x86/extable: Adjust extable handling to be shadow stack compatible

When adjusting an IRET frame to recover from a fault, and equivalent
adjustment needs making in the shadow IRET frame.

The adjustment in exception_with_ints_disabled() could in principle be an
alternative block rather than an ifdef, as the only two current users of
_PRE_EXTABLE() are IRET-to-guest instructions.  However, this is not a
fastpath, and this form is more robust to future changes.

Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>

diff --git a/xen/arch/x86/traps.c b/xen/arch/x86/traps.c
index 39a1f35bc175f09c2d5db01a4b5fcff1bb2050d7..bc5b8f8ea3975a28e86b27f0aa7f68a5a17e6fbe 100644 (file)
--- a/xen/arch/x86/traps.c
+++ b/xen/arch/x86/traps.c
@@ -363,7 +363,7 @@ static void show_guest_stack(struct vcpu *v, const struct cpu_user_regs *regs)
 }
 
 /*
- * Notes for get_stack_trace_bottom() and get_stack_dump_bottom()
+ * Notes for get_{stack,shstk}*_bottom() helpers
  *
  * Stack pages 1 - 4:
  *   These are all 1-page IST stacks.  Each of these stacks have an exception
@@ -400,6 +400,18 @@ unsigned long get_stack_trace_bottom(unsigned long sp)
     }
 }
 
+static unsigned long get_shstk_bottom(unsigned long sp)
+{
+    switch ( get_stack_page(sp) )
+    {
+#ifdef CONFIG_XEN_SHSTK
+    case 0:  return ROUNDUP(sp, IST_SHSTK_SIZE) - sizeof(unsigned long);
+    case 5:  return ROUNDUP(sp, PAGE_SIZE)      - sizeof(unsigned long);
+#endif
+    default: return sp - sizeof(unsigned long);
+    }
+}
+
 unsigned long get_stack_dump_bottom(unsigned long sp)
 {
     switch ( get_stack_page(sp) )
@@ -763,6 +775,55 @@ static void do_reserved_trap(struct cpu_user_regs *regs)
           trapnr, vec_name(trapnr), regs->error_code);
 }
 
+static void extable_shstk_fixup(struct cpu_user_regs *regs, unsigned long fixup)
+{
+    unsigned long ssp, *ptr, *base;
+
+    asm ( "rdsspq %0" : "=r" (ssp) : "0" (1) );
+    if ( ssp == 1 )
+        return;
+
+    ptr = _p(ssp);
+    base = _p(get_shstk_bottom(ssp));
+
+    for ( ; ptr < base; ++ptr )
+    {
+        /*
+         * Search for %rip.  The shstk currently looks like this:
+         *
+         *   ...  [Likely pointed to by SSP]
+         *   %cs  [== regs->cs]
+         *   %rip [== regs->rip]
+         *   SSP  [Likely points to 3 slots higher, above %cs]
+         *   ...  [call tree to this function, likely 2/3 slots]
+         *
+         * and we want to overwrite %rip with fixup.  There are two
+         * complications:
+         *   1) We cant depend on SSP values, because they won't differ by 3
+         *      slots if the exception is taken on an IST stack.
+         *   2) There are synthetic (unrealistic but not impossible) scenarios
+         *      where %rip can end up in the call tree to this function, so we
+         *      can't check against regs->rip alone.
+         *
+         * Check for both regs->rip and regs->cs matching.
+         */
+        if ( ptr[0] == regs->rip && ptr[1] == regs->cs )
+        {
+            asm ( "wrssq %[fix], %[stk]"
+                  : [stk] "=m" (ptr[0])
+                  : [fix] "r" (fixup) );
+            return;
+        }
+    }
+
+    /*
+     * We failed to locate and fix up the shadow IRET frame.  This could be
+     * due to shadow stack corruption, or bad logic above.  We cannot continue
+     * executing the interrupted context.
+     */
+    BUG();
+}
+
 static bool extable_fixup(struct cpu_user_regs *regs, bool print)
 {
     unsigned long fixup = search_exception_table(regs);
@@ -779,6 +840,9 @@ static bool extable_fixup(struct cpu_user_regs *regs, bool print)
                vec_name(regs->entry_vector), regs->error_code,
                _p(regs->rip), _p(regs->rip), _p(fixup));
 
+    if ( IS_ENABLED(CONFIG_XEN_SHSTK) )
+        extable_shstk_fixup(regs, fixup);
+
     regs->rip = fixup;
     this_cpu(last_extable_addr) = regs->rip;
 

diff --git a/xen/arch/x86/x86_64/entry.S b/xen/arch/x86/x86_64/entry.S
index f7ee3dce9170929249af272b484b1f7049c2fd43..e68d680791aab283b28b32aada8454ac8bfe817e 100644 (file)
--- a/xen/arch/x86/x86_64/entry.S
+++ b/xen/arch/x86/x86_64/entry.S
@@ -708,7 +708,16 @@ exception_with_ints_disabled:
         call  search_pre_exception_table
         testq %rax,%rax                 # no fixup code for faulting EIP?
         jz    1b
-        movq  %rax,UREGS_rip(%rsp)
+        movq  %rax,UREGS_rip(%rsp)      # fixup regular stack
+
+#ifdef CONFIG_XEN_SHSTK
+        mov    $1, %edi
+        rdsspq %rdi
+        cmp    $1, %edi
+        je     .L_exn_shstk_done
+        wrssq  %rax, 1*8(%rdi)          # fixup shadow stack
+.L_exn_shstk_done:
+#endif
         subq  $8,UREGS_rsp(%rsp)        # add ec/ev to previous stack frame
         testb $15,UREGS_rsp(%rsp)       # return %rsp is now aligned?
         jz    1f                        # then there is a pad quadword already