bitkeeper revision 1.1159.170.27 (4199eb03e15gd8adslXoDx2C8r41ow)

Merge freefall.cl.cam.ac.uk:/auto/groups/xeno/BK/xeno.bk
into freefall.cl.cam.ac.uk:/local/scratch/kaf24/xeno

linux-2.6.9-xen-sparse/arch/xen/configs/xen0_defconfig

@@ -158,7 +158,7 @@ CONFIG_MAGIC_SYSRQ=y
 # CONFIG_DEBUG_INFO is not set
 # CONFIG_DEBUG_SPINLOCK_SLEEP is not set
 # CONFIG_FRAME_POINTER is not set
-# CONFIG_4KSTACKS is not set
+CONFIG_4KSTACKS=y
 CONFIG_X86_BIOS_REBOOT=y
 CONFIG_X86_STD_RESOURCES=y
 CONFIG_PC=y

linux-2.6.9-xen-sparse/arch/xen/configs/xenU_defconfig

@@ -119,7 +119,7 @@ CONFIG_HAVE_DEC_LOCK=y
 CONFIG_EARLY_PRINTK=y
 # CONFIG_DEBUG_SPINLOCK_SLEEP is not set
 # CONFIG_FRAME_POINTER is not set
-# CONFIG_4KSTACKS is not set
+CONFIG_4KSTACKS=y
 CONFIG_X86_BIOS_REBOOT=y
 CONFIG_X86_STD_RESOURCES=y
 CONFIG_PC=y

linux-2.6.9-xen-sparse/arch/xen/i386/kernel/setup.c

@@ -1319,6 +1319,11 @@ void __init setup_arch(char **cmdline_p)
 
         unsigned long max_low_pfn;
 
+	/* Force a quick death if the kernel panics. */
+	extern int panic_timeout;
+	if ( panic_timeout == 0 )
+		panic_timeout = 1;
+
 	HYPERVISOR_vm_assist(VMASST_CMD_enable,
 			     VMASST_TYPE_4gb_segments);
 

linux-2.6.9-xen-sparse/arch/xen/i386/mm/pageattr.c

@@ -119,7 +119,7 @@ __change_page_attr(struct page *page, pgprot_t prot)
 		if ((pte_val(*kpte) & _PAGE_PSE) == 0) { 
 			pte_t old = *kpte;
 			pte_t standard = mk_pte(page, PAGE_KERNEL); 
-			set_pte_atomic(kpte, mk_pte(page, prot)); 
+			set_pte_batched(kpte, mk_pte(page, prot)); 
 			if (pte_same(old,standard))
 				get_page(kpte_page);
 		} else {
@@ -130,7 +130,7 @@ __change_page_attr(struct page *page, pgprot_t prot)
 			set_pmd_pte(kpte,address,mk_pte(split, PAGE_KERNEL));
 		}	
 	} else if ((pte_val(*kpte) & _PAGE_PSE) == 0) { 
-		set_pte_atomic(kpte, mk_pte(page, PAGE_KERNEL));
+		set_pte_batched(kpte, mk_pte(page, PAGE_KERNEL));
 		__put_page(kpte_page);
 	}
 
@@ -171,6 +171,7 @@ int change_page_attr(struct page *page, int numpages, pgprot_t prot)
 		if (err) 
 			break; 
 	} 	
+	flush_page_update_queue();
 	spin_unlock_irqrestore(&cpa_lock, flags);
 	return err;
 }

linux-2.6.9-xen-sparse/arch/xen/kernel/evtchn.c

@@ -467,6 +467,8 @@ void __init init_IRQ(void)
 {
     int i;
 
+    irq_ctx_init(0);
+
     spin_lock_init(&irq_mapping_update_lock);
 
     /* No VIRQ -> IRQ mappings. */

linux-2.6.9-xen-sparse/include/asm-xen/asm-i386/pgtable-2level.h

@@ -23,13 +23,10 @@ static inline int pgd_present(pgd_t pgd)	{ return 1; }
  * within a page table are directly modified.  Thus, the following
  * hook is made available.
  */
-#ifdef CONFIG_XEN_WRITABLE_PAGETABLES
+#define set_pte_batched(pteptr, pteval) \
+queue_l1_entry_update(pteptr, (pteval).pte_low)
 #define set_pte(pteptr, pteval) (*(pteptr) = pteval)
-#define set_pte_atomic(pteptr, pteval) (*(pteptr) = pteval)
-#else
-#define set_pte(pteptr, pteval) xen_l1_entry_update(pteptr, (pteval).pte_low)
-#define set_pte_atomic(pteptr, pteval) xen_l1_entry_update(pteptr, (pteval).pte_low)
-#endif
+#define set_pte_atomic(pteptr, pteval) set_pte(pteptr,pteval)
 /*
  * (pmds are folded into pgds so this doesn't get actually called,
  * but the define is needed for a generic inline function.)

xen/arch/x86/memory.c

@@ -1577,7 +1577,7 @@ void ptwr_flush(const int which)
 {
     unsigned long  sstat, spte, pte, *ptep, l1va;
     l1_pgentry_t  *sl1e = NULL, *pl1e, ol1e, nl1e;
-    l2_pgentry_t  *pl2e, nl2e;
+    l2_pgentry_t  *pl2e;
     int            i, cpu = smp_processor_id();
     struct domain *d = current;
 
@@ -1691,8 +1691,7 @@ void ptwr_flush(const int which)
     if ( (which == PTWR_PT_ACTIVE) && likely(!d->mm.shadow_mode) )
     {
         pl2e = &linear_l2_table[ptwr_info[cpu].ptinfo[which].l2_idx];
-        nl2e = mk_l2_pgentry(l2_pgentry_val(*pl2e) | _PAGE_PRESENT);
-        update_l2e(pl2e, *pl2e, nl2e);
+        *pl2e = mk_l2_pgentry(l2_pgentry_val(*pl2e) | _PAGE_PRESENT); 
     }
 
     /*
@@ -1711,9 +1710,9 @@ void ptwr_flush(const int which)
 /* Write page fault handler: check if guest is trying to modify a PTE. */
 int ptwr_do_page_fault(unsigned long addr)
 {
-    unsigned long    pte, pfn;
+    unsigned long    pte, pfn, l2e;
     struct pfn_info *page;
-    l2_pgentry_t    *pl2e, nl2e;
+    l2_pgentry_t    *pl2e;
     int              which, cpu = smp_processor_id();
     u32              l2_idx;
 
@@ -1739,14 +1738,34 @@ int ptwr_do_page_fault(unsigned long addr)
     if ( unlikely(l2_idx >= PGT_va_unknown) )
         domain_crash(); /* Urk! This L1 is mapped in multiple L2 slots! */
     l2_idx >>= PGT_va_shift;
-        
+
+    if ( l2_idx == (addr >> L2_PAGETABLE_SHIFT) )
+    {
+        MEM_LOG("PTWR failure! Pagetable maps itself at %08lx\n", addr);
+        domain_crash();
+    }
+
     /*
      * Is the L1 p.t. mapped into the current address space? If so we call it
      * an ACTIVE p.t., otherwise it is INACTIVE.
      */
     pl2e = &linear_l2_table[l2_idx];
-    which = (l2_pgentry_val(*pl2e) >> PAGE_SHIFT != pfn) ?
-        PTWR_PT_INACTIVE : PTWR_PT_ACTIVE;
+    l2e  = l2_pgentry_val(*pl2e);
+    which = PTWR_PT_INACTIVE;
+    if ( (l2e >> PAGE_SHIFT) == pfn )
+    {
+        /*
+         * If the PRESENT bit is clear, we may be conflicting with the current 
+         * ACTIVE p.t. (it may be the same p.t. mapped at another virt addr).
+         */
+        if ( unlikely(!(l2e & _PAGE_PRESENT)) &&
+             ptwr_info[cpu].ptinfo[PTWR_PT_ACTIVE].l1va )
+            ptwr_flush(PTWR_PT_ACTIVE);
+        
+        /* Now do a final check of the PRESENT bit to set ACTIVE. */
+        if ( likely(l2e & _PAGE_PRESENT) )
+            which = PTWR_PT_ACTIVE;
+    }
     
     PTWR_PRINTK("[%c] page_fault on l1 pt at va %08lx, pt for %08x, "
                 "pfn %08lx\n", PTWR_PRINT_WHICH,
@@ -1765,8 +1784,7 @@ int ptwr_do_page_fault(unsigned long addr)
     /* For safety, disconnect the L1 p.t. page from current space. */
     if ( (which == PTWR_PT_ACTIVE) && likely(!current->mm.shadow_mode) )
     {
-        nl2e = mk_l2_pgentry(l2_pgentry_val(*pl2e) & ~_PAGE_PRESENT);
-        update_l2e(pl2e, *pl2e, nl2e);
+        *pl2e = mk_l2_pgentry(l2e & ~_PAGE_PRESENT);
         flush_tlb(); /* XXX Multi-CPU guests? */
     }