summaryrefslogtreecommitdiffstats
path: root/include/net/fq_impl.h
Commit message (Collapse)AuthorAgeFilesLines
* wifi: mac80211: add support for restricting netdev features per vifFelix Fietkau2022-12-011-6/+10
| | | | | | | | | | | | | | | | | | | This can be used to selectively disable feature flags for checksum offload, scatter/gather or GSO by changing vif->netdev_features. Removing features from vif->netdev_features does not affect the netdev features themselves, but instead fixes up skbs in the tx path so that the offloads are not needed in the driver. Aside from making it easier to deal with vif type based hardware limitations, this also makes it possible to optimize performance on hardware without native GSO support by declaring GSO support in hw->netdev_features and removing it from vif->netdev_features. This allows mac80211 to handle GSO segmentation after the sta lookup, but before itxq enqueue, thus reducing the number of unnecessary sta lookups, as well as some other per-packet processing. Signed-off-by: Felix Fietkau <nbd@nbd.name> Link: https://lore.kernel.org/r/20221010094338.78070-1-nbd@nbd.name Signed-off-by: Johannes Berg <johannes.berg@intel.com>
* net/fq_impl: Use the bitmap API to allocate bitmapsChristophe JAILLET2022-07-111-3/+2
| | | | | | | | | | Use bitmap_zalloc()/bitmap_free() instead of hand-writing them. It is less verbose and it improves the semantic. Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr> Link: https://lore.kernel.org/r/c7bf099af07eb497b02d195906ee8c11fea3b3bd.1657377335.git.christophe.jaillet@wanadoo.fr Signed-off-by: Jakub Kicinski <kuba@kernel.org>
* net/fq_impl: do not maintain a backlog-sorted list of flowsFelix Fietkau2021-01-211-46/+67
| | | | | | | | | | | | | A sorted flow list is only needed to drop packets in the biggest flow when hitting the overmemory condition. By scanning flows only when needed, we can avoid paying the cost of maintaining the list under normal conditions In order to avoid scanning lots of empty flows and touching too many cold cache lines, a bitmap of flows with backlog is maintained Signed-off-by: Felix Fietkau <nbd@nbd.name> Link: https://lore.kernel.org/r/20201218184718.93650-3-nbd@nbd.name Signed-off-by: Johannes Berg <johannes.berg@intel.com>
* net/fq_impl: drop get_default_func, move default flow to fq_tinFelix Fietkau2021-01-211-6/+5
| | | | | | | | | Simplifies the code and prepares for a rework of scanning for flows on overmemory drop. Signed-off-by: Felix Fietkau <nbd@nbd.name> Link: https://lore.kernel.org/r/20201218184718.93650-2-nbd@nbd.name Signed-off-by: Johannes Berg <johannes.berg@intel.com>
* net/fq_impl: bulk-free packets from a flow on overmemoryFelix Fietkau2021-01-211-11/+44
| | | | | | | | | | This is similar to what sch_fq_codel does. It also amortizes the worst case cost of a follow-up patch that changes the selection of the biggest flow for dropping packets Signed-off-by: Felix Fietkau <nbd@nbd.name> Link: https://lore.kernel.org/r/20201218184718.93650-1-nbd@nbd.name Signed-off-by: Johannes Berg <johannes.berg@intel.com>
* net/fq_impl: use skb_get_hash instead of skb_get_hash_perturbFelix Fietkau2020-07-311-2/+1
| | | | | | | | | This avoids unnecessarily regenerating the skb flow hash Signed-off-by: Felix Fietkau <nbd@nbd.name> Link: https://lore.kernel.org/r/20200726130947.88145-1-nbd@nbd.name [small commit message fixup] Signed-off-by: Johannes Berg <johannes.berg@intel.com>
* net/fq_impl: Switch to kvmalloc() for memory allocationToke Høiland-Jørgensen2019-11-081-2/+2
| | | | | | | | | | | | | | | The FQ implementation used by mac80211 allocates memory using kmalloc(), which can fail; and Johannes reported that this actually happens in practice. To avoid this, switch the allocation to kvmalloc() instead; this also brings fq_impl in line with all the FQ qdiscs. Fixes: 557fc4a09803 ("fq: add fair queuing framework") Reported-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com> Link: https://lore.kernel.org/r/20191105155750.547379-1-toke@redhat.com Signed-off-by: Johannes Berg <johannes.berg@intel.com>
* net/flow_dissector: switch to siphashEric Dumazet2019-10-231-2/+2
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | UDP IPv6 packets auto flowlabels are using a 32bit secret (static u32 hashrnd in net/core/flow_dissector.c) and apply jhash() over fields known by the receivers. Attackers can easily infer the 32bit secret and use this information to identify a device and/or user, since this 32bit secret is only set at boot time. Really, using jhash() to generate cookies sent on the wire is a serious security concern. Trying to change the rol32(hash, 16) in ip6_make_flowlabel() would be a dead end. Trying to periodically change the secret (like in sch_sfq.c) could change paths taken in the network for long lived flows. Let's switch to siphash, as we did in commit df453700e8d8 ("inet: switch IP ID generator to siphash") Using a cryptographically strong pseudo random function will solve this privacy issue and more generally remove other weak points in the stack. Packet schedulers using skb_get_hash_perturb() benefit from this change. Fixes: b56774163f99 ("ipv6: Enable auto flow labels by default") Fixes: 42240901f7c4 ("ipv6: Implement different admin modes for automatic flow labels") Fixes: 67800f9b1f4e ("ipv6: Call skb_get_hash_flowi6 to get skb->hash in ip6_make_flowlabel") Fixes: cb1ce2ef387b ("ipv6: Implement automatic flow label generation on transmit") Signed-off-by: Eric Dumazet <edumazet@google.com> Reported-by: Jonathan Berger <jonathann1@walla.com> Reported-by: Amit Klein <aksecurity@gmail.com> Reported-by: Benny Pinkas <benny@pinkas.net> Cc: Tom Herbert <tom@herbertland.com> Signed-off-by: David S. Miller <davem@davemloft.net>
* treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 257Thomas Gleixner2019-06-051-2/+1
| | | | | | | | | | | | | | | | | | | | | | Based on 1 normalized pattern(s): gpl v2 extracted by the scancode license scanner the SPDX license identifier GPL-2.0-only has been chosen to replace the boilerplate/reference in 19 file(s). Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Allison Randal <allison@lohutok.net> Reviewed-by: Richard Fontana <rfontana@redhat.com> Reviewed-by: Steve Winslow <swinslow@gmail.com> Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org> Reviewed-by: Alexios Zavras <alexios.zavras@intel.com> Cc: linux-spdx@vger.kernel.org Link: https://lkml.kernel.org/r/20190529141333.108140152@linutronix.de Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
* mac80211: calculate hash for fq without holding fq->lock in itxq enqueueFelix Fietkau2019-04-261-8/+10
| | | | | | | | Reduces lock contention on enqueue/dequeue of iTXQ packets Signed-off-by: Felix Fietkau <nbd@nbd.name> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Signed-off-by: Johannes Berg <johannes.berg@intel.com>
* Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/netDavid S. Miller2017-10-301-3/+6
|\ | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Several conflicts here. NFP driver bug fix adding nfp_netdev_is_nfp_repr() check to nfp_fl_output() needed some adjustments because the code block is in an else block now. Parallel additions to net/pkt_cls.h and net/sch_generic.h A bug fix in __tcp_retransmit_skb() conflicted with some of the rbtree changes in net-next. The tc action RCU callback fixes in 'net' had some overlap with some of the recent tcf_block reworking. Signed-off-by: David S. Miller <davem@davemloft.net>
| * fq_impl: Properly enforce memory limitToke Høiland-Jørgensen2017-10-181-3/+6
| | | | | | | | | | | | | | | | | | | | The fq structure would fail to properly enforce the memory limit in the case where the packet being enqueued was bigger than the packet being removed to bring the memory usage down. So keep dropping packets until the memory usage is back below the limit. Also, fix the statistics for memory limit violations. Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk> Signed-off-by: Johannes Berg <johannes.berg@intel.com>
* | fq: support filtering a given tinJohannes Berg2017-10-111-10/+62
|/ | | | | | | | | | | | | Add to the FQ API a way to filter a given tin, in order to remove frames that fulfil certain criteria according to a filter function. This will be used by mac80211 to remove frames belonging to an AP VLAN interface that's being removed. Signed-off-by: Johannes Berg <johannes.berg@intel.com> Acked-by: Toke Høiland-Jørgensen <toke@toke.dk> Signed-off-by: Johannes Berg <johannes.berg@intel.com>
* fq.h: Port memory limit mechanism from fq_codelToke Høiland-Jørgensen2016-09-301-1/+6
| | | | | | | | | | | The reusable fairness queueing implementation (fq.h) lacks the memory usage limit that the fq_codel qdisc has. This means that small devices (e.g. WiFi routers) can run out of memory when flooded with a large number of packets. This ports the memory limit feature from fq_codel to fq.h. Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk> Signed-off-by: Johannes Berg <johannes.berg@intel.com>
* fq: split out backlog update logicMichal Kazior2016-04-281-11/+19
| | | | | | | | | | | | | | mac80211 (which will be the first user of the fq.h) recently started to support software A-MSDU aggregation. It glues skbuffs together into a single one so the backlog accounting needs to be more fine-grained. To avoid backlog sorting logic duplication split it up for re-use. Signed-off-by: Michal Kazior <michal.kazior@tieto.com> Signed-off-by: David S. Miller <davem@davemloft.net>
* fq: add fair queuing frameworkMichal Kazior2016-04-251-0/+269
This works on the same implementation principle as codel*.h, i.e. there's a generic header with structures and macros and a implementation header carrying function definitions to include in given, e.g. driver or module. The fairness logic comes from net/sched/sch_fq_codel.c but is generalized so it is more flexible and easier to re-use. Signed-off-by: Michal Kazior <michal.kazior@tieto.com> Signed-off-by: David S. Miller <davem@davemloft.net>
generated by cgit v1.2.3 (git 2.25.1) at 2024-11-26 18:17:16 +0000