net/iucv: updates 2021-01-28
Subsystems (5)
cover net/iucv: updates 2021-01-28
Hi Dave & Jakub,
please apply the following patch series for iucv to netdev's net-next tree.
This reworks & simplifies the TX notification path in af_iucv, so that we can send out SG skbs over TRANS_HIPER sockets. Also remove a noisy WARN_ONCE() in the RX path.
Thanks,
Julian
Alexander Egorenkov (1):
net/af_iucv: remove WARN_ONCE on malformed RX packets
Julian Wiedmann (4):
net/af_iucv: don't lookup the socket on TX notification net/af_iucv: count packets in the xmit path
net/af_iucv: don't track individual TX skbs for TRANS_HIPER sockets net/af_iucv: build SG skbs for TRANS_HIPER sockets
drivers/s390/net/qeth_core_main.c | 6 +- include/net/iucv/af_iucv.h | 3 +- net/iucv/af_iucv.c | 122 ++++++++++++------------------ 3 files changed, 53 insertions(+), 78 deletions(-)
Signature · 1 line
2.17.1
PATCH 1/5 net/af_iucv: remove WARN_ONCE on malformed RX packets
From: Alexander Egorenkov <[email protected]>
syzbot reported the following finding:
AF_IUCV failed to receive skb, len=0
WARNING: CPU: 0 PID: 522 at net/iucv/af_iucv.c:2039 afiucv_hs_rcv+0x174/0x190 net/iucv/af_iucv.c:2039 CPU: 0 PID: 522 Comm: syz-executor091 Not tainted 5.10.0-rc1-syzkaller-07082-g55027a88ec9f #0 Hardware name: IBM 3906 M04 701 (KVM/Linux) Call Trace: [<00000000b87ea538>] afiucv_hs_rcv+0x178/0x190 net/iucv/af_iucv.c:2039 ([<00000000b87ea534>] afiucv_hs_rcv+0x174/0x190 net/iucv/af_iucv.c:2039) [<00000000b796533e>] __netif_receive_skb_one_core+0x13e/0x188 net/core/dev.c:5315 [<00000000b79653ce>] __netif_receive_skb+0x46/0x1c0 net/core/dev.c:5429 [<00000000b79655fe>] netif_receive_skb_internal+0xb6/0x220 net/core/dev.c:5534 [<00000000b796ac3a>] netif_receive_skb+0x42/0x318 net/core/dev.c:5593 [<00000000b6fd45f4>] tun_rx_batched.isra.0+0x6fc/0x860 drivers/net/tun.c:1485 [<00000000b6fddc4e>] tun_get_user+0x1c26/0x27f0 drivers/net/tun.c:1939 [<00000000b6fe0f00>] tun_chr_write_iter+0x158/0x248 drivers/net/tun.c:1968 [<00000000b4f22bfa>] call_write_iter include/linux/fs.h:1887 [inline] [<00000000b4f22bfa>] new_sync_write+0x442/0x648 fs/read_write.c:518 [<00000000b4f238fe>] vfs_write.part.0+0x36e/0x5d8 fs/read_write.c:605 [<00000000b4f2984e>] vfs_write+0x10e/0x148 fs/read_write.c:615 [<00000000b4f29d0e>] ksys_write+0x166/0x290 fs/read_write.c:658 [<00000000b8dc4ab4>] system_call+0xe0/0x28c arch/s390/kernel/entry.S:415
Last Breaking-Event-Address:
[<00000000b8dc64d4>] __s390_indirect_jump_r14+0x0/0xc
Malformed RX packets shouldn't generate any warnings because debugging info already flows to dropmon via the kfree_skb().
--- net/iucv/af_iucv.c | 1 - 1 file changed, 1 deletion(-)
net/iucv/af_iucv.c +0 -1
@@ -2036,7 +2036,6 @@ static int afiucv_hs_rcv(struct sk_buff *skb, struct net_device *dev,char nullstring[8 ];if (!pskb_may_pull (skb,sizeof (*trans_hdr))) { -WARN_ONCE (1 ,"AF_IUCV failed to receive skb, len=%u" , skb->len);kfree_skb (skb);return NET_RX_SUCCESS ; }
Signature · 1 line
2.17.1
PATCH 2/5 net/af_iucv: don't lookup the socket on TX notification
In reply to Julian WiedmannWhoever called iucv_sk(sk)->sk_txnotify() must already know that they're dealing with an af_iucv socket.
--- net/iucv/af_iucv.c | 15 +++------------ 1 file changed, 3 insertions(+), 12 deletions(-)
net/iucv/af_iucv.c +3 -12
@@ -2134,23 +2134,14 @@ static int afiucv_hs_rcv(struct sk_buff *skb, struct net_device *dev,static void afiucv_hs_callback_txnotify (struct sk_buff *skb,enum iucv_tx_notify n) { -struct sock *isk = skb->sk; -struct sock *sk =NULL ; -struct iucv_sock *iucv =NULL ; +struct iucv_sock *iucv =iucv_sk (skb->sk); +struct sock *sk = skb->sk;struct sk_buff_head *list;struct sk_buff *list_skb;struct sk_buff *nskb;unsigned long flags; -read_lock_irqsave (&iucv_sk_list.lock, flags); -sk_for_each (sk, &iucv_sk_list.head) -if (sk == isk) { - iucv =iucv_sk (sk); -break ; - } -read_unlock_irqrestore (&iucv_sk_list.lock, flags); - -if (!iucv ||sock_flag (sk,SOCK_ZAPPED )) +if (sock_flag (sk,SOCK_ZAPPED ))return ; list = &iucv->send_skb_q;
Signature · 1 line
2.17.1
PATCH 3/5 net/af_iucv: count packets in the xmit path
In reply to Julian WiedmannThe TX code keeps track of all skbs that are in-flight but haven't actually been sent out yet. For native IUCV sockets that's not a huge deal, but with TRANS_HIPER sockets it would be much better if we didn't need to maintain a list of skb clones.
Note that we actually only care about the _count_ of skbs in this stage of the TX pipeline. So as prep work for removing the skb tracking on TRANS_HIPER sockets, keep track of the skb count in a separate variable and pair any list {enqueue, unlink} with a count {increment, decrement}.
Then replace all occurences where we currently look at the skb list's fill level.
--- include/net/iucv/af_iucv.h | 1 + net/iucv/af_iucv.c | 30 ++++++++++++++++++++++++------ 2 files changed, 25 insertions(+), 6 deletions(-)
include/net/iucv/af_iucv.h +1 -0
@@ -128,6 +128,7 @@ struct iucv_sock {u8 flags;u16 msglimit;u16 msglimit_peer; +atomic_t skbs_in_xmit;atomic_t msg_sent;atomic_t msg_recv;atomic_t pendings;
net/iucv/af_iucv.c +24 -6
@@ -182,7 +182,7 @@ static inline int iucv_below_msglim(struct sock *sk)if (sk->sk_state !=IUCV_CONNECTED )return 1 ;if (iucv->transport ==AF_IUCV_TRANS_IUCV ) -return (skb_queue_len (&iucv->send_skb_q) < iucv->path->msglim); +return (atomic_read (&iucv->skbs_in_xmit) < iucv->path->msglim);else return ((atomic_read (&iucv->msg_sent) < iucv->msglimit_peer) && (atomic_read (&iucv->pendings) <=0 ));
@@ -269,8 +269,10 @@ static int afiucv_hs_send(struct iucv_message *imsg, struct sock *sock, }skb_queue_tail (&iucv->send_skb_q, nskb); +atomic_inc (&iucv->skbs_in_xmit); err =dev_queue_xmit (skb);if (net_xmit_eval (err)) { +atomic_dec (&iucv->skbs_in_xmit);skb_unlink (nskb, &iucv->send_skb_q);kfree_skb (nskb); }else {
@@ -424,7 +426,7 @@ static void iucv_sock_close(struct sock *sk) sk->sk_state =IUCV_CLOSING ; sk->sk_state_change (sk); -if (!err && !skb_queue_empty (&iucv->send_skb_q)) { +if (!err &&atomic_read (&iucv->skbs_in_xmit) >0 ) {if (sock_flag (sk,SOCK_LINGER ) && sk->sk_lingertime) timeo = sk->sk_lingertime;else
@@ -491,6 +493,7 @@ static struct sock *iucv_sock_alloc(struct socket *sock, int proto, gfp_t prio,atomic_set (&iucv->pendings,0 ); iucv->flags =0 ; iucv->msglimit =0 ; +atomic_set (&iucv->skbs_in_xmit,0 );atomic_set (&iucv->msg_sent,0 );atomic_set (&iucv->msg_recv,0 ); iucv->path =NULL ;
@@ -1055,6 +1058,7 @@ static int iucv_sock_sendmsg(struct socket *sock, struct msghdr *msg, } }else {/* Classic VM IUCV transport */ skb_queue_tail (&iucv->send_skb_q, skb); +atomic_inc (&iucv->skbs_in_xmit);if (((iucv->path->flags &IUCV_IPRMDATA ) & iucv->flags) && skb->len <=7 ) {
@@ -1063,6 +1067,7 @@ static int iucv_sock_sendmsg(struct socket *sock, struct msghdr *msg,/* on success: there is no message_complete callback */ /* for an IPRMDATA msg; remove skb from send queue */ if (err ==0 ) { +atomic_dec (&iucv->skbs_in_xmit);skb_unlink (skb, &iucv->send_skb_q);kfree_skb (skb); }
@@ -1071,6 +1076,7 @@ static int iucv_sock_sendmsg(struct socket *sock, struct msghdr *msg,/* IUCV_IPRMDATA path flag is set... sever path */ if (err ==0x15 ) { pr_iucv->path_sever (iucv->path,NULL ); +atomic_dec (&iucv->skbs_in_xmit);skb_unlink (skb, &iucv->send_skb_q); err = -EPIPE ; goto fail;
@@ -1109,6 +1115,8 @@ static int iucv_sock_sendmsg(struct socket *sock, struct msghdr *msg, }else { err = -EPIPE ; } + +atomic_dec (&iucv->skbs_in_xmit);skb_unlink (skb, &iucv->send_skb_q); goto fail; }
@@ -1748,10 +1756,14 @@ static void iucv_callback_txdone(struct iucv_path *path, {struct sock *sk = path->private;struct sk_buff *this =NULL ; -struct sk_buff_head *list = &iucv_sk (sk)->send_skb_q; +struct sk_buff_head *list;struct sk_buff *list_skb; +struct iucv_sock *iucv;unsigned long flags; + iucv =iucv_sk (sk); + list = &iucv->send_skb_q; +bh_lock_sock (sk);spin_lock_irqsave (&list->lock, flags);
@@ -1761,8 +1773,11 @@ static void iucv_callback_txdone(struct iucv_path *path,break ; } } -if (this) +if (this) { +atomic_dec (&iucv->skbs_in_xmit);__skb_unlink (this, list); + } +spin_unlock_irqrestore (&list->lock, flags);if (this) {
@@ -1772,7 +1787,7 @@ static void iucv_callback_txdone(struct iucv_path *path, }if (sk->sk_state ==IUCV_CLOSING ) { -if (skb_queue_empty (&iucv_sk (sk)->send_skb_q)) { +if (atomic_read (&iucv->skbs_in_xmit) ==0 ) { sk->sk_state =IUCV_CLOSED ; sk->sk_state_change (sk); }
@@ -2150,6 +2165,7 @@ static void afiucv_hs_callback_txnotify(struct sk_buff *skb,if (skb_shinfo (list_skb) ==skb_shinfo (skb)) {switch (n) {case TX_NOTIFY_OK : +atomic_dec (&iucv->skbs_in_xmit);__skb_unlink (list_skb, list);kfree_skb (list_skb);iucv_sock_wake_msglim (sk);
@@ -2158,6 +2174,7 @@ static void afiucv_hs_callback_txnotify(struct sk_buff *skb,atomic_inc (&iucv->pendings);break ;case TX_NOTIFY_DELAYED_OK : +atomic_dec (&iucv->skbs_in_xmit);__skb_unlink (list_skb, list);atomic_dec (&iucv->pendings);if (atomic_read (&iucv->pendings) <=0 )
@@ -2169,6 +2186,7 @@ static void afiucv_hs_callback_txnotify(struct sk_buff *skb,case TX_NOTIFY_TPQFULL :/* not yet used */ case TX_NOTIFY_GENERALERROR :case TX_NOTIFY_DELAYED_GENERALERROR : +atomic_dec (&iucv->skbs_in_xmit);__skb_unlink (list_skb, list);kfree_skb (list_skb);if (sk->sk_state ==IUCV_CONNECTED ) {
@@ -2183,7 +2201,7 @@ static void afiucv_hs_callback_txnotify(struct sk_buff *skb,spin_unlock_irqrestore (&list->lock, flags);if (sk->sk_state ==IUCV_CLOSING ) { -if (skb_queue_empty (&iucv_sk (sk)->send_skb_q)) { +if (atomic_read (&iucv->skbs_in_xmit) ==0 ) { sk->sk_state =IUCV_CLOSED ; sk->sk_state_change (sk); }
Signature · 1 line
2.17.1
PATCH 4/5 net/af_iucv: don't track individual TX skbs for TRANS_HIPER sockets
In reply to Julian WiedmannStop maintaining the skb_send_q list for TRANS_HIPER sockets.
Not only is it extra overhead, but keeping around a list of skb clones means that we later also have to match the ->sk_txnotify() calls against these clones and free them accordingly.
The current matching logic (comparing the skbs' shinfo location) is frustratingly fragile, and breaks if the skb's head is mangled in any sort of way while passing from dev_queue_xmit() to the device's HW queue.
Also adjust the interface for ->sk_txnotify(), to make clear that we don't actually care about any skb internals.
--- drivers/s390/net/qeth_core_main.c | 6 ++- include/net/iucv/af_iucv.h | 2 +- net/iucv/af_iucv.c | 80 ++++++++----------------------- 3 files changed, 26 insertions(+), 62 deletions(-)
drivers/s390/net/qeth_core_main.c +4 -2
@@ -1407,10 +1407,12 @@ static void qeth_notify_skbs(struct qeth_qdio_out_q *q,struct sk_buff *skb; skb_queue_walk(&buf->skb_list, skb) { +struct sock *sk = skb->sk; +QETH_CARD_TEXT_ (q->card,5 ,"skbn%d" , notification);QETH_CARD_TEXT_ (q->card,5 ,"%lx" , (long ) skb); -if (skb->sk && skb->sk->sk_family ==PF_IUCV ) -iucv_sk (skb->sk)->sk_txnotify (skb, notification); +if (sk && sk->sk_family ==PF_IUCV ) +iucv_sk (sk)->sk_txnotify (sk, notification); } }
include/net/iucv/af_iucv.h +1 -1
@@ -133,7 +133,7 @@ struct iucv_sock {atomic_t msg_recv;atomic_t pendings;int transport; -void (*sk_txnotify)(struct sk_buff *skb, +void (*sk_txnotify)(struct sock *sk,enum iucv_tx_notify n); };
net/iucv/af_iucv.c +21 -59
121 lines not on this page · raw message
Signature · 1 line
2.17.1
PATCH 5/5 net/af_iucv: build SG skbs for TRANS_HIPER sockets
In reply to Julian WiedmannThe TX path no longer falls apart when some of its SG skbs are later linearized by lower layers of the stack. So enable the use of SG skbs in iucv_sock_sendmsg() again.
This effectively reverts
commit dc5367bcc556 ("net/af_iucv: don't use paged skbs for TX on HiperSockets").
--- net/iucv/af_iucv.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
net/iucv/af_iucv.c +4 -2
@@ -256,7 +256,9 @@ static int afiucv_hs_send(struct iucv_message *imsg, struct sock *sock, err = -EMSGSIZE ; goto err_free; } -skb_trim (skb, skb->dev->mtu); + err =pskb_trim (skb, skb->dev->mtu); +if (err) + goto err_free; } skb->protocol =cpu_to_be16 (ETH_P_AF_IUCV );
@@ -996,7 +998,7 @@ static int iucv_sock_sendmsg(struct socket *sock, struct msghdr *msg,if (iucv->transport ==AF_IUCV_TRANS_HIPER ) { headroom =sizeof (struct af_iucv_trans_hdr ) +LL_RESERVED_SPACE (iucv->hs_dev); - linear = len; + linear =min (len,PAGE_SIZE - headroom); }else {if (len <PAGE_SIZE ) { linear = len;
Signature · 1 line
2.17.1
This series was applied to netdev/net-next.git (refs/heads/master):
Hello:
This series was applied to netdev/net-next.git (refs/heads/master):
Quoted 9 lines · On Thu, 28 Jan 2021 12:41:03 +0100 you wrote:
Hi Dave & Jakub,
please apply the following patch series for iucv to netdev's net-next tree.
This reworks & simplifies the TX notification path in
af_iucv, so that we can send out SG skbs overTRANS_HIPERsockets. Also remove a noisyWARN_ONCE()in the RX path.[...]
Here is the summary with links:
- [net-next,1/5] net/af_iucv: remove WARN_ONCE on malformed RX packets
https://git.kernel.org/netdev/net-next/c/27e9c1de5299 - [net-next,2/5] net/af_iucv: don't lookup the socket on TX notification https://git.kernel.org/netdev/net-next/c/c464444fa2ca - [net-next,3/5] net/af_iucv: count packets in the xmit path https://git.kernel.org/netdev/net-next/c/ef6af7bdb9e6 - [net-next,4/5] net/af_iucv: don't track individual TX skbs for TRANS_HIPER sockets https://git.kernel.org/netdev/net-next/c/80bc97aa0aaa - [net-next,5/5] net/af_iucv: build SG skbs for TRANS_HIPER sockets https://git.kernel.org/netdev/net-next/c/2c3b4456c812
You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html