x86/unwind: Handle NULL pointer calls better in frame unwinder

commit f4f34e1 upstream.

When the frame unwinder is invoked for an oops caused by a call to NULL, it
currently skips the parent function because BP still points to the parent's
stack frame; the (nonexistent) current function only has the first half of
a stack frame, and BP doesn't point to it yet.

Add a special case for IP==0 that calculates a fake BP from SP, then uses
the real BP for the next frame.

Note that this handles first_frame specially: Return information about the
parent function as long as the saved IP is >=first_frame, even if the fake
BP points below it.

With an artificially-added NULL call in prctl_set_seccomp(), before this
patch, the trace is:

Call Trace:
 ? prctl_set_seccomp+0x3a/0x50
 __x64_sys_prctl+0x457/0x6f0
 ? __ia32_sys_prctl+0x750/0x750
 do_syscall_64+0x72/0x160
 entry_SYSCALL_64_after_hwframe+0x44/0xa9

After this patch, the trace is:

Call Trace:
 prctl_set_seccomp+0x3a/0x50
 __x64_sys_prctl+0x457/0x6f0
 ? __ia32_sys_prctl+0x750/0x750
 do_syscall_64+0x72/0x160
 entry_SYSCALL_64_after_hwframe+0x44/0xa9

Signed-off-by: Jann Horn <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Acked-by: Josh Poimboeuf <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: syzbot <[email protected]>
Cc: "H. Peter Anvin" <[email protected]>
Cc: Masahiro Yamada <[email protected]>
Cc: Michal Marek <[email protected]>
Cc: [email protected]
Link: https://lkml.kernel.org/r/[email protected]
Signed-off-by: Greg Kroah-Hartman <[email protected]>

Author: thejh

arch/x86/include/asm/unwind.h

@@ -23,6 +23,12 @@ struct unwind_state {
 #elif defined(CONFIG_UNWINDER_FRAME_POINTER)
 	bool got_irq;
 	unsigned long *bp, *orig_sp, ip;
+	/*
+	 * If non-NULL: The current frame is incomplete and doesn't contain a
+	 * valid BP. When looking for the next frame, use this instead of the
+	 * non-existent saved BP.
+	 */
+	unsigned long *next_bp;
 	struct pt_regs *regs;
 #else
 	unsigned long *sp;

arch/x86/kernel/unwind_frame.c

@@ -320,10 +320,14 @@ bool unwind_next_frame(struct unwind_state *state)
 	}
 
 	/* Get the next frame pointer: */
-	if (state->regs)
+	if (state->next_bp) {
+		next_bp = state->next_bp;
+		state->next_bp = NULL;
+	} else if (state->regs) {
 		next_bp = (unsigned long *)state->regs->bp;
-	else
+	} else {
 		next_bp = (unsigned long *)READ_ONCE_TASK_STACK(state->task, *state->bp);
+	}
 
 	/* Move to the next frame if it's safe: */
 	if (!update_stack_state(state, next_bp))
@@ -398,6 +402,21 @@ void __unwind_start(struct unwind_state *state, struct task_struct *task,
 
 	bp = get_frame_pointer(task, regs);
 
+	/*
+	 * If we crash with IP==0, the last successfully executed instruction
+	 * was probably an indirect function call with a NULL function pointer.
+	 * That means that SP points into the middle of an incomplete frame:
+	 * *SP is a return pointer, and *(SP-sizeof(unsigned long)) is where we
+	 * would have written a frame pointer if we hadn't crashed.
+	 * Pretend that the frame is complete and that BP points to it, but save
+	 * the real BP so that we can use it when looking for the next frame.
+	 */
+	if (regs && regs->ip == 0 &&
+	    (unsigned long *)kernel_stack_pointer(regs) >= first_frame) {
+		state->next_bp = bp;
+		bp = ((unsigned long *)kernel_stack_pointer(regs)) - 1;
+	}
+
 	/* Initialize stack info and make sure the frame data is accessible: */
 	get_stack_info(bp, state->task, &state->stack_info,
 		       &state->stack_mask);
@@ -410,7 +429,7 @@ void __unwind_start(struct unwind_state *state, struct task_struct *task,
 	 */
 	while (!unwind_done(state) &&
 	       (!on_stack(&state->stack_info, first_frame, sizeof(long)) ||
-			state->bp < first_frame))
+			(state->next_bp == NULL && state->bp < first_frame)))
 		unwind_next_frame(state);
 }
 EXPORT_SYMBOL_GPL(__unwind_start);