Various fixes to Xen stack management. Ensure the cpu0_stack

is always sufficiently aligned (requires some linker script
tricks). Small fixes to show_trace() and ensure that Xen
stack pointers are always below the 'struct cpu_info' region.

Signed-off-by: Keir Fraser <keir@xensource.com>

10 files changed, 127 insertions, 110 deletions

diff --git a/.hgignore b/.hgignore
index 766a688126..9a52e9ce12 100644
--- a/.hgignore
+++ b/.hgignore
@@ -181,6 +181,7 @@
 ^xen/TAGS$
 ^xen/arch/x86/asm-offsets\.s$
 ^xen/arch/x86/boot/mkelf32$
+^xen/arch/x86/xen\.lds$
 ^xen/ddb/.*$
 ^xen/include/asm$
 ^xen/include/asm-.*/asm-offsets\.h$
diff --git a/xen/arch/x86/Makefile b/xen/arch/x86/Makefile
index 54fcf82dae..9cdbc5798c 100644
--- a/xen/arch/x86/Makefile
+++ b/xen/arch/x86/Makefile
@@ -29,6 +29,7 @@ ifeq ($(TARGET_SUBARCH),x86_32)
 endif
 
 OBJS := $(subst $(TARGET_SUBARCH)/asm-offsets.o,,$(OBJS))
+OBJS := $(subst $(TARGET_SUBARCH)/xen.lds.o,,$(OBJS))
 
 ifneq ($(crash_debug),y)
 OBJS := $(patsubst cdb%.o,,$(OBJS))
@@ -43,22 +44,25 @@ $(TARGET): $(TARGET)-syms boot/mkelf32
 $(CURDIR)/arch.o: $(OBJS)
 	$(LD) $(LDFLAGS) -r -o $@ $(OBJS)
 
-$(TARGET)-syms: boot/$(TARGET_SUBARCH).o $(ALL_OBJS) $(TARGET_SUBARCH)/xen.lds
-	$(LD) $(LDFLAGS) -T $(TARGET_SUBARCH)/xen.lds -N \
+$(TARGET)-syms: boot/$(TARGET_SUBARCH).o $(ALL_OBJS) xen.lds
+	$(LD) $(LDFLAGS) -T xen.lds -N \
 	    boot/$(TARGET_SUBARCH).o $(ALL_OBJS) -o $@
 	$(NM) -n $@ | $(BASEDIR)/tools/symbols >$(BASEDIR)/xen-syms.S
 	$(MAKE) $(BASEDIR)/xen-syms.o
-	$(LD) $(LDFLAGS) -T $(TARGET_SUBARCH)/xen.lds -N \
+	$(LD) $(LDFLAGS) -T xen.lds -N \
 	    boot/$(TARGET_SUBARCH).o $(ALL_OBJS) $(BASEDIR)/xen-syms.o -o $@
 	$(NM) -n $@ | $(BASEDIR)/tools/symbols >$(BASEDIR)/xen-syms.S
 	$(MAKE) $(BASEDIR)/xen-syms.o
-	$(LD) $(LDFLAGS) -T $(TARGET_SUBARCH)/xen.lds -N \
+	$(LD) $(LDFLAGS) -T xen.lds -N \
 	    boot/$(TARGET_SUBARCH).o $(ALL_OBJS) $(BASEDIR)/xen-syms.o -o $@
 	rm -f $(BASEDIR)/xen-syms.S $(BASEDIR)/xen-syms.o
 
 asm-offsets.s: $(TARGET_SUBARCH)/asm-offsets.c $(HDRS)
 	$(CC) $(CFLAGS) -S -o $@ $<
 
+xen.lds: $(TARGET_SUBARCH)/xen.lds.S $(HDRS)
+	$(CC) $(CFLAGS) -P -E -Ui386 -D__ASSEMBLY__ -o $@ $<
+
 boot/mkelf32: boot/mkelf32.c
 	$(HOSTCC) $(HOSTCFLAGS) -o $@ $<
 
@@ -73,5 +77,6 @@ clean:
 	rm -f dm/*.o dm/*~ dm/core
 	rm -f genapic/*.o genapic/*~ genapic/core
 	rm -f cpu/*.o cpu/*~ cpu/core
+	rm -f xen.lds
 
 .PHONY: default clean
diff --git a/xen/arch/x86/boot/x86_32.S b/xen/arch/x86/boot/x86_32.S
index 59e68ca7ef..d5a85e77fc 100644
--- a/xen/arch/x86/boot/x86_32.S
+++ b/xen/arch/x86/boot/x86_32.S
@@ -189,7 +189,7 @@ ignore_int:
 /*** STACK LOCATION ***/
         
 ENTRY(stack_start)
-        .long cpu0_stack + STACK_SIZE - 200 - __PAGE_OFFSET
+        .long cpu0_stack + STACK_SIZE - __PAGE_OFFSET
         .long __HYPERVISOR_DS
         
 /*** DESCRIPTOR TABLES ***/
@@ -256,10 +256,6 @@ ENTRY(idle_pg_table_l2)
         .fill 1*PAGE_SIZE,1,0
 #endif
 
-#if (STACK_ORDER == 0)
-.section ".bss.page_aligned","w"
-#else
-.section ".bss.twopage_aligned","w"
-#endif
+.section ".bss.stack_aligned","w"
 ENTRY(cpu0_stack)
         .fill STACK_SIZE,1,0
diff --git a/xen/arch/x86/boot/x86_64.S b/xen/arch/x86/boot/x86_64.S
index 126850a0f8..0d82c2087f 100644
--- a/xen/arch/x86/boot/x86_64.S
+++ b/xen/arch/x86/boot/x86_64.S
@@ -219,7 +219,7 @@ idt:
         .quad   idt_table
 
 ENTRY(stack_start)
-        .quad   cpu0_stack + STACK_SIZE - 200
+        .quad   cpu0_stack + STACK_SIZE
 
 high_start:
         .quad   __high_start
@@ -265,10 +265,6 @@ ENTRY(idle_pg_table_l2)
         .org 0x4000 + PAGE_SIZE
         .code64
 
-#if (STACK_ORDER == 0)
-.section ".bss.page_aligned","w"
-#else
-.section ".bss.twopage_aligned","w"
-#endif
+.section ".bss.stack_aligned","w"
 ENTRY(cpu0_stack)
         .fill STACK_SIZE,1,0
diff --git a/xen/arch/x86/domain.c b/xen/arch/x86/domain.c
index 4c586226dd..850b88d141 100644
--- a/xen/arch/x86/domain.c
+++ b/xen/arch/x86/domain.c
@@ -98,7 +98,7 @@ void startup_cpu_idle_loop(void)
     cpu_set(smp_processor_id(), v->domain->cpumask);
     v->arch.schedule_tail = continue_idle_task;
 
-    idle_loop();
+    reset_stack_and_jump(idle_loop);
 }
 
 static long no_idt[2];
diff --git a/xen/arch/x86/setup.c b/xen/arch/x86/setup.c
index 1259e095e8..1d9b96f72f 100644
--- a/xen/arch/x86/setup.c
+++ b/xen/arch/x86/setup.c
@@ -138,10 +138,18 @@ static void __init do_initcalls(void)
         (*call)();
 }
 
-static void __init start_of_day(void)
+/* Variables handed off from __start_xen() to start_of_day(). */
+static unsigned long initial_images_start, initial_images_end;
+static multiboot_info_t *mbi;
+
+void __init start_of_day(void)
 {
     int i;
     unsigned long vgdt, gdt_pfn;
+    char *cmdline;
+    unsigned long _initrd_start = 0, _initrd_len = 0;
+    unsigned int initrdidx = 1;
+    module_t *mod = (module_t *)__va(mbi->mods_addr);
 
     early_cpu_init();
 
@@ -249,20 +257,93 @@ static void __init start_of_day(void)
     schedulers_start();
 
     watchdog_enable();
+
+    shadow_mode_init();
+
+    /* initialize access control security module */
+    acm_init(&initrdidx, mbi, initial_images_start);
+
+    /* Create initial domain 0. */
+    dom0 = do_createdomain(0, 0);
+    if ( dom0 == NULL )
+        panic("Error creating domain 0\n");
+
+    set_bit(_DOMF_privileged, &dom0->domain_flags);
+    /* post-create hooks sets security label */
+    acm_post_domain0_create(dom0->domain_id);
+
+    /* Grab the DOM0 command line. */
+    cmdline = (char *)(mod[0].string ? __va(mod[0].string) : NULL);
+    if ( cmdline != NULL )
+    {
+        static char dom0_cmdline[MAX_GUEST_CMDLINE];
+
+        /* Skip past the image name and copy to a local buffer. */
+        while ( *cmdline == ' ' ) cmdline++;
+        if ( (cmdline = strchr(cmdline, ' ')) != NULL )
+        {
+            while ( *cmdline == ' ' ) cmdline++;
+            strcpy(dom0_cmdline, cmdline);
+        }
+
+        cmdline = dom0_cmdline;
+
+        /* Append any extra parameters. */
+        if ( skip_ioapic_setup && !strstr(cmdline, "noapic") )
+            strcat(cmdline, " noapic");
+        if ( acpi_skip_timer_override &&
+             !strstr(cmdline, "acpi_skip_timer_override") )
+            strcat(cmdline, " acpi_skip_timer_override");
+        if ( (strlen(acpi_param) != 0) && !strstr(cmdline, "acpi=") )
+        {
+            strcat(cmdline, " acpi=");
+            strcat(cmdline, acpi_param);
+        }
+    }
+
+    if ( (initrdidx > 0) && (initrdidx < mbi->mods_count) )
+    {
+        _initrd_start = initial_images_start +
+            (mod[initrdidx].mod_start - mod[0].mod_start);
+        _initrd_len   = mod[initrdidx].mod_end - mod[initrdidx].mod_start;
+    }
+
+    /*
+     * We're going to setup domain0 using the module(s) that we stashed safely
+     * above our heap. The second module, if present, is an initrd ramdisk.
+     */
+    if ( construct_dom0(dom0,
+                        initial_images_start, 
+                        mod[0].mod_end-mod[0].mod_start,
+                        _initrd_start,
+                        _initrd_len,
+                        cmdline) != 0)
+        panic("Could not set up DOM0 guest OS\n");
+
+    /* Scrub RAM that is still free and so may go to an unprivileged domain. */
+    scrub_heap_pages();
+
+    init_trace_bufs();
+
+    /* Give up the VGA console if DOM0 is configured to grab it. */
+    console_endboot(cmdline && strstr(cmdline, "tty0"));
+
+    /* Hide UART from DOM0 if we're using it */
+    serial_endboot();
+
+    domain_unpause_by_systemcontroller(dom0);
+
+    startup_cpu_idle_loop();
 }
 
 #define EARLY_FAIL() for ( ; ; ) __asm__ __volatile__ ( "hlt" )
 
 static struct e820entry e820_raw[E820MAX];
 
-void __init __start_xen(multiboot_info_t *mbi)
+void __init __start_xen(multiboot_info_t *__mbi)
 {
-    char *cmdline;
-    module_t *mod = (module_t *)__va(mbi->mods_addr);
+    module_t *mod = (module_t *)__va(__mbi->mods_addr);
     unsigned long nr_pages, modules_length;
-    unsigned long initial_images_start, initial_images_end;
-    unsigned long _initrd_start = 0, _initrd_len = 0;
-    unsigned int initrdidx = 1;
     physaddr_t s, e;
     int i, e820_warn = 0, e820_raw_nr = 0, bytes = 0;
     struct ns16550_defaults ns16550 = {
@@ -271,6 +352,8 @@ void __init __start_xen(multiboot_info_t *mbi)
         .stop_bits = 1
     };
 
+    mbi = __mbi;
+
     /* Parse the command-line options. */
     if ( (mbi->flags & MBI_CMDLINE) && (mbi->cmdline != 0) )
         cmdline_parse(__va(mbi->cmdline));
@@ -486,84 +569,7 @@ void __init __start_xen(multiboot_info_t *mbi)
 
     early_boot = 0;
 
-    start_of_day();
-
-    shadow_mode_init();
-
-    /* initialize access control security module */
-    acm_init(&initrdidx, mbi, initial_images_start);
-
-    /* Create initial domain 0. */
-    dom0 = do_createdomain(0, 0);
-    if ( dom0 == NULL )
-        panic("Error creating domain 0\n");
-
-    set_bit(_DOMF_privileged, &dom0->domain_flags);
-    /* post-create hooks sets security label */
-    acm_post_domain0_create(dom0->domain_id);
-
-    /* Grab the DOM0 command line. */
-    cmdline = (char *)(mod[0].string ? __va(mod[0].string) : NULL);
-    if ( cmdline != NULL )
-    {
-        static char dom0_cmdline[MAX_GUEST_CMDLINE];
-
-        /* Skip past the image name and copy to a local buffer. */
-        while ( *cmdline == ' ' ) cmdline++;
-        if ( (cmdline = strchr(cmdline, ' ')) != NULL )
-        {
-            while ( *cmdline == ' ' ) cmdline++;
-            strcpy(dom0_cmdline, cmdline);
-        }
-
-        cmdline = dom0_cmdline;
-
-        /* Append any extra parameters. */
-        if ( skip_ioapic_setup && !strstr(cmdline, "noapic") )
-            strcat(cmdline, " noapic");
-        if ( acpi_skip_timer_override &&
-             !strstr(cmdline, "acpi_skip_timer_override") )
-            strcat(cmdline, " acpi_skip_timer_override");
-        if ( (strlen(acpi_param) != 0) && !strstr(cmdline, "acpi=") )
-        {
-            strcat(cmdline, " acpi=");
-            strcat(cmdline, acpi_param);
-        }
-    }
-
-    if ( (initrdidx > 0) && (initrdidx < mbi->mods_count) )
-    {
-        _initrd_start = initial_images_start +
-            (mod[initrdidx].mod_start - mod[0].mod_start);
-        _initrd_len   = mod[initrdidx].mod_end - mod[initrdidx].mod_start;
-    }
-
-    /*
-     * We're going to setup domain0 using the module(s) that we stashed safely
-     * above our heap. The second module, if present, is an initrd ramdisk.
-     */
-    if ( construct_dom0(dom0,
-                        initial_images_start, 
-                        mod[0].mod_end-mod[0].mod_start,
-                        _initrd_start,
-                        _initrd_len,
-                        cmdline) != 0)
-        panic("Could not set up DOM0 guest OS\n");
-
-    /* Scrub RAM that is still free and so may go to an unprivileged domain. */
-    scrub_heap_pages();
-
-    init_trace_bufs();
-
-    /* Give up the VGA console if DOM0 is configured to grab it. */
-    console_endboot(cmdline && strstr(cmdline, "tty0"));
-
-    /* Hide UART from DOM0 if we're using it */
-    serial_endboot();
-
-    domain_unpause_by_systemcontroller(dom0);
-
-    startup_cpu_idle_loop();
+    reset_stack_and_jump(start_of_day);
 }
 
 void arch_get_xen_caps(xen_capabilities_info_t info)
diff --git a/xen/arch/x86/traps.c b/xen/arch/x86/traps.c
index 40959e6758..64a6e2a5d7 100644
--- a/xen/arch/x86/traps.c
+++ b/xen/arch/x86/traps.c
@@ -192,7 +192,8 @@ static void show_trace(struct cpu_user_regs *regs)
 
     /* Bounds for range of valid frame pointer. */
     low  = (unsigned long)(ESP_BEFORE_EXCEPTION(regs) - 2);
-    high = (low & ~(STACK_SIZE - 1)) + (STACK_SIZE - sizeof(struct cpu_info));
+    high = (low & ~(STACK_SIZE - 1)) + 
+        (STACK_SIZE - sizeof(struct cpu_info) - 2*sizeof(unsigned long));
 
     /* The initial frame pointer. */
     next = regs->ebp;
@@ -200,14 +201,14 @@ static void show_trace(struct cpu_user_regs *regs)
     for ( ; ; )
     {
         /* Valid frame pointer? */
-        if ( (next < low) || (next > high) )
+        if ( (next < low) || (next >= high) )
         {
             /*
              * Exception stack frames have a different layout, denoted by an
              * inverted frame pointer.
              */
             next = ~next;
-            if ( (next < low) || (next > high) )
+            if ( (next < low) || (next >= high) )
                 break;
             frame = (unsigned long *)next;
             next  = frame[0];
diff --git a/xen/arch/x86/x86_32/xen.lds b/xen/arch/x86/x86_32/xen.lds.S
index f3c168f662..fb76cbc73d 100644
--- a/xen/arch/x86/x86_32/xen.lds
+++ b/xen/arch/x86/x86_32/xen.lds.S
@@ -2,6 +2,12 @@
  * Written by Martin Mares <mj@atrey.karlin.mff.cuni.cz>
  * Modified for i386 Xen by Keir Fraser
  */
+
+#include <xen/config.h>
+#include <asm/page.h>
+#undef ENTRY
+#undef ALIGN
+
 OUTPUT_FORMAT("elf32-i386", "elf32-i386", "elf32-i386")
 OUTPUT_ARCH(i386)
 ENTRY(start)
@@ -50,12 +56,12 @@ SECTIONS
   __initcall_start = .;
   .initcall.init : { *(.initcall.init) } :text
   __initcall_end = .;
-  . = ALIGN(8192);
+  . = ALIGN(STACK_SIZE);
   __init_end = .;
 
   __bss_start = .;		/* BSS */
   .bss : {
-	*(.bss.twopage_aligned)
+	*(.bss.stack_aligned)
 	*(.bss.page_aligned)
 	*(.bss)
 	} :text
diff --git a/xen/arch/x86/x86_64/xen.lds b/xen/arch/x86/x86_64/xen.lds.S
index 837d335f0d..d8685201ab 100644
--- a/xen/arch/x86/x86_64/xen.lds
+++ b/xen/arch/x86/x86_64/xen.lds.S
@@ -1,5 +1,11 @@
 /* Excerpts written by Martin Mares <mj@atrey.karlin.mff.cuni.cz> */
 /* Modified for x86-64 Xen by Keir Fraser */
+
+#include <xen/config.h>
+#include <asm/page.h>
+#undef ENTRY
+#undef ALIGN
+
 OUTPUT_FORMAT("elf64-x86-64", "elf64-x86-64", "elf64-x86-64")
 OUTPUT_ARCH(i386:x86-64)
 ENTRY(start)
@@ -48,12 +54,12 @@ SECTIONS
   __initcall_start = .;
   .initcall.init : { *(.initcall.init) } :text
   __initcall_end = .;
-  . = ALIGN(8192);
+  . = ALIGN(STACK_SIZE);
   __init_end = .;
 
   __bss_start = .;		/* BSS */
   .bss : {
-	*(.bss.twopage_aligned)
+	*(.bss.stack_aligned)
 	*(.bss.page_aligned)
 	*(.bss)
 	} :text
diff --git a/xen/include/asm-x86/current.h b/xen/include/asm-x86/current.h
index 7916b57ae9..2a63b75002 100644
--- a/xen/include/asm-x86/current.h
+++ b/xen/include/asm-x86/current.h
@@ -49,7 +49,7 @@ static inline struct cpu_info *get_cpu_info(void)
 #define reset_stack_and_jump(__fn)              \
     __asm__ __volatile__ (                      \
         "mov %0,%%"__OP"sp; jmp "STR(__fn)      \
-        : : "r" (guest_cpu_user_regs()) )
+        : : "r" (guest_cpu_user_regs()) : "memory" )
 
 #define schedule_tail(_ed) (((_ed)->arch.schedule_tail)(_ed))