| Age | Commit message (Collapse) | Author |
|---|
|
git://git.kernel.org/pub/scm/linux/kernel/git/jberg/mac80211-next
Some further updates for net-next:
* fix network-manager which was broken by the previous changes
* fix delete-station events, which were broken by me making the
genlmsg_end() mistake
* fix a timer left running during suspend in some race conditions
that would cause an annoying (but harmless) warning
* (less important, but in the tree already) remove 80+80 MHz rate
reporting since the spec doesn't distinguish it from 160 MHz;
as the bitrate they're both 160 MHz bandwidth
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This tc action allows you to retrieve the connection tracking mark
This action has been used heavily by openwrt for a few years now.
There are known limitations currently:
doesn't work for initial packets, since we only query the ct table.
Fine given use case is for returning packets
no implicit defrag.
frags should be rare so fix later..
won't work for more complex tasks, e.g. lookup of other extensions
since we have no means to store results
we still have a 2nd lookup later on via normal conntrack path.
This shouldn't break anything though since skb->nfct isn't altered.
V2:
remove unnecessary braces (Jiri)
change the action identifier to 14 (Jiri)
Fix some stylistic issues caught by checkpatch
V3:
Move module params to bottom (Cong)
Get rid of tcf_hashinfo_init and friends and conform to newer API (Cong)
Acked-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Implement rtnl_link_ops->get_link_net() callback so that IFLA_LINK_NETNSID is
added to rtnetlink messages.
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This patch adds a new attribute (IFLA_LINK_NETNSID) which contains the 'link'
netns id when this netns is different from the netns where the interface
stands (for example for x-net interfaces like ip tunnels).
With this attribute, it's possible to interpret correctly all advertised
information (like IFLA_LINK, etc.).
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
With this patch, a user can define an id for a peer netns by providing a FD or a
PID. These ids are local to the netns where it is added (ie valid only into this
netns).
The main function (ie the one exported to other module), peernet2id(), allows to
get the id of a peer netns. If no id has been assigned by the user, this
function allocates one.
These ids will be used in netlink messages to point to a peer netns, for example
in case of a x-netns interface.
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
In the ipip tunnel, the skb->queue_mapping is lost in ipip_rcv().
All skb will be queued to the same cell->napi_skbs. The
gro_cell_poll is pinned to one core under load. In production traffic,
we also see severe rx_dropped in the tunl iface and it is probably due to
this limit: skb_queue_len(&cell->napi_skbs) > netdev_max_backlog.
This patch is trying to alloc_percpu(struct gro_cell) and schedule
gro_cell_poll to process the skb in the same core.
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Contrary to common expectations for an "int" return, these functions
return only a positive value -- if used correctly they cannot even
return 0 because the message header will necessarily be in the skb.
This makes the very common pattern of
if (genlmsg_end(...) < 0) { ... }
be a whole bunch of dead code. Many places also simply do
return nlmsg_end(...);
and the caller is expected to deal with it.
This also commonly (at least for me) causes errors, because it is very
common to write
if (my_function(...))
/* error condition */
and if my_function() does "return nlmsg_end()" this is of course wrong.
Additionally, there's not a single place in the kernel that actually
needs the message length returned, and if anyone needs it later then
it'll be very easy to just use skb->len there.
Remove this, and make the functions void. This removes a bunch of dead
code as described above. The patch adds lines because I did
- return nlmsg_end(...);
+ nlmsg_end(...);
+ return 0;
I could have preserved all the function's return values by returning
skb->len, but instead I've audited all the places calling the affected
functions and found that none cared. A few places actually compared
the return value with <= 0 in dump functionality, but that could just
be changed to < 0 with no change in behaviour, so I opted for the more
efficient version.
One instance of the error I've made numerous times now is also present
in net/phonet/pn_netlink.c in the route_dumpit() function - it didn't
check for <0 or <=0 and thus broke out of the loop every single time.
I've preserved this since it will (I think) have caused the messages to
userspace to be formatted differently with just a single message for
every SKB returned to userspace. It's possible that this isn't needed
for the tools that actually use this, but I don't even know what they
are so couldn't test that changing this behaviour would be acceptable.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/bluetooth/bluetooth-next
Johan Hedberg says:
====================
pull request: bluetooth-next 2015-01-16
Here are some more bluetooth & ieee802154 patches intended for 3.20:
- Refactoring & cleanups of ieee802154 & 6lowpan code
- Various fixes to the btmrvl driver
- Fixes for Bluetooth Low Energy Privacy feature handling
- Added build-time sanity checks for sockaddr sizes
- Fixes for Security Manager registration on LE-only controllers
- Refactoring of broken inquiry mode handling to a generic quirk
Please let me know if there are any issues pulling. Thanks.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This patch benefits from newly introduced switchdev notifier and uses it
to propagate fdb learn events from rocker driver to bridge. That avoids
direct function calls and possible use by other listeners (ovs).
Suggested-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: Scott Feldman <sfeldma@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This patch introduces new notifier for purposes of exposing events which happen
on switch driver side. The consumers of the event messages are mainly involved
masters, namely bridge and ovs.
Suggested-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: Scott Feldman <sfeldma@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This action provides a possibility to exec custom BPF code.
Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Remove the function hci_conn_change_link_key() that is not used anywhere.
This was partially found by using a static code analysis program called
cppcheck.
Signed-off-by: Rickard Strandqvist <rickard_strandqvist@spectrumdigital.se>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/jberg/mac80211-next
Here's a big pile of changes for this round.
We have
* a lot of regulatory code changes to deal with the
way newer Intel devices handle this
* a change to drop packets while disconnecting from
an AP instead of trying to wait for them
* a new attempt at improving the tailroom accounting
to not kick in too much for performance reasons
* improvements in wireless link statistics
* many other small improvements and small fixes that
didn't seem necessary for 3.19 (e.g. in hwsim which
is testing only code)
Conflicts:
drivers/staging/rtl8723au/os_dep/ioctl_cfg80211.c
Minor overlapping changes.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
RAW sockets with hdrinc suffer from contention on rt_uncached_lock
spinlock.
One solution is to use percpu lists, since most routes are destroyed
by the cpu that created them.
It is unclear why we even have to put these routes in uncached_list,
as all outgoing packets should be freed when a device is dismantled.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Fixes: caacf05e5ad1 ("ipv4: Properly purge netdev references on uncached routes.")
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
For some reason, we made the bandwidth separate flags, which
is rather confusing - a single rate cannot have different
bandwidths at the same time.
Change this to no longer be flags but use a separate field
for the bandwidth ('bw') instead.
While at it, add support for 5 and 10 MHz rates - these are
reported as regular legacy rates with their real bitrate,
but tagged as 5/10 now to make it easier to distinguish them.
In the nl80211 API, the flags are preserved, but the code
now can also clearly only set a single one of the flags.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
These rates are treated the same as 160 MHz in the spec, so
it makes no sense to distinguish them. As no driver uses them
yet, this is also not a problem, just remove them.
In the userspace API the field remains reserved to preserve
API and ABI.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
These rates are treated the same as 160 MHz in the spec,
so it makes no sense to distinguish them. As no driver
uses them yet, this is also not a problem, just remove
them.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Pablo Neira Ayuso says:
====================
netfilter updates for net-next
The following patchset contains netfilter updates for net-next, just a
bunch of cleanups and small enhancement to selectively flush conntracks
in ctnetlink, more specifically the patches are:
1) Rise default number of buckets in conntrack from 16384 to 65536 in
systems with >= 4GBytes, patch from Marcelo Leitner.
2) Small refactor to save one level on indentation in xt_osf, from
Joe Perches.
3) Remove unnecessary sizeof(char) in nf_log, from Fabian Frederick.
4) Another small cleanup to remove redundant variable in nfnetlink,
from Duan Jiong.
5) Fix compilation warning in nfnetlink_cthelper on parisc, from
Chen Gang.
6) Fix wrong format in debugging for ctseqadj, from Gao feng.
7) Selective conntrack flushing through the mark for ctnetlink, patch
from Kristian Evensen.
8) Remove nf_ct_conntrack_flush_report() exported symbol now that is
not required anymore after the selective flushing patch, again from
Kristian.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Introduces support for the group policy extension to the VXLAN virtual
port. The extension is disabled by default and only enabled if the user
has provided the respective configuration.
ovs-vsctl add-port br0 vxlan0 -- \
set Interface vxlan0 type=vxlan options:exts=gbp
The configuration interface to enable the extension is based on a new
attribute OVS_VXLAN_EXT_GBP nested inside OVS_TUNNEL_ATTR_EXTENSION
which can carry additional extensions as needed in the future.
The group policy metadata is stored as binary blob (struct ovs_vxlan_opts)
internally just like Geneve options but transported as nested Netlink
attributes to user space.
Renames the existing TUNNEL_OPTIONS_PRESENT to TUNNEL_GENEVE_OPT with the
binary value kept intact, a new flag TUNNEL_VXLAN_OPT is introduced.
The attributes OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS and existing
OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS are implemented mutually exclusive.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
A VXLAN net_device looking for an appropriate socket may only consider
a socket which has a matching set of flags/extensions enabled. If
incompatible flags are enabled, return a conflict to have the caller
create a distinct socket with distinct port.
The OVS VXLAN port is kept unaware of extensions at this point.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Implements supports for the Group Policy VXLAN extension [0] to provide
a lightweight and simple security label mechanism across network peers
based on VXLAN. The security context and associated metadata is mapped
to/from skb->mark. This allows further mapping to a SELinux context
using SECMARK, to implement ACLs directly with nftables, iptables, OVS,
tc, etc.
The group membership is defined by the lower 16 bits of skb->mark, the
upper 16 bits are used for flags.
SELinux allows to manage label to secure local resources. However,
distributed applications require ACLs to implemented across hosts. This
is typically achieved by matching on L2-L4 fields to identify the
original sending host and process on the receiver. On top of that,
netlabel and specifically CIPSO [1] allow to map security contexts to
universal labels. However, netlabel and CIPSO are relatively complex.
This patch provides a lightweight alternative for overlay network
environments with a trusted underlay. No additional control protocol
is required.
Host 1: Host 2:
Group A Group B Group B Group A
+-----+ +-------------+ +-------+ +-----+
| lxc | | SELinux CTX | | httpd | | VM |
+--+--+ +--+----------+ +---+---+ +--+--+
\---+---/ \----+---/
| |
+---+---+ +---+---+
| vxlan | | vxlan |
+---+---+ +---+---+
+------------------------------+
Backwards compatibility:
A VXLAN-GBP socket can receive standard VXLAN frames and will assign
the default group 0x0000 to such frames. A Linux VXLAN socket will
drop VXLAN-GBP frames. The extension is therefore disabled by default
and needs to be specifically enabled:
ip link add [...] type vxlan [...] gbp
In a mixed environment with VXLAN and VXLAN-GBP sockets, the GBP socket
must run on a separate port number.
Examples:
iptables:
host1# iptables -I OUTPUT -m owner --uid-owner 101 -j MARK --set-mark 0x200
host2# iptables -I INPUT -m mark --mark 0x200 -j DROP
OVS:
# ovs-ofctl add-flow br0 'in_port=1,actions=load:0x200->NXM_NX_TUN_GBP_ID[],NORMAL'
# ovs-ofctl add-flow br0 'in_port=2,tun_gbp_id=0x200,actions=drop'
[0] https://tools.ietf.org/html/draft-smith-vxlan-group-policy
[1] http://lwn.net/Articles/204905/
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add support for remote checksum offload in VXLAN. This uses a
reserved bit to indicate that RCO is being done, and uses the low order
reserved eight bits of the VNI to hold the start and offset values in a
compressed manner.
Start is encoded in the low order seven bits of VNI. This is start >> 1
so that the checksum start offset is 0-254 using even values only.
Checksum offset (transport checksum field) is indicated in the high
order bit in the low order byte of the VNI. If the bit is set, the
checksum field is for UDP (so offset = start + 6), else checksum
field is for TCP (so offset = start + 16). Only TCP and UDP are
supported in this implementation.
Remote checksum offload for VXLAN is described in:
https://tools.ietf.org/html/draft-herbert-vxlan-rco-00
Tested by running 200 TCP_STREAM connections with VXLAN (over IPv4).
With UDP checksums and Remote Checksum Offload
IPv4
Client
11.84% CPU utilization
Server
12.96% CPU utilization
9197 Mbps
IPv6
Client
12.46% CPU utilization
Server
14.48% CPU utilization
8963 Mbps
With UDP checksums, no remote checksum offload
IPv4
Client
15.67% CPU utilization
Server
14.83% CPU utilization
9094 Mbps
IPv6
Client
16.21% CPU utilization
Server
14.32% CPU utilization
9058 Mbps
No UDP checksums
IPv4
Client
15.03% CPU utilization
Server
23.09% CPU utilization
9089 Mbps
IPv6
Client
16.18% CPU utilization
Server
26.57% CPU utilization
8954 Mbps
Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
A self-managed device will sometimes need to set its regdomain synchronously.
Notably it should be set before usermode has a chance to query it. Expose
a new API to accomplish this which requires the RTNL.
Signed-off-by: Arik Nemtsov <arikx.nemtsov@intel.com>
Reviewed-by: Ilan Peer <ilan.peer@intel.com>
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
The same macros are used for rx as well. So rename it.
Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
tc code implicitly considers skb->protocol even in case of accelerated
vlan paths and expects vlan protocol type here. However, on rx path,
if the vlan header was already stripped, skb->protocol contains value
of next header. Similar situation is on tx path.
So for skbs that use skb->vlan_tci for tagging, use skb->vlan_proto instead.
Reported-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This patch cleans up the header flags of VXLAN in anticipation of
defining some new ones:
- Move header related definitions from vxlan.c to vxlan.h
- Change VXLAN_FLAGS to be VXLAN_HF_VNI (only currently defined flag)
- Move check for unknown flags to after we find vxlan_sock, this
assumes that some flags may be processed based on tunnel
configuration
- Add a comment about why the stack treating unknown set flags as an
error instead of ignoring them
Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This patch adds this missing structure for processing the result of the
HCI Delete Stored Link Key command.
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
|
|
When the HCI Read Stored Link Keys command completes it gives useful
information of the current stored keys and maximum keys a controller
can actually store. So process this event and store these information
in hci_dev structure.
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
|
|
This patch adds the missing commmand structure and command complete
structure for the HCI Read Store Link Key command.
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
|
|
When hci_req_run() calls its provided complete function and one of the
HCI commands in the sequence fails, then provide the opcode of failing
command. In case of success HCI_OP_NOP is provided since all commands
completed.
This patch fixes the prototype of hci_req_complete_t and all its users.
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
|
|
The base for the current statistics is pretty mixed up, support
exporting RX/TX statistics for MSDUs per TID. This (currently)
covers received MSDUs, transmitted MSDUs and retries/failures
thereof.
Doing it per TID for MSDUs makes more sense than say only per AC
because it's symmetric - we could export per-AC statistics for all
frames (which AC we used for transmission can be determined also
for management frames) but per TID is better and usually data
frames are really the ones we care about. Also, on RX we can't
determine the AC - but we do know the TID for any QoS MPDU we
received.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
The current statistics we keep aren't very clear, some are on
MPDUs and some on MSDUs/MMPDUs. Clarify the descriptions based
on the counters mac80211 keeps.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Add these two values:
* BEACON_RX: number of beacons received from this peer
* BEACON_SIGNAL_AVG: signal strength average for beacons only
These can then be used for Android Lollipop's statistics request.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
This is really just duplicating the list of information that's
already available in the nl80211 attribute, so remove the list.
Two small changes are needed:
* remove STATION_INFO_ASSOC_REQ_IES complete, but the length
(assoc_req_ies_len) can be used instead
* add NL80211_STA_INFO_RX_DROP_MISC which exists internally
but not in nl80211 yet
This gets rid of the duplicate maintenance of the two lists.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
In many cases, drivers can filter things like beacons that will
skew statistics reported by mac80211. To get correct statistics
in these cases, call drivers to obtain statistics and let them
override all values, filling values from mac80211 if the driver
didn't provide them. Not all of them make sense for the driver
to fill, so some are still always done by mac80211.
Note that this doesn't currently allow a driver to say "I know
this value is wrong, don't report it at all", or to sum it up
with a mac80211 value (as could be useful for "dropped misc"),
that can be added if it turns out to be needed.
This also gets rid of the get_rssi() method as is can now be
implemented using sta_statistics().
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
When a station is removed, its statistics may be interesting to
userspace, for example for further aggregation of statistics of
all stations that ever connected to an AP.
Introduce a new cfg80211_del_sta_sinfo() function (and make the
cfg80211_del_sta() a static inline calling it) to allow passing
a struct station_info along with this, and send the data in the
nl80211 event message.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Add the time spent scanning to the survey data so it can be
reported by drivers that collect such information.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Not all devices are able to report survey data (particularly
time spent for various operations) per channel. As all these
statistics already exist in survey data, allow such devices
to report them (if userspace requested it)
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
All of the survey data is (currently) per channel anyway,
so having the word "channel" in the name does nothing. In
the next patch I'll introduce global data to the survey,
where the word "channel" is actually confusing.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
The only user of nf_ct_conntrack_flush_report() was ctnetlink_del_conntrack().
After adding support for flushing connections with a given mark, this function
is no longer called.
Signed-off-by: Kristian Evensen <kristian.evensen@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
|
|
When hw acceleration is enabled, the GENERATE_IV or PUT_IV_SPACE flags
only require headroom space. Therefore, the tailroom-needed counter can
safely be decremented for most drivers.
The older incarnation of this patch (ca34e3b5) assumed that the above
holds true for all drivers. As reported by Christopher Chavez and
researched by Christian Lamparter and Larry Finger, this isn't a valid
assumption for p54 and cw1200.
Drivers that still require tailroom for ICV/MIC even when HW encryption
is enabled can use IEEE80211_KEY_FLAG_RESERVE_TAILROOM to indicate it.
Signed-off-by: Ido Yariv <idox.yariv@intel.com>
Cc: Christopher Chavez <chrischavez@gmx.us>
Cc: Christian Lamparter <chunkeey@googlemail.com>
Cc: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Solomon Peachy <pizza@shaftnet.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
Merge mac80211.git to get some changes that would otherwise
cause conflicts with new changes coming here.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/jberg/mac80211
Here's just a single fix - a revert of a patch that broke the
p54 and cw2100 drivers (arguably due to bad assumptions there.)
Since this affects kernels since 3.17, I decided to revert for
now and we'll revisit this optimisation properly for -next.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
With the wiphy::features flag being used up this patch adds a
new field wiphy::ext_features. Considering extensibility this
new field is declared as a byte array. This extensible flag is
exposed to user-space by NL80211_ATTR_EXT_FEATURES.
Cc: Avinash Patil <patila@marvell.com>
Signed-off-by: Gautam (Gautam Kumar) Shukla <gautams@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
|
|
This work adds the possibility to define a per route/destination
congestion control algorithm. Generally, this opens up the possibility
for a machine with different links to enforce specific congestion
control algorithms with optimal strategies for each of them based
on their network characteristics, even transparently for a single
application listening on all links.
For our specific use case, this additionally facilitates deployment
of DCTCP, for example, applications can easily serve internal
traffic/dsts in DCTCP and external one with CUBIC. Other scenarios
would also allow for utilizing e.g. long living, low priority
background flows for certain destinations/routes while still being
able for normal traffic to utilize the default congestion control
algorithm. We also thought about a per netns setting (where different
defaults are possible), but given its actually a link specific
property, we argue that a per route/destination setting is the most
natural and flexible.
The administrator can utilize this through ip-route(8) by appending
"congctl [lock] <name>", where <name> denotes the name of a
congestion control algorithm and the optional lock parameter allows
to enforce the given algorithm so that applications in user space
would not be allowed to overwrite that algorithm for that destination.
The dst metric lookups are being done when a dst entry is already
available in order to avoid a costly lookup and still before the
algorithms are being initialized, thus overhead is very low when the
feature is not being used. While the client side would need to drop
the current reference on the module, on server side this can actually
even be avoided as we just got a flat-copied socket clone.
Joint work with Florian Westphal.
Suggested-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This patch adds the minimum necessary for the RTAX_CC_ALGO congestion
control metric to be set up and dumped back to user space.
While the internal representation of RTAX_CC_ALGO is handled as a u32
key, we avoided to expose this implementation detail to user space, thus
instead, we chose the netlink attribute that is being exchanged between
user space to be the actual congestion control algorithm name, similarly
as in the setsockopt(2) API in order to allow for maximum flexibility,
even for 3rd party modules.
It is a bit unfortunate that RTAX_QUICKACK used up a whole RTAX slot as
it should have been stored in RTAX_FEATURES instead, we first thought
about reusing it for the congestion control key, but it brings more
complications and/or confusion than worth it.
Joint work with Florian Westphal.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This patch adds necessary infrastructure to the congestion control
framework for later per route congestion control support.
For a per route congestion control possibility, our aim is to store
a unique u32 key identifier into dst metrics, which can then be
mapped into a tcp_congestion_ops struct. We argue that having a
RTAX key entry is the most simple, generic and easy way to manage,
and also keeps the memory footprint of dst entries lower on 64 bit
than with storing a pointer directly, for example. Having a unique
key id also allows for decoupling actual TCP congestion control
module management from the FIB layer, i.e. we don't have to care
about expensive module refcounting inside the FIB at this point.
We first thought of using an IDR store for the realization, which
takes over dynamic assignment of unused key space and also performs
the key to pointer mapping in RCU. While doing so, we stumbled upon
the issue that due to the nature of dynamic key distribution, it
just so happens, arguably in very rare occasions, that excessive
module loads and unloads can lead to a possible reuse of previously
used key space. Thus, previously stale keys in the dst metric are
now being reassigned to a different congestion control algorithm,
which might lead to unexpected behaviour. One way to resolve this
would have been to walk FIBs on the actually rare occasion of a
module unload and reset the metric keys for each FIB in each netns,
but that's just very costly.
Therefore, we argue a better solution is to reuse the unique
congestion control algorithm name member and map that into u32 key
space through jhash. For that, we split the flags attribute (as it
currently uses 2 bits only anyway) into two u32 attributes, flags
and key, so that we can keep the cacheline boundary of 2 cachelines
on x86_64 and cache the precalculated key at registration time for
the fast path. On average we might expect 2 - 4 modules being loaded
worst case perhaps 15, so a key collision possibility is extremely
low, and guaranteed collision-free on LE/BE for all in-tree modules.
Overall this results in much simpler code, and all without the
overhead of an IDR. Due to the deterministic nature, modules can
now be unloaded, the congestion control algorithm for a specific
but unloaded key will fall back to the default one, and on module
reload time it will switch back to the expected algorithm
transparently.
Joint work with Florian Westphal.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Do the nla validation earlier, outside the write lock.
This is needed by followup patch which needs to be able to call
request_module (which can sleep) if needed.
Joint work with Daniel Borkmann.
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
