Message ID | 1359727032-7999-1-git-send-email-phil.sutter@viprinet.com |
---|---|
State | Superseded, archived |
Delegated to: | David Miller |
Headers | show |
On 02/01/2013 02:57 PM, Phil Sutter wrote: > When releasing a packet socket, the routine packet_set_ring() is reused > to free rings instead of allocating them. But when calling it for the > first time, it fills req->tp_block_nr with the value of rb->pg_vec_len > which in the second invocation makes it bail out since req->tp_block_nr > is greater zero but req->tp_block_size is zero. > > This patch solves the problem by passing a zeroed auto-variable to > packet_set_ring() upon each invocation from packet_release(). > > As far as I can tell, this issue exists even since 69e3c75 (net: TX_RING > and packet mmap), i.e. the original inclusion of TX ring support into > af_packet, but applies only to sockets with both RX and TX ring > allocated, which is probably why this was unnoticed all the time. > > Signed-off-by: Phil Sutter <phil.sutter@viprinet.com> > Cc: Johann Baudy <johann.baudy@gnu-log.net> > Cc: stable@kernel.org > --- [...] > +static int packet_free_ring(struct sock *sk, int tx_ring) > +{ > + union tpacket_req_u req_u = { 0 }; > + > + return packet_set_ring(sk, &req_u, 1, tx_ring); > +} > + > /* > * Close a PACKET socket. This is fairly simple. We immediately go > * to 'closed' state and remove our protocol entry in the device list. > @@ -2338,7 +2345,6 @@ static int packet_release(struct socket *sock) > struct sock *sk = sock->sk; > struct packet_sock *po; > struct net *net; > - union tpacket_req_u req_u; > > if (!sk) > return 0; > @@ -2364,13 +2370,11 @@ static int packet_release(struct socket *sock) > > packet_flush_mclist(sk); > > - memset(&req_u, 0, sizeof(req_u)); > - > if (po->rx_ring.pg_vec) > - packet_set_ring(sk, &req_u, 1, 0); > + packet_free_ring(sk, 0); > > if (po->tx_ring.pg_vec) > - packet_set_ring(sk, &req_u, 1, 1); > + packet_free_ring(sk, 1); Good catch! Nitpicking: I think it would be easier / more readable to simply move the memset into the two ifs than introducing an extra function for just doing that. (Also don't cc stable, since David is deciding about this anyway.) -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, Feb 01, 2013 at 05:05:08PM +0100, Daniel Borkmann wrote: > On 02/01/2013 02:57 PM, Phil Sutter wrote: > > When releasing a packet socket, the routine packet_set_ring() is reused > > to free rings instead of allocating them. But when calling it for the > > first time, it fills req->tp_block_nr with the value of rb->pg_vec_len > > which in the second invocation makes it bail out since req->tp_block_nr > > is greater zero but req->tp_block_size is zero. > > > > This patch solves the problem by passing a zeroed auto-variable to > > packet_set_ring() upon each invocation from packet_release(). > > > > As far as I can tell, this issue exists even since 69e3c75 (net: TX_RING > > and packet mmap), i.e. the original inclusion of TX ring support into > > af_packet, but applies only to sockets with both RX and TX ring > > allocated, which is probably why this was unnoticed all the time. > > > > Signed-off-by: Phil Sutter <phil.sutter@viprinet.com> > > Cc: Johann Baudy <johann.baudy@gnu-log.net> > > Cc: stable@kernel.org > > --- > > [...] > > > +static int packet_free_ring(struct sock *sk, int tx_ring) > > +{ > > + union tpacket_req_u req_u = { 0 }; > > + > > + return packet_set_ring(sk, &req_u, 1, tx_ring); > > +} > > + > > /* > > * Close a PACKET socket. This is fairly simple. We immediately go > > * to 'closed' state and remove our protocol entry in the device list. > > @@ -2338,7 +2345,6 @@ static int packet_release(struct socket *sock) > > struct sock *sk = sock->sk; > > struct packet_sock *po; > > struct net *net; > > - union tpacket_req_u req_u; > > > > if (!sk) > > return 0; > > @@ -2364,13 +2370,11 @@ static int packet_release(struct socket *sock) > > > > packet_flush_mclist(sk); > > > > - memset(&req_u, 0, sizeof(req_u)); > > - > > if (po->rx_ring.pg_vec) > > - packet_set_ring(sk, &req_u, 1, 0); > > + packet_free_ring(sk, 0); > > > > if (po->tx_ring.pg_vec) > > - packet_set_ring(sk, &req_u, 1, 1); > > + packet_free_ring(sk, 1); > > Good catch! > > Nitpicking: > > I think it would be easier / more readable to simply move the memset into > the two ifs than introducing an extra function for just doing that. Yes, this was just how my fix looked like initially, but I didn't like the resulting code duplication. Indeed, the extra function adds another point of code flow redirection. On the other hand, it implicitly points out that basically the same is done for both rings. In my point of view, both ways are equally acceptable. If you prefer the other one for mainline inclusion, just let me know and I submit an appropriate patch. > (Also don't cc stable, since David is deciding about this anyway.) Oh, OK. Thanks for the hint. Every time I wonder how to do this the right way, seems I haven't found the correct documentation for it yet. Moreover, stable@kernel.org isn't even valid. I shouldn't have trusted google (http://www.kernel.org/doc/Documentation/stable_kernel_rules.txt) over my local clones. ;) Best wishes, Phil Sutter Software Engineer
On 02/01/2013 05:21 PM, Phil Sutter wrote: > On Fri, Feb 01, 2013 at 05:05:08PM +0100, Daniel Borkmann wrote: >> I think it would be easier / more readable to simply move the memset into >> the two ifs than introducing an extra function for just doing that. > > Yes, this was just how my fix looked like initially, but I didn't like > the resulting code duplication. Indeed, the extra function adds another > point of code flow redirection. On the other hand, it implicitly points > out that basically the same is done for both rings. > > In my point of view, both ways are equally acceptable. If you prefer the > other one for mainline inclusion, just let me know and I submit an > appropriate patch. Yes, that'd be good, thanks. -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/net/packet/af_packet.c b/net/packet/af_packet.c index a91fd0b..d5f519a 100644 --- a/net/packet/af_packet.c +++ b/net/packet/af_packet.c @@ -2328,6 +2328,13 @@ static int packet_sendmsg(struct kiocb *iocb, struct socket *sock, return rc; } +static int packet_free_ring(struct sock *sk, int tx_ring) +{ + union tpacket_req_u req_u = { 0 }; + + return packet_set_ring(sk, &req_u, 1, tx_ring); +} + /* * Close a PACKET socket. This is fairly simple. We immediately go * to 'closed' state and remove our protocol entry in the device list. @@ -2338,7 +2345,6 @@ static int packet_release(struct socket *sock) struct sock *sk = sock->sk; struct packet_sock *po; struct net *net; - union tpacket_req_u req_u; if (!sk) return 0; @@ -2364,13 +2370,11 @@ static int packet_release(struct socket *sock) packet_flush_mclist(sk); - memset(&req_u, 0, sizeof(req_u)); - if (po->rx_ring.pg_vec) - packet_set_ring(sk, &req_u, 1, 0); + packet_free_ring(sk, 0); if (po->tx_ring.pg_vec) - packet_set_ring(sk, &req_u, 1, 1); + packet_free_ring(sk, 1); fanout_release(sk);
When releasing a packet socket, the routine packet_set_ring() is reused to free rings instead of allocating them. But when calling it for the first time, it fills req->tp_block_nr with the value of rb->pg_vec_len which in the second invocation makes it bail out since req->tp_block_nr is greater zero but req->tp_block_size is zero. This patch solves the problem by passing a zeroed auto-variable to packet_set_ring() upon each invocation from packet_release(). As far as I can tell, this issue exists even since 69e3c75 (net: TX_RING and packet mmap), i.e. the original inclusion of TX ring support into af_packet, but applies only to sockets with both RX and TX ring allocated, which is probably why this was unnoticed all the time. Signed-off-by: Phil Sutter <phil.sutter@viprinet.com> Cc: Johann Baudy <johann.baudy@gnu-log.net> Cc: stable@kernel.org --- net/packet/af_packet.c | 14 +++++++++----- 1 file changed, 9 insertions(+), 5 deletions(-)