[XEN] Avoid taking domain biglock in the page-fault handler.

This avoids deadlock situation with the shadow_lock.
Signed-off-by: Keir Fraser <keir@xensource.com>

diff --git a/xen/arch/x86/traps.c b/xen/arch/x86/traps.c
index 4b14bb502828167f06e1155576e01c6060f4a60e..3d8023fc333f5c6141bfde2c8ecf48d1d3dc5177 100644 (file)
--- a/xen/arch/x86/traps.c
+++ b/xen/arch/x86/traps.c
@@ -704,12 +704,6 @@ void propagate_page_fault(unsigned long addr, u16 error_code)
 static int handle_gdt_ldt_mapping_fault(
     unsigned long offset, struct cpu_user_regs *regs)
 {
-    extern int map_ldt_shadow_page(unsigned int);
-
-    struct vcpu *v = current;
-    struct domain *d  = v->domain;
-    int ret;
-
     /* Which vcpu's area did we fault in, and is it in the ldt sub-area? */
     unsigned int is_ldt_area = (offset >> (GDT_LDT_VCPU_VA_SHIFT-1)) & 1;
     unsigned int vcpu_area   = (offset >> GDT_LDT_VCPU_VA_SHIFT);
@@ -723,18 +717,15 @@ static int handle_gdt_ldt_mapping_fault(
     if ( likely(is_ldt_area) )
     {
         /* LDT fault: Copy a mapping from the guest's LDT, if it is valid. */
-        LOCK_BIGLOCK(d);
-        ret = map_ldt_shadow_page(offset >> PAGE_SHIFT);
-        UNLOCK_BIGLOCK(d);
-
-        if ( unlikely(ret == 0) )
+        if ( unlikely(map_ldt_shadow_page(offset >> PAGE_SHIFT) == 0) )
         {
             /* In hypervisor mode? Leave it to the #PF handler to fix up. */
             if ( !guest_mode(regs) )
                 return 0;
             /* In guest mode? Propagate #PF to guest, with adjusted %cr2. */
             propagate_page_fault(
-                v->arch.guest_context.ldt_base + offset, regs->error_code);
+                current->arch.guest_context.ldt_base + offset,
+                regs->error_code);
         }
     }
     else
@@ -787,7 +778,7 @@ static int __spurious_page_fault(
 
 #if CONFIG_PAGING_LEVELS >= 4
     l4t = map_domain_page(mfn);
-    l4e = l4t[l4_table_offset(addr)];
+    l4e = l4e_read_atomic(&l4t[l4_table_offset(addr)]);
     mfn = l4e_get_pfn(l4e);
     unmap_domain_page(l4t);
     if ( ((l4e_get_flags(l4e) & required_flags) != required_flags) ||
@@ -800,7 +791,7 @@ static int __spurious_page_fault(
 #ifdef CONFIG_X86_PAE
     l3t += (cr3 & 0xFE0UL) >> 3;
 #endif
-    l3e = l3t[l3_table_offset(addr)];
+    l3e = l3e_read_atomic(&l3t[l3_table_offset(addr)]);
     mfn = l3e_get_pfn(l3e);
     unmap_domain_page(l3t);
 #ifdef CONFIG_X86_PAE
@@ -814,7 +805,7 @@ static int __spurious_page_fault(
 #endif
 
     l2t = map_domain_page(mfn);
-    l2e = l2t[l2_table_offset(addr)];
+    l2e = l2e_read_atomic(&l2t[l2_table_offset(addr)]);
     mfn = l2e_get_pfn(l2e);
     unmap_domain_page(l2t);
     if ( ((l2e_get_flags(l2e) & required_flags) != required_flags) ||
@@ -827,7 +818,7 @@ static int __spurious_page_fault(
     }
 
     l1t = map_domain_page(mfn);
-    l1e = l1t[l1_table_offset(addr)];
+    l1e = l1e_read_atomic(&l1t[l1_table_offset(addr)]);
     mfn = l1e_get_pfn(l1e);
     unmap_domain_page(l1t);
     if ( ((l1e_get_flags(l1e) & required_flags) != required_flags) ||
@@ -856,12 +847,16 @@ static int __spurious_page_fault(
 static int spurious_page_fault(
     unsigned long addr, struct cpu_user_regs *regs)
 {
-    struct domain *d = current->domain;
-    int            is_spurious;
+    unsigned long flags;
+    int           is_spurious;
 
-    LOCK_BIGLOCK(d);
+    /*
+     * Disabling interrupts prevents TLB flushing, and hence prevents
+     * page tables from becoming invalid under our feet during the walk.
+     */
+    local_irq_save(flags);
     is_spurious = __spurious_page_fault(addr, regs);
-    UNLOCK_BIGLOCK(d);
+    local_irq_restore(flags);
 
     return is_spurious;
 }
@@ -878,11 +873,7 @@ static int fixup_page_fault(unsigned long addr, struct cpu_user_regs *regs)
         if ( (addr >= GDT_LDT_VIRT_START) && (addr < GDT_LDT_VIRT_END) )
             return handle_gdt_ldt_mapping_fault(
                 addr - GDT_LDT_VIRT_START, regs);
-        /*
-         * Do not propagate spurious faults in the hypervisor area to the
-         * guest. It cannot fix them up.
-         */
-        return (spurious_page_fault(addr, regs) ? EXCRET_not_a_fault : 0);
+        return 0;
     }
 
     if ( VM_ASSIST(d, VMASST_TYPE_writable_pagetables) &&

diff --git a/xen/include/asm-x86/mm.h b/xen/include/asm-x86/mm.h
index 328d635b489d655cbfd6026b9e111194cbf789ea..cb7dfd7ebbebf50cba654693fb77463397860f6d 100644 (file)
--- a/xen/include/asm-x86/mm.h
+++ b/xen/include/asm-x86/mm.h
@@ -179,8 +179,8 @@ void init_frametable(void);
 
 int alloc_page_type(struct page_info *page, unsigned long type);
 void free_page_type(struct page_info *page, unsigned long type);
-extern void invalidate_shadow_ldt(struct vcpu *d);
-extern int _shadow_mode_refcounts(struct domain *d);
+void invalidate_shadow_ldt(struct vcpu *d);
+int _shadow_mode_refcounts(struct domain *d);
 
 static inline void put_page(struct page_info *page)
 {
@@ -385,4 +385,6 @@ long subarch_memory_op(int op, XEN_GUEST_HANDLE(void) arg);
 int steal_page(
     struct domain *d, struct page_info *page, unsigned int memflags);
 
+int map_ldt_shadow_page(unsigned int);
+
 #endif /* __ASM_X86_MM_H__ */

diff --git a/xen/include/asm-x86/page.h b/xen/include/asm-x86/page.h
index 905d76d57638084a773d30dc12de9708b87f817b..b9c921a5cdbda5b49c7da11800bc9f1799c30586 100644 (file)
--- a/xen/include/asm-x86/page.h
+++ b/xen/include/asm-x86/page.h
@@ -25,6 +25,18 @@
 # include <asm/x86_64/page.h>
 #endif
 
+/* Read a pte atomically from memory. */
+#define l1e_read_atomic(l1ep) l1e_from_intpte(pte_read_atomic(l1ep))
+#define l2e_read_atomic(l2ep) l2e_from_intpte(pte_read_atomic(l2ep))
+#define l3e_read_atomic(l3ep) l3e_from_intpte(pte_read_atomic(l3ep))
+#define l4e_read_atomic(l4ep) l4e_from_intpte(pte_read_atomic(l4ep))
+
+/* Write a pte atomically to memory. */
+#define l1e_write_atomic(l1ep, l1e) pte_write_atomic(l1ep, l1e_get_intpte(l1e))
+#define l2e_write_atomic(l2ep, l2e) pte_write_atomic(l2ep, l1e_get_intpte(l2e))
+#define l3e_write_atomic(l3ep, l3e) pte_write_atomic(l3ep, l1e_get_intpte(l3e))
+#define l4e_write_atomic(l4ep, l4e) pte_write_atomic(l4ep, l1e_get_intpte(l4e))
+
 /* Get direct integer representation of a pte's contents (intpte_t). */
 #define l1e_get_intpte(x)          ((x).l1)
 #define l2e_get_intpte(x)          ((x).l2)

diff --git a/xen/include/asm-x86/x86_32/page-2level.h b/xen/include/asm-x86/x86_32/page-2level.h
index e7b74dd18d507427db4b15d9645f773b1cf9b105..a022b120878128bb93d9a567f99fef4b35f8d45e 100644 (file)
--- a/xen/include/asm-x86/x86_32/page-2level.h
+++ b/xen/include/asm-x86/x86_32/page-2level.h
@@ -28,6 +28,9 @@ typedef l2_pgentry_t root_pgentry_t;
 
 #endif /* !__ASSEMBLY__ */
 
+#define pte_read_atomic(ptep)       (*(intpte_t *)(ptep))
+#define pte_write_atomic(ptep, pte) (*(intpte_t *)(ptep)) = (pte))
+
 /* root table */
 #define root_get_pfn              l2e_get_pfn
 #define root_get_flags            l2e_get_flags

diff --git a/xen/include/asm-x86/x86_32/page-3level.h b/xen/include/asm-x86/x86_32/page-3level.h
index 8450c9c1940d12defb073f565686709b5d0c07c9..08864255ad4a74c492c9d19750c78e43609d6375 100644 (file)
--- a/xen/include/asm-x86/x86_32/page-3level.h
+++ b/xen/include/asm-x86/x86_32/page-3level.h
@@ -38,6 +38,17 @@ typedef l3_pgentry_t root_pgentry_t;
 
 #endif /* !__ASSEMBLY__ */
 
+#define pte_read_atomic(ptep) ({                                            \
+    intpte_t __pte = *(intpte_t *)(ptep), __npte;                           \
+    while ( (__npte = cmpxchg((intpte_t *)(ptep), __pte, __pte)) != __pte ) \
+        __pte = __npte;                                                     \
+    __pte; })
+#define pte_write_atomic(ptep, pte) do {                                    \
+    intpte_t __pte = *(intpte_t *)(ptep), __npte;                           \
+    while ( (__npte = cmpxchg((intpte_t *)(ptep), __pte, (pte))) != __pte ) \
+        __pte = __npte;                                                     \
+} while ( 0 )
+
 /* root table */
 #define root_get_pfn              l3e_get_pfn
 #define root_get_flags            l3e_get_flags

diff --git a/xen/include/asm-x86/x86_64/page.h b/xen/include/asm-x86/x86_64/page.h
index 05c8c57baa54ec722543f258094c90b05e5ade73..4e0f3675cee1981a46ca4d5e9bbb0fd516ab3a57 100644 (file)
--- a/xen/include/asm-x86/x86_64/page.h
+++ b/xen/include/asm-x86/x86_64/page.h
@@ -41,6 +41,9 @@ typedef l4_pgentry_t root_pgentry_t;
 
 #endif /* !__ASSEMBLY__ */
 
+#define pte_read_atomic(ptep)       (*(intpte_t *)(ptep))
+#define pte_write_atomic(ptep, pte) (*(intpte_t *)(ptep)) = (pte))
+
 /* Given a virtual address, get an entry offset into a linear page table. */
 #define l1_linear_offset(_a) (((_a) & VADDR_MASK) >> L1_PAGETABLE_SHIFT)
 #define l2_linear_offset(_a) (((_a) & VADDR_MASK) >> L2_PAGETABLE_SHIFT)