- 19 Aug, 2016 22 commits
-
-
David S. Miller authored
Yuval Mintz says: ==================== qed*: Fix ethtool issues relating to link This series addresses two issues that were introduced when adding support for ethtool's link_ksettings support - the first fixes a regression and second incorrect functionallity in the submission. Although these are fixes, as the feature currently exists only in 'next-next' I'm aiming them for it. Dave, please consider applying this series to 'net-next'. ==================== Signed-off-by:David S. Miller <davem@davemloft.net>
-
Yuval Mintz authored
While '0xdead' and '0xbeef' are "great" values, we should use the correct SPEED_* values instead. Fixes: 054c67d1 ("qed*: Add support for ethtool link_ksettings callbacks") Signed-off-by:
Yuval Mintz <Yuval.Mintz@qlogic.com> Signed-off-by:
-
Yuval Mintz authored
When moving into using ethtool's link_ksetting, qed started supplying its own bitmask of speed/capabilities, but qede is still checking for the SUPPORTED value to determine whether it supports pause. Fixes: 054c67d1 ("qed*: Add support for ethtool link_ksettings callbacks") Signed-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queueDavid S. Miller authored
Jeff Kirsher says: ==================== 1GbE Intel Wired LAN Driver Updates 2016-08-18 This series contains updates to igb only. Gangfeng Huang provides all the changes in the series to update the igb driver to support advanced receive side filters that direct receive packets by flows to different hardware queues. This enables a tight control on routing a flow in the platform. First patch allows for receive network flow classification to insert and remove receive filters by ethtool. Second and third patches add the ability to insert and remove ethertype and VLAN priority filters by ethtool. Last patch just fixes an error message to return "Not supported" versus "Unknown error 524". ==================== Signed-off-by: -
Gangfeng Huang authored
Use error "rmgr: Cannot insert RX class rule: Operation not supported" is more meaningful than "rmgr: Cannot insert RX class rule: Unknown error 524" Signed-off-by:
Gangfeng Huang <gangfeng.huang@ni.com> Tested-by:
Aaron Brown <aaron.f.brown@intel.com> Signed-off-by:
Jeff Kirsher <jeffrey.t.kirsher@intel.com>
-
Gangfeng Huang authored
This patch is meant to allow for RX network flow classification to insert and remove VLAN priority filter by ethtool Example: Add an VLAN priority filter: $ ethtool -N eth0 flow-type ether vlan 0x6000 vlan-mask 0x1FFF action 2 loc 1 Show all filters: $ ethtool -n eth0 4 RX rings available Total 1 rules Filter: 1 Flow Type: Raw Ethernet Src MAC addr: 00:00:00:00:00:00 mask: FF:FF:FF:FF:FF:FF Dest MAC addr: 00:00:00:00:00:00 mask: FF:FF:FF:FF:FF:FF Ethertype: 0x0 mask: 0xFFFF VLAN EtherType: 0x0 mask: 0xffff VLAN: 0x6000 mask: 0x1fff User-defined: 0x0 mask: 0xffffffffffffffff Action: Direct to queue 2 Delete the filter by location: $ ethtool -N delete 1 Signed-off-by:
Ruhao Gao <ruhao.gao@ni.com> Signed-off-by:
-
Gangfeng Huang authored
This patch is meant to allow for RX network flow classification to insert and remove ethertype filter by ethtool Example: Add an ethertype filter: $ ethtool -N eth0 flow-type ether proto 0x88F8 action 2 Show all filters: $ ethtool -n eth0 4 RX rings available Total 1 rules Filter: 15 Flow Type: Raw Ethernet Src MAC addr: 00:00:00:00:00:00 mask: FF:FF:FF:FF:FF:FF Dest MAC addr: 00:00:00:00:00:00 mask: FF:FF:FF:FF:FF:FF Ethertype: 0x88F8 mask: 0x0 Action: Direct to queue 2 Delete the filter by location: $ ethtool -N delete 15 Signed-off-by:
-
Gangfeng Huang authored
This patch is meant to allow for RX network flow classification to insert and remove Rx filter by ethtool. Ethtool interface has it's own rules manager Show all filters: $ ethtool -n eth0 4 RX rings available Total 2 rules Signed-off-by:
-
David S. Miller authored
Jiri Kosina says: ==================== qdisc-hashtable fixes The following two patches fix all the issues that have been reported against the conversion of qdisc linked list to hashtable (currently in net-next) so far. First patch adjusts handling of singleton qdiscs to the new semantics, and is rather straightforward. The second patch, which fixes "cosmetic" issue of duplicate entries in the qdisc dump for ingress qdiscs, is a little bit more hairy; I personally would love to see all the already existing "if (ingress)"-like hacks go away (by, let's say, introducing a general TCQ_F_? flag), but that's way out of scope of this patchset (but already on my todo). Thanks a lot to Daniel Borkmann and David Ahern for reporting the issues and testing the patches promptly. ==================== Signed-off-by: -
Jiri Kosina authored
tc_dump_qdisc() performs dumping of the per-device qdiscs in two phases; first, the "standard" dev->qdisc is being dumped. Second, if there is/are ingress queue(s), they are being dumped as well. After conversion of netdevice's qdisc linked-list into hashtable, these two sets are not in two disjunctive sets/lists any more, but are both "reachable" directly from netdevice's hashtable. As a consequence, the "full-depth" dump of the ingress qdiscs results in immediately hitting the netdevice hashtable again, and duplicating the dump that has already been performed for dev->qdisc. What in fact needs to be dumped in case of ingress queue is "just" the top-level ingress qdisc, as everything else has been dumped already. Fix this by extending tc_dump_qdisc_root() in a way that it can be instructed whether it should (while performing the "full" per-netdev qdisc dump) perform the whole recursion, or just dump "additional" top-level (ingress) qdiscs without performing any kind of recursion. This fixes duplicate dumps such as qdisc mq 0: root qdisc pfifo_fast 0: parent :4 bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1 qdisc pfifo_fast 0: parent :3 bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1 qdisc pfifo_fast 0: parent :2 bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1 qdisc pfifo_fast 0: parent :1 bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1 qdisc clsact ffff: parent ffff:fff1 qdisc pfifo_fast 0: parent :4 bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1 qdisc pfifo_fast 0: parent :3 bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1 qdisc pfifo_fast 0: parent :2 bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1 qdisc pfifo_fast 0: parent :1 bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1 Fixes: 59cc1f61 ("net: sched: convert qdisc linked list to hashtable") Reported-by:
Daniel Borkmann <daniel@iogearbox.net> Tested-by:
Jiri Kosina <jkosina@suse.cz> Signed-off-by:
-
Jiri Kosina authored
qdisc_match_from_root() is now iterating over per-netdevice qdisc hashtable instead of going through a linked-list of qdiscs (independently on the actual underlying netdev), which was the case before the switch to hashtable for qdiscs. For singleton qdiscs, there is no underlying netdev associated though, and therefore dumping a singleton qdisc will panic, as qdisc_dev(root) will always be NULL. BUG: unable to handle kernel NULL pointer dereference at 0000000000000410 IP: [<ffffffff8167efac>] qdisc_match_from_root+0x2c/0x70 PGD 1aceba067 PUD 1aceb7067 PMD 0 Oops: 0000 [#1] PREEMPT SMP [ ... ] task: ffff8801ec996e00 task.stack: ffff8801ec934000 RIP: 0010:[<ffffffff8167efac>] [<ffffffff8167efac>] qdisc_match_from_root+0x2c/0x70 RSP: 0018:ffff8801ec937ab0 EFLAGS: 00010203 RAX: 0000000000000408 RBX: ffff88025e612000 RCX: ffffffffffffffd8 RDX: 0000000000000000 RSI: 00000000ffff0000 RDI: ffffffff81cf8100 RBP: ffff8801ec937ab0 R08: 000000000001c160 R09: ffff8802668032c0 R10: ffffffff81cf8100 R11: 0000000000000030 R12: 00000000ffff0000 R13: ffff88025e612000 R14: ffffffff81cf3140 R15: 0000000000000000 FS: 00007f24b9af6740(0000) GS:ffff88026f280000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000410 CR3: 00000001aceec000 CR4: 00000000001406e0 Stack: ffff8801ec937ad0 ffffffff81681210 ffff88025dd51a00 00000000fffffff1 ffff8801ec937b88 ffffffff81681e4e ffffffff81c42bc0 ffff880262431500 ffffffff81cf3140 ffff88025dd51a10 ffff88025dd51a24 00000000ec937b38 Call Trace: [<ffffffff81681210>] qdisc_lookup+0x40/0x50 [<ffffffff81681e4e>] tc_modify_qdisc+0x21e/0x550 [<ffffffff8166ae25>] rtnetlink_rcv_msg+0x95/0x220 [<ffffffff81209602>] ? __kmalloc_track_caller+0x172/0x230 [<ffffffff8166ad90>] ? rtnl_newlink+0x870/0x870 [<ffffffff816897b7>] netlink_rcv_skb+0xa7/0xc0 [<ffffffff816657c8>] rtnetlink_rcv+0x28/0x30 [<ffffffff8168919b>] netlink_unicast+0x15b/0x210 [<ffffffff81689569>] netlink_sendmsg+0x319/0x390 [<ffffffff816379f8>] sock_sendmsg+0x38/0x50 [<ffffffff81638296>] ___sys_sendmsg+0x256/0x260 [<ffffffff811b1275>] ? __pagevec_lru_add_fn+0x135/0x280 [<ffffffff811b1a90>] ? pagevec_lru_move_fn+0xd0/0xf0 [<ffffffff811b1140>] ? trace_event_raw_event_mm_lru_insertion+0x180/0x180 [<ffffffff811b1b85>] ? __lru_cache_add+0x75/0xb0 [<ffffffff817708a6>] ? _raw_spin_unlock+0x16/0x40 [<ffffffff811d8dff>] ? handle_mm_fault+0x39f/0x1160 [<ffffffff81638b15>] __sys_sendmsg+0x45/0x80 [<ffffffff81638b62>] SyS_sendmsg+0x12/0x20 [<ffffffff810038e7>] do_syscall_64+0x57/0xb0 Fix this by special-casing singleton qdiscs (those that don't have underlying netdevice) and introduce immediate handling of those rather than trying to go over an underlying netdevice. We're in the same situation in tc_dump_qdisc_root() and tc_dump_tclass_root(). Ultimately, this will have to be slightly reworked so that we are actually able to show singleton qdiscs (noop) in the dump properly; but we're not currently doing that anyway, so no regression there, and better do this in a gradual manner. Fixes: 59cc1f61 ("net: sched: convert qdisc linked list to hashtable") Reported-by:
David Ahern <dsa@cumulusnetworks.com> Tested-by:
-
David S. Miller authored
Jon Maloy says: ==================== tipc: bearer and link improvements The first commit makes it possible to set and check the 'blocked' state of a bearer from the generic bearer layer. The second commit is a small improvement to the link congestion mechanism. ==================== Signed-off-by: -
Jon Paul Maloy authored
When a link is attempted woken up after congestion, it uses a different, more generous criteria than when it was originally declared congested. This has the effect that the link, and the sending process, sometimes will be woken up unnecessarily, just to immediately return to congestion when it turns out there is not not enough space in its send queue to host the pending message. This is a waste of CPU cycles. We now change the function link_prepare_wakeup() to use exactly the same criteria as tipc_link_xmit(). However, since we are now excluding the window limit from the wakeup calculation, and the current backlog limit for the lowest level is too small to house even a single maximum-size message, we have to expand this limit. We do this by evaluating an alternative, minimum value during the setting of the importance limits. Acked-by:
Ying Xue <ying.xue@windriver.com> Signed-off-by:
Jon Maloy <jon.maloy@ericsson.com> Signed-off-by:
-
Jon Paul Maloy authored
In commit 5b7066c3 ("tipc: stricter filtering of packets in bearer layer") we introduced a method of filtering out messages while a bearer is being reset, to avoid that links may be re-created and come back in working state while we are still in the process of shutting them down. This solution works well, but is limited to only work with L2 media, which is insufficient with the increasing use of UDP as carrier media. We now replace this solution with a more generic one, by introducing a new flag "up" in the generic struct tipc_bearer. This field will be set and reset at the same locations as with the previous solution, while the packet filtering is moved to the generic code for the sending side. On the receiving side, the filtering is still done in media specific code, but now including the UDP bearer. Acked-by:
-
David S. Miller authored
Sudarsana Reddy Kalluru says: ==================== qed*: Add support for additional statistics. The patch series adds qed/qede support for new statistics. Patch (1) adds couple of statistcs for "ethtool -S" display. Patch (2) adds support for per-queue statistics to ethtool display. Patch (3) adds qed support for NCSI statistics. Please consider applying this to 'net-next' branch. ==================== Signed-off-by: -
Sudarsana Reddy Kalluru authored
The patch adds driver support for sending the NCSI statistics to the MFW. This is an asynchronous request from MFW. Upon receiving this, driver populates the required data and send it to MFW. Signed-off-by:
Sudarsana Reddy Kalluru <sudarsana.kalluru@qlogic.com> Signed-off-by:
-
Sudarsana Reddy Kalluru authored
Signed-off-by:
-
Sudarsana Reddy Kalluru authored
The patch adds driver support for capturing stats ttl0_discard and packet_too_big_discard in "ethtool -S" display. Signed-off-by:
-
Colin Ian King authored
dev->name is a char array of IFNAMSIZ elements, hence can never be null, so the null pointer check is redundant. Remove it. Signed-off-by:
Colin Ian King <colin.king@canonical.com> Signed-off-by:
-
Appana Durga Kedareswara Rao authored
For implementing this driver most of the inputs is provided by Andrew Lunn. Updating the driver with Andrew Copy right. Signed-off-by:
Kedareswara rao Appana <appanad@xilinx.com> Signed-off-by:
Andrew Lunn <andrew@lunn.ch> Signed-off-by:
-
shubhrajyoti.datta@xilinx.com authored
Some of the platforms like zynqmp ultrascale+ has a separate clock gate for the rx clock. Add an optional rx_clk so that the clock can be enabled. Signed-off-by:
Shubhrajyoti Datta <shubhrajyoti.datta@xilinx.com> Acked-by:
Nicolas Ferre <nicolas.ferre@atmel.com> Acked-by:
Rob Herring <robh@kernel.org> Signed-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queueDavid S. Miller authored
Jeff Kirsher says: ==================== 40GbE Intel Wired LAN Driver Updates 2016-08-18 This series contains updates to i40e and i40evf only. Wei Yongjun updates i40e to use list_move() instead of list_del() & list_add() operations. Anjali fixes an issue where the client->open call was not protected with the client instance mutex, which allowed client->close to be called before the open all completed. Catherine makes sure that the VLAN count (and stats) gets reset to 0 after reset. Jake provides two patches, first adds the needed rtnl lock around i40evf_set_interrupt_capability() since i40evf_init_task() does not hold the rtnl_lock. Second fixes an issue where users could reduce the number of channels (queues) below the current flow director filter rules targets. Dave fixes a problem where a static analysis tool generates a warning so eliminating the irrelevant check and redundant assignment for the value of enabled_tc. Avinash fixes an sync issue where the iWARP device open is called before the PCI register writes are completed, so ensure the register writes complete before exiting the setup function. Alan fixes a bug which causes RSS to continue to work after being disabled. Carolyn implements a feature change which allows using ethtool to set RDD hash options using less than four parameters if desired. Dan Carpenter cleans up a stray unlock. Sridhar exposes the "trust" flag to userspace via ndo_get_vf_config(). ==================== Signed-off-by:
-
- 18 Aug, 2016 17 commits
-
-
Sridhar Samudrala authored
This enables ip -d l to indicate if trust is on or off for VFs. Signed-off-by:
Sridhar Samudrala <sridhar.samudrala@intel.com> Tested-by:
Andrew Bowers <andrewx.bowers@intel.com> Signed-off-by:
-
Dan Carpenter authored
We shifted the locking around a bit but forgot to delete this unlock so now it can unlock twice. Fixes: cd3be169 ('i40e: Move the mutex lock in i40e_client_unregister') Signed-off-by:
Dan Carpenter <dan.carpenter@oracle.com> Tested-by:
-
Bimmy Pujari authored
Signed-off-by:
Bimmy Pujari <Bimmy.pujari@intel.com> Tested-by:
-
Carolyn Wyborny authored
This patch implements a feature change which allows using ethtool to set RSS hash opts using less than four parameters if desired. Change-ID: I0fbb91255d81e997c456697c21ac39cc9754821b Signed-off-by:
Carolyn Wyborny <carolyn.wyborny@intel.com> Signed-off-by:
Kiran Patil <kiran.patil@intel.com> Tested-by:
-
Mitch Williams authored
When we allocate memory, we must free it. It's simple courtesy. Change-ID: Id007294096fb53344f1a8b9a0f78eddf9853c5d6 Signed-off-by:
Mitch Williams <mitch.a.williams@intel.com> Tested-by:
-
Alan Brady authored
This patch fixes the bug which causes RSS to continue to work after being disabled. After disabling RSS, traffic would continue to be assigned to different queues instead of falling back to a single queue. Without this patch, attempting to disable RSS would not work as expected. This patch fixes the bug by clearing the lookup table used by RSS such that all traffic is assigned to a single queue. This patch also addresses the issue of reinstating the lookup table should RSS then be re-enabled. Change-ID: Ib20c7c6a7e9f1f772bb787370f8a8c664796b141 Signed-off-by:
Alan Brady <alan.brady@intel.com> Tested-by:
-
Avinash Dayanand authored
VF goes through reset path during VF creation which happens to also have notification of VF reset to client. Adding conditional check to avoid wrongly notifying VF reset during VF creation. Also changing the call order of VF enable, calling it after VF creation rather than before. Change-ID: I96eabd99deae746a2f0fc465194c886f196178ce Signed-off-by:
Avinash Dayanand <avinash.dayanand@intel.com> Tested-by:
-
Avinash Dayanand authored
This patch is a fix for the bug i.e. unable to create iwarp device in VF. This is a sync issue and the iwarp device open is called even before the PCI register writes are done. Forcing the PCI register writes to happen just before it exits the function. Change-ID: I60c6a2c709da89e845f2764cc50ce8b7373c8c44 Signed-off-by:
-
Jacob Keller authored
If a driver is unable to maintain all current user supplied settings from ethtool (or other sources), it is not ok for a user request to succeed and silently trample over previous configuration. To that end, if you change the number of channels, it must not be allowed to reduce the number of channels (queues) below the current flow director filter rules targets. In this case, return -EINVAL when a request to reduce the number of channels would do so. In addition log a warning to the kernel buffer explaining why we failed, and report the rules which prevent us from lowering the number of channels. Change-ID: If41464d63d7aab11cedf09e4f3aa1a69e21ffd88 Signed-off-by:
Jacob Keller <jacob.e.keller@intel.com> Tested-by:
-
Dave Ertman authored
This patch fixes a problem where a static analysis tool generates a warning for "INVARIANT_CONDITION: Expression 'enabled_tc' used in the condition always yields the same result." Without this patch, the driver will not pass the static analysis tool checks without generating warnings. This patch fixes the problem by eliminating the irrelevant check and redundant assignment for the value of enabled_tc. Change-ID: Ia7d44cb050f507df7de333e96369d322e08bf408 Signed-off-by:
Dave Ertman <david.m.ertman@intel.com> Tested-by:
-
Jacob Keller authored
The function calls netif_set_real_num_(tx|rx)_queues, both of which should be done only under rntl lock. Unfortunately the i40evf_init_task did not hold the rtnl_lock as necessary. This patch adds the locking needed. Change-ID: Ib72a21c3ce22b71a226b16f9bbe0f5f8cc3e849b Signed-off-by:
-
Catherine Sullivan authored
When we are resetting the pf stats we should also reset the RX csum error stat. Change-ID: I7af5ee0ec81a10f6deee1a7b8c2082ea068ef620 Signed-off-by:
Catherine Sullivan <catherine.sullivan@intel.com> Tested-by:
-
Catherine Sullivan authored
When we do a reset, all the VLAN filters get added again. Therefore we also want to reset the VLAN count to 0 or we quickly run out of filters. Change-ID: I459f26851e22204dc8b8999928ad87cde8170119 Signed-off-by:
-
Anjali Singhai Jain authored
The client->open call in this path was not protected with the client instance mutex, and hence the client->close can get initiated before the open completes. Change-Id: I0ed60c38868dd3f44966b6ed49a063d0e5b7edf5 Signed-off-by:
Anjali Singhai Jain <anjali.singhai@intel.com> Tested-by:
-
Wei Yongjun authored
Using list_move() instead of list_del() + list_add(). Signed-off-by:
Wei Yongjun <weiyj.lk@gmail.com> Tested-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/davem/netDavid S. Miller authored
Minor overlapping changes for both merge conflicts. Resolution work done by Stephen Rothwell was used as a reference. Signed-off-by: -
git://git.kernel.org/pub/scm/linux/kernel/git/davem/netLinus Torvalds authored
Pull networking fixes from David Miller: 1) Buffers powersave frame test is reversed in cfg80211, fix from Felix Fietkau. 2) Remove bogus WARN_ON in openvswitch, from Jarno Rajahalme. 3) Fix some tg3 ethtool logic bugs, and one that would cause no interrupts to be generated when rx-coalescing is set to 0. From Satish Baddipadige and Siva Reddy Kallam. 4) QLCNIC mailbox corruption and napi budget handling fix from Manish Chopra. 5) Fix fib_trie logic when walking the trie during /proc/net/route output than can access a stale node pointer. From David Forster. 6) Several sctp_diag fixes from Phil Sutter. 7) PAUSE frame handling fixes in mlxsw driver from Ido Schimmel. 8) Checksum fixup fixes in bpf from Daniel Borkmann. 9) Memork leaks in nfnetlink, from Liping Zhang. 10) Use after free in rxrpc, from David Howells. 11) Use after free in new skb_array code of macvtap driver, from Jason Wang. 12) Calipso resource leak, from Colin Ian King. 13) mediatek bug fixes (missing stats sync init, etc.) from Sean Wang. 14) Fix bpf non-linear packet write helpers, from Daniel Borkmann. 15) Fix lockdep splats in macsec, from Sabrina Dubroca. 16) hv_netvsc bug fixes from Vitaly Kuznetsov, mostly to do with VF handling. 17) Various tc-action bug fixes, from CONG Wang. * git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (116 commits) net_sched: allow flushing tc police actions net_sched: unify the init logic for act_police net_sched: convert tcf_exts from list to pointer array net_sched: move tc offload macros to pkt_cls.h net_sched: fix a typo in tc_for_each_action() net_sched: remove an unnecessary list_del() net_sched: remove the leftover cleanup_a() mlxsw: spectrum: Allow packets to be trapped from any PG mlxsw: spectrum: Unmap 802.1Q FID before destroying it mlxsw: spectrum: Add missing rollbacks in error path mlxsw: reg: Fix missing op field fill-up mlxsw: spectrum: Trap loop-backed packets mlxsw: spectrum: Add missing packet traps mlxsw: spectrum: Mark port as active before registering it mlxsw: spectrum: Create PVID vPort before registering netdevice mlxsw: spectrum: Remove redundant errors from the code mlxsw: spectrum: Don't return upon error in removal path i40e: check for and deal with non-contiguous TCs ixgbe: Re-enable ability to toggle VLAN filtering ixgbe: Force VLNCTRL.VFE to be set in all VMDq paths ...
-
- 17 Aug, 2016 1 commit
-
-
David S. Miller authored
Tom Herbert says: ==================== strp: Stream parser for messages This patch set introduces a utility for parsing application layer protocol messages in a TCP stream. This is a generalization of the mechanism implemented of Kernel Connection Multiplexor. This patch set adapts KCM to use the strparser. We expect that kTLS can use this mechanism also. RDS would probably be another candidate to use a common stream parsing mechanism. The API includes a context structure, a set of callbacks, utility functions, and a data ready function. The callbacks include a parse_msg function that is called to perform parsing (e.g. BPF parsing in case of KCM), and a rcv_msg function that is called when a full message has been completed. For strparser we specify the return codes from the parser to allow the backend to indicate that control of the socket should be transferred back to userspace to handle some exceptions in the stream: The return values are: >0 : indicates length of successfully parsed message 0 : indicates more data must be received to parse the message -ESTRPIPE : current message should not be processed by the kernel, return control of the socket to userspace which can proceed to read the messages itself other < 0 : Error is parsing, give control back to userspace assuming that synchronization is lost and the stream is unrecoverable (application expected to close TCP socket) There is one issue I haven't been able to fully resolve. If parse_msg returns ESTRPIPE (wants control back to userspace) the parser may already have consumed some bytes of the message. There is no way to put bytes back into the TCP receive queue and tcp_read_sock does not allow an easy way to peek messages. In lieu of a better solution, we return ENODATA on the socket to indicate that the data stream is unrecoverable (application needs to close socket). This condition should only happen if an application layer message header is split across two skbuffs and parsing just the first skbuff wasn't sufficient to determine the that transfer to userspace is needed. This patch set contains: - strparser implementation - changes to kcm to use strparser - strparser.txt documentation v2: - Add copyright notice to C files - Remove GPL module license from strparser.c - Add report of rxpause v3: - Restore GPL module license - Use EXPORT_SYMBOL_GPL v4: - Removed unused function, changed another to be static as suggested by davem - Rewoked data_ready to be called from upper layer, no longer requires taking over socket data_ready callback as suggested by Lance Chao Tested: - Ran a KCM thrash test for 24 hours. No behavioral or performance differences observed. ==================== Signed-off-by:
-
