)]}' { "log": [ { "commit": "512615b6b843ff3ff5ad583f34c39b3f302f5f26", "tree": "7420705a314bc691bc478225148935dc67f71904", "parents": [ "d4cada4ae1c012815f95fa507eb86a0ae9d607d7" ], "author": { "name": "Eric Dumazet", "email": "eric.dumazet@gmail.com", "time": "Sun Nov 08 10:17:58 2009 +0000" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Sun Nov 08 20:53:06 2009 -0800" }, "message": "udp: secondary hash on (local port, local address)\n\nExtends udp_table to contain a secondary hash table.\n\nsocket anchor for this second hash is free, because UDP\ndoesnt use skc_bind_node : We define an union to hold\nboth skc_bind_node \u0026 a new hlist_nulls_node udp_portaddr_node\n\nudp_lib_get_port() inserts sockets into second hash chain\n(additional cost of one atomic op)\n\nudp_lib_unhash() deletes socket from second hash chain\n(additional cost of one atomic op)\n\nNote : No spinlock lockdep annotation is needed, because\nlock for the secondary hash chain is always get after\nlock for primary hash chain.\n\nSigned-off-by: Eric Dumazet \u003ceric.dumazet@gmail.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "d4cada4ae1c012815f95fa507eb86a0ae9d607d7", "tree": "23cdfb3763c9140ae095bf8095c3e6b16f7b48f3", "parents": [ "fdcc8aa953a1123a289791dd192090651036d593" ], "author": { "name": "Eric Dumazet", "email": "eric.dumazet@gmail.com", "time": "Sun Nov 08 10:17:30 2009 +0000" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Sun Nov 08 20:53:05 2009 -0800" }, "message": "udp: split sk_hash into two u16 hashes\n\nUnion sk_hash with two u16 hashes for udp (no extra memory taken)\n\nOne 16 bits hash on (local port) value (the previous udp \u0027hash\u0027)\n\nOne 16 bits hash on (local address, local port) values, initialized\nbut not yet used. This second hash is using jenkin hash for better\ndistribution.\n\nBecause the \u0027port\u0027 is xored later, a partial hash is performed\non local address + net_hash_mix(net)\n\nSigned-off-by: Eric Dumazet \u003ceric.dumazet@gmail.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "e022f0b4a03f4fff9323b509df023b8af635716e", "tree": "f287633fd72c4f70882b768836101acff7d1bc62", "parents": [ "748879776e3b738d53e64df6dbec7394b829462a" ], "author": { "name": "Krishna Kumar", "email": "krkumar2@in.ibm.com", "time": "Mon Oct 19 23:46:20 2009 +0000" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Tue Oct 20 18:55:45 2009 -0700" }, "message": "net: Introduce sk_tx_queue_mapping\n\nIntroduce sk_tx_queue_mapping; and functions that set, test and\nget this value. Reset sk_tx_queue_mapping to -1 whenever the dst\ncache is set/reset, and in socket alloc. Setting txq to -1 and\nusing valid txq\u003d\u003c0 to n-1\u003e allows the tx path to use the value\nof sk_tx_queue_mapping directly instead of subtracting 1 on every\ntx.\n\nSigned-off-by: Krishna Kumar \u003ckrkumar2@in.ibm.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "421355de876b9f3fcc7e4cb6026e416fb12a5068", "tree": "fb814456af86c75e1eb9e330994dda461fc5b4f7", "parents": [ "aace495933a981274b6491d71b915165a61defdc", "0fe7463a35aadfaf22d1ca58325ab3851b8d757c" ], "author": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Tue Oct 13 12:55:20 2009 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Tue Oct 13 12:55:20 2009 -0700" }, "message": "Merge branch \u0027master\u0027 of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6\n" }, { "commit": "3b885787ea4112eaa80945999ea0901bf742707f", "tree": "06fc15f8e8083d5652ccb4d06653d9812dce9c0b", "parents": [ "d5e63bded6e819ca77ee1a1d97c783a31f6caf30" ], "author": { "name": "Neil Horman", "email": "nhorman@tuxdriver.com", "time": "Mon Oct 12 13:26:31 2009 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Mon Oct 12 13:26:31 2009 -0700" }, "message": "net: Generalize socket rx gap / receive queue overflow cmsg\n\nCreate a new socket level option to report number of queue overflows\n\nRecently I augmented the AF_PACKET protocol to report the number of frames lost\non the socket receive queue between any two enqueued frames. This value was\nexported via a SOL_PACKET level cmsg. AFter I completed that work it was\nrequested that this feature be generalized so that any datagram oriented socket\ncould make use of this option. As such I\u0027ve created this patch, It creates a\nnew SOL_SOCKET level option called SO_RXQ_OVFL, which when enabled exports a\nSOL_SOCKET level cmsg that reports the nubmer of times the sk_receive_queue\noverflowed between any two given frames. It also augments the AF_PACKET\nprotocol to take advantage of this new feature (as it previously did not touch\nsk-\u003esk_drops, which this patch uses to record the overflow count). Tested\nsuccessfully by me.\n\nNotes:\n\n1) Unlike my previous patch, this patch simply records the sk_drops value, which\nis not a number of drops between packets, but rather a total number of drops.\nDeltas must be computed in user space.\n\n2) While this patch currently works with datagram oriented protocols, it will\nalso be accepted by non-datagram oriented protocols. I\u0027m not sure if thats\nagreeable to everyone, but my argument in favor of doing so is that, for those\nprotocols which aren\u0027t applicable to this option, sk_drops will always be zero,\nand reporting no drops on a receive queue that isn\u0027t used for those\nnon-participating protocols seems reasonable to me. This also saves us having\nto code in a per-protocol opt in mechanism.\n\n3) This applies cleanly to net-next assuming that commit\n977750076d98c7ff6cbda51858bb5a5894a9d9ab (my af packet cmsg patch) is reverted\n\nSigned-off-by: Neil Horman \u003cnhorman@tuxdriver.com\u003e\nSigned-off-by: Eric Dumazet \u003ceric.dumazet@gmail.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "5fdb9973c10c2d2e046da0976782ece25e78dc8b", "tree": "6b3b8309595e49173d935ed95e71fcc2788d4e24", "parents": [ "53f7e35f8b7fc2f5620a863ac613bcf3080cb6ba" ], "author": { "name": "Eric Dumazet", "email": "eric.dumazet@gmail.com", "time": "Thu Oct 08 22:50:25 2009 +0000" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Sun Oct 11 23:03:52 2009 -0700" }, "message": "net: Fix struct sock bitfield annotation\n\nSince commit a98b65a3 (net: annotate struct sock bitfield), we lost\n8 bytes in struct sock on 64bit arches because of\nkmemcheck_bitfield_end(flags) misplacement.\n\nFix this by putting together sk_shutdown, sk_no_check, sk_userlocks,\nsk_protocol and sk_type in the \u0027flags\u0027 32bits bitfield\n\nSigned-off-by: Eric Dumazet \u003ceric.dumazet@gmail.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "bcdce7195e0eab55b37dbd53be53057f38006380", "tree": "b4480a2e43561adf2060d1b8da818535cc91dc04", "parents": [ "bd32cafc4707ccc1e66fafdb47fac42217569070" ], "author": { "name": "Eric Dumazet", "email": "eric.dumazet@gmail.com", "time": "Tue Oct 06 17:28:29 2009 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Tue Oct 06 17:28:29 2009 -0700" }, "message": "net: speedup sk_wake_async()\n\nAn incoming datagram must bring into cpu cache *lot* of cache lines,\nin particular : (other parts omitted (hash chains, ip route cache...))\n\nOn 32bit arches :\n\noffsetof(struct sock, sk_rcvbuf) \u003d0x30 (read)\noffsetof(struct sock, sk_lock) \u003d0x34 (rw)\n\noffsetof(struct sock, sk_sleep) \u003d0x50 (read)\noffsetof(struct sock, sk_rmem_alloc) \u003d0x64 (rw)\noffsetof(struct sock, sk_receive_queue)\u003d0x74 (rw)\n\noffsetof(struct sock, sk_forward_alloc)\u003d0x98 (rw)\n\noffsetof(struct sock, sk_callback_lock)\u003d0xcc (rw)\noffsetof(struct sock, sk_drops) \u003d0xd8 (read if we add dropcount support, rw if frame dropped)\noffsetof(struct sock, sk_filter) \u003d0xf8 (read)\n\noffsetof(struct sock, sk_socket) \u003d0x138 (read)\n\noffsetof(struct sock, sk_data_ready) \u003d0x15c (read)\n\n\nWe can avoid sk-\u003esk_socket and socket-\u003efasync_list referencing on sockets\nwith no fasync() structures. (socket-\u003efasync_list ptr is probably already in cache\nbecause it shares a cache line with socket-\u003ewait, ie location pointed by sk-\u003esk_sleep)\n\nThis avoids one cache line load per incoming packet for common cases (no fasync())\n\nWe can leave (or even move in a future patch) sk-\u003esk_socket in a cold location\n\nSigned-off-by: Eric Dumazet \u003ceric.dumazet@gmail.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "b7058842c940ad2c08dd829b21e5c92ebe3b8758", "tree": "5fe78d599fc345ca0bcd4b083b79095a54b2921b", "parents": [ "eb1cf0f8f7a9e5a6d573d5bd72c015686a042db0" ], "author": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Wed Sep 30 16:12:20 2009 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Wed Sep 30 16:12:20 2009 -0700" }, "message": "net: Make setsockopt() optlen be unsigned.\n\nThis provides safety against negative optlen at the type\nlevel instead of depending upon (sometimes non-trivial)\nchecks against this sprinkled all over the the place, in\neach and every implementation.\n\nBased upon work done by Arjan van de Ven and feedback\nfrom Linus Torvalds.\n\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "4dc6dc7162c08b9965163c9ab3f9375d4adff2c7", "tree": "f4b9959e9ca5e5b8bdb6301663a15e0f67d4c68c", "parents": [ "303d67c288319768b19ed8dbed429fef7eb7c275" ], "author": { "name": "Eric Dumazet", "email": "eric.dumazet@gmail.com", "time": "Wed Jul 15 23:13:10 2009 +0000" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Thu Jul 16 18:05:26 2009 -0700" }, "message": "net: sock_copy() fixes\n\nCommit e912b1142be8f1e2c71c71001dc992c6e5eb2ec1\n(net: sk_prot_alloc() should not blindly overwrite memory)\ntook care of not zeroing whole new socket at allocation time.\n\nsock_copy() is another spot where we should be very careful.\nWe should not set refcnt to a non null value, until\nwe are sure other fields are correctly setup, or\na lockless reader could catch this socket by mistake,\nwhile not fully (re)initialized.\n\nThis patch puts sk_node \u0026 sk_refcnt to the very beginning\nof struct sock to ease sock_copy() \u0026 sk_prot_alloc() job.\n\nWe add appropriate smp_wmb() before sk_refcnt initializations\nto match our RCU requirements (changes to sock keys should\nbe committed to memory before sk_refcnt setting)\n\nSigned-off-by: Eric Dumazet \u003ceric.dumazet@gmail.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "ad46276952f1af34cd91d46d49ba13d347d56367", "tree": "55cf35156794ab34d8a607c25fd044c37231f9e4", "parents": [ "a57de0b4336e48db2811a2030bb68dba8dd09d88" ], "author": { "name": "Jiri Olsa", "email": "jolsa@redhat.com", "time": "Wed Jul 08 12:10:31 2009 +0000" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Thu Jul 09 17:06:58 2009 -0700" }, "message": "memory barrier: adding smp_mb__after_lock\n\nAdding smp_mb__after_lock define to be used as a smp_mb call after\na lock.\n\nMaking it nop for x86, since {read|write|spin}_lock() on x86 are\nfull memory barriers.\n\nSigned-off-by: Jiri Olsa \u003cjolsa@redhat.com\u003e\nSigned-off-by: Eric Dumazet \u003ceric.dumazet@gmail.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "a57de0b4336e48db2811a2030bb68dba8dd09d88", "tree": "a01c189d5fd55c69c9e2e842241e84b46728bc60", "parents": [ "1b614fb9a00e97b1eab54d4e442d405229c059dd" ], "author": { "name": "Jiri Olsa", "email": "jolsa@redhat.com", "time": "Wed Jul 08 12:09:13 2009 +0000" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Thu Jul 09 17:06:57 2009 -0700" }, "message": "net: adding memory barrier to the poll and receive callbacks\n\nAdding memory barrier after the poll_wait function, paired with\nreceive callbacks. Adding fuctions sock_poll_wait and sk_has_sleeper\nto wrap the memory barrier.\n\nWithout the memory barrier, following race can happen.\nThe race fires, when following code paths meet, and the tp-\u003ercv_nxt\nand __add_wait_queue updates stay in CPU caches.\n\nCPU1 CPU2\n\nsys_select receive packet\n ... ...\n __add_wait_queue update tp-\u003ercv_nxt\n ... ...\n tp-\u003ercv_nxt check sock_def_readable\n ... {\n schedule ...\n if (sk-\u003esk_sleep \u0026\u0026 waitqueue_active(sk-\u003esk_sleep))\n wake_up_interruptible(sk-\u003esk_sleep)\n ...\n }\n\nIf there was no cache the code would work ok, since the wait_queue and\nrcv_nxt are opposit to each other.\n\nMeaning that once tp-\u003ercv_nxt is updated by CPU2, the CPU1 either already\npassed the tp-\u003ercv_nxt check and sleeps, or will get the new value for\ntp-\u003ercv_nxt and will return with new data mask.\nIn both cases the process (CPU1) is being added to the wait queue, so the\nwaitqueue_active (CPU2) call cannot miss and will wake up CPU1.\n\nThe bad case is when the __add_wait_queue changes done by CPU1 stay in its\ncache, and so does the tp-\u003ercv_nxt update on CPU2 side. The CPU1 will then\nendup calling schedule and sleep forever if there are no more data on the\nsocket.\n\nCalls to poll_wait in following modules were ommited:\n\tnet/bluetooth/af_bluetooth.c\n\tnet/irda/af_irda.c\n\tnet/irda/irnet/irnet_ppp.c\n\tnet/mac80211/rc80211_pid_debugfs.c\n\tnet/phonet/socket.c\n\tnet/rds/af_rds.c\n\tnet/rfkill/core.c\n\tnet/sunrpc/cache.c\n\tnet/sunrpc/rpc_pipe.c\n\tnet/tipc/socket.c\n\nSigned-off-by: Jiri Olsa \u003cjolsa@redhat.com\u003e\nSigned-off-by: Eric Dumazet \u003ceric.dumazet@gmail.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "09ce42d3167e3f20b501fa780c2415332330fac5", "tree": "2f0a6e2e3ec51187f4df78e8c75aa31602a6edd4", "parents": [ "d7ed9c05ebf56c04811276207d7110706debe09f", "7959ea254ed18faee41160b1c50b3c9664735967" ], "author": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Wed Jun 24 10:01:12 2009 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Wed Jun 24 10:01:12 2009 -0700" }, "message": "Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next-2.6\n\n* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next-2.6:\n bnx2: Fix the behavior of ethtool when ONBOOT\u003dno\n qla3xxx: Don\u0027t sleep while holding lock.\n qla3xxx: Give the PHY time to come out of reset.\n ipv4 routing: Ensure that route cache entries are usable and reclaimable with caching is off\n net: Move rx skb_orphan call to where needed\n ipv6: Use correct data types for ICMPv6 type and code\n net: let KS8842 driver depend on HAS_IOMEM\n can: let SJA1000 driver depend on HAS_IOMEM\n netxen: fix firmware init handshake\n netxen: fix build with without CONFIG_PM\n netfilter: xt_rateest: fix comparison with self\n netfilter: xt_quota: fix incomplete initialization\n netfilter: nf_log: fix direct userspace memory access in proc handler\n netfilter: fix some sparse endianess warnings\n netfilter: nf_conntrack: fix conntrack lookup race\n netfilter: nf_conntrack: fix confirmation race condition\n netfilter: nf_conntrack: death_by_timeout() fix\n" }, { "commit": "d55d87fdff8252d0e2f7c28c2d443aee17e9d70f", "tree": "a4bfb0a47525672c690b63e85bf3f47f4ab014f7", "parents": [ "d5fdd6babcfc2b0e6a8da1acf492a69fb54b4c47" ], "author": { "name": "Herbert Xu", "email": "herbert@gondor.apana.org.au", "time": "Mon Jun 22 02:25:25 2009 +0000" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Tue Jun 23 16:36:25 2009 -0700" }, "message": "net: Move rx skb_orphan call to where needed\n\nIn order to get the tun driver to account packets, we need to be\nable to receive packets with destructors set. To be on the safe\nside, I added an skb_orphan call for all protocols by default since\nsome of them (IP in particular) cannot handle receiving packets\ndestructors properly.\n\nNow it seems that at least one protocol (CAN) expects to be able\nto pass skb-\u003esk through the rx path without getting clobbered.\n\nSo this patch attempts to fix this properly by moving the skb_orphan\ncall to where it\u0027s actually needed. In particular, I\u0027ve added it\nto skb_set_owner_[rw] which is what most users of skb-\u003edestructor\ncall.\n\nThis is actually an improvement for tun too since it means that\nwe only give back the amount charged to the socket when the skb\nis passed to another socket that will also be charged accordingly.\n\nSigned-off-by: Herbert Xu \u003cherbert@gondor.apana.org.au\u003e\nTested-by: Oliver Hartkopp \u003colver@hartkopp.net\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "d2aa4550379f92e929af7ed1dd4f55e6a1e331f8", "tree": "5ef0fc69a507f0d701fd157b6652427eabd5efdd", "parents": [ "9e3e4b1d2d13bead8d52703c82a02b55f108b491", "cb2107be43d2fc5eadec58b92b54bf32c00bfff3" ], "author": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Thu Jun 18 14:07:15 2009 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Thu Jun 18 14:07:15 2009 -0700" }, "message": "Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next-2.6\n\n* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next-2.6: (55 commits)\n netxen: fix tx ring accounting\n netxen: fix detection of cut-thru firmware mode\n forcedeth: fix dma api mismatches\n atm: sk_wmem_alloc initial value is one\n net: correct off-by-one write allocations reports\n via-velocity : fix no link detection on boot\n Net / e100: Fix suspend of devices that cannot be power managed\n TI DaVinci EMAC : Fix rmmod error\n net: group address list and its count\n ipv4: Fix fib_trie rebalancing, part 2\n pkt_sched: Update drops stats in act_police\n sky2: version 1.23\n sky2: add GRO support\n sky2: skb recycling\n sky2: reduce default transmit ring\n sky2: receive counter update\n sky2: fix shutdown synchronization\n sky2: PCI irq issues\n sky2: more receive shutdown\n sky2: turn off pause during shutdown\n ...\n\nManually fix trivial conflict in net/core/skbuff.c due to kmemcheck\n" }, { "commit": "c564039fd83ea16a86a96d52632794b24849e507", "tree": "42c9f525d08ea2e7d36c0231cc628587a175a39b", "parents": [ "1d4ac5d5ef9dd965ae211ebe8acbf83dc4d9571b" ], "author": { "name": "Eric Dumazet", "email": "eric.dumazet@gmail.com", "time": "Tue Jun 16 10:12:03 2009 +0000" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Wed Jun 17 04:31:25 2009 -0700" }, "message": "net: sk_wmem_alloc has initial value of one, not zero\n\ncommit 2b85a34e911bf483c27cfdd124aeb1605145dc80\n(net: No more expensive sock_hold()/sock_put() on each tx)\nchanged initial sk_wmem_alloc value.\n\nSome protocols check sk_wmem_alloc value to determine if a timer\nmust delay socket deallocation. We must take care of the sk_wmem_alloc\nvalue being one instead of zero when no write allocations are pending.\n\nReported by Ingo Molnar, and full diagnostic from David Miller.\n\nThis patch introduces three helpers to get read/write allocations\nand a followup patch will use these helpers to report correct\nwrite allocations to user.\n\nReported-by: Ingo Molnar \u003cmingo@elte.hu\u003e\nSigned-off-by: Eric Dumazet \u003ceric.dumazet@gmail.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "b3fec0fe35a4ff048484f1408385a27695d4273b", "tree": "088c23f098421ea681d9976a83aad73d15be1027", "parents": [ "e1f5b94fd0c93c3e27ede88b7ab652d086dc960f", "722f2a6c87f34ee0fd0130a8cf45f81e0705594a" ], "author": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Tue Jun 16 13:09:51 2009 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Tue Jun 16 13:09:51 2009 -0700" }, "message": "Merge branch \u0027for-linus2\u0027 of git://git.kernel.org/pub/scm/linux/kernel/git/vegard/kmemcheck\n\n* \u0027for-linus2\u0027 of git://git.kernel.org/pub/scm/linux/kernel/git/vegard/kmemcheck: (39 commits)\n signal: fix __send_signal() false positive kmemcheck warning\n fs: fix do_mount_root() false positive kmemcheck warning\n fs: introduce __getname_gfp()\n trace: annotate bitfields in struct ring_buffer_event\n net: annotate struct sock bitfield\n c2port: annotate bitfield for kmemcheck\n net: annotate inet_timewait_sock bitfields\n ieee1394/csr1212: fix false positive kmemcheck report\n ieee1394: annotate bitfield\n net: annotate bitfields in struct inet_sock\n net: use kmemcheck bitfields API for skbuff\n kmemcheck: introduce bitfield API\n kmemcheck: add opcode self-testing at boot\n x86: unify pte_hidden\n x86: make _PAGE_HIDDEN conditional\n kmemcheck: make kconfig accessible for other architectures\n kmemcheck: enable in the x86 Kconfig\n kmemcheck: add hooks for the page allocator\n kmemcheck: add hooks for page- and sg-dma-mappings\n kmemcheck: don\u0027t track page tables\n ...\n" }, { "commit": "a98b65a3ad71e702e760bc63f57684301628e837", "tree": "6ec3d174f315e9b77faf25c354f5eda99bd31301", "parents": [ "c53bd2e1949ddbe06fe2a6079c0658d58ce25edb" ], "author": { "name": "Vegard Nossum", "email": "vegard.nossum@gmail.com", "time": "Thu Feb 26 14:46:57 2009 +0100" }, "committer": { "name": "Vegard Nossum", "email": "vegard.nossum@gmail.com", "time": "Mon Jun 15 15:49:36 2009 +0200" }, "message": "net: annotate struct sock bitfield\n\n2009/2/24 Ingo Molnar \u003cmingo@elte.hu\u003e:\n\u003e ok, this is the last warning i have from today\u0027s overnight -tip\n\u003e testruns - a 32-bit system warning in sock_init_data():\n\u003e\n\u003e [ 2.610389] NET: Registered protocol family 16\n\u003e [ 2.616138] initcall netlink_proto_init+0x0/0x170 returned 0 after 7812 usecs\n\u003e [ 2.620010] WARNING: kmemcheck: Caught 32-bit read from uninitialized memory (f642c184)\n\u003e [ 2.624002] 010000000200000000000000604990c000000000000000000000000000000000\n\u003e [ 2.634076] i i i i i i u u i i i i i i i i i i i i i i i i i i i i i i i i\n\u003e [ 2.641038] ^\n\u003e [ 2.643376]\n\u003e [ 2.644004] Pid: 1, comm: swapper Not tainted (2.6.29-rc6-tip-01751-g4d1c22c-dirty #885)\n\u003e [ 2.648003] EIP: 0060:[\u003cc07141a1\u003e] EFLAGS: 00010282 CPU: 0\n\u003e [ 2.652008] EIP is at sock_init_data+0xa1/0x190\n\u003e [ 2.656003] EAX: 0001a800 EBX: f6836c00 ECX: 00463000 EDX: c0e46fe0\n\u003e [ 2.660003] ESI: f642c180 EDI: c0b83088 EBP: f6863ed8 ESP: c0c412ec\n\u003e [ 2.664003] DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068\n\u003e [ 2.668003] CR0: 8005003b CR2: f682c400 CR3: 00b91000 CR4: 000006f0\n\u003e [ 2.672003] DR0: 00000000 DR1: 00000000 DR2: 00000000 DR3: 00000000\n\u003e [ 2.676003] DR6: ffff4ff0 DR7: 00000400\n\u003e [ 2.680002] [\u003cc07423e5\u003e] __netlink_create+0x35/0xa0\n\u003e [ 2.684002] [\u003cc07443cc\u003e] netlink_kernel_create+0x4c/0x140\n\u003e [ 2.688002] [\u003cc072755e\u003e] rtnetlink_net_init+0x1e/0x40\n\u003e [ 2.696002] [\u003cc071b601\u003e] register_pernet_operations+0x11/0x30\n\u003e [ 2.700002] [\u003cc071b72c\u003e] register_pernet_subsys+0x1c/0x30\n\u003e [ 2.704002] [\u003cc0bf3c8c\u003e] rtnetlink_init+0x4c/0x100\n\u003e [ 2.708002] [\u003cc0bf4669\u003e] netlink_proto_init+0x159/0x170\n\u003e [ 2.712002] [\u003cc0101124\u003e] do_one_initcall+0x24/0x150\n\u003e [ 2.716002] [\u003cc0bbf3c7\u003e] do_initcalls+0x27/0x40\n\u003e [ 2.723201] [\u003cc0bbf3fc\u003e] do_basic_setup+0x1c/0x20\n\u003e [ 2.728002] [\u003cc0bbfb8a\u003e] kernel_init+0x5a/0xa0\n\u003e [ 2.732002] [\u003cc0103e47\u003e] kernel_thread_helper+0x7/0x10\n\u003e [ 2.736002] [\u003cffffffff\u003e] 0xffffffff\n\nWe fix this false positive by annotating the bitfield in struct\nsock.\n\nReported-by: Ingo Molnar \u003cmingo@elte.hu\u003e\nSigned-off-by: Vegard Nossum \u003cvegard.nossum@gmail.com\u003e\n" }, { "commit": "2b85a34e911bf483c27cfdd124aeb1605145dc80", "tree": "3cea3e8a27b62de2f92e759641c27200d8bde421", "parents": [ "f2333a014c1e13ac8e1b73a6fd77731c524eff78" ], "author": { "name": "Eric Dumazet", "email": "eric.dumazet@gmail.com", "time": "Thu Jun 11 02:55:43 2009 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Thu Jun 11 02:55:43 2009 -0700" }, "message": "net: No more expensive sock_hold()/sock_put() on each tx\n\nOne of the problem with sock memory accounting is it uses\na pair of sock_hold()/sock_put() for each transmitted packet.\n\nThis slows down bidirectional flows because the receive path\nalso needs to take a refcount on socket and might use a different\ncpu than transmit path or transmit completion path. So these\ntwo atomic operations also trigger cache line bounces.\n\nWe can see this in tx or tx/rx workloads (media gateways for example),\nwhere sock_wfree() can be in top five functions in profiles.\n\nWe use this sock_hold()/sock_put() so that sock freeing\nis delayed until all tx packets are completed.\n\nAs we also update sk_wmem_alloc, we could offset sk_wmem_alloc\nby one unit at init time, until sk_free() is called.\nOnce sk_free() is called, we atomic_dec_and_test(sk_wmem_alloc)\nto decrement initial offset and atomicaly check if any packets\nare in flight.\n\nskb_set_owner_w() doesnt call sock_hold() anymore\n\nsock_wfree() doesnt call sock_put() anymore, but check if sk_wmem_alloc\nreached 0 to perform the final freeing.\n\nDrawback is that a skb-\u003etruesize error could lead to unfreeable sockets, or\neven worse, prematurely calling __sk_free() on a live socket.\n\nNice speedups on SMP. tbench for example, going from 2691 MB/s to 2711 MB/s\non my 8 cpu dev machine, even if tbench was not really hitting sk_refcnt\ncontention point. 5 % speedup on a UDP transmit workload (depends\non number of flows), lowering TX completion cpu usage.\n\nSigned-off-by: Eric Dumazet \u003ceric.dumazet@gmail.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "e70049b9e74267dd47e1ffa62302073487afcb48", "tree": "2cd000c0751ef31c9044b020d63f278cdf4f332d", "parents": [ "d18921a0e319ab512f8186b1b1142c7b8634c779", "f7e603ad8f78cd3b59e33fa72707da0cbabdf699" ], "author": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Tue Feb 24 03:50:29 2009 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Tue Feb 24 03:50:29 2009 -0800" }, "message": "Merge branch \u0027master\u0027 of /home/davem/src/GIT/linux-2.6/\n" }, { "commit": "92a0acce186cde8ead56c6915d9479773673ea1a", "tree": "97b7663f77f8274cb52d429c8a7db97c70daf745", "parents": [ "34edaa88324004baf4884fb0388f86059d9c4878" ], "author": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Tue Feb 17 21:24:05 2009 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Tue Feb 17 21:24:05 2009 -0800" }, "message": "net: Kill skb_truesize_check(), it only catches false-positives.\n\nA long time ago we had bugs, primarily in TCP, where we would modify\nskb-\u003etruesize (for TSO queue collapsing) in ways which would corrupt\nthe socket memory accounting.\n\nskb_truesize_check() was added in order to try and catch this error\nmore systematically.\n\nHowever this debugging check has morphed into a Frankenstein of sorts\nand these days it does nothing other than catch false-positives.\n\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "20d4947353be60e909e6b1a79d241457edd6833f", "tree": "939ced518fc52e57df9ee9efb0cd14b6e26a3bc4", "parents": [ "ac45f602ee3d1b6f326f68bc0c2591ceebf05ba4" ], "author": { "name": "Patrick Ohly", "email": "patrick.ohly@intel.com", "time": "Thu Feb 12 05:03:38 2009 +0000" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Sun Feb 15 22:43:35 2009 -0800" }, "message": "net: socket infrastructure for SO_TIMESTAMPING\n\nThe overlap with the old SO_TIMESTAMP[NS] options is handled so\nthat time stamping in software (net_enable_timestamp()) is\nenabled when SO_TIMESTAMP[NS] and/or SO_TIMESTAMPING_RX_SOFTWARE\nis set. It\u0027s disabled if all of these are off.\n\nSigned-off-by: Patrick Ohly \u003cpatrick.ohly@intel.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "5e30589521518bff36fd2638b3c3d69679c50436", "tree": "6ac985658a06b0787e4354d0d16d380ea9b16a5a", "parents": [ "ac178ef0ae9eb44fd527d87aa9b6394e05f56e1f", "d2f8d7ee1a9b4650b4e43325b321801264f7c37a" ], "author": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Sat Feb 14 23:12:00 2009 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Sat Feb 14 23:12:00 2009 -0800" }, "message": "Merge branch \u0027master\u0027 of /home/davem/src/GIT/linux-2.6/\n\nConflicts:\n\tdrivers/net/wireless/iwlwifi/iwl-agn.c\n\tdrivers/net/wireless/iwlwifi/iwl3945-base.c\n" }, { "commit": "99709372736a216f99eb32b76fba835a2bfc93a8", "tree": "62a84b976965436b841c65b63e52240d0b06bc29", "parents": [ "fb0886745a75ce98bde3aac421adc69fe61a1905" ], "author": { "name": "Andrew Morton", "email": "akpm@linux-foundation.org", "time": "Thu Feb 12 16:43:17 2009 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Thu Feb 12 16:43:17 2009 -0800" }, "message": "net: don\u0027t use in_atomic() in gfp_any()\n\nThe problem is that in_atomic() will return false inside spinlocks if\nCONFIG_PREEMPT\u003dn. This will lead to deadlockable GFP_KERNEL allocations\nfrom spinlocked regions.\n\nSecondly, if CONFIG_PREEMPT\u003dy, this bug solves itself because networking\nwill instead use GFP_ATOMIC from this callsite. Hence we won\u0027t get the\nmight_sleep() debugging warnings which would have informed us of the buggy\ncallsites.\n\nSolve both these problems by switching to in_interrupt(). Now, if someone\nruns a gfp_any() allocation from inside spinlock we will get the warning\nif CONFIG_PREEMPT\u003dy.\n\nI reviewed all callsites and most of them were too complex for my little\nbrain and none of them documented their interface requirements. I have no\nidea what this patch will do.\n\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "4cc7f68d65558f683c702d4fe3a5aac4c5227b97", "tree": "88f2987e503790d2015a7f0a3b5748d05a0df140", "parents": [ "9a279bcbe347496799711155ed41a89bc40f79c5" ], "author": { "name": "Herbert Xu", "email": "herbert@gondor.apana.org.au", "time": "Wed Feb 04 16:55:54 2009 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Wed Feb 04 16:55:54 2009 -0800" }, "message": "net: Reexport sock_alloc_send_pskb\n\nThe function sock_alloc_send_pskb is completely useless if not\nexported since most of the code in it won\u0027t be used as is. In\nfact, this code has already been duplicated in the tun driver.\n\nNow that we need accounting in the tun driver, we can in fact\nuse this function as is. So this patch marks it for export again.\n\nSigned-off-by: Herbert Xu \u003cherbert@gondor.apana.org.au\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "dd24c00191d5e4a1ae896aafe33c6b8095ab4bd1", "tree": "e955c09e0b288e50c706b6ee409229d5a930c80c", "parents": [ "1748376b6626acf59c24e9592ac67b3fe2a0e026" ], "author": { "name": "Eric Dumazet", "email": "dada1@cosmosbay.com", "time": "Tue Nov 25 21:17:14 2008 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Tue Nov 25 21:17:14 2008 -0800" }, "message": "net: Use a percpu_counter for orphan_count\n\nInstead of using one atomic_t per protocol, use a percpu_counter\nfor \"orphan_count\", to reduce cache line contention on\nheavy duty network servers. \n\nSigned-off-by: Eric Dumazet \u003cdada1@cosmosbay.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "1748376b6626acf59c24e9592ac67b3fe2a0e026", "tree": "65a28205daf1ca92e31389440764fc407365014e", "parents": [ "c1b56878fb68e9c14070939ea4537ad4db79ffae" ], "author": { "name": "Eric Dumazet", "email": "dada1@cosmosbay.com", "time": "Tue Nov 25 21:16:35 2008 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Tue Nov 25 21:16:35 2008 -0800" }, "message": "net: Use a percpu_counter for sockets_allocated\n\nInstead of using one atomic_t per protocol, use a percpu_counter\nfor \"sockets_allocated\", to reduce cache line contention on\nheavy duty network servers. \n\nNote : We revert commit (248969ae31e1b3276fc4399d67ce29a5d81e6fd9\nnet: af_unix can make unix_nr_socks visbile in /proc),\nsince it is not anymore used after sock_prot_inuse_add() addition\n\nSigned-off-by: Eric Dumazet \u003cdada1@cosmosbay.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "198d6ba4d7f48c94f990f4604f0b3d73925e0ded", "tree": "56bbdf8ba2553c23c8099da9344a8f1d1aba97ab", "parents": [ "9a57f7fabd383920585ed8b74eacd117c6551f2d", "7f0f598a0069d1ab072375965a4b69137233169c" ], "author": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Tue Nov 18 23:38:23 2008 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Tue Nov 18 23:38:23 2008 -0800" }, "message": "Merge branch \u0027master\u0027 of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6\n\nConflicts:\n\n\tdrivers/isdn/i4l/isdn_net.c\n\tfs/cifs/connect.c\n" }, { "commit": "88ab1932eac721c6e7336708558fa5ed02c85c80", "tree": "c8788a1e3de08100bca341fa4180adfe5d02880f", "parents": [ "bbaffaca4810de1a25e32ecaf836eeaacc7a3d11" ], "author": { "name": "Eric Dumazet", "email": "dada1@cosmosbay.com", "time": "Sun Nov 16 19:39:21 2008 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Sun Nov 16 19:39:21 2008 -0800" }, "message": "udp: Use hlist_nulls in UDP RCU code\n\nThis is a straightforward patch, using hlist_nulls infrastructure.\n\nRCUification already done on UDP two weeks ago.\n\nUsing hlist_nulls permits us to avoid some memory barriers, both\nat lookup time and delete time.\n\nPatch is large because it adds new macros to include/net/sock.h.\nThese macros will be used by TCP \u0026 DCCP in next patch.\n\nSigned-off-by: Eric Dumazet \u003cdada1@cosmosbay.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "e8f6fbf62de37cbc2e179176ac7010d5f4396b67", "tree": "ac22d719c3c93d24b7eeda395b3271be60bfc661", "parents": [ "1fa989e80a9a104bf3b81842a5f4c1867d7aa9c4" ], "author": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Wed Nov 12 01:38:36 2008 +0000" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Thu Nov 13 23:19:10 2008 -0800" }, "message": "lockdep: include/linux/lockdep.h - fix warning in net/bluetooth/af_bluetooth.c\n\nfix this warning:\n\n net/bluetooth/af_bluetooth.c:60: warning: ‘bt_key_strings’ defined but not used\n net/bluetooth/af_bluetooth.c:71: warning: ‘bt_slock_key_strings’ defined but not used\n\nthis is a lockdep macro problem in the !LOCKDEP case.\n\nWe cannot convert it to an inline because the macro works on multiple types,\nbut we can mark the parameter used.\n\n[ also clean up a misaligned tab in sock_lock_init_class_and_name() ]\n\n[ also remove #ifdefs from around af_family_clock_key strings - which\n were certainly added to get rid of the ugly build warnings. ]\n\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "2378982487c492541d17adc0a870e7e83b07ba43", "tree": "0763aee8034a2db3c3949a6b9d7952263420335e", "parents": [ "e4412cb8a6b4190cf73ae44c859fd7e3727d8b13" ], "author": { "name": "Alexey Dobriyan", "email": "adobriyan@gmail.com", "time": "Wed Nov 12 23:25:32 2008 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Wed Nov 12 23:25:32 2008 -0800" }, "message": "net: ifdef struct sock::sk_async_wait_queue\n\nEvery user is under CONFIG_NET_DMA already, so ifdef field as well.\n\nSigned-off-by: Alexey Dobriyan \u003cadobriyan@gmail.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "d5f642384e9da75393160350f75bbb9a527f7c58", "tree": "098e581485de2878ce716f7a6b45af2e89505551", "parents": [ "d2ad3ca88da02baeccd5216780f1fe983c6953ba" ], "author": { "name": "Alexey Dobriyan", "email": "adobriyan@gmail.com", "time": "Tue Nov 04 14:45:58 2008 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Tue Nov 04 14:45:58 2008 -0800" }, "message": "net: #ifdef -\u003esk_security\n\nSigned-off-by: Alexey Dobriyan \u003cadobriyan@gmail.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "a1744d3bee19d3b9cbfb825ab316a101b9c9f109", "tree": "c0e2324c09beca0eb5782eb5abf241ea2b7a4a11", "parents": [ "275f165fa970174f8a98205529750e8abb6c0a33", "a432226614c5616e3cfd211e0acffa0acfb4770c" ], "author": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Fri Oct 31 00:17:34 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Fri Oct 31 00:17:34 2008 -0700" }, "message": "Merge branch \u0027master\u0027 of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6\n\nConflicts:\n\n\tdrivers/net/wireless/p54/p54common.c\n" }, { "commit": "ad1d967c88e349c7e822ad75dd3247a2a50d2ea3", "tree": "4564a715df604f9e8816f60f7bc6ab0d211c848c", "parents": [ "194dcdba5a11a0238aef7ed91f32df77cb31505b" ], "author": { "name": "Randy Dunlap", "email": "randy.dunlap@oracle.com", "time": "Thu Oct 30 23:54:35 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Thu Oct 30 23:54:35 2008 -0700" }, "message": "net: delete excess kernel-doc notation\n\nRemove excess kernel-doc function parameters from networking header\n\u0026 driver files:\n\nWarning(include/net/sock.h:946): Excess function parameter or struct member \u0027sk\u0027 description in \u0027sk_filter_release\u0027\nWarning(include/linux/netdevice.h:1545): Excess function parameter or struct member \u0027cpu\u0027 description in \u0027netif_tx_lock\u0027\nWarning(drivers/net/wan/z85230.c:712): Excess function parameter or struct member \u0027regs\u0027 description in \u0027z8530_interrupt\u0027\n\nSigned-off-by: Randy Dunlap \u003crandy.dunlap@oracle.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "96631ed16c514cf8b28fab991a076985ce378c26", "tree": "60023b60c6eaf2acdc8fcad258585e4425ebaf91", "parents": [ "f52b5054ec108aaa9e903850d6b62af8ae3fe6ae" ], "author": { "name": "Eric Dumazet", "email": "dada1@cosmosbay.com", "time": "Wed Oct 29 11:19:58 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Wed Oct 29 11:19:58 2008 -0700" }, "message": "udp: introduce sk_for_each_rcu_safenext()\n\nCorey Minyard found a race added in commit 271b72c7fa82c2c7a795bc16896149933110672d\n(udp: RCU handling for Unicast packets.)\n\n \"If the socket is moved from one list to another list in-between the\n time the hash is calculated and the next field is accessed, and the\n socket has moved to the end of the new list, the traversal will not\n complete properly on the list it should have, since the socket will\n be on the end of the new list and there\u0027s not a way to tell it\u0027s on a\n new list and restart the list traversal. I think that this can be\n solved by pre-fetching the \"next\" field (with proper barriers) before\n checking the hash.\"\n\nThis patch corrects this problem, introducing a new\nsk_for_each_rcu_safenext() macro.\n\nSigned-off-by: Eric Dumazet \u003cdada1@cosmosbay.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "271b72c7fa82c2c7a795bc16896149933110672d", "tree": "5634b95c04b4a7ac9babf2d8ac34cfb6c38a8f83", "parents": [ "645ca708f936b2fbeb79e52d7823e3eb2c0905f8" ], "author": { "name": "Eric Dumazet", "email": "dada1@cosmosbay.com", "time": "Wed Oct 29 02:11:14 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Wed Oct 29 02:11:14 2008 -0700" }, "message": "udp: RCU handling for Unicast packets.\n\nGoals are :\n\n1) Optimizing handling of incoming Unicast UDP frames, so that no memory\n writes should happen in the fast path.\n\n Note: Multicasts and broadcasts still will need to take a lock,\n because doing a full lockless lookup in this case is difficult.\n\n2) No expensive operations in the socket bind/unhash phases :\n - No expensive synchronize_rcu() calls.\n\n - No added rcu_head in socket structure, increasing memory needs,\n but more important, forcing us to use call_rcu() calls,\n that have the bad property of making sockets structure cold.\n (rcu grace period between socket freeing and its potential reuse\n make this socket being cold in CPU cache).\n David did a previous patch using call_rcu() and noticed a 20%\n impact on TCP connection rates.\n Quoting Cristopher Lameter :\n \"Right. That results in cacheline cooldown. You\u0027d want to recycle\n the object as they are cache hot on a per cpu basis. That is screwed\n up by the delayed regular rcu processing. We have seen multiple\n regressions due to cacheline cooldown.\n The only choice in cacheline hot sensitive areas is to deal with the\n complexity that comes with SLAB_DESTROY_BY_RCU or give up on RCU.\"\n\n - Because udp sockets are allocated from dedicated kmem_cache,\n use of SLAB_DESTROY_BY_RCU can help here.\n\nTheory of operation :\n---------------------\n\nAs the lookup is lockfree (using rcu_read_lock()/rcu_read_unlock()),\nspecial attention must be taken by readers and writers.\n\nUse of SLAB_DESTROY_BY_RCU is tricky too, because a socket can be freed,\nreused, inserted in a different chain or in worst case in the same chain\nwhile readers could do lookups in the same time.\n\nIn order to avoid loops, a reader must check each socket found in a chain\nreally belongs to the chain the reader was traversing. If it finds a\nmismatch, lookup must start again at the begining. This *restart* loop\nis the reason we had to use rdlock for the multicast case, because\nwe dont want to send same message several times to the same socket.\n\nWe use RCU only for fast path.\nThus, /proc/net/udp still takes spinlocks.\n\nSigned-off-by: Eric Dumazet \u003cdada1@cosmosbay.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "645ca708f936b2fbeb79e52d7823e3eb2c0905f8", "tree": "b384696994ee3cb04759a7bfffc29a48e4bf40f6", "parents": [ "b189db5d299c6824780af5590564ff608adb3dea" ], "author": { "name": "Eric Dumazet", "email": "dada1@cosmosbay.com", "time": "Wed Oct 29 01:41:45 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Wed Oct 29 01:41:45 2008 -0700" }, "message": "udp: introduce struct udp_table and multiple spinlocks\n\nUDP sockets are hashed in a 128 slots hash table.\n\nThis hash table is protected by *one* rwlock.\n\nThis rwlock is readlocked each time an incoming UDP message is handled.\n\nThis rwlock is writelocked each time a socket must be inserted in\nhash table (bind time), or deleted from this table (close time)\n\nThis is not scalable on SMP machines :\n\n1) Even in read mode, lock() and unlock() are atomic operations and\n must dirty a contended cache line, shared by all cpus.\n\n2) A writer might be starved if many readers are \u0027in flight\u0027. This can\n happen on a machine with some NIC receiving many UDP messages. User\n process can be delayed a long time at socket creation/dismantle time.\n\nThis patch prepares RCU migration, by introducing \u0027struct udp_table\nand struct udp_hslot\u0027, and using one spinlock per chain, to reduce\ncontention on central rwlock.\n\nIntroducing one spinlock per chain reduces latencies, for port\nrandomization on heavily loaded UDP servers. This also speedup\nbindings to specific ports.\n\nudp_lib_unhash() was uninlined, becoming to big.\n\nSome cleanups were done to ease review of following patch\n(RCUification of UDP Unicast lookups)\n\nSigned-off-by: Eric Dumazet \u003cdada1@cosmosbay.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "def8b4faff5ca349beafbbfeb2c51f3602a6ef3a", "tree": "a90fbb0b6ae2a49c507465801f31df77bc5ebf9d", "parents": [ "b057efd4d226fcc3a92b0dc6d8ea8e8185ecb260" ], "author": { "name": "Alexey Dobriyan", "email": "adobriyan@gmail.com", "time": "Tue Oct 28 13:24:06 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Tue Oct 28 13:24:06 2008 -0700" }, "message": "net: reduce structures when XFRM\u003dn\n\nifdef out\n* struct sk_buff::sp\t\t(pointer)\n* struct dst_entry::xfrm\t(pointer)\n* struct sock::sk_policy\t(2 pointers)\n\nSigned-off-by: Alexey Dobriyan \u003cadobriyan@gmail.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "c57943a1c96214ee68f3890bb6772841ffbfd606", "tree": "bfe79b29240d442c8ea104a89c2e827032b2824e", "parents": [ "b339a47c370ec669f789c5989f54eec1d78574bb" ], "author": { "name": "Peter Zijlstra", "email": "a.p.zijlstra@chello.nl", "time": "Tue Oct 07 14:18:42 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Tue Oct 07 14:18:42 2008 -0700" }, "message": "net: wrap sk-\u003esk_backlog_rcv()\n\nWrap calling sk-\u003esk_backlog_rcv() in a function. This will allow extending the\ngeneric sk_backlog_rcv behaviour.\n\nSigned-off-by: Peter Zijlstra \u003ca.p.zijlstra@chello.nl\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "23542618deb77cfed312842fe8c41ed19fb16470", "tree": "9658279a6e3593846a9127c7a8bffb7ce0077d62", "parents": [ "607c4aaf03041c8bd81555a0218050c0f895088e" ], "author": { "name": "KOVACS Krisztian", "email": "hidden@sch.bme.hu", "time": "Tue Oct 07 12:41:01 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Tue Oct 07 12:41:01 2008 -0700" }, "message": "inet: Don\u0027t lookup the socket if there\u0027s a socket attached to the skb\n\nUse the socket cached in the skb if it\u0027s present.\n\nSigned-off-by: KOVACS Krisztian \u003chidden@sch.bme.hu\u003e\nAcked-by: Arnaldo Carvalho de Melo \u003cacme@redhat.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "af01d537463714e36e2c96d2da35902b76cd6827", "tree": "6972a43882b0f5809a02d1f3866726cc78c02fe8", "parents": [ "a627266570605a98c5fda5b8234d9e92015e4d14" ], "author": { "name": "Alexey Dobriyan", "email": "adobriyan@gmail.com", "time": "Thu Aug 28 02:53:51 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Thu Aug 28 02:53:51 2008 -0700" }, "message": "net: more #ifdef CONFIG_COMPAT\n\nAll users of struct proto::compat_[gs]etsockopt and\nstruct inet_connection_sock_af_ops::compat_[gs]etsockopt are under\n#ifdef already, so use it in structure definition too.\n\nSigned-off-by: Alexey Dobriyan \u003cadobriyan@gmail.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "5c52ba170f8167511bdb65b981f4582100c40675", "tree": "1dab120003eb696dfb4ca18c40861274a317739d", "parents": [ "cf1100a7a4f2573f50f9a923b53373977328e3c8" ], "author": { "name": "Pavel Emelyanov", "email": "xemul@openvz.org", "time": "Wed Jul 16 20:28:10 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Wed Jul 16 20:28:10 2008 -0700" }, "message": "sock: add net to prot-\u003eenter_memory_pressure callback\n\nThe tcp_enter_memory_pressure calls NET_INC_STATS, but doesn\u0027t\nhave where to get the net from.\n\nI decided to add a sk argument, not the net itself, only to factor\nall the required sock_net(sk) calls inside the enter_memory_pressure \ncallback itself.\n\nSigned-off-by: Pavel Emelyanov \u003cxemul@openvz.org\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "972692e0db9b0a62329ca394062b58917ddbd03c", "tree": "6060599150b57de1f94ed85f8078da098fb8eb6a", "parents": [ "5bbc1722d52ad3df062d5742a7e958276e57ebd7" ], "author": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Tue Jun 17 22:41:38 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Tue Jun 17 22:41:38 2008 -0700" }, "message": "net: Add sk_set_socket() helper.\n\nIn order to more easily grep for all things that set\nsk-\u003esk_socket, add sk_set_socket() helper inline function.\n\nSuggested (although only half-seriously) by Evgeniy Polyakov.\n\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "cb61cb9b8b5ef6c2697d84e5015e314626eb2fba", "tree": "430dd8d5b56d0a9911750ffcbd97563ce68d1b27", "parents": [ "fe2c802ab62aa63d276deafa905875f3455f2621" ], "author": { "name": "Eric Dumazet", "email": "dada1@cosmosbay.com", "time": "Tue Jun 17 21:04:56 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Tue Jun 17 21:04:56 2008 -0700" }, "message": "udp: sk_drops handling\n\nIn commits 33c732c36169d7022ad7d6eb474b0c9be43a2dc1 ([IPV4]: Add raw\ndrops counter) and a92aa318b4b369091fd80433c80e62838db8bc1c ([IPV6]:\nAdd raw drops counter), Wang Chen added raw drops counter for\n/proc/net/raw \u0026 /proc/net/raw6\n\nThis patch adds this capability to UDP sockets too (/proc/net/udp \u0026\n/proc/net/udp6).\n\nThis means that \u0027RcvbufErrors\u0027 errors found in /proc/net/snmp can be also\nbe examined for each udp socket.\n\n# grep Udp: /proc/net/snmp\nUdp: InDatagrams NoPorts InErrors OutDatagrams RcvbufErrors SndbufErrors\nUdp: 23971006 75 899420 16390693 146348 0\n\n# cat /proc/net/udp\n sl local_address rem_address st tx_queue rx_queue tr tm-\u003ewhen retrnsmt ---\nuid timeout inode ref pointer drops\n 75: 00000000:02CB 00000000:0000 07 00000000:00000000 00:00000000 00000000 ---\n 0 0 2358 2 ffff81082a538c80 0\n111: 00000000:006F 00000000:0000 07 00000000:00000000 00:00000000 00000000 ---\n 0 0 2286 2 ffff81042dd35c80 146348\n\nIn this example, only port 111 (0x006F) was flooded by messages that\nuser program could not read fast enough. 146348 messages were lost.\n\nSigned-off-by: Eric Dumazet \u003cdada1@cosmosbay.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "338db085518a8436cdecd33f7b52a06ec16d9ec1", "tree": "d8f9c2ea1fa8fe9701e3c7f85452984d98819d1a", "parents": [ "22196d3648581b253f927186b30075fb005287b0" ], "author": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Tue Jun 17 01:09:00 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Tue Jun 17 01:09:00 2008 -0700" }, "message": "net: Kill SOCK_SLEEP_PRE and SOCK_SLEEP_POST, no users.\n\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "7d06b2e053d2d536348e3a0f6bb02982a41bea37", "tree": "ec6a5c3f448e84cd431a0397b9e2a87ca25aec17", "parents": [ "4ae127d1b6c71f9240dd4245f240e6dd8fc98014" ], "author": { "name": "Brian Haley", "email": "brian.haley@hp.com", "time": "Sat Jun 14 17:04:49 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Sat Jun 14 17:04:49 2008 -0700" }, "message": "net: change proto destroy method to return void\n\nChange struct proto destroy function pointer to return void. Noticed\nby Al Viro.\n\nSigned-off-by: Brian Haley \u003cbrian.haley@hp.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "65a18ec58e5e6186103f62f720acea94dfb26f4e", "tree": "aa0e58979c343d4519ba9dfa3a1e6a416e98e16b", "parents": [ "5d1e4468a7705db7c1415a65fd16f07113afc1b2" ], "author": { "name": "Denis V. Lunev", "email": "den@openvz.org", "time": "Wed Apr 16 01:59:46 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Wed Apr 16 01:59:46 2008 -0700" }, "message": "[NETNS]: Add netns refcnt debug for kernel sockets.\n\nProtocol control sockets and netlink kernel sockets should not prevent the\nnamespace stop request. They are initialized and disposed in a special way by\nsk_change_net/sk_release_kernel.\n\nSigned-off-by: Denis V. Lunev \u003cden@openvz.org\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "43db6d65e0ef943a361cb91f8baa49132009227b", "tree": "5770f1dac28a209bb6e0c04f5948c454bb6a8733", "parents": [ "b715631fad3ed320b85d386a84a6fb0b3f86b0b9" ], "author": { "name": "Stephen Hemminger", "email": "shemminger@vyatta.com", "time": "Thu Apr 10 01:43:09 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Thu Apr 10 01:43:09 2008 -0700" }, "message": "socket: sk_filter deinline\n\nThe sk_filter function is too big to be inlined. This saves 2296 bytes\nof text on allyesconfig.\n\nSigned-off-by: Stephen Hemminger \u003cshemminger@vyatta.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "c29a0bc4dfc4d833eb702b1929cec96a3eeb9f7a", "tree": "4917ad0bd8979dcae34f626032b28e62b9e358fd", "parents": [ "8efa6e93cb2666dceafc4844057fdcb9aa324fb7" ], "author": { "name": "Pavel Emelyanov", "email": "xemul@openvz.org", "time": "Mon Mar 31 19:41:46 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Mon Mar 31 19:41:46 2008 -0700" }, "message": "[SOCK][NETNS]: Add a struct net argument to sock_prot_inuse_add and _get.\n\nThis counter is about to become per-proto-and-per-net, so we\u0027ll need \ntwo arguments to determine which cell in this \"table\" to work with.\n\nAll the places, but proc already pass proper net to it - proc will be\ntuned a bit later.\n\nSome indentation with spaces in proc files is done to keep the file\ncoding style consistent.\n\nSigned-off-by: Pavel Emelyanov \u003cxemul@openvz.org\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "bdcde3d71a67e97f25e851f3ca97c9bb5ef03e7f", "tree": "eab4a66777582718ffe2eff9038324331106ec18", "parents": [ "60e7663d462af3994f292cb3691ea4f7371a9220" ], "author": { "name": "Pavel Emelyanov", "email": "xemul@openvz.org", "time": "Fri Mar 28 16:39:33 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Fri Mar 28 16:39:33 2008 -0700" }, "message": "[SOCK]: Drop inuse pcounter from struct proto (v2).\n\nAn uppercut - do not use the pcounter on struct proto.\n\nSigned-off-by: Pavel Emelyanov \u003cxemul@openvz.org\u003e\nAcked-by: Eric Dumazet \u003cdada1@cosmosbay.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "60e7663d462af3994f292cb3691ea4f7371a9220", "tree": "84eb3c792c8f3c46d77e0296855ea37ae573b24a", "parents": [ "1338d466d9c3f8a65cc6d83c629cd906f2a989f8" ], "author": { "name": "Pavel Emelyanov", "email": "xemul@openvz.org", "time": "Fri Mar 28 16:39:10 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Fri Mar 28 16:39:10 2008 -0700" }, "message": "[SOCK]: Drop per-proto inuse init and fre functions (v2).\n\nConstructive part of the set is finished here. We have to remove the\npcounter, so start with its init and free functions.\n\nSigned-off-by: Pavel Emelyanov \u003cxemul@openvz.org\u003e\nAcked-by: Eric Dumazet \u003cdada1@cosmosbay.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "1338d466d9c3f8a65cc6d83c629cd906f2a989f8", "tree": "2f30913c1c9c6d3c0616ac593e79b0431cc150d3", "parents": [ "13ff3d6fa4e6d8b6ee7c64245a0078e6a0e6f977" ], "author": { "name": "Pavel Emelyanov", "email": "xemul@openvz.org", "time": "Fri Mar 28 16:38:43 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Fri Mar 28 16:38:43 2008 -0700" }, "message": "[SOCK]: Introduce a percpu inuse counters array (v2).\n\nAnd redirect sock_prot_inuse_add and _get to use one.\n\nAs far as the dereferences are concerned. Before the patch we made\n1 dereference to proto-\u003einuse.add call, the call itself and then\ncalled the __get_cpu_var() on a static variable. After the patch we \nmake a direct call, then one dereference to proto-\u003einuse_idx and \nthen the same __get_cpu_var() on a still static variable. So this \npatch doesn\u0027t seem to produce performance penalty on SMP.\n\nThis is not per-net yet, but I will deliberately make NET_NS\u003dy case\nseparated from NET_NS\u003dn one, since it\u0027ll cost us one-or-two more \ndereferences to get the struct net and the inuse counter.\n\nSigned-off-by: Pavel Emelyanov \u003cxemul@openvz.org\u003e\nAcked-by: Eric Dumazet \u003cdada1@cosmosbay.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "13ff3d6fa4e6d8b6ee7c64245a0078e6a0e6f977", "tree": "c52580ac89aff1b4eabbd94e6cb73af15e91051f", "parents": [ "bc578a54f0fd489d0722303f9a52508495ccaf9a" ], "author": { "name": "Pavel Emelyanov", "email": "xemul@openvz.org", "time": "Fri Mar 28 16:38:17 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Fri Mar 28 16:38:17 2008 -0700" }, "message": "[SOCK]: Enumerate struct proto-s to facilitate percpu inuse accounting (v2).\n\nThe inuse counters are going to become a per-cpu array. Introduce an\nindex for this array on the struct proto.\n\nTo handle the case of proto register-unregister-register loop the\nbitmap is used. All its bits manipulations are protected with\nproto_list_lock and a sanity check for the bitmap being exhausted is\nalso added.\n\nSigned-off-by: Pavel Emelyanov \u003cxemul@openvz.org\u003e\nAcked-by: Eric Dumazet \u003cdada1@cosmosbay.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "f5aa23fd49063745f85644dd7a9330acd706add6", "tree": "51ff4c7006fe7f7551b29f24ec1830e19dc7a278", "parents": [ "9c2f5746b9cd536f0007709196d85a7e7d0070fa" ], "author": { "name": "Denis V. Lunev", "email": "den@openvz.org", "time": "Wed Mar 26 00:48:17 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Wed Mar 26 00:48:17 2008 -0700" }, "message": "[NETNS]: Compilation warnings under CONFIG_NET_NS.\n\nRecent commits from YOSHIFUJI Hideaki \u003cyoshfuji@linux-ipv6.org\u003e\nhave been introduced a several compilation warnings\n\u0027assignment discards qualifiers from pointer target type\u0027\ndue to extra const modifier in the inline call parameters of\n{dev|sock|twsk}_net_set.\n\nDrop it.\n\nSigned-off-by: Denis V. Lunev \u003cden@openvz.org\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "3b1e0a655f8eba44ab1ee2a1068d169ccfb853b9", "tree": "09edb35f32ebcfb1b4dad904425128a110ef16ee", "parents": [ "c346dca10840a874240c78efe3f39acf4312a1f2" ], "author": { "name": "YOSHIFUJI Hideaki", "email": "yoshfuji@linux-ipv6.org", "time": "Wed Mar 26 02:26:21 2008 +0900" }, "committer": { "name": "YOSHIFUJI Hideaki", "email": "yoshfuji@linux-ipv6.org", "time": "Wed Mar 26 04:39:55 2008 +0900" }, "message": "[NET] NETNS: Omit sock-\u003esk_net without CONFIG_NET_NS.\n\nIntroduce per-sock inlines: sock_net(), sock_net_set()\nand per-inet_timewait_sock inlines: twsk_net(), twsk_net_set().\nWithout CONFIG_NET_NS, no namespace other than \u0026init_net exists.\nLet\u0027s explicitly define them to help compiler optimizations.\n\nSigned-off-by: YOSHIFUJI Hideaki \u003cyoshfuji@linux-ipv6.org\u003e\n" }, { "commit": "fc8717baa8f52dd8d1b90df9008300ef3ec794ed", "tree": "149c5e1c8db5bcf2af47caeb732f900bc722875f", "parents": [ "6ba5a3c52da00015e739469e3b00cd6d0d4c5c67" ], "author": { "name": "Pavel Emelyanov", "email": "xemul@openvz.org", "time": "Sat Mar 22 16:56:51 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Sat Mar 22 16:56:51 2008 -0700" }, "message": "[RAW]: Add raw_hashinfo member on struct proto.\n\nSorry for the patch sequence confusion :| but I found that the similar\nthing can be done for raw sockets easily too late.\n\nExpand the proto.h union with the raw_hashinfo member and use it in\nraw_prot and rawv6_prot. This allows to drop the protocol specific\nversions of hash and unhash callbacks.\n\nSigned-off-by: Pavel Emelyanov \u003cxemul@openvz.org\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "39d8cda76cfb1178455f9d196b39e773878e6c05", "tree": "b58af892ccb3fd232078a87148c1dc9622be0c84", "parents": [ "22aba383ce52f8ca8740f9a74dc66b1b68138813" ], "author": { "name": "Pavel Emelyanov", "email": "xemul@openvz.org", "time": "Sat Mar 22 16:50:58 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Sat Mar 22 16:50:58 2008 -0700" }, "message": "[SOCK]: Add udp_hash member to struct proto.\n\nInspired by the commit ab1e0a13 ([SOCK] proto: Add hashinfo member to \nstruct proto) from Arnaldo, I made similar thing for UDP/-Lite IPv4 \nand -v6 protocols.\n\nThe result is not that exciting, but it removes some levels of\nindirection in udpxxx_get_port and saves some space in code and text.\n\nThe first step is to union existing hashinfo and new udp_hash on the\nstruct proto and give a name to this union, since future initialization \nof tcpxxx_prot, dccp_vx_protinfo and udpxxx_protinfo will cause gcc \nwarning about inability to initialize anonymous member this way.\n\nSigned-off-by: Pavel Emelyanov \u003cxemul@openvz.org\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "4cd9029d25f6612302f82634620f107c65f790b1", "tree": "f98b5afba51838a8fbf47b2eecf5adceef6eb2fc", "parents": [ "28518fc1701a757a3df8aa2d2ac2e5d1efd1c3e5" ], "author": { "name": "Stephen Hemminger", "email": "shemminger@vyatta.com", "time": "Fri Mar 21 15:54:53 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Fri Mar 21 15:54:53 2008 -0700" }, "message": "socket: SOCK_DEBUG type checking\n\nUse the inline trick (same as pr_debug) to get checking of debug\nstatements even if no code is generated.\n\nSigned-off-by: Stephen Hemminger \u003cshemminger@vyatta.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "82cc1a7a56872056af0ead6c7d695aa223f36695", "tree": "0891cf854a40cac90fc4fec3bd6b5ecb213d4452", "parents": [ "a25606c845856e5ca5ed54d23cab077e3a49bf10" ], "author": { "name": "Peter P Waskiewicz Jr", "email": "peter.p.waskiewicz.jr@intel.com", "time": "Fri Mar 21 03:43:19 2008 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Fri Mar 21 03:43:19 2008 -0700" }, "message": "[NET]: Add per-connection option to set max TSO frame size\n\nUpdate: My mailer ate one of Jarek\u0027s feedback mails... Fixed the\nparameter in netif_set_gso_max_size() to be u32, not u16. Fixed the\nwhitespace issue due to a patch import botch. Changed the types from\nu32 to unsigned int to be more consistent with other variables in the\narea. Also brought the patch up to the latest net-2.6.26 tree.\n\nUpdate: Made gso_max_size container 32 bits, not 16. Moved the\nlocation of gso_max_size within netdev to be less hotpath. Made more\nconsistent names between the sock and netdev layers, and added a\ndefine for the max GSO size.\n\nUpdate: Respun for net-2.6.26 tree.\n\nUpdate: changed max_gso_frame_size and sk_gso_max_size from signed to\nunsigned - thanks Stephen!\n\nThis patch adds the ability for device drivers to control the size of\nthe TSO frames being sent to them, per TCP connection. By setting the\nnetdevice\u0027s gso_max_size value, the socket layer will set the GSO\nframe size based on that value. This will propogate into the TCP\nlayer, and send TSO\u0027s of that size to the hardware.\n\nThis can be desirable to help tune the bursty nature of TSO on a\nper-adapter basis, where one may have 1 GbE and 10 GbE devices\ncoexisting in a system, one running multiqueue and the other not, etc.\n\nThis can also be desirable for devices that cannot support full 64 KB\nTSO\u0027s, but still want to benefit from some level of segmentation\noffloading.\n\nSigned-off-by: Peter P Waskiewicz Jr \u003cpeter.p.waskiewicz.jr@intel.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "edf0208702007ec1f6a36756fdd005f771a4cf17", "tree": "f323141ddde44db532d984ea7bf769a05220aa0c", "parents": [ "9dfbec1fb2bedff6b118504055cd9f0485edba45" ], "author": { "name": "Denis V. Lunev", "email": "den@openvz.org", "time": "Fri Feb 29 11:18:32 2008 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Fri Feb 29 11:18:32 2008 -0800" }, "message": "[NET]: Make netlink_kernel_release publically available as sk_release_kernel.\n\nThis staff will be needed for non-netlink kernel sockets, which should\nalso not pin a namespace like tcp_socket and icmp_socket.\n\nSigned-off-by: Denis V. Lunev \u003cden@openvz.org\u003e\nAcked-by: Daniel Lezcano \u003cdlezcano@fr.ibm.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "31729363418ea25b01aa9410838c38e36792e44c", "tree": "a6ddbab1cc5c6f5b732401b2c3521b41cfbec72a", "parents": [ "2df96af03d97e98cbea9f9cb71fc56ea5fc544b2" ], "author": { "name": "Randy Dunlap", "email": "randy.dunlap@oracle.com", "time": "Mon Feb 18 20:52:13 2008 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Mon Feb 18 20:52:13 2008 -0800" }, "message": "net: fix kernel-doc warnings in header files\n\nAdd missing structure kernel-doc descriptions to sock.h \u0026 skbuff.h\nto fix kernel-doc warnings.\n\n(I think that Stephen H. sent a similar patch, but I can\u0027t find it.\nI just want to kill the warnings, with either patch.)\n\nSigned-off-by: Randy Dunlap \u003crandy.dunlap@oracle.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "ab1e0a13d70299e792fd0527cefd070c1405fa5b", "tree": "d470e7b94b0e33ea59b12713366f1bee0b94f78c", "parents": [ "9dc7f30e3bac329998a2a9bb814bd0abc7cb58e2" ], "author": { "name": "Arnaldo Carvalho de Melo", "email": "acme@redhat.com", "time": "Sun Feb 03 04:06:04 2008 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Sun Feb 03 04:28:52 2008 -0800" }, "message": "[SOCK] proto: Add hashinfo member to struct proto\n\nThis way we can remove TCP and DCCP specific versions of\n\nsk-\u003esk_prot-\u003eget_port: both v4 and v6 use inet_csk_get_port\nsk-\u003esk_prot-\u003ehash: inet_hash is directly used, only v6 need\n a specific version to deal with mapped sockets\nsk-\u003esk_prot-\u003eunhash: both v4 and v6 use inet_hash directly\n\nstruct inet_connection_sock_af_ops also gets a new member, bind_conflict, so\nthat inet_csk_get_port can find the per family routine.\n\nNow only the lookup routines receive as a parameter a struct inet_hashtable.\n\nWith this we further reuse code, reducing the difference among INET transport\nprotocols.\n\nEventually work has to be done on UDP and SCTP to make them share this\ninfrastructure and get as a bonus inet_diag interfaces so that iproute can be\nused with these protocols.\n\nnet-2.6/net/ipv4/inet_hashtables.c:\n struct proto\t\t\t | +8\n struct inet_connection_sock_af_ops | +8\n 2 structs changed\n __inet_hash_nolisten | +18\n __inet_hash | -210\n inet_put_port | +8\n inet_bind_bucket_create | +1\n __inet_hash_connect | -8\n 5 functions changed, 27 bytes added, 218 bytes removed, diff: -191\n\nnet-2.6/net/core/sock.c:\n proto_seq_show | +3\n 1 function changed, 3 bytes added, diff: +3\n\nnet-2.6/net/ipv4/inet_connection_sock.c:\n inet_csk_get_port | +15\n 1 function changed, 15 bytes added, diff: +15\n\nnet-2.6/net/ipv4/tcp.c:\n tcp_set_state | -7\n 1 function changed, 7 bytes removed, diff: -7\n\nnet-2.6/net/ipv4/tcp_ipv4.c:\n tcp_v4_get_port | -31\n tcp_v4_hash | -48\n tcp_v4_destroy_sock | -7\n tcp_v4_syn_recv_sock | -2\n tcp_unhash | -179\n 5 functions changed, 267 bytes removed, diff: -267\n\nnet-2.6/net/ipv6/inet6_hashtables.c:\n __inet6_hash | +8\n 1 function changed, 8 bytes added, diff: +8\n\nnet-2.6/net/ipv4/inet_hashtables.c:\n inet_unhash | +190\n inet_hash | +242\n 2 functions changed, 432 bytes added, diff: +432\n\nvmlinux:\n 16 functions changed, 485 bytes added, 492 bytes removed, diff: -7\n\n/home/acme/git/net-2.6/net/ipv6/tcp_ipv6.c:\n tcp_v6_get_port | -31\n tcp_v6_hash | -7\n tcp_v6_syn_recv_sock | -9\n 3 functions changed, 47 bytes removed, diff: -47\n\n/home/acme/git/net-2.6/net/dccp/proto.c:\n dccp_destroy_sock | -7\n dccp_unhash | -179\n dccp_hash | -49\n dccp_set_state | -7\n dccp_done | +1\n 5 functions changed, 1 bytes added, 242 bytes removed, diff: -241\n\n/home/acme/git/net-2.6/net/dccp/ipv4.c:\n dccp_v4_get_port | -31\n dccp_v4_request_recv_sock | -2\n 2 functions changed, 33 bytes removed, diff: -33\n\n/home/acme/git/net-2.6/net/dccp/ipv6.c:\n dccp_v6_get_port | -31\n dccp_v6_hash | -7\n dccp_v6_request_recv_sock | +5\n 3 functions changed, 5 bytes added, 38 bytes removed, diff: -33\n\nSigned-off-by: Arnaldo Carvalho de Melo \u003cacme@redhat.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "4a19ec5800fc3bb64e2d87c4d9fdd9e636086fe0", "tree": "610bd4e7dbcbdae25ba3806f4256745e98617825", "parents": [ "036c2e27bc3a6498afb35de017d810194032d765" ], "author": { "name": "Laszlo Attila Toth", "email": "panther@balabit.hu", "time": "Wed Jan 30 19:08:16 2008 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Thu Jan 31 19:27:19 2008 -0800" }, "message": "[NET]: Introducing socket mark socket option.\n\nA userspace program may wish to set the mark for each packets its send\nwithout using the netfilter MARK target. Changing the mark can be used\nfor mark based routing without netfilter or for packet filtering.\n\nIt requires CAP_NET_ADMIN capability.\n\nSigned-off-by: Laszlo Attila Toth \u003cpanther@balabit.hu\u003e\nAcked-by: Patrick McHardy \u003ckaber@trash.net\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "9993e7d313e80bdc005d09c7def91903e0068f07", "tree": "ed7c65b35fd3392948ab2f6b225f69225fdcbba8", "parents": [ "e186932b3d26bd975022a1e254407e95dddceae7" ], "author": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Thu Jan 10 21:56:38 2008 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Mon Jan 28 15:01:42 2008 -0800" }, "message": "[TCP]: Do not purge sk_forward_alloc entirely in tcp_delack_timer().\n\nOtherwise we beat heavily on the global tcp_memory atomics\nwhen all of the sockets in the system are slowly sending\nperioding packet clumps.\n\nNoticed and suggested by Eric Dumazet.\n\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "65f7651788e18fadb2fbb7276af935d7871e1803", "tree": "dcad32d4344d6d11d80061773d9d1dbc9ae92223", "parents": [ "571e7682026fd0e25833d103a3eeb74be29bf199" ], "author": { "name": "Eric Dumazet", "email": "dada1@cosmosbay.com", "time": "Thu Jan 03 20:46:48 2008 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Mon Jan 28 15:00:36 2008 -0800" }, "message": "[NET]: prot_inuse cleanups and optimizations\n\n1) Cleanups (all functions are prefixed by sock_prot_inuse)\n\nsock_prot_inc_use(prot) -\u003e sock_prot_inuse_add(prot,-1)\nsock_prot_dec_use(prot) -\u003e sock_prot_inuse_add(prot,-1)\nsock_prot_inuse() -\u003e sock_prot_inuse_get()\n\nNew functions :\n\nsock_prot_inuse_init() and sock_prot_inuse_free() to abstract pcounter use.\n\n2) if CONFIG_PROC_FS\u003dn, we can zap \u0027inuse\u0027 member from \"struct proto\",\nsince nobody wants to read the inuse value.\n\nThis saves 1372 bytes on i386/SMP and some cpu cycles.\n\nSigned-off-by: Eric Dumazet \u003cdada1@cosmosbay.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "3ab224be6d69de912ee21302745ea45a99274dbc", "tree": "335dcef1cfacfefe3f36c21d5f144e011bc3bfba", "parents": [ "a06b494b61de44617dd58612164bdde56fca7bfb" ], "author": { "name": "Hideo Aoki", "email": "haoki@redhat.com", "time": "Mon Dec 31 00:11:19 2007 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Mon Jan 28 15:00:18 2008 -0800" }, "message": "[NET] CORE: Introducing new memory accounting interface.\n\nThis patch introduces new memory accounting functions for each network\nprotocol. Most of them are renamed from memory accounting functions\nfor stream protocols. At the same time, some stream memory accounting\nfunctions are removed since other functions do same thing.\n\nRenaming:\n\tsk_stream_free_skb()\t\t-\u003e\tsk_wmem_free_skb()\n\t__sk_stream_mem_reclaim()\t-\u003e\t__sk_mem_reclaim()\n\tsk_stream_mem_reclaim()\t\t-\u003e\tsk_mem_reclaim()\n\tsk_stream_mem_schedule \t\t-\u003e \t__sk_mem_schedule()\n\tsk_stream_pages() \t\t-\u003e\tsk_mem_pages()\n\tsk_stream_rmem_schedule()\t-\u003e\tsk_rmem_schedule()\n\tsk_stream_wmem_schedule()\t-\u003e\tsk_wmem_schedule()\n\tsk_charge_skb()\t\t\t-\u003e\tsk_mem_charge()\n\nRemoveing\n\tsk_stream_rfree():\tconsolidates into sock_rfree()\n\tsk_stream_set_owner_r(): consolidates into skb_set_owner_r()\n\tsk_stream_mem_schedule()\n\nThe following functions are added.\n \tsk_has_account(): check if the protocol supports accounting\n\tsk_mem_uncharge(): do the opposite of sk_mem_charge()\n\nIn addition, to achieve consolidation, updating sk_wmem_queued is\nremoved from sk_mem_charge().\n\nNext, to consolidate memory accounting functions, this patch adds\nmemory accounting calls to network core functions. Moreover, present\nmemory accounting call is renamed to new accounting call.\n\nFinally we replace present memory accounting calls with new interface\nin TCP and SCTP.\n\nSigned-off-by: Takahiro Yasui \u003ctyasui@redhat.com\u003e\nSigned-off-by: Hideo Aoki \u003chaoki@redhat.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "21371f768bf7127ee45bfaadd17899df6a439e8f", "tree": "91b5debf1d861a40c942a33f1286b70031a94305", "parents": [ "b15c4bcd15741b31019379298edfca28dc78029d" ], "author": { "name": "Eric Dumazet", "email": "dada1@cosmosbay.com", "time": "Mon Dec 24 20:57:56 2007 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Mon Jan 28 15:00:05 2008 -0800" }, "message": "[SOCK] Avoid divides in sk_stream_pages() and __sk_stream_mem_reclaim()\n\nsk_forward_alloc being signed, we should take care of divides by\nSK_STREAM_MEM_QUANTUM we do in sk_stream_pages() and\n__sk_stream_mem_reclaim()\n\nThis patchs introduces SK_STREAM_MEM_QUANTUM_SHIFT, defined\nas ilog2(SK_STREAM_MEM_QUANTUM), to be able to use right\nshifts instead of plain divides.\n\nThis should help compiler to choose right shifts instead of\nexpensive divides (as seen with CONFIG_CC_OPTIMIZE_FOR_SIZE\u003dy on x86)\n\nSigned-off-by: Eric Dumazet \u003cdada1@cosmosbay.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "8df09ea3b8ccfe0c94844102d33fa46f57c08d9e", "tree": "a7e91a1e2c9eea21b1e1a6a67bd855899627aafe", "parents": [ "ce55dd3610f7ac29bf8d159c2e2ace9aaf2c3038" ], "author": { "name": "Eric Dumazet", "email": "dada1@cosmosbay.com", "time": "Fri Dec 21 03:07:41 2007 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Mon Jan 28 14:59:59 2008 -0800" }, "message": "[SOCK] Avoid integer divides where not necessary in include/net/sock.h\n\nBecause sk_wmem_queued, sk_sndbuf are signed, a divide per two\nmay force compiler to use an integer divide.\n\nWe can instead use a right shift.\n\nSigned-off-by: Eric Dumazet \u003cdada1@cosmosbay.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "41380930d2cbdc0abf7513a675864258b7ac973d", "tree": "bded49b68abc1d1414acd7bb10dadc01b84fd232", "parents": [ "8b7817f3a959ed99d7443afc12f78a7e1fcc2063" ], "author": { "name": "Harvey Harrison", "email": "harvey.harrison@gmail.com", "time": "Wed Dec 12 10:46:51 2007 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Mon Jan 28 14:57:23 2008 -0800" }, "message": "[NET]: Remove FASTCALL macro\n\nX86_32 was the last user of the FASTCALL macro, now that it\nuses regparm(3) by default, this macro expands to nothing.\n\nSigned-off-by: Harvey Harrison \u003charvey.harrison@gmail.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "33eb9cfc700ae9ce621d47d6ca6d6b4ad7cd97f3", "tree": "d5e3c0e9644249fa327c613571613c43ec474c6b", "parents": [ "7e2e109cef0d59abcb9aca8b82993e304ed8970c" ], "author": { "name": "Pavel Emelyanov", "email": "xemul@openvz.org", "time": "Wed Dec 05 01:37:34 2007 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Mon Jan 28 14:56:26 2008 -0800" }, "message": "[NET]: Isolate the net/core/ sysctl table\n\nUsing ctl paths we can put all the stuff, related to net/core/\nsysctl table, into one file and remove all the references on it.\n\nAs a good side effect this hides the \"core_table\" name from\nthe global scope :)\n\nSigned-off-by: Pavel Emelyanov \u003cxemul@openvz.org\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "0eeb8ffcfeaa0d909ce39147f7b8fdd6cef1aacd", "tree": "8e252aebdb09e023fa7bde266c1081f08c17a7f9", "parents": [ "6ac552fdc6e96bf2f39c18d6e66b8c8080bbb06e" ], "author": { "name": "Denis V. Lunev", "email": "den@openvz.org", "time": "Tue Dec 04 01:15:45 2007 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Mon Jan 28 14:55:51 2008 -0800" }, "message": "[NET]: netns compilation speedup\n\nThis patch speedups compilation when net_namespace.h is changed.\n\nSigned-off-by: Denis V. Lunev \u003cden@openvz.org\u003e\nAcked-by: \"Eric W. Biederman\" \u003cebiederm@xmission.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "df97c708d5e6eebdd9ded1fa588eae09acf53793", "tree": "35bcca504e629f3b9635fd52870a8add0846f333", "parents": [ "f561d0f27d6283c49359bb96048f8ac3728c812c" ], "author": { "name": "Pavel Emelyanov", "email": "xemul@openvz.org", "time": "Thu Nov 29 21:22:33 2007 +1100" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Mon Jan 28 14:55:08 2008 -0800" }, "message": "[NET]: Eliminate unused argument from sk_stream_alloc_pskb\n\nThe 3rd argument is always zero (according to grep :) Eliminate\nit and merge the function with sk_stream_alloc_skb.\n\nThis saves 44 more bytes, and together with the previous patch\nwe have:\n\nadd/remove: 1/0 grow/shrink: 0/8 up/down: 183/-751 (-568)\nfunction old new delta\nsk_stream_alloc_skb - 183 +183\nip_rt_init 529 525 -4\narp_ignore 112 107 -5\n__inet_lookup_listener 284 274 -10\ntcp_sendmsg 2583 2481 -102\ntcp_sendpage 1449 1300 -149\ntso_fragment 417 258 -159\ntcp_fragment 1149 988 -161\n__tcp_push_pending_frames 1998 1837 -161\n\nSigned-off-by: Pavel Emelyanov \u003cxemul@openvz.org\u003e\nSigned-off-by: Herbert Xu \u003cherbert@gondor.apana.org.au\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "f561d0f27d6283c49359bb96048f8ac3728c812c", "tree": "95665f72ac84fc9c91f5c1ee568f169984631887", "parents": [ "3015a347dcd470fcc8becf1f84b0502391a3c0e0" ], "author": { "name": "Pavel Emelyanov", "email": "xemul@openvz.org", "time": "Thu Nov 29 20:28:50 2007 +1100" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Mon Jan 28 14:55:07 2008 -0800" }, "message": "[NET]: Uninline the sk_stream_alloc_pskb\n\nThis function seems too big for inlining. Indeed, it saves\nhalf-a-kilo when uninlined:\n\nadd/remove: 1/0 grow/shrink: 0/7 up/down: 195/-719 (-524)\nfunction old new delta\nsk_stream_alloc_pskb - 195 +195\nip_rt_init 529 525 -4\n__inet_lookup_listener 284 274 -10\ntcp_sendmsg 2583 2486 -97\ntcp_sendpage 1449 1305 -144\ntso_fragment 417 267 -150\ntcp_fragment 1149 992 -157\n__tcp_push_pending_frames 1998 1841 -157\n\nSigned-off-by: Pavel Emelyanov \u003cxemul@openvz.org\u003e\nSigned-off-by: Herbert Xu \u003cherbert@gondor.apana.org.au\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "ebb53d75657f86587ac8cf3e38ab0c860a8e3d4f", "tree": "2b1d63cb8ee54b19589d6d13a693147001e642ad", "parents": [ "de4d1db369785c29d68915edfee0cb70e8199f4c" ], "author": { "name": "Arnaldo Carvalho de Melo", "email": "acme@redhat.com", "time": "Wed Nov 21 22:08:50 2007 +0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Mon Jan 28 14:54:40 2008 -0800" }, "message": "[NET] proto: Use pcounters for the inuse field\n\nSigned-off-by: Arnaldo Carvalho de Melo \u003cacme@redhat.com\u003e\nSigned-off-by: Herbert Xu \u003cherbert@gondor.apana.org.au\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "c0ef877b2c9f543e9fb7953bfe1a0cd3a4eae362", "tree": "8e97df63b9e1b3cac8040461c263c0c938f8b994", "parents": [ "de0fa95c14bc4d4b545fae26439371ebfdcb8534" ], "author": { "name": "Pavel Emelyanov", "email": "xemul@openvz.org", "time": "Thu Nov 15 03:03:19 2007 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Mon Jan 28 14:54:00 2008 -0800" }, "message": "[NET]: Move sock_valbool_flag to socket.c\n\nThe sock_valbool_flag() helper is used in setsockopt to\nset or reset some flag on the sock. This helper is required\nin the net/socket.c only, so move it there.\n\nBesides, patch two places in sys_setsockopt() that repeat\nthis helper functionality manually.\n\nSince this is not a bugfix, but a trivial cleanup, I\nprepared this patch against net-2.6.25, but it also\napplies (with a single offset) to the latest net-2.6.\n\nSigned-off-by: Pavel Emelyanov \u003cxemul@openvz.org\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "33c732c36169d7022ad7d6eb474b0c9be43a2dc1", "tree": "a9ac2b8a488cc288bfd2f407d8ad9dd592b29055", "parents": [ "6ff7751d06f63131830102ffa0c9a38b313f100e" ], "author": { "name": "Wang Chen", "email": "wangchen@cn.fujitsu.com", "time": "Tue Nov 13 20:30:01 2007 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Mon Jan 28 14:53:33 2008 -0800" }, "message": "[IPV4]: Add raw drops counter.\n\nAdd raw drops counter for IPv4 in /proc/net/raw .\n\nSigned-off-by: Wang Chen \u003cwangchen@cn.fujitsu.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "9d3e44425e3498eb33f25d94392b4fd0d56a5176", "tree": "453e71e0dfae127df026c64e2b1a24aa4f629b3d", "parents": [ "0f99be0d115a5716292c58dfdb20d2eddd0f3387" ], "author": { "name": "Eric Dumazet", "email": "dada1@cosmosbay.com", "time": "Tue Jan 08 23:41:28 2008 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Tue Jan 08 23:41:28 2008 -0800" }, "message": "[SOCK]: Adds a rcu_dereference() in sk_filter\n\nIt seems commit fda9ef5d679b07c9d9097aaf6ef7f069d794a8f9 introduced a RCU \nprotection for sk_filter(), without a rcu_dereference()\n\nEither we need a rcu_dereference(), either a comment should explain why we \ndont need it. I vote for the former.\n\nSigned-off-by: Eric Dumazet \u003cdada1@cosmosbay.com\u003e\nAcked-by: Herbert Xu \u003cherbert@gondor.apana.org.au\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "21df56c6e2372e09c916111efb6c14c372a5ab2e", "tree": "c5b8e2c6f5a9951b267137229cb18e2b53878d01", "parents": [ "483b23ffa3a5f44767038b0a676d757e0668437e" ], "author": { "name": "Herbert Xu", "email": "herbert@gondor.apana.org.au", "time": "Sun Nov 18 18:48:08 2007 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Sun Nov 18 18:48:08 2007 -0800" }, "message": "[TCP]: Fix TCP header misalignment\n\nIndeed my previous change to alloc_pskb has made it possible\nfor the TCP header to be misaligned iff the MTU is not a multiple\nof 4 (and less than a page). So I suspect the optimised IPsec\nMTU calculation is giving you just such an MTU :)\n\nThis patch fixes it by changing alloc_pskb to make sure that\nthe size is at least 32-bit aligned. This does not cause the\nproblem fixed by the previous patch because max_header is always\n32-bit aligned which means that in the SG/NOTSO case this will\nbe a no-op.\n\nI thought about putting this in the callers but all the current\ncallers are from TCP. If and when we get a non-TCP caller we\ncan always create a TCP wrapper for this function and move the\nalignment over there.\n\nSigned-off-by: Herbert Xu \u003cherbert@gondor.apana.org.au\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "fb93134dfc2a6e6fbedc7c270a31da03fce88db9", "tree": "92c6d19049162f2f3eb9348a3159b5d0874a9daf", "parents": [ "18b2b7bd09811779309592a10080fe9ffb93144d" ], "author": { "name": "Herbert Xu", "email": "herbert@gondor.apana.org.au", "time": "Wed Nov 14 15:45:21 2007 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Wed Nov 14 15:45:21 2007 -0800" }, "message": "[TCP]: Fix size calculation in sk_stream_alloc_pskb\n\nWe round up the header size in sk_stream_alloc_pskb so that\nTSO packets get zero tail room. Unfortunately this rounding\nup is not coordinated with the select_size() function used by\nTCP to calculate the second parameter of sk_stream_alloc_pskb.\n\nAs a result, we may allocate more than a page of data in the\nnon-TSO case when exactly one page is desired.\n\nIn fact, rounding up the head room is detrimental in the non-TSO\ncase because it makes memory that would otherwise be available to\nthe payload head room. TSO doesn\u0027t need this either, all it wants\nis the guarantee that there is no tail room.\n\nSo this patch fixes this by adjusting the skb_reserve call so that\nexactly the requested amount (which all callers have calculated in\na precise way) is made available as tail room.\n\nSigned-off-by: Herbert Xu \u003cherbert@gondor.apana.org.au\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "286ab3d46058840d68e5d7d52e316c1f7e98c59f", "tree": "1d70e7895c49d2b148e026aa047efe186697fff9", "parents": [ "91781004b9c029ee55b7aa9ef950a373ba865dc6" ], "author": { "name": "Eric Dumazet", "email": "dada1@cosmosbay.com", "time": "Mon Nov 05 23:38:39 2007 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@sunset.davemloft.net", "time": "Wed Nov 07 04:08:57 2007 -0800" }, "message": "[NET]: Define infrastructure to keep \u0027inuse\u0027 changes in an efficent SMP/NUMA way.\n\n\"struct proto\" currently uses an array stats[NR_CPUS] to track change on\n\u0027inuse\u0027 sockets per protocol.\n\nIf NR_CPUS is big, this means we use a big memory area for this.\nMoreover, all this memory area is located on a single node on NUMA\nmachines, increasing memory pressure on the boot node.\n\nIn this patch, I tried to :\n\n- Keep a fast !CONFIG_SMP implementation\n- Keep a fast CONFIG_SMP implementation for often used protocols\n(tcp,udp,raw,...)\n- Introduce a NUMA efficient implementation\n\nSome helper macros are defined in include/net/sock.h\nThese macros take into account CONFIG_SMP\n\nIf a \"struct proto\" is declared without using DEFINE_PROTO_INUSE /\nREF_PROTO_INUSE\nmacros, it will automatically use a default implementation, using a\ndynamically allocated percpu zone.\nThis default implementation will be NUMA efficient, but might use 32/64\nbytes per possible cpu\nbecause of current alloc_percpu() implementation.\nHowever it still should be better than previous implementation based on\nstats[NR_CPUS] field.\n\nWhen a \"struct proto\" is changed to use the new macros, we use a single\nstatic \"int\" percpu variable,\nlowering the memory and cpu costs, still preserving NUMA efficiency.\n\nSigned-off-by: Eric Dumazet \u003cdada1@cosmosbay.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "6257ff2177ff02d7f260a7a501876aa41cb9a9f6", "tree": "9d9f80ccf16f3d4ef062e896f62974c5496193ad", "parents": [ "154adbc8469ff21fbf5c958446ee92dbaab01be1" ], "author": { "name": "Pavel Emelyanov", "email": "xemul@openvz.org", "time": "Thu Nov 01 00:39:31 2007 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Thu Nov 01 00:39:31 2007 -0700" }, "message": "[NET]: Forget the zero_it argument of sk_alloc()\n\nFinally, the zero_it argument can be completely removed from\nthe callers and from the function prototype.\n\nBesides, fix the checkpatch.pl warnings about using the\nassignments inside if-s.\n\nThis patch is rather big, and it is a part of the previous one.\nI splitted it wishing to make the patches more readable. Hope \nthis particular split helped.\n\nSigned-off-by: Pavel Emelyanov \u003cxemul@openvz.org\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "f1a6c4da14c365d3ee0b5de43a93f7470982637c", "tree": "929b046ef28e1600cf08667459bd8c9a86069dc9", "parents": [ "261ab365fadd53ddc1b292b1663800e11fbf3e71" ], "author": { "name": "Pavel Emelyanov", "email": "xemul@openvz.org", "time": "Thu Nov 01 00:29:45 2007 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Thu Nov 01 00:29:45 2007 -0700" }, "message": "[NET]: Move the sock_copy() from the header\n\nThe sock_copy() call is not used outside the sock.c file,\nso just move it into a sock.c\n\nSigned-off-by: Pavel Emelyanov \u003cxemul@openvz.org\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "47e958eac280c263397582d5581e868c3227a1bd", "tree": "67fc50a26b7861b3330eae7f1874df851c0ef740", "parents": [ "d3904b739928edd83d117f1eb5bfa69f18d6f046" ], "author": { "name": "Pavel Emelyanov", "email": "xemul@openvz.org", "time": "Wed Oct 17 21:22:42 2007 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Wed Oct 17 21:22:42 2007 -0700" }, "message": "[NET]: Fix the race between sk_filter_(de|at)tach and sk_clone()\n\nThe proposed fix is to delay the reference counter decrement\nuntil the quiescent state pass. This will give sk_clone() a\nchance to get the reference on the cloned filter.\n\nRegular sk_filter_uncharge can happen from the sk_free() only\nand there\u0027s no need in delaying the put - the socket is dead\nanyway and is to be release itself.\n\nSigned-off-by: Pavel Emelyanov \u003cxemul@openvz.org\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "309dd5fc872448e35634d510049642312ebc170d", "tree": "5284d7b48980482ecba06b760ac16d9f94e66ea9", "parents": [ "55b333253d5bcafbe187b50474e40789301c53c6" ], "author": { "name": "Pavel Emelyanov", "email": "xemul@openvz.org", "time": "Wed Oct 17 21:21:51 2007 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@davemloft.net", "time": "Wed Oct 17 21:21:51 2007 -0700" }, "message": "[NET]: Move the filter releasing into a separate call\n\nThis is done merely as a preparation for the fix.\n\nThe sk_filter_uncharge() unaccounts the filter memory and calls\nthe sk_filter_release(), which in turn decrements the refcount\nanf frees the filter.\n\nThe latter function will be required separately.\n\nSigned-off-by: Pavel Emelyanov \u003cxemul@openvz.org\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "cfcabdcc2d5a810208e5bb3974121b7ed60119aa", "tree": "1aed711eeecc5a303b57f1fc47e1b5746e8a72c2", "parents": [ "de83c058af25aa97ed4864abab11e90e8dead6e2" ], "author": { "name": "Stephen Hemminger", "email": "shemminger@linux-foundation.org", "time": "Tue Oct 09 01:59:42 2007 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@sunset.davemloft.net", "time": "Wed Oct 10 16:54:48 2007 -0700" }, "message": "[NET]: sparse warning fixes\n\nFix a bunch of sparse warnings. Mostly about 0 used as\nNULL pointer, and shadowed variable declarations.\nOne notable case was that hash size should have been unsigned.\n\nSigned-off-by: Stephen Hemminger \u003cshemminger@linux-foundation.org\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "1b8d7ae42d02e483ad94035cca851e4f7fbecb40", "tree": "81f8cc0ee49ef99cc67dfed3dc7b7ecb510abf8b", "parents": [ "457c4cbc5a3dde259d2a1f15d5f9785290397267" ], "author": { "name": "Eric W. Biederman", "email": "ebiederm@xmission.com", "time": "Mon Oct 08 23:24:22 2007 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@sunset.davemloft.net", "time": "Wed Oct 10 16:49:07 2007 -0700" }, "message": "[NET]: Make socket creation namespace safe.\n\nThis patch passes in the namespace a new socket should be created in\nand has the socket code do the appropriate reference counting. By\nvirtue of this all socket create methods are touched. In addition\nthe socket create methods are modified so that they will fail if\nyou attempt to create a socket in a non-default network namespace.\n\nFailing if we attempt to create a socket outside of the default\nnetwork namespace ensures that as we incrementally make the network stack\nnetwork namespace aware we will not export functionality that someone\nhas not audited and made certain is network namespace safe.\nAllowing us to partially enable network namespaces before all of the\nexotic protocols are supported.\n\nAny protocol layers I have missed will fail to compile because I now\npass an extra parameter into the socket creation code.\n\n[ Integrated AF_IUCV build fixes from Andrew Morton... -DaveM ]\n\nSigned-off-by: Eric W. Biederman \u003cebiederm@xmission.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "07feaebfcc10cd35e745c7073667935246494bee", "tree": "297f212ebca1c6916a8b82ba176ee4ee508ccb08", "parents": [ "4a1c537113cdc688aabc3fb9bb6ed18ec821c779" ], "author": { "name": "Eric W. Biederman", "email": "ebiederm@xmission.com", "time": "Wed Sep 12 11:58:02 2007 +0200" }, "committer": { "name": "David S. Miller", "email": "davem@sunset.davemloft.net", "time": "Wed Oct 10 16:49:05 2007 -0700" }, "message": "[NET]: Add a network namespace parameter to struct sock\n\nSockets need to get a reference to their network namespace,\nor possibly a simple hold if someone registers on the network\nnamespace notifier and will free the sockets when the namespace\nis going to be destroyed.\n\nSigned-off-by: Eric W. Biederman \u003cebiederm@xmission.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "d2e9117c7aa9544d910634e17e3519fd67155229", "tree": "531507ed5d2031fa80775036fec1b57584a236cb", "parents": [ "02b3d34631831a19ee691516e233756b270eac6d" ], "author": { "name": "John Heffner", "email": "jheffner@psc.edu", "time": "Wed Sep 12 10:44:19 2007 +0200" }, "committer": { "name": "David S. Miller", "email": "davem@sunset.davemloft.net", "time": "Wed Oct 10 16:49:01 2007 -0700" }, "message": "[NET]: Change type of owner in sock_lock_t to int, rename\n\nThe type of owner in sock_lock_t is currently (struct sock_iocb *),\npresumably for historical reasons. It is never used as this type, only\ntested as NULL or set to (void *)1. For clarity, this changes it to type\nint, and renames to owned, to avoid any possible type casting errors.\n\nSigned-off-by: John Heffner \u003cjheffner@psc.edu\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "172589ccdde41b59861c92c4a971b95514ef24e3", "tree": "ae775e6db4cb85aad1c74b6d93ba359f9dfe88ea", "parents": [ "c45248c70125cc374fdf264659643276c72801bf" ], "author": { "name": "Ilpo Järvinen", "email": "ilpo.jarvinen@helsinki.fi", "time": "Tue Aug 28 15:50:33 2007 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@sunset.davemloft.net", "time": "Wed Oct 10 16:48:37 2007 -0700" }, "message": "[NET]: DIV_ROUND_UP cleanup (part two)\n\nHopefully captured all single statement cases under net/. I\u0027m\nnot too sure if there is some policy about #includes that are\n\"guaranteed\" (ie., in the current tree) to be available through\nsome other #included header, so I just added linux/kernel.h to\neach changed file that didn\u0027t #include it previously.\n\nSigned-off-by: Ilpo Järvinen \u003cilpo.jarvinen@helsinki.fi\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "4e07a91c37c69ec1647c218214591ee4fe3408fe", "tree": "abb4e5900b114168d57218ebb74cf0c624a38375", "parents": [ "be02097cf6342eb0426833f54c95e0fb4c9bca45" ], "author": { "name": "Arnaldo Carvalho de Melo", "email": "acme@redhat.com", "time": "Tue May 29 13:17:47 2007 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@sunset.davemloft.net", "time": "Thu May 31 01:23:32 2007 -0700" }, "message": "[SOCK]: Shrink struct sock by 8 bytes on 64-bit.\n\nSigned-off-by: Arnaldo Carvalho de Melo \u003cacme@redhat.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "6272e2667965dfb5b59199f462cd0f001fb304a6", "tree": "a14a4537dcd7af09863cc3a1c19a3efe386d67ab", "parents": [ "039b6b3ed84e45a6f8316358dd2bfdc83d59fc45" ], "author": { "name": "Christoph Hellwig", "email": "hch@lst.de", "time": "Tue May 08 00:29:21 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue May 08 11:15:09 2007 -0700" }, "message": "cleanup compat ioctl handling\n\nMerge all compat ioctl handling into compat_ioctl.c instead of splitting it\nover compat.c and compat_ioctl.c. This also allows to get rid of ioctl32.h\n\nSigned-off-by: Christoph Hellwig \u003chch@lst.de\u003e\nLooks-good-to: Andi Kleen \u003cak@suse.de\u003e\nAcked-by: Arnd Bergmann \u003carnd@arndb.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "9958089a43ae8a9af07402461c0b2b7548c7341e", "tree": "6d07cf01f9f6cc647dd6b01d0a16757f025845f8", "parents": [ "4ac02bab77438b484a5cf855a002fb6a1d592894" ], "author": { "name": "Andi Kleen", "email": "ak@suse.de", "time": "Fri Apr 20 17:12:43 2007 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@sunset.davemloft.net", "time": "Wed Apr 25 22:29:26 2007 -0700" }, "message": "[NET]: Move sk_setup_caps() out of line.\n\nIt is far too large to be an inline and not in any hot paths.\n\nSigned-off-by: Andi Kleen \u003cak@suse.de\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "92f37fd2ee805aa77925c1e64fd56088b46094fc", "tree": "8251c38b83ab362116dac89d94412ce229b42831", "parents": [ "c7a3c5da35055e2fa97ed4f0da3eec4bd0ef4c38" ], "author": { "name": "Eric Dumazet", "email": "dada1@cosmosbay.com", "time": "Sun Mar 25 22:14:49 2007 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@sunset.davemloft.net", "time": "Wed Apr 25 22:24:21 2007 -0700" }, "message": "[NET]: Adding SO_TIMESTAMPNS / SCM_TIMESTAMPNS support\n\nNow that network timestamps use ktime_t infrastructure, we can add a new\nSOL_SOCKET sockopt SO_TIMESTAMPNS.\n\nThis command is similar to SO_TIMESTAMP, but permits transmission of\na \u0027timespec struct\u0027 instead of a \u0027timeval struct\u0027 control message.\n(nanosecond resolution instead of microsecond)\n\nControl message is labelled SCM_TIMESTAMPNS instead of SCM_TIMESTAMP\n\nA socket cannot mix SO_TIMESTAMP and SO_TIMESTAMPNS : the two modes are\nmutually exclusive.\n\nsock_recv_timestamp() became too big to be fully inlined so I added a\n__sock_recv_timestamp() helper function.\n\nSigned-off-by: Eric Dumazet \u003cdada1@cosmosbay.com\u003e\nCC: linux-arch@vger.kernel.org\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "a2a316fd068c455c609ecc155dcfaa7e208d29fe", "tree": "9839e2c7e05e9d25241fa7c7b27fd06214120292", "parents": [ "ae40eb1ef30ab4120bd3c8b7e3da99ee53d27a23" ], "author": { "name": "Stephen Hemminger", "email": "shemminger@linux-foundation.org", "time": "Thu Mar 08 20:41:08 2007 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@sunset.davemloft.net", "time": "Wed Apr 25 22:24:05 2007 -0700" }, "message": "[NET]: Replace CONFIG_NET_DEBUG with sysctl.\n\nCovert network warning messages from a compile time to runtime choice.\nRemoves kernel config option and replaces it with new /proc/sys/net/core/warnings.\n\nSigned-off-by: Stephen Hemminger \u003cshemminger@linux-foundation.org\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "ae40eb1ef30ab4120bd3c8b7e3da99ee53d27a23", "tree": "b5ae288b3c27d13bde9648c41d7db3cfe1884bc2", "parents": [ "cb69cc52364690d7789940c480b3a9490784b680" ], "author": { "name": "Eric Dumazet", "email": "dada1@cosmosbay.com", "time": "Sun Mar 18 17:33:16 2007 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@sunset.davemloft.net", "time": "Wed Apr 25 22:24:04 2007 -0700" }, "message": "[NET]: Introduce SIOCGSTAMPNS ioctl to get timestamps with nanosec resolution\n\nNow network timestamps use ktime_t infrastructure, we can add a new\nioctl() SIOCGSTAMPNS command to get timestamps in \u0027struct timespec\u0027.\nUser programs can thus access to nanosecond resolution.\n\nSigned-off-by: Eric Dumazet \u003cdada1@cosmosbay.com\u003e\nCC: Stephen Hemminger \u003cshemminger@linux-foundation.org\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "fe067e8ab5e0dc5ca3c54634924c628da92090b4", "tree": "98f5a6ebbb770f16682cfc52caea2da1e7eeb73b", "parents": [ "02ea4923b4997d7e1310c027081f46d584b9d714" ], "author": { "name": "David S. Miller", "email": "davem@sunset.davemloft.net", "time": "Wed Mar 07 12:12:44 2007 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@sunset.davemloft.net", "time": "Wed Apr 25 22:24:02 2007 -0700" }, "message": "[TCP]: Abstract out all write queue operations.\n\nThis allows the write queue implementation to be changed,\nfor example, to one which allows fast interval searching.\n\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "b7aa0bf70c4afb9e38be25f5c0922498d0f8684c", "tree": "4bc9d61031f4eb40d73887d6bde09e7d6bf2b259", "parents": [ "3927f2e8f9afa3424bb51ca81f7abac01ffd0005" ], "author": { "name": "Eric Dumazet", "email": "dada1@cosmosbay.com", "time": "Thu Apr 19 16:16:32 2007 -0700" }, "committer": { "name": "David S. Miller", "email": "davem@sunset.davemloft.net", "time": "Wed Apr 25 22:23:34 2007 -0700" }, "message": "[NET]: convert network timestamps to ktime_t\n\nWe currently use a special structure (struct skb_timeval) and plain\n\u0027struct timeval\u0027 to store packet timestamps in sk_buffs and struct\nsock.\n\nThis has some drawbacks :\n- Fixed resolution of micro second.\n- Waste of space on 64bit platforms where sizeof(struct timeval)\u003d16\n\nI suggest using ktime_t that is a nice abstraction of high resolution\ntime services, currently capable of nanosecond resolution.\n\nAs sizeof(ktime_t) is 8 bytes, using ktime_t in \u0027struct sock\u0027 permits\na 8 byte shrink of this structure on 64bit architectures. Some other\nstructures also benefit from this size reduction (struct ipq in\nipv4/ip_fragment.c, struct frag_queue in ipv6/reassembly.c, ...)\n\nOnce this ktime infrastructure adopted, we can more easily provide\nnanosecond resolution on top of it. (ioctl SIOCGSTAMPNS and/or\nSO_TIMESTAMPNS/SCM_TIMESTAMPNS)\n\nNote : this patch includes a bug correction in\ncompat_sock_get_timestamp() where a \"err \u003d 0;\" was missing (so this\nsyscall returned -ENOENT instead of 0)\n\nSigned-off-by: Eric Dumazet \u003cdada1@cosmosbay.com\u003e\nCC: Stephen Hemminger \u003cshemminger@linux-foundation.org\u003e\nCC: John find \u003clinux.kernel@free.fr\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "fa438ccfdfd3f6db02c13b61b21454eb81cd6a13", "tree": "a1759259d7543586185e2fb9db21461147944f18", "parents": [ "e317f6f69cb95527799d308a9421b7dc1252989a" ], "author": { "name": "Eric Dumazet", "email": "dada1@cosmosbay.com", "time": "Sun Mar 04 16:05:44 2007 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@sunset.davemloft.net", "time": "Wed Apr 25 22:23:27 2007 -0700" }, "message": "[NET]: Keep sk_backlog near sk_lock\n\nsk_backlog is a critical field of struct sock. (known famous words)\n\nIt is (ab)used in hot paths, in particular in release_sock(), tcp_recvmsg(),\ntcp_v4_rcv(), sk_receive_skb().\n\nIt really makes sense to place it next to sk_lock, because sk_backlog is only\nused after sk_lock locked (and thus memory cache line in L1 cache). This\nshould reduce cache misses and sk_lock acquisition time.\n\n(In theory, we could only move the head pointer near sk_lock, and leaving tail\nfar away, because \u0027tail\u0027 is normally not so hot, but keep it simple :) )\n\nSigned-off-by: Eric Dumazet \u003cdada1@cosmosbay.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "64a146513f8f12ba204b7bf5cb7e9505594ead42", "tree": "849f875501ac6f08cf309c5b13cf21ed687974e2", "parents": [ "187f5f84ef14a07dccf5f4503099708d60f6a724" ], "author": { "name": "David S. Miller", "email": "davem@sunset.davemloft.net", "time": "Tue Mar 06 11:21:05 2007 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@sunset.davemloft.net", "time": "Tue Mar 06 11:21:05 2007 -0800" }, "message": "[NET]: Revert incorrect accept queue backlog changes.\n\nThis reverts two changes:\n\n8488df894d05d6fa41c2bd298c335f944bb0e401\n248f06726e866942b3d8ca8f411f9067713b7ff8\n\nA backlog value of N really does mean allow \"N + 1\" connections\nto queue to a listening socket. This allows one to specify\n\"0\" as the backlog and still get 1 connection.\n\nNoticed by Gerrit Renker and Rick Jones.\n\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "8488df894d05d6fa41c2bd298c335f944bb0e401", "tree": "2d79c5c57a71833a1c55f1d39e6628f486ed2539", "parents": [ "bb648a0d22908116b4ef168935a160d7f17c4e6d" ], "author": { "name": "Wei Dong", "email": "weid@np.css.fujitsu.com", "time": "Fri Mar 02 12:37:26 2007 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@sunset.davemloft.net", "time": "Fri Mar 02 20:37:33 2007 -0800" }, "message": "[NET]: Fix bugs in \"Whether sock accept queue is full\" checking\n\n\twhen I use linux TCP socket, and find there is a bug in function\nsk_acceptq_is_full().\n\n\tWhen a new SYN comes, TCP module first checks its validation. If valid,\nsend SYN,ACK to the client and add the sock to the syn hash table. Next\ntime if received the valid ACK for SYN,ACK from the client. server will\naccept this connection and increase the sk-\u003esk_ack_backlog -- which is\ndone in function tcp_check_req().We check wether acceptq is full in\nfunction tcp_v4_syn_recv_sock().\n\nConsider an example:\n\n After listen(sockfd, 1) system call, sk-\u003esk_max_ack_backlog is set to\n1. As we know, sk-\u003esk_ack_backlog is initialized to 0. Assuming accept()\nsystem call is not invoked now.\n\n1. 1st connection comes. invoke sk_acceptq_is_full(). sk-\n\u003esk_ack_backlog\u003d0 sk-\u003esk_max_ack_backlog\u003d1, function return 0 accept\nthis connection. Increase the sk-\u003esk_ack_backlog\n2. 2nd connection comes. invoke sk_acceptq_is_full(). sk-\n\u003esk_ack_backlog\u003d1 sk-\u003esk_max_ack_backlog\u003d1, function return 0 accept\nthis connection. Increase the sk-\u003esk_ack_backlog\n3. 3rd connection comes. invoke sk_acceptq_is_full(). sk-\n\u003esk_ack_backlog\u003d2 sk-\u003esk_max_ack_backlog\u003d1, function return 1. Refuse\nthis connection.\n\nI think it has bugs. after listen system call. sk-\u003esk_max_ack_backlog\u003d1\nbut now it can accept 2 connections.\n\nSigned-off-by: Wei Dong \u003cweid@np.css.fujitsu.com\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" }, { "commit": "4498121ca3acbf928681b71261227d28dc29b6f6", "tree": "f6947b8bf2cee22498fcea4e9c447c4eed4b6998", "parents": [ "6548cda289b549ed60c35f16a0051609aeee2fd0" ], "author": { "name": "Patrick McHardy", "email": "kaber@trash.net", "time": "Tue Feb 27 09:56:42 2007 -0800" }, "committer": { "name": "David S. Miller", "email": "davem@sunset.davemloft.net", "time": "Wed Feb 28 09:42:13 2007 -0800" }, "message": "[NET]: Handle disabled preemption in gfp_any()\n\nctnetlink uses netlink_unicast from an atomic_notifier_chain\n(which is called within a RCU read side critical section)\nwithout holding further locks. netlink_unicast calls netlink_trim\nwith the result of gfp_any() for the gfp flags, which are passed\ndown to pskb_expand_header. gfp_any() only checks for softirq\ncontext and returns GFP_KERNEL, resulting in this warning:\n\nBUG: sleeping function called from invalid context at mm/slab.c:3032\nin_atomic():1, irqs_disabled():0\nno locks held by rmmod/7010.\n\nCall Trace:\n [\u003cffffffff8109467f\u003e] debug_show_held_locks+0x9/0xb\n [\u003cffffffff8100b0b4\u003e] __might_sleep+0xd9/0xdb\n [\u003cffffffff810b5082\u003e] __kmalloc+0x68/0x110\n [\u003cffffffff811ba8f2\u003e] pskb_expand_head+0x4d/0x13b\n [\u003cffffffff81053147\u003e] netlink_broadcast+0xa5/0x2e0\n [\u003cffffffff881cd1d7\u003e] :nfnetlink:nfnetlink_send+0x83/0x8a\n [\u003cffffffff8834f6a6\u003e] :nf_conntrack_netlink:ctnetlink_conntrack_event+0x94c/0x96a\n [\u003cffffffff810624d6\u003e] notifier_call_chain+0x29/0x3e\n [\u003cffffffff8106251d\u003e] atomic_notifier_call_chain+0x32/0x60\n [\u003cffffffff881d266d\u003e] :nf_conntrack:destroy_conntrack+0xa5/0x1d3\n [\u003cffffffff881d194e\u003e] :nf_conntrack:nf_ct_cleanup+0x8c/0x12c\n [\u003cffffffff881d4614\u003e] :nf_conntrack:kill_l3proto+0x0/0x13\n [\u003cffffffff881d482a\u003e] :nf_conntrack:nf_conntrack_l3proto_unregister+0x90/0x94\n [\u003cffffffff883551b3\u003e] :nf_conntrack_ipv4:nf_conntrack_l3proto_ipv4_fini+0x2b/0x5d\n [\u003cffffffff8109d44f\u003e] sys_delete_module+0x1b5/0x1e6\n [\u003cffffffff8105f245\u003e] trace_hardirqs_on_thunk+0x35/0x37\n [\u003cffffffff8105911e\u003e] system_call+0x7e/0x83\n\nSince netlink_unicast is supposed to be callable from within RCU\nread side critical sections, make gfp_any() check for in_atomic()\ninstead of in_softirq().\n\nAdditionally nfnetlink_send needs to use gfp_any() as well for the\ncall to netlink_broadcast).\n\nSigned-off-by: Patrick McHardy \u003ckaber@trash.net\u003e\nSigned-off-by: David S. Miller \u003cdavem@davemloft.net\u003e\n" } ], "next": "ed07536ed6731775219c1df7fa26a7588753e693" }