[PATCH 5/8] stackleak: rework stack high bound handling

[Mark Rutland]

Prior to returning to userpace, we reset current->lowest_stack to a
reasonable high bound. Currently we do this by subtracting the arbitrary
value `THREAD_SIZE/64` from the top of the stack, for reasons lost to
history.

Looking at configurations today:

* On i386 where THREAD_SIZE is 8K, the bound will be 128 bytes. The
  pt_regs at the top of the stack is 68 bytes (with 0 to 16 bytes of
  padding above), and so this covers an additional portion of 44 to 60
  bytes.

* On x86_64 where THREAD_SIZE is at least 16K (up to 32K with KASAN) the
  bound will be at least 256 bytes (up to 512 with KASAN). The pt_regs
  at the top of the stack is 168 bytes, and so this cover an additional
  88 bytes of stack (up to 344 with KASAN).

* On arm64 where THREAD_SIZE is at least 16K (up to 64K with 64K pages
  and VMAP_STACK), the bound will be at least 256 bytes (up to 1024 with
  KASAN). The pt_regs at the top of the stack is 336 bytes, so this can
  fall within the pt_regs, or can cover an additional 688 bytes of
  stack.

Clearly the `THREAD_SIZE/64` value doesn't make much sense -- in the
worst case, this will cause more than 600 bytes of stack to be erased
for every syscall, even if actual stack usage were substantially
smaller.

This patches makes this slightly less nonsensical by consistently
resetting current->lowest_stack to the base of the task pt_regs. For
clarity and for consistency with the handling of the low bound, the
generation of the high bound is split into a helper with commentary
explaining why.

Since the pt_regs at the top of the stack will be clobbered upon the
next exception entry, we don't need to poison these at exception exit.
By using task_pt_regs() as the high stack boundary instead of
current_top_of_stack() we avoid some redundant poisoning, and the
compiler can share the address generation between the poisoning and
restting of `current->lowest_stack`, making the generated code more
optimal.

It's not clear to me whether the existing `THREAD_SIZE/64` offset was a
dodgy heuristic to skip the pt_regs, or whether it was attempting to
minimize the number of times stackleak_check_stack() would have to
update `current->lowest_stack` when stack usage was shallow at the cost
of unconditionally poisoning a small portion of the stack for every exit
to userspace.

For now I've simply removed the offset, and if we need/want to minimize
updates for shallow stack usage it should be easy to add a better
heuristic atop, with appropriate commentary so we know what's going on.

Signed-off-by: Mark Rutland <mark.rutland@xxxxxxx>
Cc: Alexander Popov <alex.popov@xxxxxxxxx>
Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
Cc: Andy Lutomirski <luto@xxxxxxxxxx>
Cc: Kees Cook <keescook@xxxxxxxxxxxx>
---
 include/linux/stackleak.h | 14 ++++++++++++++
 kernel/stackleak.c        | 19 ++++++++++++++-----
 2 files changed, 28 insertions(+), 5 deletions(-)

diff --git a/include/linux/stackleak.h b/include/linux/stackleak.h
index 67430faa5c518..467661aeb4136 100644
--- a/include/linux/stackleak.h
+++ b/include/linux/stackleak.h
@@ -28,6 +28,20 @@ stackleak_task_low_bound(const struct task_struct *tsk)
 	return (unsigned long)end_of_stack(tsk) + sizeof(unsigned long);
 }
 
+/*
+ * The address immediately after the highest address on tsk's stack which we
+ * can plausibly erase.
+ */
+static __always_inline unsigned long
+stackleak_task_high_bound(const struct task_struct *tsk)
+{
+	/*
+	 * The task's pt_regs lives at the top of the task stack and will be
+	 * overwritten by exception entry, so there's no need to erase them.
+	 */
+	return (unsigned long)task_pt_regs(tsk);
+}
+
 static inline void stackleak_task_init(struct task_struct *t)
 {
 	t->lowest_stack = stackleak_task_low_bound(t);
diff --git a/kernel/stackleak.c b/kernel/stackleak.c
index 8fbc1e4c21435..f597d3323729d 100644
--- a/kernel/stackleak.c
+++ b/kernel/stackleak.c
@@ -73,6 +73,7 @@ late_initcall(stackleak_sysctls_init);
 static __always_inline void __stackleak_erase(void)
 {
 	const unsigned long task_stack_low = stackleak_task_low_bound(current);
+	const unsigned long task_stack_high = stackleak_task_high_bound(current);
 	unsigned long erase_low = current->lowest_stack;
 	unsigned long erase_high;
 	unsigned int poison_count = 0;
@@ -97,14 +98,22 @@ static __always_inline void __stackleak_erase(void)
 #endif
 
 	/*
-	 * Now write the poison value to the kernel stack between 'erase_low'
-	 * and 'erase_high'. We assume that the stack pointer doesn't change
-	 * when we write poison.
+	 * Write poison to the task's stack between 'erase_low' and
+	 * 'erase_high'.
+	 *
+	 * If we're running on a different stack (e.g. an entry trampoline
+	 * stack) we can erase everything below the pt_regs at the top of the
+	 * task stack.
+	 *
+	 * If we're running on the task stack itself, we must not clobber any
+	 * stack used by this function and its caller. We assume that this
+	 * function has a fixed-size stack frame, and the current stack pointer
+	 * doesn't change while we write poison.
 	 */
 	if (on_thread_stack())
 		erase_high = current_stack_pointer;
 	else
-		erase_high = current_top_of_stack();
+		erase_high = task_stack_high;
 
 	while (erase_low < erase_high) {
 		*(unsigned long *)erase_low = STACKLEAK_POISON;
@@ -112,7 +121,7 @@ static __always_inline void __stackleak_erase(void)
 	}
 
 	/* Reset the 'lowest_stack' value for the next syscall */
-	current->lowest_stack = current_top_of_stack() - THREAD_SIZE/64;
+	current->lowest_stack = task_stack_high;
 }
 
 asmlinkage void noinstr stackleak_erase(void)
-- 
2.30.2