Linux immediately returns SYNACK on (spurious) SYN retransmits, but
keeps the SYNACK timer running independently. Thus the timer may
fire right after the SYNACK retransmit and causes a SYN-SYNACK
cross-fire burst.
Adopt the fast retransmit/recovery idea in established state by
re-arming the SYNACK timer after the fast (SYNACK) retransmit. The
timer may fire late up to 500ms due to the current SYNACK timer wheel,
but it's OK to be conservative when network is congested. Eric's new
listener design should address this issue.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add MIB counters for checksum errors in IP layer,
and TCP/UDP/ICMP layers, to help diagnose problems.
$ nstat -a | grep Csum
IcmpInCsumErrors 72 0.0
TcpInCsumErrors 382 0.0
UdpInCsumErrors 463221 0.0
Icmp6InCsumErrors 75 0.0
Udp6InCsumErrors 173442 0.0
IpExtInCsumErrors 10884 0.0
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Current tg3 shows hardware timestamping support for all devices
when that is true only for the hardware with PTP_CAPABLE flag
present.
Signed-off-by: Flavio Leitner <fbl@redhat.com>
Acked-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Instead of feeding net_secret[] at boot time, defer the init
at the point first socket is created.
This permits some platforms to use better entropy sources than
the ones available at boot time.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Lancer FW requires that the first cmd issued by the host-driver be an FLR.
So, re-order be_probe() to move be_cmd_function_reset() ahead of
be_cmd_fw_init().
Signed-off-by: Kalesh AP <kalesh.purayil@emulex.com>
Signed-off-by: Somnath Kotur <somnath.kotur@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Skyhawk-R requires V1 version of GET_FUNCTION_CONFIG cmd to be used for
querrying resources available per function.
Signed-off-by: Kalesh AP <kalesh.purayil@emulex.com>
Signed-off-by: Somnath Kotur <somnath.kotur@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pablo Neira Ayuso says:
====================
The following patchset contains relevant updates for the Netfilter
tree, they are:
* Enhancements for ipset: Add the counter extension for sets, this
information can be used from the iptables set match, to change
the matching behaviour. Jozsef required to add the extension
infrastructure and moved the existing timeout support upon it.
This also includes a change in net/sched/em_ipset to adapt it to
the new extension structure.
* Enhancements for performance boosting in nfnetlink_queue: Add new
configuration flags that allows user-space to receive big packets (GRO)
and to disable checksumming calculation. This were proposed by Eric
Dumazet during the Netfilter Workshop 2013 in Copenhagen. Florian
Westphal was kind enough to find the time to materialize the proposal.
* A sparse fix from Simon, he noticed it in the SCTP NAT helper, the fix
required a change in the interface of sctp_end_cksum.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Add support for another ARM member of the R-Car family, R-Car M1A, also known as
R8A77781 -- it will share the code with previously added R8A77790.
Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Change the type of the crc32 parameter of sctp_end_cksum()
from __be32 to __u32 to reflect that fact that it is passed
to cpu_to_le32().
There are five in-tree users of sctp_end_cksum().
The following four had warnings flagged by sparse which are
no longer present with this change.
net/netfilter/ipvs/ip_vs_proto_sctp.c:sctp_nat_csum()
net/netfilter/ipvs/ip_vs_proto_sctp.c:sctp_csum_check()
net/sctp/input.c:sctp_rcv_checksum()
net/sctp/output.c:sctp_packet_transmit()
The fifth user is net/netfilter/nf_nat_proto_sctp.c:sctp_manip_pkt().
It has been updated to pass a __u32 instead of a __be32,
the value in question was already calculated in cpu byte-order.
net/netfilter/nf_nat_proto_sctp.c:sctp_manip_pkt() has also
been updated to assign the return value of sctp_end_cksum()
directly to a variable of type __le32, matching the
type of the return value. Previously the return value
was assigned to a variable of type __be32 and then that variable
was finally assigned to another variable of type __le32.
Problems flagged by sparse.
Compile and sparse tested only.
Signed-off-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Userspace can now indicate that it can cope with larger-than-mtu sized
packets and packets that have invalid ipv4/tcp checksums.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Once we allow userspace to receive gso/gro packets, userspace
needs to be able to determine when checksums appear to be
broken, but are not.
NFQA_SKB_CSUMNOTREADY means 'checksums will be fixed in kernel
later, pretend they are ok'.
NFQA_SKB_GSO could be used for statistics, or to determine when
packet size exceeds mtu.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
skb_gso_segment is expensive, so it would be nice if we could
avoid it in the future. However, userspace needs to be prepared
to receive larger-than-mtu-packets (which will also have incorrect
l3/l4 checksums), so we cannot simply remove it.
The plan is to add a per-queue feature flag that userspace can
set when binding the queue.
The problem is that in nf_queue, we only have a queue number,
not the queue context/configuration settings.
This patch should have no impact other than the skb_gso_segment
call now being in a function that has access to the queue config
data.
A new size attribute in nf_queue_entry is needed so
nfnetlink_queue can duplicate the entry of the gso skb
when segmenting the skb while also copying the route key.
The follow up patch adds switch to disable skb_gso_segment when
queue config says so.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
required by future patch that will need to duplicate the
nf_queue_entry, bumping refcounts of the copy.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
The new revision of the set match supports to match the counters
and to suppress updating the counters at matching too.
At the set:list types, the updating of the subcounters can be
suppressed as well.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Introduce extensions to elements in the core and prepare timeout as
the first one.
This patch also modifies the em_ipset classifier to use the new
extension struct layout.
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Fixed: warning: cast from pointer to integer of different size
The Hyper-V hosts always use 64 bit request id. The guests can have 32 or 64
bit pointers which equal to the ulong type size. So we cast it to ulong type.
And, assigning 32bit integer to 64 bit variable works fine.
The VMBus returns the same id in the completion packet. But the value has no
effect on the host side.
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Petko Manolov says:
====================
This series of patches is fixing a bug related to multiple control URB
submissions (noted by Sarah Sharp), optimizes read and write_mii_word
routines and removes socket buffer pool used in the receive path.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Pegasus driver used single callback for sync and async control URBs.
Special flags were employed to distinguish between both, but due to flawed
logic it didn't always work. As a result of this change
[get|set]_registers() are now much simpler. Async write is also leaner
and does not use single, statically allocated memory for usb_ctrlrequest,
which is another potential race when asynchronously submitting URBs.
Signed-off-by: Petko Manolov <petkan@nucleusys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Duplicated code in routines reading and writing MII registers is now
packed in __mii_op().
Signed-off-by: Petko Manolov <petkan@nucleusys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The socket buffer pool for the receive path is now gone. It's existence
didn't make much difference (performance-wise) and the code is better off
without the spinlocks protecting it.
Signed-off-by: Petko Manolov <petkan@nucleusys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Following patch adds icmp-registration module for ipv6. It allows
ipv6 protocol to register icmp_sender which is used for sending
ipv6 icmp msgs. This extra layer allows us to kill ipv6 dependency
for sending icmp packets.
This patch also fixes ip_tunnel compilation problem when ip_tunnel
is statically compiled in kernel but ipv6 is module
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Increase fragmentation hash bucket size to 1024 from old 64 elems.
After we increased the frag mem limits commit c2a93660 (net: increase
fragment memory usage limits) the hash size of 64 elements is simply
too small. Also considering the mem limit is per netns and the hash
table is shared for all netns.
For the embedded people, note that this increase will change the hash
table/array from using approx 1 Kbytes to 16 Kbytes.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
ARM cannot handle udelay for more than 2 miliseconds, and
it is rather bad style to block the cpu for 16ms anyway,
so let's use msleep instead.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Chas Williams <chas@cmf.nrl.navy.mil>
Cc: linux-atm-general@lists.sourceforge.net
Cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
Nicolas Dichtel says:
====================
The goal of this patchset is to be able to get all infos exported via the
/proc/net/packet and also beeing able to get filter associated to af_packet
sockets.
As usual, the patch against iproute2 will be sent once the patches are included
and net-next merged. I can send it on demand.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch allows to dump BPF filters attached to a socket with
SO_ATTACH_FILTER.
Note that we check CAP_SYS_ADMIN before allowing to dump this info.
For now, only AF_PACKET sockets use this feature.
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
sk_rmem_alloc is disclosed via /proc/net/packet but not via netlink messages.
The goal is to have the same level of information.
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This value is disclosed via /proc/net/packet but not via netlink messages.
The goal is to have the same level of information.
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Testing like this for TP_STATUS_AVAILABLE clearly is a stupid bug
since it always returns true. Fix this by only checking for flags
where the kernel owns the packet and negate this result, since we
also could run into the non-zero status TP_STATUS_WRONG_FORMAT
and need to reclaim frames.
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Allow configuring the default destination port on a per-device basis.
Adds new netlink paramater IFLA_VXLAN_PORT to allow setting destination
port when creating new vxlan.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Rather than computing source port and returning it in host order
then swapping later, go ahead and compute it in network order to
start with. Cleaner and less error prone.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Source compatiability for build iproute2 was broken by:
commit c7995c43fa
Author: Atzm Watanabe <atzm@stratosphere.co.jp>
vxlan: Allow setting destination to unicast address.
Since this commit has not made it upstream (still net-next),
and better to avoid gratitious changes to exported API's;
go back to original definition, and add a comment.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The NDA_PORT attribute was added, but the author wasn't careful
about width (port is 16 bits), or byte order. The attribute was
being dumped as 16 bits, but only 32 bit value would be accepted
when setting up a device. Also, the remote port is in network
byte order and was being compared with default port in host byte
order.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The default port for VXLAN is not same as IANA value.
Document this.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Current bridge fdb update code does not seem to update the port
during fdb update. This patch adds a check for fdb dst (port)
change during fdb update. Also rearranges the call to
fdb_notify to send only one notification for create and update.
Changelog:
v2 - Change notify flag to bool
Signed-off-by: Roopa Prabhu <roopa@cumulusnetworks.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jeff Kirsher says:
====================
This series contains updates to e1000e, igb and ixgbe.
There are 2 patches in this series which could be applied to net,
but since Linus is so very close to releasing 3.9, I do not think
it prudent to try and push these into net at this time. I have CC'd
stable on these patches so that they can queue them up as soon as
3.9 gets released.
The 2 patches are:
e1000e: fix numeric overflow in phc settime method
ixgbe: fix EICR write in ixgbe_msix_other
Richard provides a fix for e1000e by using a helper function from time.h
to resolve a unintended overflow in the PTP settime function.
Bruce provides a fix to wait for NAPI to be done with the current context
after disabling interrupts and then disable NAPI when the interface
is going down. This fixes a possible "unable to handle kernel paging
request" panic in net-next.
Andi Kleen provides a patch for igb to use mdelay instead of udelay
when we needed 100000us.
Jacob provides a fix for ixgbe to simply mask the lower 16bits off so that
ixgbe_msix_other does not write them in the EICR, which causes them to
remain high and be properly handled by the clean_rings interrupt routine
as normal.
Emil cleans up the logic in ixgbe_setup_loopback_test() to only access
registers applicable to the MAC type. In addition, removes majority
of the AUTOC register reads by using a cached value instead to avoid
writing corrupted values to AUTOC due to bad FW. Emil also add support
for disabling link during boot time. Lastly, he provides a patch which
adds the MAC type to the version in ethtool_regs which will make it
easier to check the MAC type when dumping registers with ethtool.
There is a separate ethtool tool patch which is dependent upon Emil's
last patch of the series to add the MAC type to the version in
ethtool_regs, which will be sent separately.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Or Gerlitz says:
====================
This series adds support for the SRIOV ndo_set_vf callbacks to the mlx4 driver.
Series done against the net-next tree as of commit 37fe066098 "net:
fix address check in rtnl_fdb_del"
We have successfully tested the series on net-next, except for getting
the VF link info issue I have reported earlier today on netdev, we
see the problem for both ixgbe and mlx4 VFs. Just to make sure get
VF config is working OK with patch #6 - we have run it over 3.8.8 too.
We added to the V1 series two patches that disable HW timestamping
when running over a VF, as this isn't supported yet.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Support getting VF config.
Signed-off-by: Rony Efraim <ronye@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>