Re: [PATCH 2/5] rxrpc: Fix uninitialized error code in rxrpc_send_data_packet()

From: Arnd Bergmann
Date: Wed May 29 2019 - 07:53:37 EST


On Tue, May 28, 2019 at 4:24 PM Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote:
>
> With gcc 4.1:
>
> net/rxrpc/output.c: In function ârxrpc_send_data_packetâ:
> net/rxrpc/output.c:338: warning: âretâ may be used uninitialized in this function
>
> Indeed, if the first jump to the send_fragmentable label is made, and
> the address family is not handled in the switch() statement, ret will be
> used uninitialized.
>
> Fix this by initializing err to zero before the jump, like is already
> done for the jump to the done label.
>
> Fixes: 5a924b8951f835b5 ("rxrpc: Don't store the rxrpc header in the Tx queue sk_buffs")
> Signed-off-by: Geert Uytterhoeven <geert@xxxxxxxxxxxxxx>
> ---
> While this is not a real false-positive, I believe it cannot cause harm
> in practice, as AF_RXRPC cannot be used with other transport families
> than IPv4 and IPv6.

This looks like a variant of the infamous bug
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=18501

What I don't understand is why clang fails to warn about it with
-Wsometimes-uninitialized.
(cc clang-built-linux mailing list).

Arnd

> net/rxrpc/output.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/net/rxrpc/output.c b/net/rxrpc/output.c
> index 004c762c2e8d063c..1473d774d67100c5 100644
> --- a/net/rxrpc/output.c
> +++ b/net/rxrpc/output.c
> @@ -403,8 +403,10 @@ int rxrpc_send_data_packet(struct rxrpc_call *call, struct sk_buff *skb,
>
> /* send the packet with the don't fragment bit set if we currently
> * think it's small enough */
> - if (iov[1].iov_len >= call->peer->maxdata)
> + if (iov[1].iov_len >= call->peer->maxdata) {
> + ret = 0;
> goto send_fragmentable;
> + }
>
> down_read(&conn->params.local->defrag_sem);
>