[PATCH v2 bpf-next] bpf: fix use-after-free of bpf_link when priming half-fails

[Andrii Nakryiko <andriin@fb.com>]

If bpf_link_prime() succeeds to allocate new anon file, but then fails to
allocate ID for it, link priming is considered to be failed and user is
supposed ot be able to directly kfree() bpf_link, because it was never exposed
to user-space.

But at that point file already keeps a pointer to bpf_link and will eventually
call bpf_link_release(), so if bpf_link was kfree()'d by caller, that would
lead to use-after-free.

Fix this by first allocating ID and only then allocating file. Adding ID to
link_idr is ok, because link at that point still doesn't have its ID set, so
no user-space process can create a new FD for it.

Suggested-by: Martin KaFai Lau <kafai@fb.com>
Fixes: a3b80e107894 ("bpf: Allocate ID for bpf_link")
Reported-by: syzbot+39b64425f91b5aab714d@syzkaller.appspotmail.com
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
---
 kernel/bpf/syscall.c | 13 +++++++------
 1 file changed, 7 insertions(+), 6 deletions(-)

diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
index c75b2dd2459c..108c8051dff2 100644
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c
@@ -2348,19 +2348,20 @@ int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer)
 	if (fd < 0)
 		return fd;
 
-	file = anon_inode_getfile("bpf_link", &bpf_link_fops, link, O_CLOEXEC);
-	if (IS_ERR(file)) {
-		put_unused_fd(fd);
-		return PTR_ERR(file);
-	}
 
 	id = bpf_link_alloc_id(link);
 	if (id < 0) {
 		put_unused_fd(fd);
-		fput(file);
 		return id;
 	}
 
+	file = anon_inode_getfile("bpf_link", &bpf_link_fops, link, O_CLOEXEC);
+	if (IS_ERR(file)) {
+		bpf_link_free_id(id);
+		put_unused_fd(fd);
+		return PTR_ERR(file);
+	}
+
 	primer->link = link;
 	primer->file = file;
 	primer->fd = fd;
-- 
2.24.1