Manual merge.

buildconfigs/Rules.mk

@@ -115,7 +115,7 @@ config-update-pae:
 ifeq ($(XEN_TARGET_X86_PAE),y)
 	sed -e 's!^CONFIG_HIGHMEM4G=y$$!\# CONFIG_HIGHMEM4G is not set!;s!^\# CONFIG_HIGHMEM64G is not set$$!CONFIG_HIGHMEM64G=y!' $(CONFIG_FILE) > $(CONFIG_FILE)- && mv $(CONFIG_FILE)- $(CONFIG_FILE)
 else
-	@: # do nothing yet
+	grep '^CONFIG_HIGHMEM64G=y' $(CONFIG_FILE) >/dev/null && ( sed -e 's!^CONFIG_HIGHMEM64G=y$$!\# CONFIG_HIGHMEM64G is not set!;s!^\# CONFIG_HIGHMEM4G is not set$$!CONFIG_HIGHMEM4G=y!' $(CONFIG_FILE) > $(CONFIG_FILE)- && mv $(CONFIG_FILE)- $(CONFIG_FILE) ) || true
 endif
 
 # never delete any intermediate files.

linux-2.4-xen-sparse/arch/xen/Makefile

@@ -61,7 +61,6 @@ SUBDIRS += arch/xen/drivers/console
 SUBDIRS += arch/xen/drivers/evtchn
 SUBDIRS += arch/xen/drivers/blkif
 SUBDIRS += arch/xen/drivers/netif
-#SUBDIRS += arch/xen/drivers/usbif
 SUBDIRS += arch/xen/drivers/balloon
 ifdef CONFIG_XEN_PRIVILEGED_GUEST
 SUBDIRS += arch/xen/drivers/dom0 
@@ -72,7 +71,6 @@ CORE_FILES += arch/xen/drivers/evtchn/drv.o
 CORE_FILES += arch/xen/drivers/console/drv.o
 DRIVERS += arch/xen/drivers/blkif/drv.o
 DRIVERS += arch/xen/drivers/netif/drv.o
-DRIVERS += arch/xen/drivers/usbif/drv.o
 ifdef CONFIG_XEN_PRIVILEGED_GUEST
 CORE_FILES += arch/xen/drivers/dom0/drv.o
 endif

linux-2.4-xen-sparse/arch/xen/config.in

@@ -16,14 +16,10 @@ mainmenu_option next_comment
 comment 'Xen'
 bool 'Support for privileged operations (domain 0)' CONFIG_XEN_PRIVILEGED_GUEST
 bool 'Device-driver domain (physical device access)' CONFIG_XEN_PHYSDEV_ACCESS
-if [ "$CONFIG_XEN_PHYSDEV_ACCESS" = "y" ]; then
-    bool 'USB-device backend driver' CONFIG_XEN_USB_BACKEND
-fi
 bool 'Scrub memory before freeing it to Xen' CONFIG_XEN_SCRUB_PAGES
 bool 'Network-device frontend driver' CONFIG_XEN_NETDEV_FRONTEND
 bool 'Block-device frontend driver' CONFIG_XEN_BLKDEV_FRONTEND
 bool 'Block-device uses grant tables' CONFIG_XEN_BLKDEV_GRANT
-bool 'USB-device frontend driver' CONFIG_XEN_USB_FRONTEND
 endmenu
 # The IBM S/390 patch needs this.
 define_bool CONFIG_NO_IDLE_HZ y
@@ -267,7 +263,7 @@ fi
 
 source drivers/char/Config.in
 
-if [ "$CONFIG_XEN_PHYSDEV_ACCESS" = "y" -o "$CONFIG_XEN_USB_FRONTEND" = "y" ]; then
+if [ "$CONFIG_XEN_PHYSDEV_ACCESS" = "y" ]; then
    source drivers/media/Config.in
 fi
 
@@ -302,14 +298,8 @@ if [ "$CONFIG_XEN_PHYSDEV_ACCESS" = "y" ]; then
    endmenu
 fi
 
-if [ "$CONFIG_XEN_PHYSDEV_ACCESS" = "y" -o "$CONFIG_XEN_USB_FRONTEND" = "y" ]; then
-   if [ "$CONFIG_XEN_USB_FRONTEND" = "y" -o "$CONFIG_XEN_USB_BACKEND" = "y" ]; then
-       define_bool CONFIG_USB y
-   fi
-   source drivers/usb/Config.in
-fi
-
 if [ "$CONFIG_XEN_PHYSDEV_ACCESS" = "y" ]; then
+   source drivers/usb/Config.in
    source net/bluetooth/Config.in
 fi
 

linux-2.6-xen-sparse/arch/xen/configs/xen0_defconfig_x86_32

@@ -1,7 +1,7 @@
 #
 # Automatically generated make config: don't edit
 # Linux kernel version: 2.6.12-xen0
-# Mon Jul 25 09:48:34 2005
+# Wed Aug  3 09:54:56 2005
 #
 CONFIG_XEN=y
 CONFIG_ARCH_XEN=y
@@ -18,8 +18,8 @@ CONFIG_XEN_BLKDEV_GRANT=y
 CONFIG_XEN_NETDEV_BACKEND=y
 CONFIG_XEN_BLKDEV_FRONTEND=y
 CONFIG_XEN_NETDEV_FRONTEND=y
-#CONFIG_XEN_NETDEV_GRANT_TX=y
-#CONFIG_XEN_NETDEV_GRANT_RX=y
+# CONFIG_XEN_NETDEV_GRANT_TX is not set
+# CONFIG_XEN_NETDEV_GRANT_RX is not set
 # CONFIG_XEN_NETDEV_FRONTEND_PIPELINED_TRANSMITTER is not set
 # CONFIG_XEN_BLKDEV_TAP is not set
 # CONFIG_XEN_SHADOW_MODE is not set
@@ -93,11 +93,11 @@ CONFIG_GENERIC_IOMAP=y
 # CONFIG_M586 is not set
 # CONFIG_M586TSC is not set
 # CONFIG_M586MMX is not set
-# CONFIG_M686 is not set
+CONFIG_M686=y
 # CONFIG_MPENTIUMII is not set
 # CONFIG_MPENTIUMIII is not set
 # CONFIG_MPENTIUMM is not set
-CONFIG_MPENTIUM4=y
+# CONFIG_MPENTIUM4 is not set
 # CONFIG_MK6 is not set
 # CONFIG_MK7 is not set
 # CONFIG_MK8 is not set
@@ -112,15 +112,15 @@ CONFIG_MPENTIUM4=y
 # CONFIG_X86_GENERIC is not set
 CONFIG_X86_CMPXCHG=y
 CONFIG_X86_XADD=y
-CONFIG_X86_L1_CACHE_SHIFT=7
+CONFIG_X86_L1_CACHE_SHIFT=5
 CONFIG_RWSEM_XCHGADD_ALGORITHM=y
 CONFIG_GENERIC_CALIBRATE_DELAY=y
+CONFIG_X86_PPRO_FENCE=y
 CONFIG_X86_WP_WORKS_OK=y
 CONFIG_X86_INVLPG=y
 CONFIG_X86_BSWAP=y
 CONFIG_X86_POPAD_OK=y
 CONFIG_X86_GOOD_APIC=y
-CONFIG_X86_INTEL_USERCOPY=y
 CONFIG_X86_USE_PPRO_CHECKSUM=y
 # CONFIG_HPET_TIMER is not set
 # CONFIG_HPET_EMULATE_RTC is not set

linux-2.6-xen-sparse/arch/xen/configs/xen0_defconfig_x86_64

@@ -1,7 +1,7 @@
 #
 # Automatically generated make config: don't edit
 # Linux kernel version: 2.6.12-xen0
-# Wed Jun 29 10:01:20 2005
+# Tue Aug  2 23:55:35 2005
 #
 CONFIG_XEN=y
 CONFIG_ARCH_XEN=y
@@ -18,6 +18,8 @@ CONFIG_XEN_BLKDEV_GRANT=y
 CONFIG_XEN_NETDEV_BACKEND=y
 CONFIG_XEN_BLKDEV_FRONTEND=y
 CONFIG_XEN_NETDEV_FRONTEND=y
+# CONFIG_XEN_NETDEV_GRANT_TX is not set
+# CONFIG_XEN_NETDEV_GRANT_RX is not set
 # CONFIG_XEN_NETDEV_FRONTEND_PIPELINED_TRANSMITTER is not set
 # CONFIG_XEN_BLKDEV_TAP is not set
 # CONFIG_XEN_SHADOW_MODE is not set

linux-2.6-xen-sparse/arch/xen/configs/xenU_defconfig_x86_32

@@ -1,7 +1,7 @@
 #
 # Automatically generated make config: don't edit
 # Linux kernel version: 2.6.12-xenU
-# Mon Jul 25 10:06:06 2005
+# Wed Aug  3 09:57:44 2005
 #
 CONFIG_XEN=y
 CONFIG_ARCH_XEN=y
@@ -15,8 +15,8 @@ CONFIG_NO_IDLE_HZ=y
 CONFIG_XEN_BLKDEV_GRANT=y
 CONFIG_XEN_BLKDEV_FRONTEND=y
 CONFIG_XEN_NETDEV_FRONTEND=y
-#CONFIG_XEN_NETDEV_GRANT_TX=y
-#CONFIG_XEN_NETDEV_GRANT_RX=y
+# CONFIG_XEN_NETDEV_GRANT_TX is not set
+# CONFIG_XEN_NETDEV_GRANT_RX is not set
 # CONFIG_XEN_NETDEV_FRONTEND_PIPELINED_TRANSMITTER is not set
 # CONFIG_XEN_BLKDEV_TAP is not set
 # CONFIG_XEN_SHADOW_MODE is not set
@@ -90,11 +90,11 @@ CONFIG_GENERIC_IOMAP=y
 # CONFIG_M586 is not set
 # CONFIG_M586TSC is not set
 # CONFIG_M586MMX is not set
-# CONFIG_M686 is not set
+CONFIG_M686=y
 # CONFIG_MPENTIUMII is not set
 # CONFIG_MPENTIUMIII is not set
 # CONFIG_MPENTIUMM is not set
-CONFIG_MPENTIUM4=y
+# CONFIG_MPENTIUM4 is not set
 # CONFIG_MK6 is not set
 # CONFIG_MK7 is not set
 # CONFIG_MK8 is not set
@@ -109,15 +109,15 @@ CONFIG_MPENTIUM4=y
 # CONFIG_X86_GENERIC is not set
 CONFIG_X86_CMPXCHG=y
 CONFIG_X86_XADD=y
-CONFIG_X86_L1_CACHE_SHIFT=7
+CONFIG_X86_L1_CACHE_SHIFT=5
 CONFIG_RWSEM_XCHGADD_ALGORITHM=y
 CONFIG_GENERIC_CALIBRATE_DELAY=y
+CONFIG_X86_PPRO_FENCE=y
 CONFIG_X86_WP_WORKS_OK=y
 CONFIG_X86_INVLPG=y
 CONFIG_X86_BSWAP=y
 CONFIG_X86_POPAD_OK=y
 CONFIG_X86_GOOD_APIC=y
-CONFIG_X86_INTEL_USERCOPY=y
 CONFIG_X86_USE_PPRO_CHECKSUM=y
 # CONFIG_HPET_TIMER is not set
 # CONFIG_HPET_EMULATE_RTC is not set

linux-2.6-xen-sparse/arch/xen/configs/xenU_defconfig_x86_64

@@ -1,7 +1,7 @@
 #
 # Automatically generated make config: don't edit
 # Linux kernel version: 2.6.12-xenU
-# Thu Jul  7 11:43:14 2005
+# Tue Aug  2 23:56:13 2005
 #
 CONFIG_XEN=y
 CONFIG_ARCH_XEN=y
@@ -15,6 +15,8 @@ CONFIG_NO_IDLE_HZ=y
 CONFIG_XEN_BLKDEV_GRANT=y
 CONFIG_XEN_BLKDEV_FRONTEND=y
 CONFIG_XEN_NETDEV_FRONTEND=y
+# CONFIG_XEN_NETDEV_GRANT_TX is not set
+# CONFIG_XEN_NETDEV_GRANT_RX is not set
 # CONFIG_XEN_NETDEV_FRONTEND_PIPELINED_TRANSMITTER is not set
 # CONFIG_XEN_BLKDEV_TAP is not set
 # CONFIG_XEN_SHADOW_MODE is not set

linux-2.6-xen-sparse/arch/xen/i386/kernel/smpboot.c

@@ -1529,7 +1529,7 @@ void __init smp_cpus_done(unsigned int max_cpus)
 extern irqreturn_t smp_reschedule_interrupt(int, void *, struct pt_regs *);
 extern irqreturn_t smp_call_function_interrupt(int, void *, struct pt_regs *);
 
-void __init smp_intr_init(void)
+void smp_intr_init(void)
 {
 	int cpu = smp_processor_id();
 
@@ -1546,3 +1546,28 @@ void __init smp_intr_init(void)
 	                   smp_call_function_interrupt,
 	                   SA_INTERRUPT, callfunc_name[cpu], NULL));
 }
+
+static void smp_intr_exit(void)
+{
+	int cpu = smp_processor_id();
+
+	free_irq(per_cpu(resched_irq, cpu), NULL);
+	unbind_ipi_from_irq(RESCHEDULE_VECTOR);
+
+	free_irq(per_cpu(callfunc_irq, cpu), NULL);
+	unbind_ipi_from_irq(CALL_FUNCTION_VECTOR);
+}
+
+void smp_suspend(void)
+{
+	/* XXX todo: take down time and ipi's on all cpus */
+	local_teardown_timer_irq();
+	smp_intr_exit();
+}
+
+void smp_resume(void)
+{
+	/* XXX todo: restore time and ipi's on all cpus */
+	smp_intr_init();
+	local_setup_timer_irq();
+}

linux-2.6-xen-sparse/arch/xen/i386/kernel/time.c

@@ -860,6 +860,8 @@ void start_hz_timer(void)
 void time_suspend(void)
 {
 	/* nothing */
+	teardown_irq(per_cpu(timer_irq, 0), &irq_timer);
+	unbind_virq_from_irq(VIRQ_TIMER);
 }
 
 /* No locking required. We are only CPU running, and interrupts are off. */
@@ -874,10 +876,25 @@ void time_resume(void)
 	processed_system_time =
 		per_cpu(shadow_time, smp_processor_id()).system_timestamp;
 	per_cpu(processed_system_time, 0) = processed_system_time;
+
+	per_cpu(timer_irq, 0) = bind_virq_to_irq(VIRQ_TIMER);
+	(void)setup_irq(per_cpu(timer_irq, 0), &irq_timer);
 }
 
 #ifdef CONFIG_SMP
 static char timer_name[NR_CPUS][15];
+void local_setup_timer_irq(void)
+{
+	int cpu = smp_processor_id();
+
+	if (cpu == 0)
+		return;
+	per_cpu(timer_irq, cpu) = bind_virq_to_irq(VIRQ_TIMER);
+	sprintf(timer_name[cpu], "timer%d", cpu);
+	BUG_ON(request_irq(per_cpu(timer_irq, cpu), timer_interrupt,
+	                   SA_INTERRUPT, timer_name[cpu], NULL));
+}
+
 void local_setup_timer(void)
 {
 	int seq, cpu = smp_processor_id();
@@ -888,10 +905,17 @@ void local_setup_timer(void)
 			per_cpu(shadow_time, cpu).system_timestamp;
 	} while (read_seqretry(&xtime_lock, seq));
 
-	per_cpu(timer_irq, cpu) = bind_virq_to_irq(VIRQ_TIMER);
-	sprintf(timer_name[cpu], "timer%d", cpu);
-	BUG_ON(request_irq(per_cpu(timer_irq, cpu), timer_interrupt,
-	                   SA_INTERRUPT, timer_name[cpu], NULL));
+	local_setup_timer_irq();
+}
+
+void local_teardown_timer_irq(void)
+{
+	int cpu = smp_processor_id();
+
+	if (cpu == 0)
+		return;
+	free_irq(per_cpu(timer_irq, cpu), NULL);
+	unbind_virq_from_irq(VIRQ_TIMER);
 }
 #endif
 

linux-2.6-xen-sparse/arch/xen/kernel/evtchn.c

@@ -284,7 +284,7 @@ void unbind_ipi_from_irq(int ipi)
     evtchn_op_t op;
     int cpu    = smp_processor_id();
     int evtchn = per_cpu(ipi_to_evtchn, cpu)[ipi];
-    int irq    = irq_to_evtchn[evtchn];
+    int irq    = evtchn_to_irq[evtchn];
 
     spin_lock(&irq_mapping_update_lock);
 

linux-2.6-xen-sparse/arch/xen/kernel/reboot.c

@@ -16,6 +16,7 @@ static int errno;
 #include <asm-xen/xen-public/dom0_ops.h>
 #include <asm-xen/linux-public/suspend.h>
 #include <asm-xen/queues.h>
+#include <asm-xen/xenbus.h>
 
 void machine_restart(char * __unused)
 {
@@ -90,6 +91,10 @@ static void __do_suspend(void)
 #define gnttab_resume()  do{}while(0)
 #endif
 
+#ifdef CONFIG_SMP
+    extern void smp_suspend(void);
+    extern void smp_resume(void);
+#endif
     extern void time_suspend(void);
     extern void time_resume(void);
     extern unsigned long max_pfn;
@@ -114,6 +119,12 @@ static void __do_suspend(void)
 
     time_suspend();
 
+#ifdef CONFIG_SMP
+    smp_suspend();
+#endif
+
+    xenbus_suspend();
+
     ctrl_if_suspend();
 
     irq_suspend();
@@ -153,6 +164,12 @@ static void __do_suspend(void)
 
     ctrl_if_resume();
 
+    xenbus_resume();
+
+#ifdef CONFIG_SMP
+    smp_resume();
+#endif
+
     time_resume();
 
     blkdev_resume();

linux-2.6-xen-sparse/drivers/xen/balloon/balloon.c

@@ -84,9 +84,6 @@ static struct timer_list balloon_timer;
 /* Flag for dom0 xenstore workaround */
 static int balloon_xenbus_init=0;
 
-/* Init Function */
-void balloon_init_watcher(void);
-
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0)
 /* Use the private and mapping fields of struct page as a list. */
 #define PAGE_TO_LIST(p) ( (struct list_head *)&p->private )
@@ -354,27 +351,21 @@ static void watch_target(struct xenbus_watch *watch, const char *node)
     
 }
 
-/* 
-   Try to set up our watcher, if not already set
-   
-*/
-void balloon_init_watcher(void) 
+/* Init Function - Try to set up our watcher, if not already set. */
+void balloon_init_watcher(void)
 {
     int err;
 
-    if(!xen_start_info.store_evtchn)
-    {
+    if (!xen_start_info.store_evtchn) {
         IPRINTK("Delaying watcher init until xenstore is available\n");
         return;
     }
 
     down(&xenbus_lock);
 
-    if(! balloon_xenbus_init) 
-    {
+    if (!balloon_xenbus_init) {
         err = register_xenbus_watch(&xb_watch);
-        if(err) 
-        {
+        if (err) {
             /* BIG FAT FIXME: dom0 sequencing workaround
              * dom0 can't set a watch on memory/target until
              * after the tools create it.  So, we have to watch
@@ -384,16 +375,13 @@ void balloon_init_watcher(void)
              * non-existant keys
              */
             register_xenbus_watch(&root_watch);
-        } 
-        else
-        {
+        } else {
             IPRINTK("Balloon xenbus watcher initialized\n");
             balloon_xenbus_init = 1;
         }
     }
 
     up(&xenbus_lock);
-
 }
 
 EXPORT_SYMBOL(balloon_init_watcher);

linux-2.6-xen-sparse/drivers/xen/xenbus/xenbus_comms.c

@@ -36,6 +36,8 @@
 #include <linux/err.h>
 #include "xenbus_comms.h"
 
+static unsigned int xb_irq;
+
 #define RINGBUF_DATASIZE ((PAGE_SIZE / 2) - sizeof(struct ringbuf_head))
 struct ringbuf_head
 {
@@ -202,14 +204,17 @@ int xb_read(void *data, unsigned len)
 	return 0;
 }
 
-/* Set up interrpt handler off store event channel. */
+/* Set up interrupt handler off store event channel. */
 int xb_init_comms(void)
 {
-	int err, irq;
+	int err;
 
-	irq = bind_evtchn_to_irq(xen_start_info.store_evtchn);
+	if (!xen_start_info.store_evtchn)
+		return 0;
 
-	err = request_irq(irq, wake_waiting, SA_SHIRQ, "xenbus", &xb_waitq);
+	xb_irq = bind_evtchn_to_irq(xen_start_info.store_evtchn);
+
+	err = request_irq(xb_irq, wake_waiting, 0, "xenbus", &xb_waitq);
 	if (err) {
 		printk(KERN_ERR "XENBUS request irq failed %i\n", err);
 		unbind_evtchn_from_irq(xen_start_info.store_evtchn);
@@ -222,3 +227,13 @@ int xb_init_comms(void)
 
 	return 0;
 }
+
+void xb_suspend_comms(void)
+{
+
+	if (!xen_start_info.store_evtchn)
+		return;
+
+	free_irq(xb_irq, &xb_waitq);
+	unbind_evtchn_from_irq(xen_start_info.store_evtchn);
+}

linux-2.6-xen-sparse/drivers/xen/xenbus/xenbus_comms.h

@@ -3,6 +3,7 @@
 #define _XENBUS_COMMS_H
 int xs_init(void);
 int xb_init_comms(void);
+void xb_suspend_comms(void);
 
 /* Low level routines. */
 int xb_write(const void *data, unsigned len);

linux-2.6-xen-sparse/drivers/xen/xenbus/xenbus_probe.c

@@ -29,6 +29,7 @@
 
 #include <asm-xen/hypervisor.h>
 #include <asm-xen/xenbus.h>
+#include <asm-xen/balloon.h>
 #include <linux/kernel.h>
 #include <linux/err.h>
 #include <linux/string.h>
@@ -295,6 +296,19 @@ static struct xenbus_watch dev_watch = {
 	.callback = dev_changed,
 };
 
+void xenbus_suspend(void)
+{
+	/* We keep lock, so no comms can happen as page moves. */
+	down(&xenbus_lock);
+	xb_suspend_comms();
+}
+
+void xenbus_resume(void)
+{
+	xb_init_comms();
+	up(&xenbus_lock);
+}
+
 /* called from a thread in privcmd/privcmd.c */
 int do_xenbus_probe(void *unused)
 {

linux-2.6-xen-sparse/include/asm-generic/pgtable.h

@@ -37,7 +37,7 @@ do {				  					\
  */
 #define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \
 do {				  					  \
-	set_pte_at((__vma)>vm_mm, (__address), __ptep, __entry);	  \
+	set_pte_at((__vma)->vm_mm, (__address), __ptep, __entry);	  \
 	flush_tlb_page(__vma, __address);				  \
 } while (0)
 #endif

linux-2.6-xen-sparse/include/asm-xen/asm-i386/mach-xen/irq_vectors.h

@@ -129,6 +129,7 @@
 extern int  bind_virq_to_irq(int virq);
 extern void unbind_virq_from_irq(int virq);
 extern int  bind_ipi_to_irq(int ipi);
+extern int  bind_ipi_on_cpu_to_irq(int ipi);
 extern void unbind_ipi_from_irq(int ipi);
 extern int  bind_evtchn_to_irq(int evtchn);
 extern void unbind_evtchn_from_irq(int evtchn);

linux-2.6-xen-sparse/include/asm-xen/asm-x86_64/mach-xen/irq_vectors.h

@@ -127,6 +127,7 @@
 extern int  bind_virq_to_irq(int virq);
 extern void unbind_virq_from_irq(int virq);
 extern int  bind_ipi_to_irq(int ipi);
+extern int  bind_ipi_on_cpu_to_irq(int ipi);
 extern void unbind_ipi_from_irq(int ipi);
 extern int  bind_evtchn_to_irq(int evtchn);
 extern void unbind_evtchn_from_irq(int evtchn);

linux-2.6-xen-sparse/include/asm-xen/balloon.h

@@ -48,4 +48,7 @@ extern spinlock_t balloon_lock;
 #define balloon_lock(__flags)   spin_lock_irqsave(&balloon_lock, __flags)
 #define balloon_unlock(__flags) spin_unlock_irqrestore(&balloon_lock, __flags)
 
+/* Init Function - Try to set up our watcher, if not already set. */
+void balloon_init_watcher(void);
+
 #endif /* __ASM_BALLOON_H__ */

linux-2.6-xen-sparse/include/asm-xen/xenbus.h

@@ -115,4 +115,8 @@ struct xenbus_watch
 int register_xenbus_watch(struct xenbus_watch *watch);
 void unregister_xenbus_watch(struct xenbus_watch *watch);
 
+/* Called from xen core code. */
+void xenbus_suspend(void);
+void xenbus_resume(void);
+
 #endif /* _ASM_XEN_XENBUS_H */

tools/libxc/xc.h

@@ -279,9 +279,12 @@ int xc_linux_save(int xc_handle, int fd, u32 dom);
  * @parm fd the file descriptor to restore a domain from
  * @parm dom the id of the domain
  * @parm nr_pfns the number of pages
+ * @parm store_evtchn the store event channel for this domain to use
+ * @parm store_mfn returned with the mfn of the store page
  * @return 0 on success, -1 on failure
  */
-int xc_linux_restore(int xc_handle, int io_fd, u32 dom, unsigned long nr_pfns);
+int xc_linux_restore(int xc_handle, int io_fd, u32 dom, unsigned long nr_pfns,
+		     unsigned int store_evtchn, unsigned long *store_mfn);
 
 int xc_linux_build(int xc_handle,
                    u32 domid,

tools/libxc/xc_linux_restore.c

@@ -48,7 +48,8 @@ read_exact(int fd, void *buf, size_t count)
     return r;
 }
 
-int xc_linux_restore(int xc_handle, int io_fd, u32 dom, unsigned long nr_pfns)
+int xc_linux_restore(int xc_handle, int io_fd, u32 dom, unsigned long nr_pfns,
+		     unsigned int store_evtchn, unsigned long *store_mfn)
 {
     dom0_op_t op;
     int rc = 1, i, n, k;
@@ -464,10 +465,13 @@ int xc_linux_restore(int xc_handle, int io_fd, u32 dom, unsigned long nr_pfns)
     }
     ctxt.user_regs.esi = mfn = pfn_to_mfn_table[pfn];
     p_srec = xc_map_foreign_range(
-        xc_handle, dom, PAGE_SIZE, PROT_WRITE, mfn);
+        xc_handle, dom, PAGE_SIZE, PROT_READ | PROT_WRITE, mfn);
     p_srec->resume_info.nr_pages    = nr_pfns;
     p_srec->resume_info.shared_info = shared_info_frame << PAGE_SHIFT;
     p_srec->resume_info.flags       = 0;
+    *store_mfn = p_srec->resume_info.store_mfn   =
+	pfn_to_mfn_table[p_srec->resume_info.store_mfn];
+    p_srec->resume_info.store_evtchn = store_evtchn;
     munmap(p_srec, PAGE_SIZE);
 
     /* Uncanonicalise each GDT frame number. */

tools/libxc/xc_linux_save.c

@@ -20,7 +20,7 @@
 #define DEBUG 0
 
 #if 1
-#define ERR(_f, _a...) fprintf ( stderr, _f , ## _a )
+#define ERR(_f, _a...) do { fprintf(stderr, _f , ## _a); fflush(stderr); } while (0)
 #else
 #define ERR(_f, _a...) ((void)0)
 #endif
@@ -643,6 +643,22 @@ int xc_linux_save(int xc_handle, int io_fd, u32 dom)
         goto out;
     }
 
+    /* Map the suspend-record MFN to pin it. The page must be owned by 
+       dom for this to succeed. */
+    p_srec = xc_map_foreign_range(xc_handle, dom,
+                                   sizeof(*p_srec), PROT_READ | PROT_WRITE, 
+                                   ctxt.user_regs.esi);
+    if (!p_srec){
+        ERR("Couldn't map suspend record");
+        goto out;
+    }
+
+    /* Canonicalize store mfn. */
+    if ( !translate_mfn_to_pfn(&p_srec->resume_info.store_mfn) ) {
+	ERR("Store frame is not in range of pseudophys map");
+	goto out;
+    }
+
     print_stats( xc_handle, dom, 0, &stats, 0 );
 
     /* Now write out each data page, canonicalising page tables as we go... */
@@ -983,16 +999,6 @@ int xc_linux_save(int xc_handle, int io_fd, u32 dom)
 	}
     }
 
-    /* Map the suspend-record MFN to pin it. The page must be owned by 
-       dom for this to succeed. */
-    p_srec = xc_map_foreign_range(xc_handle, dom,
-                                   sizeof(*p_srec), PROT_READ, 
-                                   ctxt.user_regs.esi);
-    if (!p_srec){
-        ERR("Couldn't map suspend record");
-        goto out;
-    }
-
     if (nr_pfns != p_srec->nr_pfns )
     {
 	ERR("Suspend record nr_pfns unexpected (%ld != %ld)",

tools/python/xen/lowlevel/xs/xs.c

@@ -254,11 +254,11 @@ static PyObject *xspy_get_permissions(PyObject *self, PyObject *args,
     val = PyList_New(perms_n);
     for (i = 0; i < perms_n; i++, perms++) {
         PyObject *p = Py_BuildValue("{s:i,s:i,s:i,s:i,s:i}",
-                                    "dom",    perms->id,
-                                    "read",   (perms->perms & XS_PERM_READ),
-                                    "write",  (perms->perms & XS_PERM_WRITE),
-                                    "create", (perms->perms & XS_PERM_CREATE),
-                                    "owner",  (perms->perms & XS_PERM_OWNER));
+                                    "dom",   perms->id,
+                                    "read",  (perms->perms & XS_PERM_READ),
+                                    "write", (perms->perms & XS_PERM_WRITE),
+                                    "exist", (perms->perms & XS_PERM_ENOENT_OK),
+                                    "owner", (perms->perms & XS_PERM_OWNER));
         PyList_SetItem(val, i, p);
     }
  exit:

tools/python/xen/xend/XendCheckpoint.py

@@ -6,6 +6,7 @@
 
 import errno
 import os
+import re
 import select
 import sxp
 from string import join
@@ -64,6 +65,13 @@ def save(xd, fd, dominfo):
                 if l.rstrip() == "suspend":
                     log.info("suspending %d" % dominfo.id)
                     xd.domain_shutdown(dominfo.id, reason='suspend')
+                    if dominfo.store_channel:
+                        try:
+                            dominfo.db.releaseDomain(dominfo.id)
+                        except Exception, ex:
+                            log.warning("error in domain release on xenstore: %s",
+                                        ex)
+                            pass
                     dominfo.state_wait("suspended")
                     log.info("suspend %d done" % dominfo.id)
                     child.tochild.write("done\n")
@@ -76,6 +84,11 @@ def save(xd, fd, dominfo):
     if child.wait() != 0:
         raise XendError("xc_save failed: %s" % lasterr)
 
+    if dominfo.store_channel:
+        dominfo.store_channel.close()
+        dominfo.db['store_channel'].delete()
+        dominfo.db.saveDB(save=True)
+        dominfo.store_channel = None
     xd.domain_destroy(dominfo.id)
     return None
 
@@ -107,8 +120,13 @@ def restore(xd, fd):
         raise XendError(
             "not a valid guest state file: pfn count out of range")
 
+    if dominfo.store_channel:
+        evtchn = dominfo.store_channel.port2
+    else:
+        evtchn = 0
+
     cmd = [PATH_XC_RESTORE, str(xc.handle()), str(fd),
-           str(dominfo.id), str(nr_pfns)]
+           str(dominfo.id), str(nr_pfns), str(evtchn)]
     log.info("[xc_restore] " + join(cmd))
     child = xPopen3(cmd, True, -1, [fd, xc.handle()])
     child.tochild.close()
@@ -128,7 +146,21 @@ def restore(xd, fd):
                 lasterr = l.rstrip()
             if fd == child.fromchild.fileno():
                 l = child.fromchild.readline()
-                log.info(l.rstrip())
+                while l:
+                    m = re.match(r"^(store-mfn) (\d+)\n$", l)
+                    if m:
+                        if dominfo.store_channel:
+                            dominfo.store_mfn = int(m.group(2))
+                            if dominfo.store_mfn >= 0:
+                                dominfo.db.introduceDomain(dominfo.id,
+                                                           dominfo.store_mfn,
+                                                           dominfo.store_channel)
+                            dominfo.exportToDB(save=True, sync=True)
+                    log.info(l.rstrip())
+                    try:
+                        l = child.fromchild.readline()
+                    except:
+                        l = None
         if filter(lambda (fd, event): event & select.POLLHUP, r):
             break
 

tools/xcutils/xc_restore.c

@@ -16,15 +16,23 @@
 int
 main(int argc, char **argv)
 {
-    unsigned int xc_fd, io_fd, domid, nr_pfns;
+    unsigned int xc_fd, io_fd, domid, nr_pfns, evtchn;
+    int ret;
+    unsigned long mfn;
 
-    if (argc != 5)
-	errx(1, "usage: %s xcfd iofd domid nr_pfns", argv[0]);
+    if (argc != 6)
+	errx(1, "usage: %s xcfd iofd domid nr_pfns evtchn", argv[0]);
 
     xc_fd = atoi(argv[1]);
     io_fd = atoi(argv[2]);
     domid = atoi(argv[3]);
     nr_pfns = atoi(argv[4]);
+    evtchn = atoi(argv[5]);
 
-    return xc_linux_restore(xc_fd, io_fd, domid, nr_pfns);
+    ret = xc_linux_restore(xc_fd, io_fd, domid, nr_pfns, evtchn, &mfn);
+    if (ret == 0) {
+	printf("store-mfn %li\n", mfn);
+	fflush(stdout);
+    }
+    return ret;
 }

tools/xenstore/testsuite/07watch.sh

@@ -160,3 +160,22 @@ async 2 write /test3 create contents
 1 waitwatch' | ./xs_test 2>&1`" = "1:/test2/foo:token
 1:contents2
 1:waitwatch timeout" ]
+
+# We can watch something which doesn't exist.
+[ "`echo '1 watch /dir/subdir token
+2 mkdir /dir/subdir
+1 waitwatch' | ./xs_test 2>&1`" = "1:/dir/subdir:token" ]
+
+# If we don't have permission, we won't see event (rm).
+[ "`echo '1 setid 1
+1 watch /dir/subdir token
+setperm /dir 0 NONE
+rm /dir/subdir
+1 waitwatch' | ./xs_test 2>&1`" = "1:waitwatch timeout" ]
+
+# If we don't have permission, we won't see event (create).
+[ "`echo '1 setid 1
+1 watch /dir/subdir token
+mkdir /dir/subdir
+write /dir/subdir/entry create contents
+1 waitwatch' | ./xs_test 2>&1`" = "1:waitwatch timeout" ]

tools/xenstore/testsuite/test.sh

@@ -14,7 +14,10 @@ run_test()
 	PID=`cat /tmp/pid`
 	rm /tmp/pid
     else
-	PID=`./xenstored_test --output-pid`
+	./xenstored_test --output-pid --trace-file=testsuite/tmp/trace --no-fork > /tmp/pid 2> testsuite/tmp/xenstored_errors &
+	while [ ! -s /tmp/pid ]; do sleep 0; done
+	PID=`cat /tmp/pid`
+	rm /tmp/pid
     fi
     if sh -e $2 $1; then
 	if [ -s testsuite/tmp/vgout ]; then

tools/xenstore/xenstored_core.c

@@ -504,11 +504,13 @@ void send_error(struct connection *conn, int error)
 {
 	unsigned int i;
 
-	for (i = 0; error != xsd_errors[i].errnum; i++)
-		if (i == ARRAY_SIZE(xsd_errors) - 1)
-			corrupt(conn, "Unknown error %i (%s)", error,
-				strerror(error));
-
+	for (i = 0; error != xsd_errors[i].errnum; i++) {
+		if (i == ARRAY_SIZE(xsd_errors) - 1) {
+			eprintf("xenstored: error %i untranslatable", error);
+			i = 0; 	/* EINVAL */
+			break;
+		}
+	}
 	send_reply(conn, XS_ERROR, xsd_errors[i].errstring,
 			  strlen(xsd_errors[i].errstring) + 1);
 }
@@ -705,7 +707,7 @@ static enum xs_perm_type perm_for_id(domid_t id,
 
 	/* Owners and tools get it all... */
 	if (!id || perms[0].id == id)
-		return XS_PERM_READ|XS_PERM_WRITE|XS_PERM_CREATE|XS_PERM_OWNER;
+		return XS_PERM_READ|XS_PERM_WRITE|XS_PERM_OWNER;
 
 	for (i = 1; i < num; i++)
 		if (perms[i].id == id)
@@ -714,20 +716,13 @@ static enum xs_perm_type perm_for_id(domid_t id,
 	return perms[0].perms;
 }
 
-/* We have a weird permissions system.  You can allow someone into a
- * specific node without allowing it in the parents.  If it's going to
- * fail, however, we don't want the errno to indicate any information
- * about the node. */
-static int check_with_parents(struct connection *conn, const char *node,
-			      int errnum)
+/* What do parents say? */
+static enum xs_perm_type ask_parents(struct connection *conn,
+				     const char *node)
 {
 	struct xs_permissions *perms;
 	unsigned int num;
 
-	/* We always tell them about memory failures. */
-	if (errnum == ENOMEM)
-		return errnum;
-
 	do {
 		node = get_parent(node);
 		perms = get_perms(conn->transaction, node, &num);
@@ -739,10 +734,23 @@ static int check_with_parents(struct connection *conn, const char *node,
 	if (!perms)
 		corrupt(conn, "No permissions file at root");
 
-	if (!(perm_for_id(conn->id, perms, num) & XS_PERM_READ))
-		return EACCES;
+	return perm_for_id(conn->id, perms, num);
+}
 
-	return errnum;
+/* We have a weird permissions system.  You can allow someone into a
+ * specific node without allowing it in the parents.  If it's going to
+ * fail, however, we don't want the errno to indicate any information
+ * about the node. */
+static int errno_from_parents(struct connection *conn, const char *node,
+			      int errnum)
+{
+	/* We always tell them about memory failures. */
+	if (errnum == ENOMEM)
+		return errnum;
+
+	if (ask_parents(conn, node) & XS_PERM_READ)
+		return errnum;
+	return EACCES;
 }
 
 char *canonicalize(struct connection *conn, const char *node)
@@ -774,24 +782,26 @@ bool check_node_perms(struct connection *conn, const char *node,
 	}
 
 	perms = get_perms(conn->transaction, node, &num);
-	/* No permissions.  If we want to create it and
-	 * it doesn't exist, check parent directory. */
-	if (!perms && errno == ENOENT && (perm & XS_PERM_CREATE)) {
-		char *parent = get_parent(node);
-		if (!parent)
-			return false;
 
-		perms = get_perms(conn->transaction, parent, &num);
-	}
-	if (!perms) {
-		errno = check_with_parents(conn, node, errno);
+	if (perms) {
+		if (perm_for_id(conn->id, perms, num) & perm)
+			return true;
+		errno = EACCES;
 		return false;
 	}
 
-	if (perm_for_id(conn->id, perms, num) & perm)
-		return true;
+	/* If it's OK not to exist, we consult parents. */
+	if (errno == ENOENT && (perm & XS_PERM_ENOENT_OK)) {
+		if (ask_parents(conn, node) & perm)
+			return true;
+		/* Parents say they should not know. */
+		errno = EACCES;
+		return false;
+	}
 
-	errno = check_with_parents(conn, node, EACCES);
+	/* They might not have permission to even *see* this node, in
+	 * which case we return EACCES even if it's ENOENT or EIO. */
+	errno = errno_from_parents(conn, node, errno);
 	return false;
 }
 
@@ -928,9 +938,9 @@ static void do_write(struct connection *conn, struct buffered_data *in)
 	if (streq(vec[1], XS_WRITE_NONE))
 		mode = XS_PERM_WRITE;
 	else if (streq(vec[1], XS_WRITE_CREATE))
-		mode = XS_PERM_WRITE|XS_PERM_CREATE;
+		mode = XS_PERM_WRITE|XS_PERM_ENOENT_OK;
 	else if (streq(vec[1], XS_WRITE_CREATE_EXCL))
-		mode = XS_PERM_WRITE|XS_PERM_CREATE;
+		mode = XS_PERM_WRITE|XS_PERM_ENOENT_OK;
 	else {
 		send_error(conn, EINVAL);
 		return;
@@ -949,7 +959,7 @@ static void do_write(struct connection *conn, struct buffered_data *in)
 		}
 
 		/* Not going to create it? */
-		if (!(mode & XS_PERM_CREATE)) {
+		if (streq(vec[1], XS_WRITE_NONE)) {
 			send_error(conn, ENOENT);
 			return;
 		}
@@ -983,7 +993,7 @@ static void do_write(struct connection *conn, struct buffered_data *in)
 static void do_mkdir(struct connection *conn, const char *node)
 {
 	node = canonicalize(conn, node);
-	if (!check_node_perms(conn, node, XS_PERM_WRITE|XS_PERM_CREATE)) {
+	if (!check_node_perms(conn, node, XS_PERM_WRITE|XS_PERM_ENOENT_OK)) {
 		send_error(conn, errno);
 		return;
 	}

tools/xenstore/xenstored_watch.c

@@ -95,10 +95,19 @@ static int destroy_watch_event(void *_event)
 	return 0;
 }
 
-static void add_event(struct watch *watch, const char *node)
+static void add_event(struct connection *conn,
+		      struct watch *watch, const char *node)
 {
 	struct watch_event *event;
 
+	/* Check read permission: no permission, no watch event.
+	 * If it doesn't exist, we need permission to read parent.
+	 */
+	if (!check_node_perms(conn, node, XS_PERM_READ|XS_PERM_ENOENT_OK)) {
+		fprintf(stderr, "No permission for %s\n", node);
+		return;
+	}
+
 	if (watch->relative_path) {
 		node += strlen(watch->relative_path);
 		if (*node == '/') /* Could be "" */
@@ -132,9 +141,9 @@ void fire_watches(struct connection *conn, const char *node, bool recurse)
 
 		list_for_each_entry(watch, &i->watches, list) {
 			if (is_child(node, watch->node))
-				add_event(watch, node);
+				add_event(i, watch, node);
 			else if (recurse && is_child(watch->node, node))
-				add_event(watch, watch->node);
+				add_event(i, watch, watch->node);
 			else
 				continue;
 			/* If connection not doing anything, queue this. */
@@ -206,7 +215,7 @@ void do_watch(struct connection *conn, struct buffered_data *in)
 
 	relative = !strstarts(vec[0], "/");
 	vec[0] = canonicalize(conn, vec[0]);
-	if (!check_node_perms(conn, vec[0], XS_PERM_READ)) {
+	if (!is_valid_nodename(vec[0])) {
 		send_error(conn, errno);
 		return;
 	}

tools/xenstore/xs_lib.h

@@ -30,7 +30,7 @@ enum xs_perm_type {
 	XS_PERM_READ = 1,
 	XS_PERM_WRITE = 2,
 	/* Internal use. */
-	XS_PERM_CREATE = 4,
+	XS_PERM_ENOENT_OK = 4,
 	XS_PERM_OWNER = 8,
 };