[PATCH] bpf: fix false positive kmemleak report in bpf_ringbuf_area_alloc()

kmemleak scans struct page, but it does not scan the page content.
if we allocate some memory with kmalloc(), then allocate page with
alloc_page(), and if we put kmalloc pointer somewhere inside that page,
kmemleak will report kmalloc pointer as a false positive.

we can instruct kmemleak to scan the memory area by calling
kmemleak_alloc()/kmemleak_free(), but part of struct bpf_ringbuf is
mmaped to user space, and if struct bpf_ringbuf changes we would have to
revisit and review size argument in kmemleak_alloc(), because we do not
want kmemleak to scan the user space memory.
let's simplify things and use kmemleak_not_leak() here.

Link: https://lore.kernel.org/lkml/YNTAqiE7CWJhOK2M@nuc10/
Link: https://lore.kernel.org/lkml/20210615101515.GC26027@arm.com/
Link: https://syzkaller.appspot.com/bug?extid=5d895828587f49e7fe9b
Reported-and-tested-by: syzbot+5d895828587f49e7fe9b@syzkaller.appspotmail.com
Signed-off-by: Rustam Kovhaev <redacted>
---
 kernel/bpf/ringbuf.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/kernel/bpf/ringbuf.c b/kernel/bpf/ringbuf.c
index 84b3b35fc0d0..9e0c10c6892a 100644
--- a/kernel/bpf/ringbuf.c
+++ b/kernel/bpf/ringbuf.c

@@ -8,6 +8,7 @@
 #include <linux/vmalloc.h>
 #include <linux/wait.h>
 #include <linux/poll.h>
+#include <linux/kmemleak.h>
 #include <uapi/linux/btf.h>
 
 #define RINGBUF_CREATE_FLAG_MASK (BPF_F_NUMA_NODE)

@@ -105,6 +106,7 @@ static struct bpf_ringbuf *bpf_ringbuf_area_alloc(size_t data_sz, int numa_node)
 	rb = vmap(pages, nr_meta_pages + 2 * nr_data_pages,
 		  VM_ALLOC | VM_USERMAP, PAGE_KERNEL);
 	if (rb) {
+		kmemleak_not_leak(pages);
 		rb->pages = pages;
 		rb->nr_pages = nr_pages;
 		return rb;

-- 
2.30.2

On 6/26/21 8:11 PM, Rustam Kovhaev wrote:

kmemleak scans struct page, but it does not scan the page content.
if we allocate some memory with kmalloc(), then allocate page with
alloc_page(), and if we put kmalloc pointer somewhere inside that page,
kmemleak will report kmalloc pointer as a false positive.

we can instruct kmemleak to scan the memory area by calling
kmemleak_alloc()/kmemleak_free(), but part of struct bpf_ringbuf is
mmaped to user space, and if struct bpf_ringbuf changes we would have to
revisit and review size argument in kmemleak_alloc(), because we do not
want kmemleak to scan the user space memory.
let's simplify things and use kmemleak_not_leak() here.

Link: https://lore.kernel.org/lkml/YNTAqiE7CWJhOK2M@nuc10/
Link: https://lore.kernel.org/lkml/20210615101515.GC26027@arm.com/
Link: https://syzkaller.appspot.com/bug?extid=5d895828587f49e7fe9b
Reported-and-tested-by: syzbot+5d895828587f49e7fe9b@syzkaller.appspotmail.com
Signed-off-by: Rustam Kovhaev <redacted>

Applied, thanks! (Also included Andrii's prior analysis as well to the commit
log so there's a bit more context if we need to revisit in future [0].)

   [0] https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next.git/commit/?id=ccff81e1d028bbbf8573d3364a87542386c707bf