mm: add a NO_INHERIT flag to the PR_SET_MDWE prctl

This extends the current PR_SET_MDWE prctl arg with a bit to indicate that
the process doesn't want MDWE protection to propagate to children.

To implement this no-inherit mode, the tag in current->mm->flags must be
absent from MMF_INIT_MASK. This means that the encoding for "MDWE but
without inherit" is different in the prctl than in the mm flags. This
leads to a bit of bit-mangling in the prctl implementation.

Link: https://lkml.kernel.org/r/[e-mail]
Signed-off-by: Florent Revest <[e-mail]>
Reviewed-by: Kees Cook <[e-mail]>
Reviewed-by: Catalin Marinas <[e-mail]>
Cc: Alexey Izbyshev <[e-mail]>
Cc: Anshuman Khandual <[e-mail]>
Cc: Ayush Jain <[e-mail]>
Cc: David Hildenbrand <[e-mail]>
Cc: Greg Thelen <[e-mail]>
Cc: Joey Gouly <[e-mail]>
Cc: KP Singh <[e-mail]>
Cc: Mark Brown <[e-mail]>
Cc: Michal Hocko <[e-mail]>
Cc: Peter Xu <[e-mail]>
Cc: Ryan Roberts <[e-mail]>
Cc: Szabolcs Nagy <[e-mail]>
Cc: Topi Miettinen <[e-mail]>
Signed-off-by: Andrew Morton <[e-mail]>

include/linux/sched/coredump.h

···
 				 MMF_DISABLE_THP_MASK | MMF_HAS_MDWE_MASK)
 
 #define MMF_VM_MERGE_ANY	29
+#define MMF_HAS_MDWE_NO_INHERIT	30
+
+static inline unsigned long mmf_init_flags(unsigned long flags)
+{
+	if (flags & (1UL << MMF_HAS_MDWE_NO_INHERIT))
+		flags &= ~((1UL << MMF_HAS_MDWE) |
+			   (1UL << MMF_HAS_MDWE_NO_INHERIT));
+	return flags & MMF_INIT_MASK;
+}
+
 #endif /* _LINUX_SCHED_COREDUMP_H */

include/uapi/linux/prctl.h

···
 /* Memory deny write / execute */
 #define PR_SET_MDWE			65
 # define PR_MDWE_REFUSE_EXEC_GAIN	(1UL << 0)
+# define PR_MDWE_NO_INHERIT		(1UL << 1)
 
 #define PR_GET_MDWE			66
 

kernel/fork.c

···
 	hugetlb_count_init(mm);
 
 	if (current->mm) {
-		mm->flags = current->mm->flags & MMF_INIT_MASK;
+		mm->flags = mmf_init_flags(current->mm->flags);
 		mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK;
 	} else {
 		mm->flags = default_dump_filter;

kernel/sys.c

···
 }
 #endif /* CONFIG_ANON_VMA_NAME */
 
+static inline unsigned long get_current_mdwe(void)
+{
+	unsigned long ret = 0;
+
+	if (test_bit(MMF_HAS_MDWE, &current->mm->flags))
+		ret |= PR_MDWE_REFUSE_EXEC_GAIN;
+	if (test_bit(MMF_HAS_MDWE_NO_INHERIT, &current->mm->flags))
+		ret |= PR_MDWE_NO_INHERIT;
+
+	return ret;
+}
+
 static inline int prctl_set_mdwe(unsigned long bits, unsigned long arg3,
 				 unsigned long arg4, unsigned long arg5)
 {
+	unsigned long current_bits;
+
 	if (arg3 || arg4 || arg5)
 		return -EINVAL;
 
-	if (bits & ~(PR_MDWE_REFUSE_EXEC_GAIN))
+	if (bits & ~(PR_MDWE_REFUSE_EXEC_GAIN | PR_MDWE_NO_INHERIT))
 		return -EINVAL;
 
+	/* NO_INHERIT only makes sense with REFUSE_EXEC_GAIN */
+	if (bits & PR_MDWE_NO_INHERIT && !(bits & PR_MDWE_REFUSE_EXEC_GAIN))
+		return -EINVAL;
+
+	current_bits = get_current_mdwe();
+	if (current_bits && current_bits != bits)
+		return -EPERM; /* Cannot unset the flags */
+
+	if (bits & PR_MDWE_NO_INHERIT)
+		set_bit(MMF_HAS_MDWE_NO_INHERIT, &current->mm->flags);
 	if (bits & PR_MDWE_REFUSE_EXEC_GAIN)
 		set_bit(MMF_HAS_MDWE, &current->mm->flags);
-	else if (test_bit(MMF_HAS_MDWE, &current->mm->flags))
-		return -EPERM; /* Cannot unset the flag */
 
 	return 0;
 }
···
 {
 	if (arg2 || arg3 || arg4 || arg5)
 		return -EINVAL;
-
-	return test_bit(MMF_HAS_MDWE, &current->mm->flags) ?
-		PR_MDWE_REFUSE_EXEC_GAIN : 0;
+	return get_current_mdwe();
 }
 
 static int prctl_get_auxv(void __user *addr, unsigned long len)

tools/include/uapi/linux/prctl.h

···
 /* Memory deny write / execute */
 #define PR_SET_MDWE			65
 # define PR_MDWE_REFUSE_EXEC_GAIN	(1UL << 0)
+# define PR_MDWE_NO_INHERIT		(1UL << 1)
 
 #define PR_GET_MDWE			66