io_uring: bump poll refs to full 31-bits

[ upstream commmit e2c0cb7c0cc72939b61a7efee376206725796625 ]

The previous commit:

1bc84c40088 ("io_uring: remove poll entry from list when canceling all")

removed a potential overflow condition for the poll references. They
are currently limited to 20-bits, even if we have 31-bits available. The
upper bit is used to mark for cancelation.

Bump the poll ref space to 31-bits, making that kind of situation much
harder to trigger in general. We'll separately add overflow checking
and handling.

Fixes: aa43477b0402 ("io_uring: poll rework")
Signed-off-by: Jens Axboe <axboe@kernel.dk>
[pavel: backport]
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

fs/io_uring.c

@@ -5314,7 +5314,7 @@ struct io_poll_table {
 };
 
 #define IO_POLL_CANCEL_FLAG	BIT(31)
-#define IO_POLL_REF_MASK	((1u << 20)-1)
+#define IO_POLL_REF_MASK	GENMASK(30, 0)
 
 /*
  * If refs part of ->poll_refs (see IO_POLL_REF_MASK) is 0, it's free. We can