[IA64] Change vCPU initialization to avoid domVTi privregs memory leak

1) This patch moved some processing from vcpu_initialise() and
   added a new function vcpu_late_initialise().
   It executes the following initializations for VCPU of
   dom0/domU.
    - Allocate the VHPT
    - Allocate the privregs area and assign these pages into
      guest pseudo physical address space.
    - Set the tlbflush_timestamp.

   It is executed in the following sequence.

   dom0:
     start_kernel()
       ->domain_create()
       ->alloc_vcpu(VCPU0)
         ->alloc_vcpu_struct(VCPU0)
         ->vcpu_initialise(VCPU0)
       ->vcpu_late_initialise(VCPU0)

       ->construct_dom0
         ->alloc_vcpu(othe VCPUs)
           ->alloc_vcpu_struct(other VCPUs)
           ->vcpu_initialise(other VCPUs)

     ia64_hypercall(FW_HYPERCALL_IPI)
       ->fw_hypercall_ipi(XEN_SAL_BOOT_RENDEZ_VEC)
         ->arch_set_info_guest(other VCPUs)
           ->vcpu_late_initialise(other VCPUs)

   domU:
     do_domctl(XEN_DOMCTL_createdomain)
       ->domain_create()

     do_domctl(XEN_DOMCTL_max_vcpus)
       ->alloc_vcpu(all VCPUs)
         ->alloc_vcpu_struct(all VCPUs)
         ->vcpu_initialise(all VCPUs)

     do_domctl(XEN_DOMCTL_setvcpucontext)
       ->set_info_guest(VCPU0)
         ->arch_set_info_guest(VCPU0)
           ->vcpu_late_initialise(VCPU0)

     ia64_hypercall(FW_HYPERCALL_IPI)
       ->fw_hypercall_ipi(XEN_SAL_BOOT_RENDEZ_VEC)
         ->arch_set_info_guest(other VCPUs)
           ->vcpu_late_initialise(other VCPUs)


2) This patch modified the domain_set_shared_info_va().
   Currently, initialization of arch.privregs->interrupt_mask_addr
   of all VCPUs is executed in domain_set_shared_info_va().
   However, allocation of privregs area is late by modified of 1).
   Therefore, this patch modified initialization of
   arch.privregs->interrupt_mask_addr to the following sequence.

   dom0 and domU:
     ia64_hypercall(FW_HYPERCALL_SET_SHARED_INFO_VA)
       ->domain_set_shared_info_va()
         Initialize interrupt_mask_addr of VCPU0

     ia64_hypercall(FW_HYPERCALL_IPI)
       ->fw_hypercall_ipi(XEN_SAL_BOOT_RENDEZ_VEC)
         ->arch_set_info_guest(other VCPUs)
           ->vcpu_late_initialise(other VCPUs)
           Initialize interrupt_mask_addr of other VCPUs


Signed-off-by: Masaki Kanno <kanno.masaki@jp.fujitsu.com>

xen/arch/ia64/xen/dom0_ops.c

@@ -103,16 +103,6 @@ long arch_do_domctl(xen_domctl_t *op, XEN_GUEST_HANDLE(xen_domctl_t) u_domctl)
                     ret = -EINVAL;
                     break;
                 }
-                if (!d->arch.is_vti) {
-                    struct vcpu *v;
-                    for_each_vcpu(d, v) {
-                        BUG_ON(v->arch.privregs == NULL);
-                        free_domheap_pages(virt_to_page(v->arch.privregs),
-                                      get_order_from_shift(XMAPPEDREGS_SHIFT));
-                        v->arch.privregs = NULL;
-                        relinquish_vcpu_resources(v);
-                    }
-                }
                 d->arch.is_vti = 1;
                 vmx_setup_platform(d);
             }

xen/arch/ia64/xen/domain.c

@@ -305,24 +305,14 @@ void hlt_timer_fn(void *data)
 
 void relinquish_vcpu_resources(struct vcpu *v)
 {
-    if (HAS_PERVCPU_VHPT(v->domain))
-        pervcpu_vhpt_free(v);
-    if (v->arch.privregs != NULL) {
-        // this might be called by arch_do_domctl() with XEN_DOMCTL_arch_setup()
-        // for domVTi.
-        if (!(atomic_read(&v->domain->refcnt) & DOMAIN_DESTROYED)) {
-            unsigned long i;
-            for (i = 0; i < XMAPPEDREGS_SIZE; i += PAGE_SIZE)
-                guest_physmap_remove_page(v->domain,
-                    IA64_XMAPPEDREGS_PADDR(v->vcpu_id) + i,
-                    virt_to_maddr(v->arch.privregs + i));
-        }
-
-        free_xenheap_pages(v->arch.privregs,
-                           get_order_from_shift(XMAPPEDREGS_SHIFT));
-        v->arch.privregs = NULL;
-    }
-    kill_timer(&v->arch.hlt_timer);
+	if (HAS_PERVCPU_VHPT(v->domain))
+		pervcpu_vhpt_free(v);
+	if (v->arch.privregs != NULL) {
+		free_xenheap_pages(v->arch.privregs,
+		                   get_order_from_shift(XMAPPEDREGS_SHIFT));
+		v->arch.privregs = NULL;
+	}
+	kill_timer(&v->arch.hlt_timer);
 }
 
 struct vcpu *alloc_vcpu_struct(void)
@@ -361,36 +351,8 @@ void free_vcpu_struct(struct vcpu *v)
 int vcpu_initialise(struct vcpu *v)
 {
 	struct domain *d = v->domain;
-	int rc, order, i;
 
 	if (!is_idle_domain(d)) {
-	    if (!d->arch.is_vti) {
-		if (HAS_PERVCPU_VHPT(d))
-			if ((rc = pervcpu_vhpt_alloc(v)) != 0)
-				return rc;
-
-		/* Create privregs page only if not VTi. */
-		order = get_order_from_shift(XMAPPEDREGS_SHIFT);
-		v->arch.privregs = alloc_xenheap_pages(order);
-		BUG_ON(v->arch.privregs == NULL);
-		memset(v->arch.privregs, 0, 1 << XMAPPEDREGS_SHIFT);
-		for (i = 0; i < (1 << order); i++)
-		    share_xen_page_with_guest(virt_to_page(v->arch.privregs) +
-		                              i, d, XENSHARE_writable);
-		/*
-		 * XXX IA64_XMAPPEDREGS_PADDR
-		 * assign these pages into guest pseudo physical address
-		 * space for dom0 to map this page by gmfn.
-		 * this is necessary for domain save, restore and dump-core.
-		 */
-		for (i = 0; i < XMAPPEDREGS_SIZE; i += PAGE_SIZE)
-		    assign_domain_page(d, IA64_XMAPPEDREGS_PADDR(v->vcpu_id) + i,
-                                      virt_to_maddr(v->arch.privregs + i));
-
-		tlbflush_update_time(&v->arch.tlbflush_timestamp,
-		                     tlbflush_current_time());
-	    }
-
 	    v->arch.metaphysical_rr0 = d->arch.metaphysical_rr0;
 	    v->arch.metaphysical_rr4 = d->arch.metaphysical_rr4;
 	    v->arch.metaphysical_saved_rr0 = d->arch.metaphysical_rr0;
@@ -420,6 +382,41 @@ int vcpu_initialise(struct vcpu *v)
 	return 0;
 }
 
+int vcpu_late_initialise(struct vcpu *v)
+{
+	struct domain *d = v->domain;
+	int rc, order, i;
+
+	if (HAS_PERVCPU_VHPT(d)) {
+		rc = pervcpu_vhpt_alloc(v);
+		if (rc != 0)
+			return rc;
+	}
+
+	/* Create privregs page. */
+	order = get_order_from_shift(XMAPPEDREGS_SHIFT);
+	v->arch.privregs = alloc_xenheap_pages(order);
+	BUG_ON(v->arch.privregs == NULL);
+	memset(v->arch.privregs, 0, 1 << XMAPPEDREGS_SHIFT);
+	for (i = 0; i < (1 << order); i++)
+		share_xen_page_with_guest(virt_to_page(v->arch.privregs) + i,
+		                          d, XENSHARE_writable);
+	/*
+	 * XXX IA64_XMAPPEDREGS_PADDR
+	 * assign these pages into guest pseudo physical address
+	 * space for dom0 to map this page by gmfn.
+	 * this is necessary for domain save, restore and dump-core.
+	 */
+	for (i = 0; i < XMAPPEDREGS_SIZE; i += PAGE_SIZE)
+		assign_domain_page(d, IA64_XMAPPEDREGS_PADDR(v->vcpu_id) + i,
+		                   virt_to_maddr(v->arch.privregs + i));
+
+	tlbflush_update_time(&v->arch.tlbflush_timestamp,
+	                     tlbflush_current_time());
+
+	return 0;
+}
+
 void vcpu_destroy(struct vcpu *v)
 {
 	if (v->domain->arch.is_vti)
@@ -553,6 +550,7 @@ int arch_set_info_guest(struct vcpu *v, vcpu_guest_context_u c)
 {
 	struct pt_regs *regs = vcpu_regs (v);
 	struct domain *d = v->domain;
+	int rc;
 	
 	*regs = c.nat->user_regs;
  	
@@ -582,16 +580,25 @@ int arch_set_info_guest(struct vcpu *v, vcpu_guest_context_u c)
 		v->arch.event_callback_ip = er->event_callback_ip;
 		v->arch.dcr = er->dcr;
 		v->arch.iva = er->iva;
-  	}
-	
-  	if ( test_bit(_VCPUF_initialised, &v->vcpu_flags) )
- 		return 0;
- 	if (d->arch.is_vti)
- 		vmx_final_setup_guest(v);
-	
- 	/* This overrides some registers.  */
-  	vcpu_init_regs(v);
-  
+	}
+
+	if (test_bit(_VCPUF_initialised, &v->vcpu_flags))
+		return 0;
+
+	if (d->arch.is_vti)
+		vmx_final_setup_guest(v);
+	else {
+		rc = vcpu_late_initialise(v);
+		if (rc != 0)
+			return rc;
+		VCPU(v, interrupt_mask_addr) = 
+			(unsigned char *) d->arch.shared_info_va +
+			INT_ENABLE_OFFSET(v);
+	}
+
+	/* This overrides some registers. */
+	vcpu_init_regs(v);
+
 	/* Don't redo final setup */
 	set_bit(_VCPUF_initialised, &v->vcpu_flags);
 	return 0;
@@ -683,7 +690,6 @@ domain_set_shared_info_va (unsigned long va)
 {
 	struct vcpu *v = current;
 	struct domain *d = v->domain;
-	struct vcpu *v1;
 
 	/* Check virtual address:
 	   must belong to region 7,
@@ -699,10 +705,8 @@ domain_set_shared_info_va (unsigned long va)
 	printk ("Domain set shared_info_va to 0x%016lx\n", va);
 	d->arch.shared_info_va = va;
 
-	for_each_vcpu (d, v1) {
-		VCPU(v1, interrupt_mask_addr) = 
-			(unsigned char *)va + INT_ENABLE_OFFSET(v1);
-	}
+	VCPU(v, interrupt_mask_addr) = (unsigned char *)va +
+	                               INT_ENABLE_OFFSET(v);
 
 	__ia64_per_cpu_var(current_psr_ic_addr) = (int *)(va + XSI_PSR_IC_OFS);
 
@@ -1106,7 +1110,7 @@ int construct_dom0(struct domain *d,
 	printk ("Dom0 max_vcpus=%d\n", dom0_max_vcpus);
 	for ( i = 1; i < dom0_max_vcpus; i++ )
 	    if (alloc_vcpu(d, i, i) == NULL)
-		printk ("Cannot allocate dom0 vcpu %d\n", i);
+		panic("Cannot allocate dom0 vcpu %d\n", i);
 
 	/* Copy the OS image. */
 	loaddomainelfimage(d,image_start);

xen/arch/ia64/xen/xensetup.c

@@ -249,6 +249,7 @@ void start_kernel(void)
     unsigned long dom0_initrd_start, dom0_initrd_size;
     unsigned long md_end, relo_start, relo_end, relo_size = 0;
     struct domain *idle_domain;
+    struct vcpu *dom0_vcpu0;
     efi_memory_desc_t *kern_md, *last_md, *md;
 #ifdef CONFIG_SMP
     int i;
@@ -503,8 +504,11 @@ printk("num_online_cpus=%d, max_cpus=%d\n",num_online_cpus(),max_cpus);
 
     /* Create initial domain 0. */
     dom0 = domain_create(0, 0);
-    if ( (dom0 == NULL) || (alloc_vcpu(dom0, 0, 0) == NULL) )
+    if (dom0 == NULL)
         panic("Error creating domain 0\n");
+    dom0_vcpu0 = alloc_vcpu(dom0, 0, 0);
+    if (dom0_vcpu0 == NULL || vcpu_late_initialise(dom0_vcpu0) != 0)
+        panic("Cannot allocate dom0 vcpu 0\n");
 
     dom0->is_privileged = 1;
 

xen/include/asm-ia64/domain.h

@@ -21,6 +21,7 @@ struct tlb_track;
 extern void domain_relinquish_resources(struct domain *);
 struct vcpu;
 extern void relinquish_vcpu_resources(struct vcpu *v);
+extern int vcpu_late_initialise(struct vcpu *v);
 
 /* given a current domain metaphysical address, return the physical address */
 extern unsigned long translate_domain_mpaddr(unsigned long mpaddr,