x86: improve output resulting from sending '0' over serial

While the original logic already implied that the kernel part of the
guest's address space is identical on all vCPU-s (i.e. for all guest
processes), it didn't fully leverage the potential here: As long as
the top page table currently active is owned by the subject domain
(currently only Dom0), the stack dump can be done without extra
effort.

For x86-64, additionally add page table traversal so that the stack
can be dumped in all cases (unless it's invalid or user space).

I left the 32-bit variant of do_page_walk() unimplemented for the
moment as I couldn't convince myself using map_domain_page() there is
a good idea, and didn't want to introduce new fixmap entries either.

Signed-off-by: Jan Beulich <jbeulich@novell.com>

xen/arch/x86/traps.c

@@ -129,18 +129,18 @@ boolean_param("ler", opt_ler);
 #define ESP_BEFORE_EXCEPTION(regs) ((unsigned long *)regs->rsp)
 #endif
 
-static void show_guest_stack(struct cpu_user_regs *regs)
+static void show_guest_stack(struct vcpu *v, struct cpu_user_regs *regs)
 {
     int i;
-    struct vcpu *curr = current;
     unsigned long *stack, addr;
+    unsigned long mask = STACK_SIZE;
 
-    if ( is_hvm_vcpu(curr) )
+    if ( is_hvm_vcpu(v) )
         return;
 
-    if ( is_pv_32on64_vcpu(curr) )
+    if ( is_pv_32on64_vcpu(v) )
     {
-        compat_show_guest_stack(regs, debug_stack_lines);
+        compat_show_guest_stack(v, regs, debug_stack_lines);
         return;
     }
 
@@ -156,11 +156,42 @@ static void show_guest_stack(struct cpu_user_regs *regs)
         printk("Guest stack trace from "__OP"sp=%p:\n  ", stack);
     }
 
+    if ( !access_ok(stack, sizeof(*stack)) )
+    {
+        printk("Guest-inaccessible memory.\n");
+        return;
+    }
+
+    if ( v != current )
+    {
+        struct vcpu *vcpu;
+
+        ASSERT(guest_kernel_mode(v, regs));
+#ifndef __x86_64__
+        addr = read_cr3();
+        for_each_vcpu( v->domain, vcpu )
+            if ( vcpu->arch.cr3 == addr )
+                break;
+#else
+        vcpu = maddr_get_owner(read_cr3()) == v->domain ? v : NULL;
+#endif
+        if ( !vcpu )
+        {
+            stack = do_page_walk(v, (unsigned long)stack);
+            if ( (unsigned long)stack < PAGE_SIZE )
+            {
+                printk("Inaccessible guest memory.\n");
+                return;
+            }
+            mask = PAGE_SIZE;
+        }
+    }
+
     for ( i = 0; i < (debug_stack_lines*stack_words_per_line); i++ )
     {
-        if ( ((long)stack & (STACK_SIZE-BYTES_PER_LONG)) == 0 )
+        if ( (((long)stack - 1) ^ ((long)(stack + 1) - 1)) & mask )
             break;
-        if ( get_user(addr, stack) )
+        if ( __get_user(addr, stack) )
         {
             if ( i != 0 )
                 printk("\n    ");
@@ -264,7 +295,7 @@ void show_stack(struct cpu_user_regs *regs)
     int i;
 
     if ( guest_mode(regs) )
-        return show_guest_stack(regs);
+        return show_guest_stack(current, regs);
 
     printk("Xen stack trace from "__OP"sp=%p:\n  ", stack);
 
@@ -346,10 +377,8 @@ void vcpu_show_execution_state(struct vcpu *v)
     vcpu_pause(v); /* acceptably dangerous */
 
     vcpu_show_registers(v);
-    /* Todo: map arbitrary vcpu's top guest stack page here. */
-    if ( (v->domain == current->domain) &&
-         guest_kernel_mode(v, &v->arch.guest_context.user_regs) )
-        show_guest_stack(&v->arch.guest_context.user_regs);
+    if ( guest_kernel_mode(v, &v->arch.guest_context.user_regs) )
+        show_guest_stack(v, &v->arch.guest_context.user_regs);
 
     vcpu_unpause(v);
 }

xen/arch/x86/x86_32/mm.c

@@ -63,6 +63,11 @@ l2_pgentry_t *virt_to_xen_l2e(unsigned long v)
     return &idle_pg_table_l2[l2_linear_offset(v)];
 }
 
+void *do_page_walk(struct vcpu *v, unsigned long addr)
+{
+    return NULL;
+}
+
 void __init paging_init(void)
 {
     unsigned long v;

xen/arch/x86/x86_64/compat/traps.c

@@ -5,18 +5,46 @@
 #include <compat/callback.h>
 #include <compat/arch-x86_32.h>
 
-void compat_show_guest_stack(struct cpu_user_regs *regs, int debug_stack_lines)
+void compat_show_guest_stack(struct vcpu *v, struct cpu_user_regs *regs,
+                             int debug_stack_lines)
 {
-    unsigned int i, *stack, addr;
+    unsigned int i, *stack, addr, mask = STACK_SIZE;
 
     stack = (unsigned int *)(unsigned long)regs->_esp;
     printk("Guest stack trace from esp=%08lx:\n ", (unsigned long)stack);
 
+    if ( !__compat_access_ok(v->domain, stack, sizeof(*stack)) )
+    {
+        printk("Guest-inaccessible memory.\n");
+        return;
+    }
+
+    if ( v != current )
+    {
+        struct vcpu *vcpu;
+
+        ASSERT(guest_kernel_mode(v, regs));
+        addr = read_cr3() >> PAGE_SHIFT;
+        for_each_vcpu( v->domain, vcpu )
+            if ( pagetable_get_pfn(vcpu->arch.guest_table) == addr )
+                break;
+        if ( !vcpu )
+        {
+            stack = do_page_walk(v, (unsigned long)stack);
+            if ( (unsigned long)stack < PAGE_SIZE )
+            {
+                printk("Inaccessible guest memory.\n");
+                return;
+            }
+            mask = PAGE_SIZE;
+        }
+    }
+
     for ( i = 0; i < debug_stack_lines * 8; i++ )
     {
-        if ( (((long)stack + 3) & (STACK_SIZE - 4)) == 0 )
+        if ( (((long)stack - 1) ^ ((long)(stack + 1) - 1)) & mask )
             break;
-        if ( get_user(addr, stack) )
+        if ( __get_user(addr, stack) )
         {
             if ( i != 0 )
                 printk("\n    ");

xen/arch/x86/x86_64/mm.c

@@ -103,6 +103,47 @@ l2_pgentry_t *virt_to_xen_l2e(unsigned long v)
     return l3e_to_l2e(*pl3e) + l2_table_offset(v);
 }
 
+void *do_page_walk(struct vcpu *v, unsigned long addr)
+{
+    unsigned long mfn = pagetable_get_pfn(v->arch.guest_table);
+    l4_pgentry_t l4e, *l4t;
+    l3_pgentry_t l3e, *l3t;
+    l2_pgentry_t l2e, *l2t;
+    l1_pgentry_t l1e, *l1t;
+
+    if ( is_hvm_vcpu(v) )
+        return NULL;
+
+    l4t = mfn_to_virt(mfn);
+    l4e = l4t[l4_table_offset(addr)];
+    mfn = l4e_get_pfn(l4e);
+    if ( !(l4e_get_flags(l4e) & _PAGE_PRESENT) )
+        return NULL;
+
+    l3t = mfn_to_virt(mfn);
+    l3e = l3t[l3_table_offset(addr)];
+    mfn = l3e_get_pfn(l3e);
+    if ( !(l3e_get_flags(l3e) & _PAGE_PRESENT) || !mfn_valid(mfn) )
+    if ( (l3e_get_flags(l3e) & _PAGE_PSE) )
+        return mfn_to_virt(mfn) + (addr & ((1UL << L3_PAGETABLE_SHIFT) - 1));
+
+    l2t = mfn_to_virt(mfn);
+    l2e = l2t[l2_table_offset(addr)];
+    mfn = l2e_get_pfn(l2e);
+    if ( !(l2e_get_flags(l2e) & _PAGE_PRESENT) || !mfn_valid(mfn) )
+        return NULL;
+    if ( (l2e_get_flags(l2e) & _PAGE_PSE) )
+        return mfn_to_virt(mfn) + (addr & ((1UL << L2_PAGETABLE_SHIFT) - 1));
+
+    l1t = mfn_to_virt(mfn);
+    l1e = l1t[l1_table_offset(addr)];
+    mfn = l1e_get_pfn(l1e);
+    if ( !(l1e_get_flags(l1e) & _PAGE_PRESENT) || !mfn_valid(mfn) )
+        return NULL;
+
+    return mfn_to_virt(mfn) + (addr & ~PAGE_MASK);
+}
+
 void __init paging_init(void)
 {
     unsigned long i, mpt_size, va;

xen/include/asm-x86/mm.h

@@ -475,6 +475,7 @@ int new_guest_cr3(unsigned long pfn);
 void make_cr3(struct vcpu *v, unsigned long mfn);
 void update_cr3(struct vcpu *v);
 void propagate_page_fault(unsigned long addr, u16 error_code);
+void *do_page_walk(struct vcpu *v, unsigned long addr);
 
 int __sync_lazy_execstate(void);
 

xen/include/asm-x86/processor.h

@@ -536,9 +536,9 @@ void show_page_walk(unsigned long addr);
 asmlinkage void fatal_trap(int trapnr, struct cpu_user_regs *regs);
 
 #ifdef CONFIG_COMPAT
-void compat_show_guest_stack(struct cpu_user_regs *, int lines);
+void compat_show_guest_stack(struct vcpu *, struct cpu_user_regs *, int lines);
 #else
-#define compat_show_guest_stack(regs, lines) ((void)0)
+#define compat_show_guest_stack(vcpu, regs, lines) ((void)0)
 #endif
 
 extern void mtrr_ap_init(void);

xen/include/asm-x86/x86_64/uaccess.h

@@ -27,11 +27,14 @@ DECLARE_PER_CPU(char, compat_arg_xlat[COMPAT_ARG_XLAT_SIZE]);
 #define array_access_ok(addr, count, size) \
     (access_ok(addr, (count)*(size)))
 
-#define __compat_addr_ok(addr) \
-    ((unsigned long)(addr) < HYPERVISOR_COMPAT_VIRT_START(current->domain))
+#define __compat_addr_ok(d, addr) \
+    ((unsigned long)(addr) < HYPERVISOR_COMPAT_VIRT_START(d))
+
+#define __compat_access_ok(d, addr, size) \
+    __compat_addr_ok(d, (unsigned long)(addr) + ((size) ? (size) - 1 : 0))
 
 #define compat_access_ok(addr, size) \
-    __compat_addr_ok((unsigned long)(addr) + ((size) ? (size) - 1 : 0))
+    __compat_access_ok(current->domain, addr, size)
 
 #define compat_array_access_ok(addr,count,size) \
     (likely((count) < (~0U / (size))) && \