               page ? page_to_mfn(page) : INVALID_MFN);
>  }
>  
> +static uint32_t *get_apic_assist_word(struct vcpu *v)
> +{
> +    if ( !(viridian_feature_mask(v->domain) & HVMPV_apic_assist) )
> +        return NULL;
> +
> +    return v->arch.hvm_vcpu.viridian.apic_assist.va;

Does this need to be conditional, i.e. isn't the field NULL anyway
when the flag is not set?

> +void viridian_start_apic_assist(struct vcpu *v, int vector)
> +{
> +    uint32_t *va = get_apic_assist_word(v);
> +
> +    if ( !va )
> +        return;
> +
> +    /*
> +     * If there is already an assist pending then something has gone
> +     * wrong and the VM will most likely hang so force a crash now
> +     * to make the problem clear.
> +     */
> +    if ( v->arch.hvm_vcpu.viridian.apic_assist.vector >= 0 )
> +        domain_crash(v->domain);
> +
> +    v->arch.hvm_vcpu.viridian.apic_assist.vector = vector;
> +    *va |= 1u;

Are the u suffixes here really good for anything?

> +bool_t viridian_complete_apic_assist(struct vcpu *v, int *vector)
> +{
> +    uint32_t *va = get_apic_assist_word(v);
> +
> +    if ( !va )
> +        return 0;
> +
> +    if ( *va & 1u )
> +        return 0; /* Interrupt not yet processed by the guest */

Missing full stop.

> @@ -1170,10 +1183,27 @@ int vlapic_has_pending_irq(struct vcpu *v)
>           !nestedhvm_vcpu_in_guestmode(v) )
>          return irr;
>  
> +    /*
> +     * If APIC assist was used then there may have been no EOI so
> +     * we need to clear the requisite bit from the ISR here, before
> +     * comparing with the IRR.
> +     */
> +    if ( viridian_complete_apic_assist(v, &vector) &&
> +         vector != -1 )

Afaict "vector" is uninitialized here when initialize_apic_assist()
didn't run for that vCPU yet (which includes the case where no
Viridian emulation is active at all).

> @@ -1181,13 +1211,29 @@ int vlapic_has_pending_irq(struct vcpu *v)
>  int vlapic_ack_pending_irq(struct vcpu *v, int vector, bool_t force_ack)
>  {
>      struct vlapic *vlapic = vcpu_vlapic(v);
> +    int isr;
>  
> -    if ( force_ack || !vlapic_virtual_intr_delivery_enabled() )
> -    {
> -        vlapic_set_vector(vector, &vlapic->regs->data[APIC_ISR]);
> -        vlapic_clear_irr(vector, vlapic);
> -    }
> +    if ( !force_ack &&
> +         vlapic_virtual_intr_delivery_enabled() )
> +        return 1;
> +
> +    if ( vlapic_test_vector(vector, &vlapic->regs->data[APIC_TMR]) )
> +        goto done;
> +
> +    isr = vlapic_find_lowest_vector(&vlapic->regs->data[APIC_ISR]);
> +    if ( isr >= 0 && isr < vector )
> +        goto done;

These two new extra operations aren't exactly expensive, but
they also aren't for free. Would it make sense to at least avoid
the second one if the respective Viridian feature bit is clear?

> +    /*
> +     * This vector is edge triggered and there are no lower priority
> +     * vectors pending, so we can use APIC assist to avoid exiting
> +     * for EOI.
> +     */
> +    viridian_start_apic_assist(v, vector);
>  
> +done:

Labels indented by at least one space please.

> +    vlapic_set_vector(vector, &vlapic->regs->data[APIC_ISR]);
> +    vlapic_clear_irr(vector, vlapic);
>      return 1;
>  }




_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: [PATCH v5 2/2] x86/hvm/viridian: Enable APIC assist enlightenment

[Paul Durrant]

> -----Original Message-----
> From: Jan Beulich [mailto:JBeulich@suse.com]
> Sent: 17 March 2016 16:43
> To: Paul Durrant
> Cc: Andrew Cooper; Wei Liu; Ian Jackson; Stefano Stabellini; xen-
> devel@lists.xenproject.org; Keir (Xen.org)
> Subject: Re: [PATCH v5 2/2] x86/hvm/viridian: Enable APIC assist
> enlightenment
> 
> >>> On 17.03.16 at 13:13, <paul.durrant@citrix.com> wrote:
> > @@ -262,6 +258,59 @@ static void initialize_apic_assist(struct vcpu *v)
> >               page ? page_to_mfn(page) : INVALID_MFN);
> >  }
> >
> > +static uint32_t *get_apic_assist_word(struct vcpu *v)
> > +{
> > +    if ( !(viridian_feature_mask(v->domain) & HVMPV_apic_assist) )
> > +        return NULL;
> > +
> > +    return v->arch.hvm_vcpu.viridian.apic_assist.va;
> 
> Does this need to be conditional, i.e. isn't the field NULL anyway
> when the flag is not set?

No. Because Windows 8 will barf on boot if we don't handle set and get of the MSR the page is always mapped and zeroed. However, I could optimize this test by unmapping the page after handling the MSR write if the flag is not set, then the va will indeed be zero if the enlightenment is not in use. I'll do that.

> 
> > +void viridian_start_apic_assist(struct vcpu *v, int vector)
> > +{
> > +    uint32_t *va = get_apic_assist_word(v);
> > +
> > +    if ( !va )
> > +        return;
> > +
> > +    /*
> > +     * If there is already an assist pending then something has gone
> > +     * wrong and the VM will most likely hang so force a crash now
> > +     * to make the problem clear.
> > +     */
> > +    if ( v->arch.hvm_vcpu.viridian.apic_assist.vector >= 0 )
> > +        domain_crash(v->domain);
> > +
> > +    v->arch.hvm_vcpu.viridian.apic_assist.vector = vector;
> > +    *va |= 1u;
> 
> Are the u suffixes here really good for anything?
> 

Probably not, but *va is unsigned so I was just trying to be consistent.

> > +bool_t viridian_complete_apic_assist(struct vcpu *v, int *vector)
> > +{
> > +    uint32_t *va = get_apic_assist_word(v);
> > +
> > +    if ( !va )
> > +        return 0;
> > +
> > +    if ( *va & 1u )
> > +        return 0; /* Interrupt not yet processed by the guest */
> 
> Missing full stop.
> 

Ok.

> > @@ -1170,10 +1183,27 @@ int vlapic_has_pending_irq(struct vcpu *v)
> >           !nestedhvm_vcpu_in_guestmode(v) )
> >          return irr;
> >
> > +    /*
> > +     * If APIC assist was used then there may have been no EOI so
> > +     * we need to clear the requisite bit from the ISR here, before
> > +     * comparing with the IRR.
> > +     */
> > +    if ( viridian_complete_apic_assist(v, &vector) &&
> > +         vector != -1 )
> 
> Afaict "vector" is uninitialized here when initialize_apic_assist()
> didn't run for that vCPU yet (which includes the case where no
> Viridian emulation is active at all).
> 

Yes, vector will be uninitialized in that case but viridian _complete_apic_assist() will return 0 (because the va will be zero) and so the second clause of the if will not be evaluated.

> > @@ -1181,13 +1211,29 @@ int vlapic_has_pending_irq(struct vcpu *v)
> >  int vlapic_ack_pending_irq(struct vcpu *v, int vector, bool_t force_ack)
> >  {
> >      struct vlapic *vlapic = vcpu_vlapic(v);
> > +    int isr;
> >
> > -    if ( force_ack || !vlapic_virtual_intr_delivery_enabled() )
> > -    {
> > -        vlapic_set_vector(vector, &vlapic->regs->data[APIC_ISR]);
> > -        vlapic_clear_irr(vector, vlapic);
> > -    }
> > +    if ( !force_ack &&
> > +         vlapic_virtual_intr_delivery_enabled() )
> > +        return 1;
> > +
> > +    if ( vlapic_test_vector(vector, &vlapic->regs->data[APIC_TMR]) )
> > +        goto done;
> > +
> > +    isr = vlapic_find_lowest_vector(&vlapic->regs->data[APIC_ISR]);
> > +    if ( isr >= 0 && isr < vector )
> > +        goto done;
> 
> These two new extra operations aren't exactly expensive, but
> they also aren't for free. Would it make sense to at least avoid
> the second one if the respective Viridian feature bit is clear?
> 

Yes, that's worth doing.

> > +    /*
> > +     * This vector is edge triggered and there are no lower priority
> > +     * vectors pending, so we can use APIC assist to avoid exiting
> > +     * for EOI.
> > +     */
> > +    viridian_start_apic_assist(v, vector);
> >
> > +done:
> 
> Labels indented by at least one space please.

OK, sorry, emacs keeps moving them back.

  Paul

> 
> > +    vlapic_set_vector(vector, &vlapic->regs->data[APIC_ISR]);
> > +    vlapic_clear_irr(vector, vlapic);
> >      return 1;
> >  }
> 
> 


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: [PATCH v5 2/2] x86/hvm/viridian: Enable APIC assist enlightenment

[Jan Beulich]

>>> On 18.03.16 at 11:06, <Paul.Durrant@citrix.com> wrote:
>> From: Jan Beulich [mailto:JBeulich@suse.com]
>> Sent: 17 March 2016 16:43
>> >>> On 17.03.16 at 13:13, <paul.durrant@citrix.com> wrote:
>> > @@ -1170,10 +1183,27 @@ int vlapic_has_pending_irq(struct vcpu *v)
>> >           !nestedhvm_vcpu_in_guestmode(v) )
>> >          return irr;
>> >
>> > +    /*
>> > +     * If APIC assist was used then there may have been no EOI so
>> > +     * we need to clear the requisite bit from the ISR here, before
>> > +     * comparing with the IRR.
>> > +     */
>> > +    if ( viridian_complete_apic_assist(v, &vector) &&
>> > +         vector != -1 )
>> 
>> Afaict "vector" is uninitialized here when initialize_apic_assist()
>> didn't run for that vCPU yet (which includes the case where no
>> Viridian emulation is active at all).
>> 
> 
> Yes, vector will be uninitialized in that case but viridian 
> _complete_apic_assist() will return 0 (because the va will be zero) and so 
> the second clause of the if will not be evaluated.

Ah, true. But raises the question why viridian_complete_apic_assist()
doesn't return the vector then rather then using indirection.

>> > +    /*
>> > +     * This vector is edge triggered and there are no lower priority
>> > +     * vectors pending, so we can use APIC assist to avoid exiting
>> > +     * for EOI.
>> > +     */
>> > +    viridian_start_apic_assist(v, vector);
>> >
>> > +done:
>> 
>> Labels indented by at least one space please.
> 
> OK, sorry, emacs keeps moving them back.

That's very unfriendly of it.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: [PATCH v5 2/2] x86/hvm/viridian: Enable APIC assist enlightenment

[Paul Durrant]

> -----Original Message-----
> From: Jan Beulich [mailto:JBeulich@suse.com]
> Sent: 18 March 2016 10:25
> To: Paul Durrant
> Cc: Andrew Cooper; Ian Jackson; Stefano Stabellini; Wei Liu; xen-
> devel@lists.xenproject.org; Keir (Xen.org)
> Subject: RE: [PATCH v5 2/2] x86/hvm/viridian: Enable APIC assist
> enlightenment
> 
> >>> On 18.03.16 at 11:06, <Paul.Durrant@citrix.com> wrote:
> >> From: Jan Beulich [mailto:JBeulich@suse.com]
> >> Sent: 17 March 2016 16:43
> >> >>> On 17.03.16 at 13:13, <paul.durrant@citrix.com> wrote:
> >> > @@ -1170,10 +1183,27 @@ int vlapic_has_pending_irq(struct vcpu *v)
> >> >           !nestedhvm_vcpu_in_guestmode(v) )
> >> >          return irr;
> >> >
> >> > +    /*
> >> > +     * If APIC assist was used then there may have been no EOI so
> >> > +     * we need to clear the requisite bit from the ISR here, before
> >> > +     * comparing with the IRR.
> >> > +     */
> >> > +    if ( viridian_complete_apic_assist(v, &vector) &&
> >> > +         vector != -1 )
> >>
> >> Afaict "vector" is uninitialized here when initialize_apic_assist()
> >> didn't run for that vCPU yet (which includes the case where no
> >> Viridian emulation is active at all).
> >>
> >
> > Yes, vector will be uninitialized in that case but viridian
> > _complete_apic_assist() will return 0 (because the va will be zero) and so
> > the second clause of the if will not be evaluated.
> 
> Ah, true. But raises the question why viridian_complete_apic_assist()
> doesn't return the vector then rather then using indirection.
> 

Yes, I guess that would be more obvious.

> >> > +    /*
> >> > +     * This vector is edge triggered and there are no lower priority
> >> > +     * vectors pending, so we can use APIC assist to avoid exiting
> >> > +     * for EOI.
> >> > +     */
> >> > +    viridian_start_apic_assist(v, vector);
> >> >
> >> > +done:
> >>
> >> Labels indented by at least one space please.
> >
> > OK, sorry, emacs keeps moving them back.
> 
> That's very unfriendly of it.
> 

Indeed.

  Paul

> Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: [PATCH v5 2/2] x86/hvm/viridian: Enable APIC assist enlightenment

[Ian Jackson]

> > > OK, sorry, emacs keeps moving them back.
> > 
> > That's very unfriendly of it.
> 
> Indeed.

I think this could be fixed by adjusting the parameters provided at
the bottom of each file in an Emacs `Local variables' block.

See the variable help for c-file-offsets and c-offsets-alist.

I think you want to set a minimum of 1 for `label' and perhaps some
other entries.

Ian.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel