x86: Update vfork to pop shadow stack

The shadow stack prevents us from pushing the saved return PC onto
the stack and returning normally.  Instead we pop the shadow stack
and return directly.  This is the safest way to return and ensures
any stack manipulations done by the vfork'd child doesn't cause the
parent to terminate when CET is enabled.

Reviewed-by: Carlos O'Donell <carlos@redhat.com>

	* sysdeps/unix/sysv/linux/i386/vfork.S (SYSCALL_ERROR_HANDLER):
	Redefine if shadow stack is enabled.
	(SYSCALL_ERROR_LABEL): Likewise.
	(__vfork): Pop shadow stack and jump back to to caller directly
	when shadow stack is in use.
	* sysdeps/unix/sysv/linux/x86_64/vfork.S (SYSCALL_ERROR_HANDLER):
	Redefine if shadow stack is enabled.
	(SYSCALL_ERROR_LABEL): Likewise.
	(__vfork): Pop shadow stack and jump back to to caller directly
	when shadow stack is in use.

ChangeLog

@@ -1,3 +1,16 @@
+2018-07-24  H.J. Lu  <hongjiu.lu@intel.com>
+
+	* sysdeps/unix/sysv/linux/i386/vfork.S (SYSCALL_ERROR_HANDLER):
+	Redefine if shadow stack is enabled.
+	(SYSCALL_ERROR_LABEL): Likewise.
+	(__vfork): Pop shadow stack and jump back to to caller directly
+	when shadow stack is in use.
+	* sysdeps/unix/sysv/linux/x86_64/vfork.S (SYSCALL_ERROR_HANDLER):
+	Redefine if shadow stack is enabled.
+	(SYSCALL_ERROR_LABEL): Likewise.
+	(__vfork): Pop shadow stack and jump back to to caller directly
+	when shadow stack is in use.
+
 2018-07-24  H.J. Lu  <hongjiu.lu@intel.com>
 
 	* sysdeps/x86_64/tst-quadmod1.S (func): Add endbr64 if IBT is

sysdeps/unix/sysv/linux/i386/vfork.S

@@ -21,6 +21,38 @@
 #include <bits/errno.h>
 #include <tcb-offsets.h>
 
+#if SHSTK_ENABLED
+/* The shadow stack prevents us from pushing the saved return PC onto
+   the stack and returning normally.  Instead we pop the shadow stack
+   and return directly.  This is the safest way to return and ensures
+   any stack manipulations done by the vfork'd child doesn't cause the
+   parent to terminate when CET is enabled.  */
+# undef SYSCALL_ERROR_HANDLER
+# ifdef PIC
+#  define SYSCALL_ERROR_HANDLER				\
+0:							\
+  calll .L1;						\
+.L1:							\
+  popl %edx;						\
+.L2:							\
+  addl $_GLOBAL_OFFSET_TABLE_ + (.L2 - .L1), %edx;	\
+  movl __libc_errno@gotntpoff(%edx), %edx;		\
+  negl %eax;						\
+  movl %eax, %gs:(%edx);				\
+  orl $-1, %eax;					\
+  jmp 1b;
+# else
+#  define SYSCALL_ERROR_HANDLER				\
+0:							\
+  movl __libc_errno@indntpoff, %edx;			\
+  negl %eax;						\
+  movl %eax, %gs:(%edx);				\
+  orl $-1, %eax;					\
+  jmp 1b;
+# endif
+# undef SYSCALL_ERROR_LABEL
+# define SYSCALL_ERROR_LABEL 0f
+#endif
 
 /* Clone the calling process, but without copying the whole address space.
    The calling process is suspended until the new process exits or is
@@ -38,16 +70,41 @@ ENTRY (__vfork)
 	movl	$SYS_ify (vfork), %eax
 	int	$0x80
 
+#if !SHSTK_ENABLED
 	/* Jump to the return PC.  Don't jump directly since this
 	   disturbs the branch target cache.  Instead push the return
 	   address back on the stack.  */
 	pushl	%ecx
 	cfi_adjust_cfa_offset (4)
+#endif
 
 	cmpl	$-4095, %eax
 	/* Branch forward if it failed.  */
 	jae	SYSCALL_ERROR_LABEL
 
+#if SHSTK_ENABLED
+1:
+	/* Check if shadow stack is in use.  */
+	xorl	%edx, %edx
+	rdsspd	%edx
+	testl	%edx, %edx
+	/* Normal return if shadow stack isn't in use.  */
+	je	L(no_shstk)
+
+	/* Pop return address from shadow stack and jump back to caller
+	   directly.  */
+	movl	$1, %edx
+	incsspd	%edx
+	jmp	*%ecx
+
+L(no_shstk):
+	/* Jump to the return PC.  Don't jump directly since this
+	   disturbs the branch target cache.  Instead push the return
+	   address back on the stack.  */
+	pushl	%ecx
+	cfi_adjust_cfa_offset (4)
+#endif
+
 	ret
 
 PSEUDO_END (__vfork)

sysdeps/unix/sysv/linux/x86_64/vfork.S

@@ -20,6 +20,21 @@
 #include <bits/errno.h>
 #include <tcb-offsets.h>
 
+#if SHSTK_ENABLED
+/* The shadow stack prevents us from pushing the saved return PC onto
+   the stack and returning normally.  Instead we pop the shadow stack
+   and return directly.  This is the safest way to return and ensures
+   any stack manipulations done by the vfork'd child doesn't cause the
+   parent to terminate when CET is enabled.  */
+# undef SYSCALL_ERROR_HANDLER
+# define SYSCALL_ERROR_HANDLER			\
+0:						\
+  SYSCALL_SET_ERRNO;				\
+  or $-1, %RAX_LP;				\
+  jmp 1b;
+# undef SYSCALL_ERROR_LABEL
+# define SYSCALL_ERROR_LABEL 0f
+#endif
 
 /* Clone the calling process, but without copying the whole address space.
    The calling process is suspended until the new process exits or is
@@ -38,13 +53,36 @@ ENTRY (__vfork)
 	movl	$SYS_ify (vfork), %eax
 	syscall
 
+#if !SHSTK_ENABLED
 	/* Push back the return PC.  */
 	pushq	%rdi
 	cfi_adjust_cfa_offset(8)
+#endif
 
 	cmpl	$-4095, %eax
 	jae SYSCALL_ERROR_LABEL		/* Branch forward if it failed.  */
 
+#if SHSTK_ENABLED
+1:
+	/* Check if shadow stack is in use.  */
+	xorl	%esi, %esi
+	rdsspq	%rsi
+	testq	%rsi, %rsi
+	/* Normal return if shadow stack isn't in use.  */
+	je	L(no_shstk)
+
+	/* Pop return address from shadow stack and jump back to caller
+	   directly.  */
+	movl	$1, %esi
+	incsspq	%rsi
+	jmp	*%rdi
+
+L(no_shstk):
+	/* Push back the return PC.  */
+	pushq	%rdi
+	cfi_adjust_cfa_offset(8)
+#endif
+
 	/* Normal return.  */
 	ret