A pointed out by Shin Hong, IPVS doesn't always use atomic operations
in an atomic manner. While this seems unlikely to be manifest in
strange behaviour, it seems appropriate to clean this up.
Cc: shin hong <hongshin@gmail.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Consitfy nlmsghdr arguments to a couple of functions as preparation
for the next patch, which will constify the netlink message data in
all nfnetlink users.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Log packets dropped by helpers using the netfilter logging API. This
is useful in combination with nfnetlink_log to analyze those packets
in userspace for debugging.
Signed-off-by: Patrick McHardy <kaber@trash.net>
commit f216f082b2
([NETFILTER]: bridge netfilter: deal with martians correctly)
added a refcount leak on in_dev.
Instead of using in_dev_get(), we can use __in_dev_get_rcu(),
as netfilter hooks are running under rcu_read_lock(), as pointed
by Patrick.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Kernel 2.6.30 introduced a patch [1] for the persistent option in the
netfilter SNAT target. This is exactly what we need here so I had a quick look
at the code and noticed that the patch is wrong. The logic is simply inverted.
The patch below fixes this.
Also note that because of this the default behavior of the SNAT target has
changed since kernel 2.6.30 as it now ignores the destination IP in choosing
the source IP for nating (which should only be the case if the persistent
option is set).
[1] http://git.eu.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=98d500d66cb7940747b424b245fc6a51ecfbf005
Signed-off-by: Maximilian Engelhardt <maxi@daemonizer.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
The inputted table is never modified, so should be considered const.
Signed-off-by: Jan Engelhardt <jengelh@medozas.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
This adds the second check that Rusty wanted to have a long time ago. :-)
Base chain policies must have absolute verdicts that cease processing
in the table, otherwise rule execution may continue in an unexpected
spurious fashion (e.g. next chain that follows in memory).
Signed-off-by: Jan Engelhardt <jengelh@medozas.de>
This adds a check that iptables's original author Rusty set forth in
a FIXME comment.
Underflows in iptables are better known as chain policies, and are
required to be unconditional or there would be a stochastical chance
for the policy rule to be skipped if it does not match. If that were
to happen, rule execution would continue in an unexpected spurious
fashion.
Signed-off-by: Jan Engelhardt <jengelh@medozas.de>
The "hook_entry" and "underflow" array contains values even for hooks
not provided, such as PREROUTING in conjunction with the "filter"
table. Usually, the values point to whatever the next rule is. For
the upcoming unconditionality and underflow checking patches however,
we must not inspect that arbitrary rule.
Skipping unassigned hooks seems like a good idea, also because
newinfo->hook_entry and newinfo->underflow will then continue to have
the poison value for detecting abnormalities.
Signed-off-by: Jan Engelhardt <jengelh@medozas.de>
Instead of inspecting each u32/char open-coded, clean up and make use
of memcmp. On some arches, memcmp is implemented as assembly or GCC's
__builtin_memcmp which can possibly take advantages of known
alignment.
Signed-off-by: Jan Engelhardt <jengelh@medozas.de>
When IPv4 and IPv6 matches were unified approx. 3.5 years ago, they
received new header filenames (e.g. xt_CLASSIFY.h). Let's remove the
old ones now.
Signed-off-by: Jan Engelhardt <jengelh@medozas.de>
ebt_log uses its own implementation of print_mac to print MAC addresses.
This patch converts it to use the %pM conversion specifier for printk.
Signed-off-by: Tobias Klauser <klto@zhaw.ch>
Signed-off-by: Patrick McHardy <kaber@trash.net>
dev_queue_xmit enqueue's a skb and calls qdisc_run which
dequeue's the skb and xmits it. In most cases, the skb that
is enqueue'd is the same one that is dequeue'd (unless the
queue gets stopped or multiple cpu's write to the same queue
and ends in a race with qdisc_run). For default qdiscs, we
can remove the redundant enqueue/dequeue and simply xmit the
skb since the default qdisc is work-conserving.
The patch uses a new flag - TCQ_F_CAN_BYPASS to identify the
default fast queue. The controversial part of the patch is
incrementing qlen when a skb is requeued - this is to avoid
checks like the second line below:
+ } else if ((q->flags & TCQ_F_CAN_BYPASS) && !qdisc_qlen(q) &&
>> !q->gso_skb &&
+ !test_and_set_bit(__QDISC_STATE_RUNNING, &q->state)) {
Results of a 2 hour testing for multiple netperf sessions (1,
2, 4, 8, 12 sessions on a 4 cpu system-X). The BW numbers are
aggregate Mb/s across iterations tested with this version on
System-X boxes with Chelsio 10gbps cards:
----------------------------------
Size | ORG BW NEW BW |
----------------------------------
128K | 156964 159381 |
256K | 158650 162042 |
----------------------------------
Changes from ver1:
1. Move sch_direct_xmit declaration from sch_generic.h to
pkt_sched.h
2. Update qdisc basic statistics for direct xmit path.
3. Set qlen to zero in qdisc_reset.
4. Changed some function names to more meaningful ones.
Signed-off-by: Krishna Kumar <krkumar2@in.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We use 1:1 mapping between QPs and SRQs on receive side,
so additional indirection level not required. Allocated the receive
buffers for the RSS QPs.
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is no point in using more QPs then actual number of receive rings.
If the RSS function for two streams gives the same result modulo number
of rings, they will arrive to the same RX ring anyway.
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
If the net device is identified as "sender" (number of sent packets
is higher then the number of received packets and the incoming packets are
small), set the moderation time to its low limit.
We do it because the incoming packets are acks, and we don't want to delay them
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
Hibernation is unsupported for now, which meets the actual
implementation in the driver. For free/thaw, the chip's D2 state should
be entered.
Signed-off-by: Daniel Mack <daniel@caiaq.de>
Acked-by: <steve.glendinning@smsc.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The datasheet say "When not enabling packing, the MinFrag value must
remain at 0". Do not set value to RxFragSize register if
TC35815_USE_PACKEDBUFFER disabled.
This is not a bugfix. No real problem reported on this.
Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
The Miss_Cnt register is cleared by reading. Accumulate its value to
rx_missed_errors count.
Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
The current timeout value is too short for very high-load condition
which jiffies might jump up in busy-loop.
Also add minimum delay before checking completion of MDIO.
Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
Clear Int_BLEx / Int_FDAEx after (not before) processing Rx interrupt.
This will reduce number of unnecessary interrupts.
Also print rx error messages only if netif_msg_rx_err() enabled.
Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
It seems Rx_StripCRC cause trouble on recovering from the BLEx (Buffer
List Exhaust) or FDAEx (Free Descriptor Area Exhaust) condition.
Do not use it.
Also bump version number up.
Reported-by: Ralf Roesch <ralf.roesch@rw-gmbh.de>
Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
Elsewhere the sin_family field holds a value with a name of the form
AF_..., so it seems reasonable to do so here as well. Also the values of
PF_INET and AF_INET are the same.
The semantic patch that makes this change is as follows:
(http://coccinelle.lip6.fr/)
// <smpl>
@@
struct sockaddr_in sip;
@@
(
sip.sin_family ==
- PF_INET
+ AF_INET
|
sip.sin_family !=
- PF_INET
+ AF_INET
|
sip.sin_family =
- PF_INET
+ AF_INET
)
// </smpl>
Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>
commit 88045b3cf0
cxgb3: fix mac index mapping
Override the mac index computation for the gen2 adapter,
as each port is expected to use index 0.
introduces a regression on 2 port 1G adapter
as its xauicfg vpd value is null.
Add a check on the device id.
Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Update the driver version number for any bug reports from end users.
Signed-off-by: Benjamin LaHaise <ben.lahaise@neterion.com>
Signed-off-by: Ramkrishna Vepa <ram.vepa@neterion.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch implements prefetching of skb->data from a copy of the pointer
in the descriptor (which is already in the L1 cache at this point). This
improves netperf rx performance (netperf -L 0,0 -c -H 192.168.254.2 -- -M
131072 -m 131072) by 4.9% on a P4 Xeon host.
Signed-off-by: Benjamin LaHaise <ben.lahaise@neterion.com>
Signed-off-by: Sreenivasa Honnur <sreenivasa.honnur@neterion.com>
Signed-off-by: Ramkrishna Vepa <ram.vepa@neterion.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch prefetches RxD descriptors which helps to lower the latency of a
cache miss in vxge_hw_ring_rxd_next_completed. This lowers the % of CPU
time used by vxge_hw_ring_rxd_next_completed() where the descriptor is
accessed in profiling netperf on a P4 Xeon from 1.5% to 1.0%.
Signed-off-by: Benjamin LaHaise <ben.lahaise@neterion.com>
Signed-off-by: Sreenivasa Honnur <sreenivasa.honnur@neterion.com>
Signed-off-by: Ramkrishna Vepa <ram.vepa@neterion.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
wmb() is extremely heavy on x86. The semantics required in the driver are
provided by mmiowb(), so use that and improve tx performance on P4 Xeons by
5-10%.
Signed-off-by: Benjamin LaHaise <ben.lahaise@neterion.com>
Signed-off-by: Sreenivasa Honnur <sreenivasa.honnur@neterion.com>
Signed-off-by: Ramkrishna Vepa <ram.vepa@neterion.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch fixes a case in the transmit completion code which was resulting
in pktgen hanging at the end of a run. The cause is due to the fact that
the ->cb[] area of an skb cannot be used in a network driver's transmit
path, as that area belongs to the network protocol. Pktgen hangs, as it
sends out the same packet multiple times, and vxge's use of this area of
the skb for a temporary list can only add the packet to the temporary list
once (while it may be on the queue many times). The fix is to remove this
abuse of skb->cb[]. Instead, skb pointers are placed into a temporary
stack array, and then free outside of the tx lock. This retains the smp
optimization of doing dev_kfree_skb() outside of the tx lock.
Signed-off-by: Benjamin LaHaise <ben.lahaise@neterion.com>
Signed-off-by: Sreenivasa Honnur <sreenivasa.honnur@neterion.com>
Signed-off-by: Ramkrishna Vepa <ram.vepa@neterion.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The vxge driver will drop a packet in its transmit function if the number
of TxDs available hits 0. Instead of doing that, simply stop the transmit
queue when transmitting a packet with the last available TxD.
Signed-off-by: Benjamin LaHaise <ben.lahaise@neterion.com>
Signed-off-by: Sreenivasa Honnur <sreenivasa.honnur@neterion.com>
Signed-off-by: Ramkrishna Vepa <ram.vepa@neterion.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The skb priority to vlan-qos egress mapping that can be configured using
set_egress_map with vconfig is overriden by the DCB code in the driver.
This patch allows this existing mechanism to work and will increase the
configuration flexibility of DCB mode on Linux.
A hierarchy of configuration is:
1. Modifies the ixgbe_select_queue() routine for DCB mode to return the
priority value from the VLAN tag. It will normally be zero, unless the egress
priority map has modified it. This will get packets into the correct queue and
result in the queue_mapping field being set correctly.
2. Any tc filter which modifies queue_mapping will be honored, as the filters
are handled after the vlan egress map is handled.
Signed-off-by: Lucy Liu <lucy.liu@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch is a bug fix to avoid system going into a bad state when
driver is loaded in context of kdump kernel. The patch fixes the issue
by performing a soft reset of pci function at probe time.
Signed-off-by: sarveshwarb <sarveshwarb@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This sockopt goes in line with SO_TYPE and SO_PROTOCOL. It makes it
possible for userspace programs to pass around file descriptors — I
am referring to arguments-to-functions, but it may even work for the
fd passing over UNIX sockets — without needing to also pass the
auxiliary information (PF_INET6/IPPROTO_TCP).
Signed-off-by: Jan Engelhardt <jengelh@medozas.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
