[PATCH v3 1/6] ioreq-server: centralize access to ioreq structures

[Paul Durrant <paul.durrant@citrix.com>]

To simplify creation of the ioreq server abstraction in a
subsequent patch, this patch centralizes all use of the shared
ioreq structure and the buffered ioreq ring to the source module
xen/arch/x86/hvm/hvm.c.

This patch also adds some missing emacs boilerplate in the places where I
needed it.

Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
---
 xen/arch/x86/hvm/emulate.c        |   38 +++++--------
 xen/arch/x86/hvm/hvm.c            |  113 ++++++++++++++++++++++++++++++++++++-
 xen/arch/x86/hvm/io.c             |   99 ++------------------------------
 xen/arch/x86/hvm/stdvga.c         |    2 +-
 xen/arch/x86/hvm/vmx/vvmx.c       |   13 ++++-
 xen/include/asm-x86/hvm/hvm.h     |   15 ++++-
 xen/include/asm-x86/hvm/io.h      |    2 +-
 xen/include/asm-x86/hvm/support.h |   19 ++++---
 8 files changed, 165 insertions(+), 136 deletions(-)

diff --git a/xen/arch/x86/hvm/emulate.c b/xen/arch/x86/hvm/emulate.c
index 868aa1d..0ba2020 100644
--- a/xen/arch/x86/hvm/emulate.c
+++ b/xen/arch/x86/hvm/emulate.c
@@ -57,24 +57,11 @@ static int hvmemul_do_io(
     int value_is_ptr = (p_data == NULL);
     struct vcpu *curr = current;
     struct hvm_vcpu_io *vio;
-    ioreq_t *p = get_ioreq(curr);
-    ioreq_t _ioreq;
+    ioreq_t p[1];
     unsigned long ram_gfn = paddr_to_pfn(ram_gpa);
     p2m_type_t p2mt;
     struct page_info *ram_page;
     int rc;
-    bool_t has_dm = 1;
-
-    /*
-     * Domains without a backing DM, don't have an ioreq page.  Just
-     * point to a struct on the stack, initialising the state as needed.
-     */
-    if ( !p )
-    {
-        has_dm = 0;
-        p = &_ioreq;
-        p->state = STATE_IOREQ_NONE;
-    }
 
     /* Check for paged out page */
     ram_page = get_page_from_gfn(curr->domain, ram_gfn, &p2mt, P2M_UNSHARE);
@@ -173,15 +160,6 @@ static int hvmemul_do_io(
         return X86EMUL_UNHANDLEABLE;
     }
 
-    if ( p->state != STATE_IOREQ_NONE )
-    {
-        gdprintk(XENLOG_WARNING, "WARNING: io already pending (%d)?\n",
-                 p->state);
-        if ( ram_page )
-            put_page(ram_page);
-        return X86EMUL_UNHANDLEABLE;
-    }
-
     vio->io_state =
         (p_data == NULL) ? HVMIO_dispatched : HVMIO_awaiting_completion;
     vio->io_size = size;
@@ -233,7 +211,7 @@ static int hvmemul_do_io(
         break;
     case X86EMUL_UNHANDLEABLE:
         /* If there is no backing DM, just ignore accesses */
-        if ( !has_dm )
+        if ( !hvm_has_dm(curr->domain) )
         {
             rc = X86EMUL_OKAY;
             vio->io_state = HVMIO_none;
@@ -241,7 +219,7 @@ static int hvmemul_do_io(
         else
         {
             rc = X86EMUL_RETRY;
-            if ( !hvm_send_assist_req(curr) )
+            if ( !hvm_send_assist_req(curr, p) )
                 vio->io_state = HVMIO_none;
             else if ( p_data == NULL )
                 rc = X86EMUL_OKAY;
@@ -1292,3 +1270,13 @@ struct segment_register *hvmemul_get_seg_reg(
         hvm_get_segment_register(current, seg, &hvmemul_ctxt->seg_reg[seg]);
     return &hvmemul_ctxt->seg_reg[seg];
 }
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
diff --git a/xen/arch/x86/hvm/hvm.c b/xen/arch/x86/hvm/hvm.c
index 9e85c13..0b2e57e 100644
--- a/xen/arch/x86/hvm/hvm.c
+++ b/xen/arch/x86/hvm/hvm.c
@@ -345,9 +345,27 @@ void hvm_migrate_pirqs(struct vcpu *v)
     spin_unlock(&d->event_lock);
 }
 
+static ioreq_t *get_ioreq(struct vcpu *v)
+{
+    struct domain *d = v->domain;
+    shared_iopage_t *p = d->arch.hvm_domain.ioreq.va;
+    ASSERT((v == current) || spin_is_locked(&d->arch.hvm_domain.ioreq.lock));
+    return p ? &p->vcpu_ioreq[v->vcpu_id] : NULL;
+}
+
+bool_t hvm_io_pending(struct vcpu *v)
+{
+    ioreq_t *p = get_ioreq(v);
+
+    if ( !p )
+         return 0;
+
+    return ( p->state != STATE_IOREQ_NONE );
+}
+
 void hvm_do_resume(struct vcpu *v)
 {
-    ioreq_t *p;
+    ioreq_t *p = get_ioreq(v);
 
     check_wakeup_from_wait();
 
@@ -355,7 +373,7 @@ void hvm_do_resume(struct vcpu *v)
         pt_restore_timer(v);
 
     /* NB. Optimised for common case (p->state == STATE_IOREQ_NONE). */
-    if ( !(p = get_ioreq(v)) )
+    if ( !p )
         goto check_inject_trap;
 
     while ( p->state != STATE_IOREQ_NONE )
@@ -1407,7 +1425,87 @@ void hvm_vcpu_down(struct vcpu *v)
     }
 }
 
-bool_t hvm_send_assist_req(struct vcpu *v)
+int hvm_buffered_io_send(struct domain *d, const ioreq_t *p)
+{
+    struct hvm_ioreq_page *iorp = &d->arch.hvm_domain.buf_ioreq;
+    buffered_iopage_t *pg = iorp->va;
+    buf_ioreq_t bp;
+    /* Timeoffset sends 64b data, but no address. Use two consecutive slots. */
+    int qw = 0;
+
+    /* Ensure buffered_iopage fits in a page */
+    BUILD_BUG_ON(sizeof(buffered_iopage_t) > PAGE_SIZE);
+
+    /*
+     * Return 0 for the cases we can't deal with:
+     *  - 'addr' is only a 20-bit field, so we cannot address beyond 1MB
+     *  - we cannot buffer accesses to guest memory buffers, as the guest
+     *    may expect the memory buffer to be synchronously accessed
+     *  - the count field is usually used with data_is_ptr and since we don't
+     *    support data_is_ptr we do not waste space for the count field either
+     */
+    if ( (p->addr > 0xffffful) || p->data_is_ptr || (p->count != 1) )
+        return 0;
+
+    bp.type = p->type;
+    bp.dir  = p->dir;
+    switch ( p->size )
+    {
+    case 1:
+        bp.size = 0;
+        break;
+    case 2:
+        bp.size = 1;
+        break;
+    case 4:
+        bp.size = 2;
+        break;
+    case 8:
+        bp.size = 3;
+        qw = 1;
+        break;
+    default:
+        gdprintk(XENLOG_WARNING, "unexpected ioreq size: %u\n", p->size);
+        return 0;
+    }
+
+    bp.data = p->data;
+    bp.addr = p->addr;
+
+    spin_lock(&iorp->lock);
+
+    if ( (pg->write_pointer - pg->read_pointer) >=
+         (IOREQ_BUFFER_SLOT_NUM - qw) )
+    {
+        /* The queue is full: send the iopacket through the normal path. */
+        spin_unlock(&iorp->lock);
+        return 0;
+    }
+
+    pg->buf_ioreq[pg->write_pointer % IOREQ_BUFFER_SLOT_NUM] = bp;
+
+    if ( qw )
+    {
+        bp.data = p->data >> 32;
+        pg->buf_ioreq[(pg->write_pointer+1) % IOREQ_BUFFER_SLOT_NUM] = bp;
+    }
+
+    /* Make the ioreq_t visible /before/ write_pointer. */
+    wmb();
+    pg->write_pointer += qw ? 2 : 1;
+
+    notify_via_xen_event_channel(d, d->arch.hvm_domain.params[HVM_PARAM_BUFIOREQ_EVTCHN]);
+    spin_unlock(&iorp->lock);
+
+    return 1;
+}
+
+bool_t hvm_has_dm(struct domain *d)
+{
+    return !!d->arch.hvm_domain.ioreq.va;
+}
+
+bool_t hvm_send_assist_req(struct vcpu *v, const ioreq_t *proto_p)
 {
     ioreq_t *p;
 
@@ -1425,6 +1523,15 @@ bool_t hvm_send_assist_req(struct vcpu *v)
         return 0;
     }
 
+    p->dir = proto_p->dir;
+    p->data_is_ptr = proto_p->data_is_ptr;
+    p->type = proto_p->type;
+    p->size = proto_p->size;
+    p->addr = proto_p->addr;
+    p->count = proto_p->count;
+    p->df = proto_p->df;
+    p->data = proto_p->data;
+
     prepare_wait_on_xen_event_channel(v->arch.hvm_vcpu.xen_port);
 
     /*
diff --git a/xen/arch/x86/hvm/io.c b/xen/arch/x86/hvm/io.c
index bf6309d..ba50c53 100644
--- a/xen/arch/x86/hvm/io.c
+++ b/xen/arch/x86/hvm/io.c
@@ -46,87 +46,9 @@
 #include <xen/iocap.h>
 #include <public/hvm/ioreq.h>
 
-int hvm_buffered_io_send(ioreq_t *p)
-{
-    struct vcpu *v = current;
-    struct hvm_ioreq_page *iorp = &v->domain->arch.hvm_domain.buf_ioreq;
-    buffered_iopage_t *pg = iorp->va;
-    buf_ioreq_t bp;
-    /* Timeoffset sends 64b data, but no address. Use two consecutive slots. */
-    int qw = 0;
-
-    /* Ensure buffered_iopage fits in a page */
-    BUILD_BUG_ON(sizeof(buffered_iopage_t) > PAGE_SIZE);
-
-    /*
-     * Return 0 for the cases we can't deal with:
-     *  - 'addr' is only a 20-bit field, so we cannot address beyond 1MB
-     *  - we cannot buffer accesses to guest memory buffers, as the guest
-     *    may expect the memory buffer to be synchronously accessed
-     *  - the count field is usually used with data_is_ptr and since we don't
-     *    support data_is_ptr we do not waste space for the count field either
-     */
-    if ( (p->addr > 0xffffful) || p->data_is_ptr || (p->count != 1) )
-        return 0;
-
-    bp.type = p->type;
-    bp.dir  = p->dir;
-    switch ( p->size )
-    {
-    case 1:
-        bp.size = 0;
-        break;
-    case 2:
-        bp.size = 1;
-        break;
-    case 4:
-        bp.size = 2;
-        break;
-    case 8:
-        bp.size = 3;
-        qw = 1;
-        break;
-    default:
-        gdprintk(XENLOG_WARNING, "unexpected ioreq size: %u\n", p->size);
-        return 0;
-    }
-    
-    bp.data = p->data;
-    bp.addr = p->addr;
-    
-    spin_lock(&iorp->lock);
-
-    if ( (pg->write_pointer - pg->read_pointer) >=
-         (IOREQ_BUFFER_SLOT_NUM - qw) )
-    {
-        /* The queue is full: send the iopacket through the normal path. */
-        spin_unlock(&iorp->lock);
-        return 0;
-    }
-    
-    memcpy(&pg->buf_ioreq[pg->write_pointer % IOREQ_BUFFER_SLOT_NUM],
-           &bp, sizeof(bp));
-    
-    if ( qw )
-    {
-        bp.data = p->data >> 32;
-        memcpy(&pg->buf_ioreq[(pg->write_pointer+1) % IOREQ_BUFFER_SLOT_NUM],
-               &bp, sizeof(bp));
-    }
-
-    /* Make the ioreq_t visible /before/ write_pointer. */
-    wmb();
-    pg->write_pointer += qw ? 2 : 1;
-
-    notify_via_xen_event_channel(v->domain,
-            v->domain->arch.hvm_domain.params[HVM_PARAM_BUFIOREQ_EVTCHN]);
-    spin_unlock(&iorp->lock);
-    
-    return 1;
-}
-
 void send_timeoffset_req(unsigned long timeoff)
 {
+    struct vcpu *curr = current;
     ioreq_t p[1];
 
     if ( timeoff == 0 )
@@ -142,33 +64,22 @@ void send_timeoffset_req(unsigned long timeoff)
 
     p->state = STATE_IOREQ_READY;
 
-    if ( !hvm_buffered_io_send(p) )
+    if ( !hvm_buffered_io_send(curr->domain, p) )
         printk("Unsuccessful timeoffset update\n");
 }
 
 /* Ask ioemu mapcache to invalidate mappings. */
 void send_invalidate_req(void)
 {
-    struct vcpu *v = current;
-    ioreq_t *p = get_ioreq(v);
-
-    if ( !p )
-        return;
-
-    if ( p->state != STATE_IOREQ_NONE )
-    {
-        gdprintk(XENLOG_ERR, "WARNING: send invalidate req with something "
-                 "already pending (%d)?\n", p->state);
-        domain_crash(v->domain);
-        return;
-    }
+    struct vcpu *curr = current;
+    ioreq_t p[1];
 
     p->type = IOREQ_TYPE_INVALIDATE;
     p->size = 4;
     p->dir = IOREQ_WRITE;
     p->data = ~0UL; /* flush all */
 
-    (void)hvm_send_assist_req(v);
+    (void)hvm_send_assist_req(curr, p);
 }
 
 int handle_mmio(void)
diff --git a/xen/arch/x86/hvm/stdvga.c b/xen/arch/x86/hvm/stdvga.c
index 19e80ed..9e2d28e 100644
--- a/xen/arch/x86/hvm/stdvga.c
+++ b/xen/arch/x86/hvm/stdvga.c
@@ -580,7 +580,7 @@ static int stdvga_intercept_mmio(ioreq_t *p)
         buf = (p->dir == IOREQ_WRITE);
     }
 
-    rc = (buf && hvm_buffered_io_send(p));
+    rc = (buf && hvm_buffered_io_send(d, p));
 
     spin_unlock(&s->lock);
 
diff --git a/xen/arch/x86/hvm/vmx/vvmx.c b/xen/arch/x86/hvm/vmx/vvmx.c
index 40167d6..0421623 100644
--- a/xen/arch/x86/hvm/vmx/vvmx.c
+++ b/xen/arch/x86/hvm/vmx/vvmx.c
@@ -1394,7 +1394,6 @@ void nvmx_switch_guest(void)
     struct vcpu *v = current;
     struct nestedvcpu *nvcpu = &vcpu_nestedhvm(v);
     struct cpu_user_regs *regs = guest_cpu_user_regs();
-    const ioreq_t *ioreq = get_ioreq(v);
 
     /*
      * A pending IO emulation may still be not finished. In this case, no
@@ -1404,7 +1403,7 @@ void nvmx_switch_guest(void)
      * don't want to continue as this setup is not implemented nor supported
      * as of right now.
      */
-    if ( !ioreq || ioreq->state != STATE_IOREQ_NONE )
+    if ( hvm_io_pending(v) )
         return;
     /*
      * a softirq may interrupt us between a virtual vmentry is
@@ -2522,3 +2521,13 @@ void nvmx_set_cr_read_shadow(struct vcpu *v, unsigned int cr)
     /* nvcpu.guest_cr is what L2 write to cr actually. */
     __vmwrite(read_shadow_field, v->arch.hvm_vcpu.nvcpu.guest_cr[cr]);
 }
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
diff --git a/xen/include/asm-x86/hvm/hvm.h b/xen/include/asm-x86/hvm/hvm.h
index dcc3483..08a62ea 100644
--- a/xen/include/asm-x86/hvm/hvm.h
+++ b/xen/include/asm-x86/hvm/hvm.h
@@ -26,6 +26,7 @@
 #include <asm/hvm/asid.h>
 #include <public/domctl.h>
 #include <public/hvm/save.h>
+#include <public/hvm/ioreq.h>
 #include <asm/mm.h>
 
 /* Interrupt acknowledgement sources. */
@@ -227,7 +228,7 @@ int prepare_ring_for_helper(struct domain *d, unsigned long gmfn,
                             struct page_info **_page, void **_va);
 void destroy_ring_for_helper(void **_va, struct page_info *page);
 
-bool_t hvm_send_assist_req(struct vcpu *v);
+bool_t hvm_send_assist_req(struct vcpu *v, const ioreq_t *p);
 
 void hvm_get_guest_pat(struct vcpu *v, u64 *guest_pat);
 int hvm_set_guest_pat(struct vcpu *v, u64 guest_pat);
@@ -339,6 +340,8 @@ static inline unsigned long hvm_get_shadow_gs_base(struct vcpu *v)
 void hvm_cpuid(unsigned int input, unsigned int *eax, unsigned int *ebx,
                                    unsigned int *ecx, unsigned int *edx);
 void hvm_migrate_timers(struct vcpu *v);
+bool_t hvm_has_dm(struct domain *d);
+bool_t hvm_io_pending(struct vcpu *v);
 void hvm_do_resume(struct vcpu *v);
 void hvm_migrate_pirqs(struct vcpu *v);
 
@@ -522,3 +525,13 @@ bool_t nhvm_vmcx_hap_enabled(struct vcpu *v);
 enum hvm_intblk nhvm_interrupt_blocked(struct vcpu *v);
 
 #endif /* __ASM_X86_HVM_HVM_H__ */
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
diff --git a/xen/include/asm-x86/hvm/io.h b/xen/include/asm-x86/hvm/io.h
index 86db58d..bfd28c2 100644
--- a/xen/include/asm-x86/hvm/io.h
+++ b/xen/include/asm-x86/hvm/io.h
@@ -92,7 +92,7 @@ static inline int hvm_buffered_io_intercept(ioreq_t *p)
 }
 
 int hvm_mmio_intercept(ioreq_t *p);
-int hvm_buffered_io_send(ioreq_t *p);
+int hvm_buffered_io_send(struct domain *d, const ioreq_t *p);
 
 static inline void register_portio_handler(
     struct domain *d, unsigned long addr,
diff --git a/xen/include/asm-x86/hvm/support.h b/xen/include/asm-x86/hvm/support.h
index 3529499..05ef5c5 100644
--- a/xen/include/asm-x86/hvm/support.h
+++ b/xen/include/asm-x86/hvm/support.h
@@ -22,19 +22,10 @@
 #define __ASM_X86_HVM_SUPPORT_H__
 
 #include <xen/types.h>
-#include <public/hvm/ioreq.h>
 #include <xen/sched.h>
 #include <xen/hvm/save.h>
 #include <asm/processor.h>
 
-static inline ioreq_t *get_ioreq(struct vcpu *v)
-{
-    struct domain *d = v->domain;
-    shared_iopage_t *p = d->arch.hvm_domain.ioreq.va;
-    ASSERT((v == current) || spin_is_locked(&d->arch.hvm_domain.ioreq.lock));
-    return p ? &p->vcpu_ioreq[v->vcpu_id] : NULL;
-}
-
 #define HVM_DELIVER_NO_ERROR_CODE  -1
 
 #ifndef NDEBUG
@@ -142,3 +133,13 @@ int hvm_mov_to_cr(unsigned int cr, unsigned int gpr);
 int hvm_mov_from_cr(unsigned int cr, unsigned int gpr);
 
 #endif /* __ASM_X86_HVM_SUPPORT_H__ */
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
-- 
1.7.10.4