summaryrefslogtreecommitdiffstats
path: root/net
Commit message (Collapse)AuthorAgeFilesLines
* ipv4: new ip_no_pmtu_disc mode to always discard incoming frag needed msgsHannes Frederic Sowa2013-12-181-1/+3
| | | | | | | | | | This new mode discards all incoming fragmentation-needed notifications as I guess was originally intended with this knob. To not break backward compatibility too much, I only added a special case for mode 2 in the receiving path. Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org> Signed-off-by: David S. Miller <davem@davemloft.net>
* inet: make no_pmtu_disc per namespace and kill ipv4_configHannes Frederic Sowa2013-12-185-14/+11
| | | | | | | | The other field in ipv4_config, log_martians, was converted to a per-interface setting, so we can just remove the whole structure. Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org> Signed-off-by: David S. Miller <davem@davemloft.net>
* Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/netDavid S. Miller2013-12-1831-157/+311
|\ | | | | | | | | | | | | | | | | | | Conflicts: drivers/net/ethernet/intel/i40e/i40e_main.c drivers/net/macvtap.c Both minor merge hassles, simple overlapping changes. Signed-off-by: David S. Miller <davem@davemloft.net>
| * neigh: Netlink notification for administrative NUD state changeBob Gilligan2013-12-171-0/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The neighbour code sends up an RTM_NEWNEIGH netlink notification if the NUD state of a neighbour cache entry is changed by a timer (e.g. from REACHABLE to STALE), even if the lladdr of the entry has not changed. But an administrative change to the the NUD state of a neighbour cache entry that does not change the lladdr (e.g. via "ip -4 neigh change ... nud ...") does not trigger a netlink notification. This means that netlink listeners will not hear about administrative NUD state changes such as from a resolved state to PERMANENT. This patch changes the neighbor code to generate an RTM_NEWNEIGH message when the NUD state of an entry is changed administratively. Signed-off-by: Bob Gilligan <gilligan@aristanetworks.com> Acked-by: Nicolas Dichtel <nicolas.dichtel@6wind.com> Signed-off-by: David S. Miller <davem@davemloft.net>
| * Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nfDavid S. Miller2013-12-173-1/+3
| |\ | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Pablo Neira Ayuso says: ==================== The following patchset contains two Netfilter fixes for your net tree, they are: * Fix endianness in nft_reject, the NFTA_REJECT_TYPE netlink attributes was not converted to network byte order as needed by all nfnetlink subsystems, from Eric Leblond. * Restrict SYNPROXY target to INPUT and FORWARD chains, this avoid a possible crash due to misconfigurations, from Patrick McHardy. ==================== Signed-off-by: David S. Miller <davem@davemloft.net>
| | * netfilter: nft_reject: fix endianness in dump functionEric Leblond2013-12-121-1/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The dump function in nft_reject_ipv4 was not converting a u32 field to network order before sending it to userspace, this needs to happen for consistency with other nf_tables and nfnetlink subsystems. Signed-off-by: Eric Leblond <eric@regit.org> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
| | * netfilter: SYNPROXY target: restrict to INPUT/FORWARDPatrick McHardy2013-12-112-0/+2
| | | | | | | | | | | | | | | | | | | | | | | | | | | Fix a crash in synproxy_send_tcp() when using the SYNPROXY target in the PREROUTING chain caused by missing routing information. Reported-by: Nicki P. <xastx@gmx.de> Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
| * | net: unix: allow bind to fail on mutex lockSasha Levin2013-12-171-2/+6
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This is similar to the set_peek_off patch where calling bind while the socket is stuck in unix_dgram_recvmsg() will block and cause a hung task spew after a while. This is also the last place that did a straightforward mutex_lock(), so there shouldn't be any more of these patches. Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
| * | udp: ipv4: do not use sk_dst_lock from softirq contextEric Dumazet2013-12-171-9/+4
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Using sk_dst_lock from softirq context is not supported right now. Instead of adding BH protection everywhere, udp_sk_rx_dst_set() can instead use xchg(), as suggested by David. Reported-by: Fengguang Wu <fengguang.wu@intel.com> Fixes: 975022310233 ("udp: ipv4: must add synchronization in udp_sk_rx_dst_set()") Signed-off-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
| * | sctp: loading sctp when load sctp_probewangweidong2013-12-161-1/+16
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | when I modprobe sctp_probe, it failed with "FATAL: ". I found that sctp should load before sctp_probe register jprobe. So I add a sctp_setup_jprobe for loading 'sctp' when first failed to register jprobe, just do this similar to dccp_probe. v2: add MODULE_SOFTDEP and check of request_module, as suggested by Neil Signed-off-by: Wang Weidong <wangweidong1@huawei.com> Acked-by: Neil Horman <nhorman@tuxdriver.com> Signed-off-by: David S. Miller <davem@davemloft.net>
| * | Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/netLinus Torvalds2013-12-1559-291/+472
| |\ \ | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Pull networking fixes from David Miller: 1) Revert CHECKSUM_COMPLETE optimization in pskb_trim_rcsum(), I can't figure out why it breaks things. 2) Fix comparison in netfilter ipset's hash_netnet4_data_equal(), it was basically doing "x == x", from Dave Jones. 3) Freescale FEC driver was DMA mapping the wrong number of bytes, from Sebastian Siewior. 4) Blackhole and prohibit routes in ipv6 were not doing the right thing because their ->input and ->output methods were not being assigned correctly. Now they behave properly like their ipv4 counterparts. From Kamala R. 5) Several drivers advertise the NETIF_F_FRAGLIST capability, but really do not support this feature and will send garbage packets if fed fraglist SKBs. From Eric Dumazet. 6) Fix long standing user triggerable BUG_ON over loopback in RDS protocol stack, from Venkat Venkatsubra. 7) Several not so common code paths can potentially try to invoke packet scheduler actions that might be NULL without checking. Shore things up by either 1) defining a method as mandatory and erroring on registration if that method is NULL 2) defininig a method as optional and the registration function hooks up a default implementation when NULL is seen. From Jamal Hadi Salim. 8) Fix fragment detection in xen-natback driver, from Paul Durrant. 9) Kill dangling enter_memory_pressure method in cg_proto ops, from Eric W Biederman. 10) SKBs that traverse namespaces should have their local_df cleared, from Hannes Frederic Sowa. 11) IOCB file position is not being updated by macvtap_aio_read() and tun_chr_aio_read(). From Zhi Yong Wu. 12) Don't free virtio_net netdev before releasing all of the NAPI instances. From Andrey Vagin. 13) Procfs entry leak in xt_hashlimit, from Sergey Popovich. 14) IPv6 routes that are no cached routes should not count against the garbage collection limits. We had this almost right, but were missing handling addrconf generated routes properly. From Hannes Frederic Sowa. 15) fib{4,6}_rule_suppress() have to consider potentially seeing NULL route info when they are called, from Stefan Tomanek. 16) TUN and MACVTAP have had truncated packet signalling for some time, fix from Jason Wang. 17) Fix use after frrr in __udp4_lib_rcv(), from Eric Dumazet. 18) xen-netback does not interpret the NAPI budget properly for TX work, fix from Paul Durrant. * git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (132 commits) igb: Fix for issue where values could be too high for udelay function. i40e: fix null dereference xen-netback: fix gso_prefix check net: make neigh_priv_len in struct net_device 16bit instead of 8bit drivers: net: cpsw: fix for cpsw crash when build as modules xen-netback: napi: don't prematurely request a tx event xen-netback: napi: fix abuse of budget sch_tbf: use do_div() for 64-bit divide udp: ipv4: must add synchronization in udp_sk_rx_dst_set() net:fec: remove duplicate lines in comment about errata ERR006358 Revert "8390 : Replace ei_debug with msg_enable/NETIF_MSG_* feature" 8390 : Replace ei_debug with msg_enable/NETIF_MSG_* feature xen-netback: make sure skb linear area covers checksum field net: smc91x: Fix device tree based configuration so it's usable udp: ipv4: fix potential use after free in udp_v4_early_demux() macvtap: signal truncated packets tun: unbreak truncated packet signalling net: sched: htb: fix the calculation of quantum net: sched: tbf: fix the calculation of max_size micrel: add support for KSZ8041RNLI ...
| | * | sch_tbf: use do_div() for 64-bit divideYang Yingliang2013-12-111-2/+4
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | It's doing a 64-bit divide which is not supported on 32-bit architectures in psched_ns_t2l(). The correct way to do this is to use do_div(). It's introduced by commit cc106e441a63 ("net: sched: tbf: fix the calculation of max_size") Reported-by: kbuild test robot <fengguang.wu@intel.com> Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Signed-off-by: David S. Miller <davem@davemloft.net>
| | * | udp: ipv4: must add synchronization in udp_sk_rx_dst_set()Eric Dumazet2013-12-111-6/+16
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Unlike TCP, UDP input path does not hold the socket lock. Before messing with sk->sk_rx_dst, we must use a spinlock, otherwise multiple cpus could leak a refcount. This patch also takes care of renewing a stale dst entry. (When the sk->sk_rx_dst would not be used by IP early demux) Fixes: 421b3885bf6d ("udp: ipv4: Add udp early demux") Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Shawn Bohrer <sbohrer@rgmadvisors.com> Signed-off-by: David S. Miller <davem@davemloft.net>
| | * | udp: ipv4: fix potential use after free in udp_v4_early_demux()Eric Dumazet2013-12-111-3/+6
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | pskb_may_pull() can reallocate skb->head, we need to move the initialization of iph and uh pointers after its call. Fixes: 421b3885bf6d ("udp: ipv4: Add udp early demux") Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Shawn Bohrer <sbohrer@rgmadvisors.com> Signed-off-by: David S. Miller <davem@davemloft.net>
| | * | net: sched: htb: fix the calculation of quantumYang Yingliang2013-12-111-8/+12
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Now, 32bit rates may be not the true rate. So use rate_bytes_ps which is from max(rate32, rate64) to calcualte quantum. Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Acked-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
| | * | net: sched: tbf: fix the calculation of max_sizeYang Yingliang2013-12-111-45/+70
| | |/ | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Current max_size is caluated from rate table. Now, the rate table has been replaced and it's wrong to caculate max_size based on this rate table. It can lead wrong calculation of max_size. The burst in kernel may be lower than user asked, because burst may gets some loss when transform it to buffer(E.g. "burst 40kb rate 30mbit/s") and it seems we cannot avoid this loss. Burst's value(max_size) based on rate table may be equal user asked. If a packet's length is max_size, this packet will be stalled in tbf_dequeue() because its length is above the burst in kernel so that it cannot get enough tokens. The max_size guards against enqueuing packet sizes above q->buffer "time" in tbf_enqueue(). To make consistent with the calculation of tokens, this patch add a helper psched_ns_t2l() to calculate burst(max_size) directly to fix this problem. After this fix, we can support to using 64bit rates to calculate burst as well. Signed-off-by: Yang Yingliang <yangyingliang@huawei.com> Acked-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
| | * udp: ipv4: fix an use after free in __udp4_lib_rcv()Eric Dumazet2013-12-101-10/+6
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Dave Jones reported a use after free in UDP stack : [ 5059.434216] ========================= [ 5059.434314] [ BUG: held lock freed! ] [ 5059.434420] 3.13.0-rc3+ #9 Not tainted [ 5059.434520] ------------------------- [ 5059.434620] named/863 is freeing memory ffff88005e960000-ffff88005e96061f, with a lock still held there! [ 5059.434815] (slock-AF_INET){+.-...}, at: [<ffffffff8149bd21>] udp_queue_rcv_skb+0xd1/0x4b0 [ 5059.435012] 3 locks held by named/863: [ 5059.435086] #0: (rcu_read_lock){.+.+..}, at: [<ffffffff8143054d>] __netif_receive_skb_core+0x11d/0x940 [ 5059.435295] #1: (rcu_read_lock){.+.+..}, at: [<ffffffff81467a5e>] ip_local_deliver_finish+0x3e/0x410 [ 5059.435500] #2: (slock-AF_INET){+.-...}, at: [<ffffffff8149bd21>] udp_queue_rcv_skb+0xd1/0x4b0 [ 5059.435734] stack backtrace: [ 5059.435858] CPU: 0 PID: 863 Comm: named Not tainted 3.13.0-rc3+ #9 [loadavg: 0.21 0.06 0.06 1/115 1365] [ 5059.436052] Hardware name: /D510MO, BIOS MOPNV10J.86A.0175.2010.0308.0620 03/08/2010 [ 5059.436223] 0000000000000002 ffff88007e203ad8 ffffffff8153a372 ffff8800677130e0 [ 5059.436390] ffff88007e203b10 ffffffff8108cafa ffff88005e960000 ffff88007b00cfc0 [ 5059.436554] ffffea00017a5800 ffffffff8141c490 0000000000000246 ffff88007e203b48 [ 5059.436718] Call Trace: [ 5059.436769] <IRQ> [<ffffffff8153a372>] dump_stack+0x4d/0x66 [ 5059.436904] [<ffffffff8108cafa>] debug_check_no_locks_freed+0x15a/0x160 [ 5059.437037] [<ffffffff8141c490>] ? __sk_free+0x110/0x230 [ 5059.437147] [<ffffffff8112da2a>] kmem_cache_free+0x6a/0x150 [ 5059.437260] [<ffffffff8141c490>] __sk_free+0x110/0x230 [ 5059.437364] [<ffffffff8141c5c9>] sk_free+0x19/0x20 [ 5059.437463] [<ffffffff8141cb25>] sock_edemux+0x25/0x40 [ 5059.437567] [<ffffffff8141c181>] sock_queue_rcv_skb+0x81/0x280 [ 5059.437685] [<ffffffff8149bd21>] ? udp_queue_rcv_skb+0xd1/0x4b0 [ 5059.437805] [<ffffffff81499c82>] __udp_queue_rcv_skb+0x42/0x240 [ 5059.437925] [<ffffffff81541d25>] ? _raw_spin_lock+0x65/0x70 [ 5059.438038] [<ffffffff8149bebb>] udp_queue_rcv_skb+0x26b/0x4b0 [ 5059.438155] [<ffffffff8149c712>] __udp4_lib_rcv+0x152/0xb00 [ 5059.438269] [<ffffffff8149d7f5>] udp_rcv+0x15/0x20 [ 5059.438367] [<ffffffff81467b2f>] ip_local_deliver_finish+0x10f/0x410 [ 5059.438492] [<ffffffff81467a5e>] ? ip_local_deliver_finish+0x3e/0x410 [ 5059.438621] [<ffffffff81468653>] ip_local_deliver+0x43/0x80 [ 5059.438733] [<ffffffff81467f70>] ip_rcv_finish+0x140/0x5a0 [ 5059.438843] [<ffffffff81468926>] ip_rcv+0x296/0x3f0 [ 5059.438945] [<ffffffff81430b72>] __netif_receive_skb_core+0x742/0x940 [ 5059.439074] [<ffffffff8143054d>] ? __netif_receive_skb_core+0x11d/0x940 [ 5059.442231] [<ffffffff8108c81d>] ? trace_hardirqs_on+0xd/0x10 [ 5059.442231] [<ffffffff81430d83>] __netif_receive_skb+0x13/0x60 [ 5059.442231] [<ffffffff81431c1e>] netif_receive_skb+0x1e/0x1f0 [ 5059.442231] [<ffffffff814334e0>] napi_gro_receive+0x70/0xa0 [ 5059.442231] [<ffffffffa01de426>] rtl8169_poll+0x166/0x700 [r8169] [ 5059.442231] [<ffffffff81432bc9>] net_rx_action+0x129/0x1e0 [ 5059.442231] [<ffffffff810478cd>] __do_softirq+0xed/0x240 [ 5059.442231] [<ffffffff81047e25>] irq_exit+0x125/0x140 [ 5059.442231] [<ffffffff81004241>] do_IRQ+0x51/0xc0 [ 5059.442231] [<ffffffff81542bef>] common_interrupt+0x6f/0x6f We need to keep a reference on the socket, by using skb_steal_sock() at the right place. Note that another patch is needed to fix a race in udp_sk_rx_dst_set(), as we hold no lock protecting the dst. Fixes: 421b3885bf6d ("udp: ipv4: Add udp early demux") Reported-by: Dave Jones <davej@redhat.com> Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Shawn Bohrer <sbohrer@rgmadvisors.com> Signed-off-by: David S. Miller <davem@davemloft.net>
| | * sctp: fix up a spacingwangweidong2013-12-101-5/+2
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | fix up spacing of proc_sctp_do_hmac_alg for according to the proc_sctp_do_rto_min[max] in sysctl.c Suggested-by: Daniel Borkmann <dborkman@redhat.com> Signed-off-by: Wang Weidong <wangweidong1@huawei.com> Acked-by: Vlad Yasevich <vyasevich@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
| | * sctp: add check rto_min and rto_max in sysctlwangweidong2013-12-101-4/+65
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | rto_min should be smaller than rto_max while rto_max should be larger than rto_min. Add two proc_handler for the checking. Suggested-by: Vlad Yasevich <vyasevich@gmail.com> Signed-off-by: Wang Weidong <wangweidong1@huawei.com> Acked-by: Vlad Yasevich <vyasevich@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
| | * sctp: check the rto_min and rto_max in setsockoptwangweidong2013-12-101-10/+22
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | When we set 0 to rto_min or rto_max, just not change the value. Also we should check the rto_min > rto_max. Suggested-by: Vlad Yasevich <vyasevich@gmail.com> Signed-off-by: Wang Weidong <wangweidong1@huawei.com> Acked-by: Vlad Yasevich <vyasevich@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
| | * ipv6: do not erase dst address with flow label destinationFlorent Fourcot2013-12-106-6/+0
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This patch is following b579035ff766c9412e2b92abf5cab794bff102b6 "ipv6: remove old conditions on flow label sharing" Since there is no reason to restrict a label to a destination, we should not erase the destination value of a socket with the value contained in the flow label storage. This patch allows to really have the same flow label to more than one destination. Signed-off-by: Florent Fourcot <florent.fourcot@enst-bretagne.fr> Reviewed-by: Hannes Frederic Sowa <hannes@stressinduktion.org> Signed-off-by: David S. Miller <davem@davemloft.net>
| | * sctp: properly latch and use autoclose value from sock to associationNeil Horman2013-12-104-11/+13
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Currently, sctp associations latch a sockets autoclose value to an association at association init time, subject to capping constraints from the max_autoclose sysctl value. This leads to an odd situation where an application may set a socket level autoclose timeout, but sliently sctp will limit the autoclose timeout to something less than that. Fix this by modifying the autoclose setsockopt function to check the limit, cap it and warn the user via syslog that the timeout is capped. This will allow getsockopt to return valid autoclose timeout values that reflect what subsequent associations actually use. While were at it, also elimintate the assoc->autoclose variable, it duplicates whats in the timeout array, which leads to multiple sources for the same information, that may differ (as the former isn't subject to any capping). This gives us the timeout information in a canonical place and saves some space in the association structure as well. Signed-off-by: Neil Horman <nhorman@tuxdriver.com> Acked-by: Vlad Yasevich <vyasevich@gmail.com> CC: Wang Weidong <wangweidong1@huawei.com> CC: David Miller <davem@davemloft.net> CC: Vlad Yasevich <vyasevich@gmail.com> CC: netdev@vger.kernel.org Signed-off-by: David S. Miller <davem@davemloft.net>
| | * tipc: protect handler_enabled variable with qitem_lock spin lockYing Xue2013-12-101-3/+8
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | 'handler_enabled' is a global flag indicating whether the TIPC signal handling service is enabled or not. The lack of lock protection for this flag incurs a risk for contention, so that a tipc_k_signal() call might queue a signal handler to a destroyed signal queue, with unpredictable results. To correct this, we let the already existing 'qitem_lock' protect the flag, as it already does with the queue itself. This way, we ensure that the flag always is consistent across all cores. Signed-off-by: Ying Xue <ying.xue@windriver.com> Reviewed-by: Paul Gortmaker <paul.gortmaker@windriver.com> Signed-off-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: David S. Miller <davem@davemloft.net>
| | * tipc: correct the order of stopping services at rmmodJon Paul Maloy2013-12-101-3/+4
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The 'signal handler' service in TIPC is a mechanism that makes it possible to postpone execution of functions, by launcing them into a job queue for execution in a separate tasklet, independent of the launching execution thread. When we do rmmod on the tipc module, this service is stopped after the network service. At the same time, the stopping of the network service may itself launch jobs for execution, with the risk that these functions may be scheduled for execution after the data structures meant to be accessed by the job have already been deleted. We have seen this happen, most often resulting in an oops. This commit ensures that the signal handler is the very first to be stopped when TIPC is shut down, so there are no surprises during the cleanup of the other services. Signed-off-by: Jon Maloy <jon.maloy@ericsson.com> Reviewed-by: Paul Gortmaker <paul.gortmaker@windriver.com> Signed-off-by: David S. Miller <davem@davemloft.net>
| | * net: unix: allow set_peek_off to failSasha Levin2013-12-102-3/+7
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | unix_dgram_recvmsg() will hold the readlock of the socket until recv is complete. In the same time, we may try to setsockopt(SO_PEEK_OFF) which will hang until unix_dgram_recvmsg() will complete (which can take a while) without allowing us to break out of it, triggering a hung task spew. Instead, allow set_peek_off to fail, this way userspace will not hang. Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Acked-by: Pavel Emelyanov <xemul@parallels.com> Signed-off-by: David S. Miller <davem@davemloft.net>
| | * inet: fix NULL pointer Oops in fib(6)_rule_suppressStefan Tomanek2013-12-102-2/+9
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This changes ensures that the routing entry investigated by the suppress function actually does point to a device struct before following that pointer, fixing a possible kernel oops situation when verifying the interface group associated with a routing table entry. According to Daniel Golle, this Oops can be triggered by a user process trying to establish an outgoing IPv6 connection while having no real IPv6 connectivity set up (only autoassigned link-local addresses). Fixes: 6ef94cfafba15 ("fib_rules: add route suppression based on ifgroup") Reported-by: Daniel Golle <daniel.golle@gmail.com> Tested-by: Daniel Golle <daniel.golle@gmail.com> Signed-off-by: Stefan Tomanek <stefan.tomanek@wertarbyte.de> Signed-off-by: David S. Miller <davem@davemloft.net>
| | * net: drop_monitor: fix the value of maxattrChangli Gao2013-12-091-1/+0
| | | | | | | | | | | | | | | | | | | | | | | | maxattr in genl_family should be used to save the max attribute type, but not the max command type. Drop monitor doesn't support any attributes, so we should leave it as zero. Signed-off-by: David S. Miller <davem@davemloft.net>
| | * ipv6: don't count addrconf generated routes against gc limitHannes Frederic Sowa2013-12-091-5/+3
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Brett Ciphery reported that new ipv6 addresses failed to get installed because the addrconf generated dsts where counted against the dst gc limit. We don't need to count those routes like we currently don't count administratively added routes. Because the max_addresses check enforces a limit on unbounded address generation first in case someone plays with router advertisments, we are still safe here. Reported-by: Brett Ciphery <brett.ciphery@windriver.com> Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org> Signed-off-by: David S. Miller <davem@davemloft.net>
| | * Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nfDavid S. Miller2013-12-093-28/+45
| | |\ | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Pablo Neira Ayuso says: ==================== The following patchset contains three Netfilter fixes for your net tree, they are: * fix incorrect comparison in the new netnet hash ipset type, from Dave Jones. * fix splat in hashlimit due to missing removal of the content of its proc entry in netnamespaces, from Sergey Popovich. * fix missing rule flushing operation by table in nf_tables. Table flushing was already discussed back in October but this got lost and no patch has hit the tree to address this issue so far, from me. ==================== Signed-off-by: David S. Miller <davem@davemloft.net>
| | | * netfilter: nf_tables: fix missing rules flushing per tablePablo Neira Ayuso2013-12-071-13/+33
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This patch allows you to atomically remove all rules stored in a table via the NFT_MSG_DELRULE command. You only need to indicate the specific table and no chain to flush all rules stored in that table. Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
| | | * netfilter: xt_hashlimit: fix proc entry leak in netns destroy pathSergey Popovich2013-12-071-14/+11
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | In (32263dd1b netfilter: xt_hashlimit: fix namespace destroy path) the hashlimit_net_exit() function is always called right before hashlimit_mt_destroy() to release netns data. If you use xt_hashlimit with IPv4 and IPv6 together, this produces the following splat via netconsole in the netns destroy path: Pid: 9499, comm: kworker/u:0 Tainted: G WC O 3.2.0-5-netctl-amd64-core2 Call Trace: [<ffffffff8104708d>] ? warn_slowpath_common+0x78/0x8c [<ffffffff81047139>] ? warn_slowpath_fmt+0x45/0x4a [<ffffffff81144a99>] ? remove_proc_entry+0xd8/0x22e [<ffffffff810ebbaa>] ? kfree+0x5b/0x6c [<ffffffffa043c501>] ? hashlimit_net_exit+0x45/0x8d [xt_hashlimit] [<ffffffff8128ab30>] ? ops_exit_list+0x1c/0x44 [<ffffffff8128b28e>] ? cleanup_net+0xf1/0x180 [<ffffffff810369fc>] ? should_resched+0x5/0x23 [<ffffffff8105b8f9>] ? process_one_work+0x161/0x269 [<ffffffff8105aea5>] ? cwq_activate_delayed_work+0x3c/0x48 [<ffffffff8105c8c2>] ? worker_thread+0xc2/0x145 [<ffffffff8105c800>] ? manage_workers.isra.25+0x15b/0x15b [<ffffffff8105fa01>] ? kthread+0x76/0x7e [<ffffffff813581f4>] ? kernel_thread_helper+0x4/0x10 [<ffffffff8105f98b>] ? kthread_worker_fn+0x139/0x139 [<ffffffff813581f0>] ? gs_change+0x13/0x13 ---[ end trace d8c3cc0ad163ef79 ]--- ------------[ cut here ]------------ WARNING: at /usr/src/linux-3.2.52/debian/build/source_netctl/fs/proc/generic.c:849 remove_proc_entry+0x217/0x22e() Hardware name: remove_proc_entry: removing non-empty directory 'net/ip6t_hashlimit', leaking at least 'IN-REJECT' This is due to lack of removal net/ip6t_hashlimit/* entries in hashlimit_proc_net_exit(), since only IPv4 entries are deleted. Fix it by always removing the IPv4 and IPv6 entries and their parent directories in the netns destroy path. Signed-off-by: Sergey Popovich <popovich_sergei@mail.ru> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
| | | * netfilter: ipset: fix incorret comparison in hash_netnet4_data_equal()Dave Jones2013-11-251-1/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Both sides of the comparison are the same, looks like a cut-and-paste error. Spotted by Coverity. Signed-off-by: Dave Jones <davej@fedoraproject.org> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
| * | | Merge tag 'nfs-for-3.13-3' of git://git.linux-nfs.org/projects/trondmy/linux-nfsLinus Torvalds2013-12-051-1/+1
| |\ \ \ | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Pull NFS client bugfixes from Trond Myklebust: - Stable fix for a NFSv4.1 delegation and state recovery deadlock - Stable fix for a loop on irrecoverable errors when returning delegations - Fix a 3-way deadlock between layoutreturn, open, and state recovery - Update the MAINTAINERS file with contact information for Trond Myklebust - Close needs to handle NFS4ERR_ADMIN_REVOKED - Enabling v4.2 should not recompile nfsd and lockd - Fix a couple of compile warnings * tag 'nfs-for-3.13-3' of git://git.linux-nfs.org/projects/trondmy/linux-nfs: nfs: fix do_div() warning by instead using sector_div() MAINTAINERS: Update contact information for Trond Myklebust NFSv4.1: Prevent a 3-way deadlock between layoutreturn, open and state recovery SUNRPC: do not fail gss proc NULL calls with EACCES NFSv4: close needs to handle NFS4ERR_ADMIN_REVOKED NFSv4: Update list of irrecoverable errors on DELEGRETURN NFSv4 wait on recovery for async session errors NFS: Fix a warning in nfs_setsecurity NFS: Enabling v4.2 should not recompile nfsd and lockd
| | * | | SUNRPC: do not fail gss proc NULL calls with EACCESAndy Adamson2013-11-261-1/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Otherwise RPCSEC_GSS_DESTROY messages are not sent. Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
* | | | | net_sched: convert tcf_proto_ops to use struct list_headWANG Cong2013-12-181-10/+8
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | We don't need to maintain our own singly linked list code. Cc: Jamal Hadi Salim <jhs@mojatatu.com> Cc: David S. Miller <davem@davemloft.net> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com> Signed-off-by: David S. Miller <davem@davemloft.net>
* | | | | net_sched: convert tc_action_ops to use struct list_headWANG Cong2013-12-181-11/+9
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | We don't need to maintain our own singly linked list code. Cc: Jamal Hadi Salim <jhs@mojatatu.com> Cc: David S. Miller <davem@davemloft.net> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com> Signed-off-by: David S. Miller <davem@davemloft.net>
* | | | | net_sched: convert tcf_hashinfo to hlist and use spinlockWANG Cong2013-12-182-66/+48
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | So that we don't need to play with singly linked list, and since the code is not on hot path, we can use spinlock instead of rwlock. Cc: Jamal Hadi Salim <jhs@mojatatu.com> Cc: David S. Miller <davem@davemloft.net> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com> Signed-off-by: David S. Miller <davem@davemloft.net>
* | | | | net_sched: init struct tcf_hashinfo at register timeWANG Cong2013-12-1810-94/+80
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | It looks weird to store the lock out of the struct but still points to a static variable. Just move them into the struct. Cc: Jamal Hadi Salim <jhs@mojatatu.com> Cc: David S. Miller <davem@davemloft.net> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com> Signed-off-by: David S. Miller <davem@davemloft.net>
* | | | | net_sched: cls: refactor out struct tcf_ext_mapWANG Cong2013-12-1810-101/+59
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | These information can be saved in tcf_exts, and this will simplify the code. Cc: Jamal Hadi Salim <jhs@mojatatu.com> Cc: David S. Miller <davem@davemloft.net> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com> Signed-off-by: David S. Miller <davem@davemloft.net>
* | | | | net_sched: act: use standard struct list_headWANG Cong2013-12-1811-89/+83
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Currently actions are chained by a singly linked list, therefore it is a bit hard to add and remove a specific entry. Convert it to struct list_head so that in the latter patch we can remove an action without finding its head. Cc: Jamal Hadi Salim <jhs@mojatatu.com> Cc: David S. Miller <davem@davemloft.net> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com> Signed-off-by: David S. Miller <davem@davemloft.net>
* | | | | net_sched: remove get_stats from tc_action_opsWANG Cong2013-12-181-4/+0
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | It is not used. Cc: Jamal Hadi Salim <jhs@mojatatu.com> Cc: David S. Miller <davem@davemloft.net> Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com> Signed-off-by: David S. Miller <davem@davemloft.net>
* | | | | packet: deliver VLAN TPID to userspaceAtzm Watanabe2013-12-181-4/+10
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This enables userspace to get VLAN TPID as well as the VLAN TCI. Signed-off-by: Atzm Watanabe <atzm@stratosphere.co.jp> Acked-by: Daniel Borkmann <dborkman@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
* | | | | packet: fill the gap of TPACKET_ALIGNMENT with zerosAtzm Watanabe2013-12-181-1/+2
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | struct tpacket{2,3}_hdr is aligned to a multiple of TPACKET_ALIGNMENT. Explicitly defining and zeroing the gap of this makes additional changes easier. Signed-off-by: Atzm Watanabe <atzm@stratosphere.co.jp> Acked-by: Daniel Borkmann <dborkman@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
* | | | | packet: make aligned size of struct tpacket{2,3}_hdr clearAtzm Watanabe2013-12-181-0/+7
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | struct tpacket{2,3}_hdr is aligned to a multiple of TPACKET_ALIGNMENT. We may add members to them until current aligned size without forcing userspace to call getsockopt(..., PACKET_HDRLEN, ...). Signed-off-by: Atzm Watanabe <atzm@stratosphere.co.jp> Acked-by: Daniel Borkmann <dborkman@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
* | | | | net: Add utility function to copy skb hashTom Herbert2013-12-171-2/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Adds skb_copy_hash to copy rxhash and l4_rxhash from one skb to another. Signed-off-by: Tom Herbert <therbert@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
* | | | | net: Add utility functions to clear rxhashTom Herbert2013-12-173-9/+8
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | In several places 'skb->rxhash = 0' is being done to clear the rxhash value in an skb. This does not clear l4_rxhash which could still be set so that the rxhash wouldn't be recalculated on subsequent call to skb_get_rxhash. This patch adds an explict function to clear all the rxhash related information in the skb properly. skb_clear_hash_if_not_l4 clears the rxhash only if it is not marked as l4_rxhash. Fixed up places where 'skb->rxhash = 0' was being called. Signed-off-by: Tom Herbert <therbert@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
* | | | | net: Change skb_get_rxhash to skb_get_hashTom Herbert2013-12-176-10/+10
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Changing name of function as part of making the hash in skbuff to be generic property, not just for receive path. Signed-off-by: Tom Herbert <therbert@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
* | | | | net/hsr: using kfree_rcu() to simplify the codeWei Yongjun2013-12-171-9/+4
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The callback function of call_rcu() just calls a kfree(), so we can use kfree_rcu() instead of call_rcu() + callback function. Signed-off-by: Wei Yongjun <yongjun_wei@trendmicro.com.cn> Acked-by: Arvid Brodin <arvid.brodin@alten.se> Signed-off-by: David S. Miller <davem@davemloft.net>
* | | | | pkt_sched: fq: more robust memory allocationEric Dumazet2013-12-171-6/+28
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This patch brings NUMA support and automatic fallback to vmalloc() in case kmalloc() failed to allocate FQ hash table. NUMA support depends on XPS being setup for the device before qdisc allocation. After a XPS change, it might be worth creating qdisc hierarchy again. Signed-off-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
* | | | | tcp: refine TSO splitsEric Dumazet2013-12-171-39/+54
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | While investigating performance problems on small RPC workloads, I noticed linux TCP stack was always splitting the last TSO skb into two parts (skbs). One being a multiple of MSS, and a small one with the Push flag. This split is done even if TCP_NODELAY is set, or if no small packet is in flight. Example with request/response of 4K/4K IP A > B: . ack 68432 win 2783 <nop,nop,timestamp 6524593 6525001> IP A > B: . 65537:68433(2896) ack 69632 win 2783 <nop,nop,timestamp 6524593 6525001> IP A > B: P 68433:69633(1200) ack 69632 win 2783 <nop,nop,timestamp 6524593 6525001> IP B > A: . ack 68433 win 2768 <nop,nop,timestamp 6525001 6524593> IP B > A: . 69632:72528(2896) ack 69633 win 2768 <nop,nop,timestamp 6525001 6524593> IP B > A: P 72528:73728(1200) ack 69633 win 2768 <nop,nop,timestamp 6525001 6524593> IP A > B: . ack 72528 win 2783 <nop,nop,timestamp 6524593 6525001> IP A > B: . 69633:72529(2896) ack 73728 win 2783 <nop,nop,timestamp 6524593 6525001> IP A > B: P 72529:73729(1200) ack 73728 win 2783 <nop,nop,timestamp 6524593 6525001> We can avoid this split by including the Nagle tests at the right place. Note : If some NIC had trouble sending TSO packets with a partial last segment, we would have hit the problem in GRO/forwarding workload already. tcp_minshall_update() is moved to tcp_output.c and is updated as we might feed a TSO packet with a partial last segment. This patch tremendously improves performance, as the traffic now looks like : IP A > B: . ack 98304 win 2783 <nop,nop,timestamp 6834277 6834685> IP A > B: P 94209:98305(4096) ack 98304 win 2783 <nop,nop,timestamp 6834277 6834685> IP B > A: . ack 98305 win 2768 <nop,nop,timestamp 6834686 6834277> IP B > A: P 98304:102400(4096) ack 98305 win 2768 <nop,nop,timestamp 6834686 6834277> IP A > B: . ack 102400 win 2783 <nop,nop,timestamp 6834279 6834686> IP A > B: P 98305:102401(4096) ack 102400 win 2783 <nop,nop,timestamp 6834279 6834686> IP B > A: . ack 102401 win 2768 <nop,nop,timestamp 6834687 6834279> IP B > A: P 102400:106496(4096) ack 102401 win 2768 <nop,nop,timestamp 6834687 6834279> IP A > B: . ack 106496 win 2783 <nop,nop,timestamp 6834280 6834687> IP A > B: P 102401:106497(4096) ack 106496 win 2783 <nop,nop,timestamp 6834280 6834687> IP B > A: . ack 106497 win 2768 <nop,nop,timestamp 6834688 6834280> IP B > A: P 106496:110592(4096) ack 106497 win 2768 <nop,nop,timestamp 6834688 6834280> Before : lpq83:~# nstat >/dev/null;perf stat ./super_netperf 200 -t TCP_RR -H lpq84 -l 20 -- -r 4K,4K 280774 Performance counter stats for './super_netperf 200 -t TCP_RR -H lpq84 -l 20 -- -r 4K,4K': 205719.049006 task-clock # 9.278 CPUs utilized 8,449,968 context-switches # 0.041 M/sec 1,935,997 CPU-migrations # 0.009 M/sec 160,541 page-faults # 0.780 K/sec 548,478,722,290 cycles # 2.666 GHz [83.20%] 455,240,670,857 stalled-cycles-frontend # 83.00% frontend cycles idle [83.48%] 272,881,454,275 stalled-cycles-backend # 49.75% backend cycles idle [66.73%] 166,091,460,030 instructions # 0.30 insns per cycle # 2.74 stalled cycles per insn [83.39%] 29,150,229,399 branches # 141.699 M/sec [83.30%] 1,943,814,026 branch-misses # 6.67% of all branches [83.32%] 22.173517844 seconds time elapsed lpq83:~# nstat | egrep "IpOutRequests|IpExtOutOctets" IpOutRequests 16851063 0.0 IpExtOutOctets 23878580777 0.0 After patch : lpq83:~# nstat >/dev/null;perf stat ./super_netperf 200 -t TCP_RR -H lpq84 -l 20 -- -r 4K,4K 280877 Performance counter stats for './super_netperf 200 -t TCP_RR -H lpq84 -l 20 -- -r 4K,4K': 107496.071918 task-clock # 4.847 CPUs utilized 5,635,458 context-switches # 0.052 M/sec 1,374,707 CPU-migrations # 0.013 M/sec 160,920 page-faults # 0.001 M/sec 281,500,010,924 cycles # 2.619 GHz [83.28%] 228,865,069,307 stalled-cycles-frontend # 81.30% frontend cycles idle [83.38%] 142,462,742,658 stalled-cycles-backend # 50.61% backend cycles idle [66.81%] 95,227,712,566 instructions # 0.34 insns per cycle # 2.40 stalled cycles per insn [83.43%] 16,209,868,171 branches # 150.795 M/sec [83.20%] 874,252,952 branch-misses # 5.39% of all branches [83.37%] 22.175821286 seconds time elapsed lpq83:~# nstat | egrep "IpOutRequests|IpExtOutOctets" IpOutRequests 11239428 0.0 IpExtOutOctets 23595191035 0.0 Indeed, the occupancy of tx skbs (IpExtOutOctets/IpOutRequests) is higher : 2099 instead of 1417, thus helping GRO to be more efficient when using FQ packet scheduler. Many thanks to Neal for review and ideas. Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Yuchung Cheng <ycheng@google.com> Cc: Neal Cardwell <ncardwell@google.com> Cc: Nandita Dukkipati <nanditad@google.com> Cc: Van Jacobson <vanj@google.com> Acked-by: Neal Cardwell <ncardwell@google.com> Tested-by: Neal Cardwell <ncardwell@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
OpenPOWER on IntegriCloud