Orphaning skb in dev_hard_start_xmit() makes bonding behavior
unfriendly for applications sending big UDP bursts : Once packets
pass the bonding device and come to real device, they might hit a full
qdisc and be dropped. Without orphaning, the sender is automatically
throttled because sk->sk_wmemalloc reaches sk->sk_sndbuf (assuming
sk_sndbuf is not too big)
We could try to defer the orphaning adding another test in
dev_hard_start_xmit(), but all this seems of little gain,
now that BQL tends to make packets more likely to be parked
in Qdisc queues instead of NIC TX ring, in cases where performance
matters.
Reverts commits :
fc6055a5ba net: Introduce skb_orphan_try()
87fd308cfc net: skb_tx_hash() fix relative to skb_orphan_try()
and removes SKBTX_DRV_NEEDS_SK_REF flag
Reported-and-bisected-by: Jean-Michel Hautbois <jhautbois@gmail.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Tested-by: Oliver Hartkopp <socketcan@hartkopp.net>
Acked-by: Oliver Hartkopp <socketcan@hartkopp.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
AF_IUCV sockets offer a shutdown function. This patch makes sure
shutdown works for HS transport as well.
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In case of transport through HiperSockets the underlying network
interface may switch to DOWN state or the underlying network device
may recover. In both cases the socket must change to IUCV_DISCONN
state. If the interface goes down, af_iucv has a chance to notify
its connection peer in addition.
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For HS transport the maximum message size depends on the MTU-size
of the HS-device bound to the AF_IUCV socket. This patch adds a
getsockopt option MSGSIZE returning the maximum message size that
can be handled for this AF_IUCV socket.
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch saves the net_device in the iucv_sock structure during
bind in order to fasten skb sending.
In addition some other small improvements are made for HS transport:
- error checking when sending skbs
- locking changes in afiucv_hs_callback_txnotify
- skb freeing in afiucv_hs_callback_txnotify
And finally it contains code cleanup to get rid of iucv_skb_queue_purge.
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When polling on an AF_IUCV socket, writing should be blocked if the
number of pending messages exceeds a defined limit.
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A SEVER is missing in the callback of a receiving SEVERED. This may
inhibit z/VM to remove the corresponding IUCV-path completely.
This patch adds a SEVER in iucv_callback_connrej (together with
additional locking.
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
af_iucv differs unnecessarily between state IUCV_SEVERED and
IUCV_DISCONN. This patch removes state IUCV_SEVERED.
While simplifying af_iucv, this patch removes the 2nd invocation of
cpcmd as well.
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
af_iucv contains timer infrastructure which is not exploited.
This patch removes the timer related code parts.
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For HiperSockets transport skbs sent are bound to one of the
available HiperSockets devices. Add missing release of reference to
a HiperSockets device before freeing an skb.
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Closing an af_iucv socket may wait for confirmation of outstanding
send requests. This patch adds confirmation code for the new
HiperSockets transport.
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The AF_IUCV address family offers support for ancillary data.
This patch enables usage of ancillary data with the new
HiperSockets transport.
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
skb truesize currently accounts for sk_buff struct and part of skb head.
kmalloc() roundings are also ignored.
Considering that skb_shared_info is larger than sk_buff, its time to
take it into account for better memory accounting.
This patch introduces SKB_TRUESIZE(X) macro to centralize various
assumptions into a single place.
At skb alloc phase, we put skb_shared_info struct at the exact end of
skb head, to allow a better use of memory (lowering number of
reallocations), since kmalloc() gives us power-of-two memory blocks.
Unless SLUB/SLUB debug is active, both skb->head and skb_shared_info are
aligned to cache lines, as before.
Note: This patch might trigger performance regressions because of
misconfigured protocol stacks, hitting per socket or global memory
limits that were previously not reached. But its a necessary step for a
more accurate memory accounting.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Andi Kleen <ak@linux.intel.com>
CC: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The current transport mechanism for af_iucv is the z/VM offered
communications facility IUCV. To provide equivalent support when
running Linux in an LPAR, HiperSockets transport is added to the
AF_IUCV address family. It requires explicit binding of an AF_IUCV
socket to a HiperSockets device. A new packet_type ETH_P_AF_IUCV
is announced. An af_iucv specific transport header is defined
preceding the skb data. A small protocol is implemented for
connecting and for flow control/congestion management.
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Reviewed-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Code cleanup making make use of local variable for struct iucv_sock.
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For future af_iucv extensions the module should be able to run in LPAR
mode too. For this we use the new dynamic loading iucv interface.
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
-Wunused-but-set-variable generates compile warnings. The affected
variables are removed.
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a spin_unlock missing on the error path. There seems like no reason
why the lock should continue to be held if the kzalloc fail.
The semantic match that finds this problem is as follows:
(http://coccinelle.lip6.fr/)
// <smpl>
@@
expression E1;
@@
* spin_lock(E1,...);
<+... when != E1
if (...) {
... when != E1
* return ...;
}
...+>
* spin_unlock(E1,...);
// </smpl>
Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch removes from net/ (but not any netfilter files)
all the unnecessary return; statements that precede the
last closing brace of void functions.
It does not remove the returns that are immediately
preceded by a label as gcc doesn't like that.
Done via:
$ grep -rP --include=*.[ch] -l "return;\n}" net/ | \
xargs perl -i -e 'local $/ ; while (<>) { s/\n[ \t\n]+return;\n}/\n}/g; print; }'
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
sk_callback_lock rwlock actually protects sk->sk_sleep pointer, so we
need two atomic operations (and associated dirtying) per incoming
packet.
RCU conversion is pretty much needed :
1) Add a new structure, called "struct socket_wq" to hold all fields
that will need rcu_read_lock() protection (currently: a
wait_queue_head_t and a struct fasync_struct pointer).
[Future patch will add a list anchor for wakeup coalescing]
2) Attach one of such structure to each "struct socket" created in
sock_alloc_inode().
3) Respect RCU grace period when freeing a "struct socket_wq"
4) Change sk_sleep pointer in "struct sock" by sk_wq, pointer to "struct
socket_wq"
5) Change sk_sleep() function to use new sk->sk_wq instead of
sk->sk_sleep
6) Change sk_has_sleeper() to wq_has_sleeper() that must be used inside
a rcu_read_lock() section.
7) Change all sk_has_sleeper() callers to :
- Use rcu_read_lock() instead of read_lock(&sk->sk_callback_lock)
- Use wq_has_sleeper() to eventually wakeup tasks.
- Use rcu_read_unlock() instead of read_unlock(&sk->sk_callback_lock)
8) sock_wake_async() is modified to use rcu protection as well.
9) Exceptions :
macvtap, drivers/net/tun.c, af_unix use integrated "struct socket_wq"
instead of dynamically allocated ones. They dont need rcu freeing.
Some cleanups or followups are probably needed, (possible
sk_callback_lock conversion to a spinlock for example...).
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Define a new function to return the waitqueue of a "struct sock".
static inline wait_queue_head_t *sk_sleep(struct sock *sk)
{
return sk->sk_sleep;
}
Change all read occurrences of sk_sleep by a call to this function.
Needed for a future RCU conversion. sk_sleep wont be a field directly
available.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The generic __sock_create function has a kern argument which allows the
security system to make decisions based on if a socket is being created by
the kernel or by userspace. This patch passes that flag to the
net_proto_family specific create function, so it can do the same thing.
Signed-off-by: Eric Paris <eparis@redhat.com>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove duplicate sock_set_flag(sk, SOCK_ZAPPED) in iucv_sock_close,
which has been overlooked in September-commit
7514bab04e.
Cc: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Instead of modifying sk->sk_ack_backlog directly, use respective
socket functions.
Signed-off-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
All usages of structure net_proto_ops should be declared const.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This provides safety against negative optlen at the type
level instead of depending upon (sometimes non-trivial)
checks against this sprinkled all over the the place, in
each and every implementation.
Based upon work done by Arjan van de Ven and feedback
from Linus Torvalds.
Signed-off-by: David S. Miller <davem@davemloft.net>
After resuming from suspend, all af_iucv sockets are disconnected.
Ensure that iucv_accept_dequeue() can handle disconnected sockets
which are not yet accepted.
Signed-off-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Moving prepare_to_wait before the condition to avoid a race between
schedule_timeout and wake up.
The race can appear during iucv_sock_connect() and iucv_callback_connack().
Signed-off-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Adding memory barrier after the poll_wait function, paired with
receive callbacks. Adding fuctions sock_poll_wait and sk_has_sleeper
to wrap the memory barrier.
Without the memory barrier, following race can happen.
The race fires, when following code paths meet, and the tp->rcv_nxt
and __add_wait_queue updates stay in CPU caches.
CPU1 CPU2
sys_select receive packet
... ...
__add_wait_queue update tp->rcv_nxt
... ...
tp->rcv_nxt check sock_def_readable
... {
schedule ...
if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
wake_up_interruptible(sk->sk_sleep)
...
}
If there was no cache the code would work ok, since the wait_queue and
rcv_nxt are opposit to each other.
Meaning that once tp->rcv_nxt is updated by CPU2, the CPU1 either already
passed the tp->rcv_nxt check and sleeps, or will get the new value for
tp->rcv_nxt and will return with new data mask.
In both cases the process (CPU1) is being added to the wait queue, so the
waitqueue_active (CPU2) call cannot miss and will wake up CPU1.
The bad case is when the __add_wait_queue changes done by CPU1 stay in its
cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1 will then
endup calling schedule and sleep forever if there are no more data on the
socket.
Calls to poll_wait in following modules were ommited:
net/bluetooth/af_bluetooth.c
net/irda/af_irda.c
net/irda/irnet/irnet_ppp.c
net/mac80211/rc80211_pid_debugfs.c
net/phonet/socket.c
net/rds/af_rds.c
net/rfkill/core.c
net/sunrpc/cache.c
net/sunrpc/rpc_pipe.c
net/tipc/socket.c
Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next-2.6: (43 commits)
via-velocity: Fix velocity driver unmapping incorrect size.
mlx4_en: Remove redundant refill code on RX
mlx4_en: Removed redundant check on lso header size
mlx4_en: Cancel port_up check in transmit function
mlx4_en: using stop/start_all_queues
mlx4_en: Removed redundant skb->len check
mlx4_en: Counting all the dropped packets on the TX side
usbnet cdc_subset: fix issues talking to PXA gadgets
Net: qla3xxx, remove sleeping in atomic
ipv4: fix NULL pointer + success return in route lookup path
isdn: clean up documentation index
cfg80211: validate station settings
cfg80211: allow setting station parameters in mesh
cfg80211: allow adding/deleting stations on mesh
ath5k: fix beacon_int handling
MAINTAINERS: Fix Atheros pattern paths
ath9k: restore PS mode, before we put the chip into FULL SLEEP state.
ath9k: wait for beacon frame along with CAB
acer-wmi: fix rfkill conversion
ath5k: avoid PCI FATAL interrupts by restoring RETRY_TIMEOUT disabling
...
If the iucv message limit for a communication path is exceeded,
sendmsg() returns -EAGAIN instead of -EPIPE.
The calling application can then handle this error situtation,
e.g. to try again after waiting some time.
For blocking sockets, sendmsg() waits up to the socket timeout
before returning -EAGAIN. For the new wait condition, a macro
has been introduced and the iucv_sock_wait_state() has been
refactored to this macro.
Signed-off-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Change the if condition to exit sendmsg() if the socket in not connected.
Signed-off-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Patch establishes a dummy afiucv-device to make sure af_iucv is
notified as iucv-bus device about suspend/resume.
The PM freeze callback severs all iucv pathes of connected af_iucv sockets.
The PM thaw/restore callback switches the state of all previously connected
sockets to IUCV_DISCONN.
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
From: Ursula Braun <ubraun@linux.vnet.ibm.com>
net/iucv/af_iucv.c in net-next-2.6 is almost correct. 4 lines should
still be deleted. These are the remaining changes:
Signed-off-by: David S. Miller <davem@davemloft.net>
The SO_MSGLIMIT socket option modifies the message limit for new
IUCV communication paths.
The message limit specifies the maximum number of outstanding messages
that are allowed for connections. This setting can be lowered by z/VM
when an IUCV connection is established.
Expects an integer value in the range of 1 to 65535.
The default value is 65535.
The message limit must be set before calling connect() or listen()
for sockets.
If sockets are already connected or in state listen, changing the message
limit is not supported.
For reading the message limit value, unconnected sockets return the limit
that has been set or the default limit. For connected sockets, the actual
message limit is returned. The actual message limit is assigned by z/VM
for each connection and it depends on IUCV MSGLIMIT authorizations
specified for the z/VM guest virtual machine.
Signed-off-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If the skb cannot be copied to user iovec, always return -EFAULT.
The skb is enqueued again, except MSG_PEEK flag is set, to allow user space
applications to correct its iovec pointer.
Signed-off-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch provides the socket type SOCK_SEQPACKET in addition to
SOCK_STREAM.
AF_IUCV sockets of type SOCK_SEQPACKET supports an 1:1 mapping of
socket read or write operations to complete IUCV messages.
Socket data or IUCV message data is not fragmented as this is the
case for SOCK_STREAM sockets.
The intention is to help application developers who write
applications or device drivers using native IUCV interfaces
(Linux kernel or z/VM IUCV interfaces).
Signed-off-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Allow 'classification' of socket data that is sent or received over
an af_iucv socket. For classification of data, the target class of an
(native) iucv message is used.
This patch provides the cmsg interface for iucv_sock_recvmsg() and
iucv_sock_sendmsg(). Applications can use the msg_control field of
struct msghdr to set or get the target class as a
"socket control message" (SCM/CMSG).
Signed-off-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The patch allows to send and receive data in the parameter list of an
iucv message.
The parameter list is an arry of 8 bytes that are used by af_iucv as
follows:
0..6 7 bytes for socket data and
7 1 byte to store the data length.
Instead of storing the data length directly, the difference
between 0xFF and the data length is used.
This convention does not interfere with the existing use of PRM
messages for shutting down the send direction of an AF_IUCV socket
(shutdown() operation). Data lenghts greater than 7 (or PRM message
byte 8 is less than 0xF8) denotes to special messages.
Currently, the special SEND_SHUTDOWN message is supported only.
To use IPRM messages, both communicators must set the IUCV_IPRMDATA
flag during path negotiation, i.e. in iucv_connect() and
path_pending().
To be compatible to older af_iucv implementations, sending PRM
messages is controlled by the socket option SO_IPRMDATA_MSG.
Receiving PRM messages does not depend on the socket option (but
requires the IUCV_IPRMDATA path flag to be set).
Sending/Receiving data in the parameter list improves performance for
small amounts of data by reducing message_completion() interrupts and
memory copy operations.
Signed-off-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Provide the socket operations getsocktopt() and setsockopt() to enable/disable
sending of data in the parameter list of IUCV messages.
The patch sets respective flag only.
Signed-off-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If the af_iucv communication partner quiesces the path to shutdown its
receive direction, provide a quiesce callback implementation to shutdown
the (local) send direction. This ensures that both sides are synchronized.
Signed-off-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
AF_IUCV runs into a race when queuing incoming iucv messages
and receiving the resulting backlog.
If the Linux system is under pressure (high load or steal time),
the message queue grows up, but messages are not received and queued
onto the backlog queue. In that case, applications do not
receive any data with recvmsg() even if AF_IUCV puts incoming
messages onto the message queue.
The race can be avoided if the message queue spinlock in the
message_pending callback is spreaded across the entire callback
function.
Signed-off-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add few more sk states in iucv_sock_shutdown().
Signed-off-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Reject incoming iucv messages if the receive direction has been shut down.
It avoids that the queue of outstanding messages increases and exceeds the
message limit of the iucv communication path.
Signed-off-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
