[PATCH] stacktrace: Force USER_DS for stack_trace_save_user()

[Peter Zijlstra]

On Wed, Jul 17, 2019 at 10:09:45AM +0200, Vegard Nossum wrote:
> On 7/17/19 10:07 AM, Peter Zijlstra wrote:

> > Does something like the below help?

> Yes.

Thanks!

---
Subject: stacktrace: Force USER_DS for stack_trace_save_user()
From: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
Date: Thu Jul 18 10:47:47 CEST 2019

When walking userspace stacks, we should set USER_DS, otherwise
access_ok() will not function as expected.

Reported-by: Vegard Nossum <vegard.nossum@xxxxxxxxxx>
Tested-by: Vegard Nossum <vegard.nossum@xxxxxxxxxx>
Reported-by: Eiichi Tsukata <devel@xxxxxxxxxxxx>
Signed-off-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
---
--- a/kernel/stacktrace.c
+++ b/kernel/stacktrace.c
@@ -226,12 +226,17 @@ unsigned int stack_trace_save_user(unsig
 		.store	= store,
 		.size	= size,
 	};
+	mm_segment_t fs;
 
 	/* Trace user stack if not a kernel thread */
 	if (current->flags & PF_KTHREAD)
 		return 0;
 
+	fs = get_fs();
+	set_fs(USER_DS);
 	arch_stack_walk_user(consume_entry, &c, task_pt_regs(current));
+	set_fs(fs);
+
 	return c.len;
 }
 #endif