to avoid a LOR on the multicast list lock in the freemoptions routines.
As it turns out, tcp_usr_detach can acquire the tcbinfo lock readonly.
Trying to wunlock the pcbinfo lock in that context has caused a number
of reported crashes.
This change unclutters in_pcbfree and moves the handling of wunlock vs
runlock of pcbinfo to the freemoptions routine.
Reported by: mjg@, bde@, o.hartmann at walstatt.org
Approved by: sbruno
Multicast incorrectly calls in to drivers with a mutex held causing drivers
to have to go through all manner of contortions to use a non sleepable lock.
Serialize multicast updates instead.
Submitted by: mmacy <mmacy@mattmacy.io>
Reviewed by: shurd, sbruno
Sponsored by: Limelight Networks
Differential Revision: https://reviews.freebsd.org/D14969
This patch adds a new socket option, SO_REUSEPORT_LB, which allow multiple
programs or threads to bind to the same port and incoming connections will be
load balanced using a hash function.
Most of the code was copied from a similar patch for DragonflyBSD.
However, in DragonflyBSD, load balancing is a global on/off setting and can not
be set per socket. This patch allows for simultaneous use of both the current
SO_REUSEPORT and the new SO_REUSEPORT_LB options on the same system.
Required changes to structures
Globally change so_options from 16 to 32 bit value to allow for more options.
Add hashtable in pcbinfo to hold all SO_REUSEPORT_LB sockets.
Limitations
As DragonflyBSD, a load balance group is limited to 256 pcbs
(256 programs or threads sharing the same socket).
Submitted by: Johannes Lundberg <johanlun0@gmail.com>
Sponsored by: Limelight Networks
Differential Revision: https://reviews.freebsd.org/D11003
It is the forerunner/foundational work of bringing in both Rack and BBR
which use hpts for pacing out packets. The feature is optional and requires
the TCPHPTS option to be enabled before the feature will be active. TCP
modules that use it must assure that the base component is compile in
the kernel in which they are loaded.
MFC after: Never
Sponsored by: Netflix Inc.
Differential Revision: https://reviews.freebsd.org/D15020
assumption is violated, "bad things" could follow.
I believe such an assert would have detected some of the problems jch@
was chasing in PR 203175 (see r307551). We also use it in our internal
TCP development efforts. And, in case a bug does slip through to
released code, this change silently ignores subsequent calls to
in_pcbfree().
Reviewed by: rrs
Sponsored by: Netflix, Inc.
Differential Revision: https://reviews.freebsd.org/D14990
through the lock-switching hoops.
A few of the INP lookup operations that lock INPs after the lookup do
so using this mechanism (to maintain lock ordering):
1. Lock lookup structure.
2. Find INP.
3. Acquire reference on INP.
4. Drop lock on lookup structure.
5. Acquire INP lock.
6. Drop reference on INP.
This change provides a slightly shorter path for cases where the INP
lock is uncontested:
1. Lock lookup structure.
2. Find INP.
3. Try to acquire the INP lock.
4. If successful, drop lock on lookup structure.
Of course, if the INP lock is contested, the functions will need to
revert to the previous way of switching locks safely.
This saves a few atomic operations when the INP lock is uncontested.
Discussed with: gallatin, rrs, rwatson
MFC after: 2 weeks
Sponsored by: Netflix, Inc.
Differential Revision: https://reviews.freebsd.org/D12911
Add a new macro to clear both the L3 and L2 route caches, to
hopefully prevent future instances where only the L3 cache was
cleared when both should have been.
MFC after: 1 week
Sponsored by: Dell EMC Isilon
Differential Revision: https://reviews.freebsd.org/D13989
Reviewed by: karels
Mainly focus on files that use BSD 3-Clause license.
The Software Package Data Exchange (SPDX) group provides a specification
to make it easier for automated tools to detect and summarize well known
opensource licenses. We are gradually adopting the specification, noting
that the tags are considered only advisory and do not, in any way,
superceed or replace the license texts.
Special thanks to Wind River for providing access to "The Duke of
Highlander" tool: an older (2014) run over FreeBSD tree was useful as a
starting point.
transmit queues aswell as non-ratelimited ones.
Add the required structure bits in order to support a backpressure
indication with ratelimited connections aswell as non-ratelimited
ones. The backpressure indicator is a value between zero and 65535
inclusivly, indicating if the destination transmit queue is empty or
full respectivly. Applications can use this value as a decision point
for when to stop transmitting data to avoid endless ENOBUFS error
codes upon transmitting an mbuf. This indicator is also useful to
reduce the latency for ratelimited queues.
Reviewed by: gallatin, kib, gnn
Differential Revision: https://reviews.freebsd.org/D11518
Sponsored by: Mellanox Technologies
considering cache line hits and misses. Put the lock and hash list
glue into the first cache line, put inp_refcount inp_flags inp_socket
into the second cache line.
o On allocation zero out entire structure except the lock and list entries,
including inp_route inp_lle inp_gencnt. When inp_route and inp_lle were
introduced, they were added below inp_zero_size, resulting on not being
cleared after free/alloc. This definitely was a source of bugs with route
caching. Could be that r315956 has just fixed one of them.
The inp_gencnt is reinitialized on every alloc, so it is safe to clear it.
This has been proved to improve TCP performance at Netflix.
Obtained from: rrs
Differential Revision: D10686
function (they used to say UMA_ZONE_NOFREE), so flag parameter goes away.
The zone_fini parameter also goes away. Previously no protocols (except
divert) supplied zone_fini function, so inpcb locks were leaked with slabs.
This was okay while zones were allocated with UMA_ZONE_NOFREE flag, but now
this is a leak. Fix that by suppling inpcb_fini() function as fini method
for all inpcb zones.
ip_forward, TCP/IPv6, and probably SCTP leaked references to L2 cache
entry because they used their own routes on the stack, not in_pcb routes.
The original model for route caching was callers that provided a route
structure to ip{,6}input() would keep the route, and this model was used
for L2 caching as well. Instead, change L2 caching to be done by default
only when using a route structure in the in_pcb; the pcb deallocation
code frees L2 as well as L3 cacches. A separate change will add route
caching to TCP/IPv6.
Another suggestion was to have the transport protocols indicate willingness
to use L2 caching, but this approach keeps the changes in the network
level
Reviewed by: ae gnn
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D10059
and those below, will be ignored--
> Description of fields to fill in above: 76 columns --|
> PR: If and which Problem Report is related.
> Submitted by: If someone else sent in the change.
> Reported by: If someone else reported the issue.
> Reviewed by: If someone else reviewed your modification.
> Approved by: If you needed approval for this commit.
> Obtained from: If the change is from a third party.
> MFC after: N [day[s]|week[s]|month[s]]. Request a reminder email.
> MFH: Ports tree branch name. Request approval for merge.
> Relnotes: Set to 'yes' for mention in release notes.
> Security: Vulnerability reference (one per line) or description.
> Sponsored by: If the change was sponsored by an organization.
> Differential Revision: https://reviews.freebsd.org/D### (*full* phabric URL needed).
> Empty fields above will be automatically removed.
M netinet/in_pcb.c
M netinet/ip_output.c
M netinet6/ip6_output.c
This is a painful change, but it is needed. On the one hand, we avoid
modifying them, and this slows down some ideas, on the other hand we still
eventually modify them and tools like netstat(1) never work on next version of
FreeBSD. We maintain a ton of spares in them, and we already got some ifdef
hell at the end of tcpcb.
Details:
- Hide struct inpcb, struct tcpcb under _KERNEL || _WANT_FOO.
- Make struct xinpcb, struct xtcpcb pure API structures, not including
kernel structures inpcb and tcpcb inside. Export into these structures
the fields from inpcb and tcpcb that are known to be used, and put there
a ton of spare space.
- Make kernel and userland utilities compilable after these changes.
- Bump __FreeBSD_version.
Reviewed by: rrs, gnn
Differential Revision: D10018
Unfortunately they will have different integer value due to Linux value being already assigned in FreeBSD.
The patch is similar to IP_RECVDSTADDR but also provides the destination port value to the application.
This allows/improves implementation of transparent proxies on UDP sockets due to having the whole information on forwarded packets.
Reviewed by: adrian, aw
Approved by: ae (mentor)
Sponsored by: rsync.net
Differential Revision: D9235
Renumber cluase 4 to 3, per what everybody else did when BSD granted
them permission to remove clause 3. My insistance on keeping the same
numbering for legal reasons is too pedantic, so give up on that point.
Submitted by: Jan Schaumann <jschauma@stevens.edu>
Pull Request: https://github.com/freebsd/freebsd/pull/96
Unfortunately they will have different integer value due to Linux value being already assigned in FreeBSD.
The patch is similar to IP_RECVDSTADDR but also provides the destination port value to the application.
This allows/improves implementation of transparent proxies on UDP sockets due to having the whole information on forwarded packets.
Sponsored-by: rsync.net
Differential Revision: D9235
Reviewed-by: adrian
Small summary
-------------
o Almost all IPsec releated code was moved into sys/netipsec.
o New kernel modules added: ipsec.ko and tcpmd5.ko. New kernel
option IPSEC_SUPPORT added. It enables support for loading
and unloading of ipsec.ko and tcpmd5.ko kernel modules.
o IPSEC_NAT_T option was removed. Now NAT-T support is enabled by
default. The UDP_ENCAP_ESPINUDP_NON_IKE encapsulation type
support was removed. Added TCP/UDP checksum handling for
inbound packets that were decapsulated by transport mode SAs.
setkey(8) modified to show run-time NAT-T configuration of SA.
o New network pseudo interface if_ipsec(4) added. For now it is
build as part of ipsec.ko module (or with IPSEC kernel).
It implements IPsec virtual tunnels to create route-based VPNs.
o The network stack now invokes IPsec functions using special
methods. The only one header file <netipsec/ipsec_support.h>
should be included to declare all the needed things to work
with IPsec.
o All IPsec protocols handlers (ESP/AH/IPCOMP protosw) were removed.
Now these protocols are handled directly via IPsec methods.
o TCP_SIGNATURE support was reworked to be more close to RFC.
o PF_KEY SADB was reworked:
- now all security associations stored in the single SPI namespace,
and all SAs MUST have unique SPI.
- several hash tables added to speed up lookups in SADB.
- SADB now uses rmlock to protect access, and concurrent threads
can do SA lookups in the same time.
- many PF_KEY message handlers were reworked to reflect changes
in SADB.
- SADB_UPDATE message was extended to support new PF_KEY headers:
SADB_X_EXT_NEW_ADDRESS_SRC and SADB_X_EXT_NEW_ADDRESS_DST. They
can be used by IKE daemon to change SA addresses.
o ipsecrequest and secpolicy structures were cardinally changed to
avoid locking protection for ipsecrequest. Now we support
only limited number (4) of bundled SAs, but they are supported
for both INET and INET6.
o INPCB security policy cache was introduced. Each PCB now caches
used security policies to avoid SP lookup for each packet.
o For inbound security policies added the mode, when the kernel does
check for full history of applied IPsec transforms.
o References counting rules for security policies and security
associations were changed. The proper SA locking added into xform
code.
o xform code was also changed. Now it is possible to unregister xforms.
tdb_xxx structures were changed and renamed to reflect changes in
SADB/SPDB, and changed rules for locking and refcounting.
Reviewed by: gnn, wblock
Obtained from: Yandex LLC
Relnotes: yes
Sponsored by: Yandex LLC
Differential Revision: https://reviews.freebsd.org/D9352
- Add RATELIMIT kernel configuration keyword which must be set to
enable the new functionality.
- Add support for hardware driven, Receive Side Scaling, RSS aware, rate
limited sendqueues and expose the functionality through the already
established SO_MAX_PACING_RATE setsockopt(). The API support rates in
the range from 1 to 4Gbytes/s which are suitable for regular TCP and
UDP streams. The setsockopt(2) manual page has been updated.
- Add rate limit function callback API to "struct ifnet" which supports
the following operations: if_snd_tag_alloc(), if_snd_tag_modify(),
if_snd_tag_query() and if_snd_tag_free().
- Add support to ifconfig to view, set and clear the IFCAP_TXRTLMT
flag, which tells if a network driver supports rate limiting or not.
- This patch also adds support for rate limiting through VLAN and LAGG
intermediate network devices.
- How rate limiting works:
1) The userspace application calls setsockopt() after accepting or
making a new connection to set the rate which is then stored in the
socket structure in the kernel. Later on when packets are transmitted
a check is made in the transmit path for rate changes. A rate change
implies a non-blocking ifp->if_snd_tag_alloc() call will be made to the
destination network interface, which then sets up a custom sendqueue
with the given rate limitation parameter. A "struct m_snd_tag" pointer is
returned which serves as a "snd_tag" hint in the m_pkthdr for the
subsequently transmitted mbufs.
2) When the network driver sees the "m->m_pkthdr.snd_tag" different
from NULL, it will move the packets into a designated rate limited sendqueue
given by the snd_tag pointer. It is up to the individual drivers how the rate
limited traffic will be rate limited.
3) Route changes are detected by the NIC drivers in the ifp->if_transmit()
routine when the ifnet pointer in the incoming snd_tag mismatches the
one of the network interface. The network adapter frees the mbuf and
returns EAGAIN which causes the ip_output() to release and clear the send
tag. Upon next ip_output() a new "snd_tag" will be tried allocated.
4) When the PCB is detached the custom sendqueue will be released by a
non-blocking ifp->if_snd_tag_free() call to the currently bound network
interface.
Reviewed by: wblock (manpages), adrian, gallatin, scottl (network)
Differential Revision: https://reviews.freebsd.org/D3687
Sponsored by: Mellanox Technologies
MFC after: 3 months
but removed due to other changes in the system. Restore the llentry pointer
to the "struct route", and use it to cache the L2 lookup (ARP or ND6) as
appropriate.
Submitted by: Mike Karels
Differential Revision: https://reviews.freebsd.org/D6262
route caching for TCP, with some improvements. In particular, invalidate
the route cache if a new route is added, which might be a better match.
The cache is automatically invalidated if the old route is deleted.
Submitted by: Mike Karels
Reviewed by: gnn
Differential Revision: https://reviews.freebsd.org/D4306
only for read locks on pcbs. The same race can happen with write
lock semantics as well.
The race scenario:
- Two threads (1 and 2) locate pcb with writer semantics (INPLOOKUP_WLOCKPCB)
and do in_pcbref() on it.
- 1 and 2 both drop the inp hash lock.
- Another thread (3) grabs the inp hash lock. Then it runs in_pcbfree(),
which wlocks the pcb. They must happen faster than 1 or 2 come INP_WLOCK()!
- 1 and 2 congest in INP_WLOCK().
- 3 does in_pcbremlists(), drops hash lock, and runs in_pcbrele_wlocked(),
which doesn't free the pcb due to two references on it.
Then it unlocks the pcb.
- 1 (or 2) gets wlock on the pcb, runs in_pcbrele_wlocked(), which doesn't
report inp as freed, due to 2 (or 1) still helding extra reference on it.
The thread tries to do smth with a disconnected pcb and crashes.
Submitted by: emeric.poupon@stormshield.eu
Reviewed by: gleb@
MFC after: 1 week
Sponsored by: Stormshield
Tested by: Cassiano Peixoto, Stormshield
Avoid too strict INP_INFO_RLOCK_ASSERT checks due to
tcp_notify() being called from in6_pcbnotify().
Reported by: Larry Rosenman <ler@lerctr.org>
Submitted by: markj, jch
- The existing TCP INP_INFO lock continues to protect the global inpcb list
stability during full list traversal (e.g. tcp_pcblist()).
- A new INP_LIST lock protects inpcb list actual modifications (inp allocation
and free) and inpcb global counters.
It allows to use TCP INP_INFO_RLOCK lock in critical paths (e.g. tcp_input())
and INP_INFO_WLOCK only in occasional operations that walk all connections.
PR: 183659
Differential Revision: https://reviews.freebsd.org/D2599
Reviewed by: jhb, adrian
Tested by: adrian, nitroboost-gmail.com
Sponsored by: Verisign, Inc.
Both are used to protect access to IP addresses lists and they can be
acquired for reading several times per packet. To reduce lock contention
it is better to use rmlock here.
Reviewed by: gnn (previous version)
Obtained from: Yandex LLC
Sponsored by: Yandex LLC
Differential Revision: https://reviews.freebsd.org/D3149
years for head. However, it is continuously misused as the mpsafe argument
for callout_init(9). Deprecate the flag and clean up callout_init() calls
to make them more consistent.
Differential Revision: https://reviews.freebsd.org/D2613
Reviewed by: jhb
MFC after: 2 weeks
bits.
The motivation here is to eventually teach netisr and potentially
other networking subsystems a bit more about how RSS work queues / buckets
are configured so things have a hope of auto-configuring in the future.
* net/rss_config.[ch] takes care of the generic bits for doing
configuration, hash function selection, etc;
* topelitz.[ch] is now in net/ rather than netinet/;
* (and would be in libkern if it didn't directly include RSS_KEYSIZE;
that's a later thing to fix up.)
* netinet/in_rss.[ch] now just contains the IPv4 specific methods;
* and netinet/in6_rss.[ch] now just contains the IPv6 specific methods.
This should have no functional impact on anyone currently using
the RSS support.
Differential Revision: D1383
Reviewed by: gnn, jfv (intel driver bits)
have chosen different (and more traditional) stateless/statuful
NAT64 as translation mechanism. Last non-trivial commits to both
faith(4) and faithd(8) happened more than 12 years ago, so I assume
it is time to drop RFC3142 in FreeBSD.
No objections from: net@
fibs. Use the mbuf's or the socket's fib instead of RT_ALL_FIBS. Fixes PR
187553. Also fixes netperf's UDP_STREAM test on a nondefault fib.
sys/netinet/ip_output.c
In ip_output, lookup the source address using the mbuf's fib instead
of RT_ALL_FIBS.
sys/netinet/in_pcb.c
in in_pcbladdr, lookup the source address using the socket's fib,
because we don't seem to have the mbuf fib. They should be the same,
though.
tests/sys/net/fibs_test.sh
Clear the expected failure on udp_dontroute.
PR: 187553
CR: https://reviews.freebsd.org/D772
MFC after: 3 weeks
Sponsored by: Spectra Logic
ifa_ifwithdstaddr. For the sake of backwards compatibility, the new
arguments were added to new functions named ifa_ifwithnet_fib and
ifa_ifwithdstaddr_fib, while the old functions became wrappers around the
new ones that passed RT_ALL_FIBS for the fib argument. However, the
backwards compatibility is not desired for FreeBSD 11, because there are
numerous other incompatible changes to the ifnet(9) API. We therefore
decided to remove it from head but leave it in place for stable/9 and
stable/10. In addition, this commit adds the fib argument to
ifa_ifwithbroadaddr for consistency's sake.
sys/sys/param.h
Increment __FreeBSD_version
sys/net/if.c
sys/net/if_var.h
sys/net/route.c
Add fibnum argument to ifa_ifwithbroadaddr, and remove the _fib
versions of ifa_ifwithdstaddr, ifa_ifwithnet, and ifa_ifwithroute.
sys/net/route.c
sys/net/rtsock.c
sys/netinet/in_pcb.c
sys/netinet/ip_options.c
sys/netinet/ip_output.c
sys/netinet6/nd6.c
Fixup calls of modified functions.
share/man/man9/ifnet.9
Document changed API.
CR: https://reviews.freebsd.org/D458
MFC after: Never
Sponsored by: Spectra Logic
awareness.
* Introduce IP_BINDMULTI - indicating that it's okay to bind multiple
sockets on the same bind details.
Although the PCB code has been taught about this (see below) this patch
doesn't introduce the rest of the PCB changes necessary to distribute
lookups among multiple PCB entries in the global wildcard table.
* Introduce IP_RSS_LISTEN_BUCKET - placing an listen socket into the
given RSS bucket (and thus a single PCBGROUP hash.)
* Modify the PCB add path to be aware of IP_BINDMULTI:
+ Only allow further PCB entries to be added if the owner credentials
and IP_BINDMULTI has been specified. Ie, only allow further
IP_BINDMULTI sockets to appear if the first bind() was IP_BINDMULTI.
* Teach the PCBGROUP code about IP_RSS_LISTE_BUCKET marked PCB entries.
Instead of using the wildcard logic and hashing, these sockets are
simply placed into the PCBGROUP and _not_ in the wildcard hash.
* When doing a PCBGROUP lookup, also do a wildcard match as well.
This allows for an RSS bucket PCB entry to appear in a PCBGROUP
rather than having to exist in the wildcard list.
Tested:
* TCP IPv4 server testing with igb(4)
* TCP IPv4 server testing with ix(4)
TODO:
* The pcbgroup lookup code duplicated the wildcard and wildcard-PCB
logic. This could be refactored into a single function.
* This doesn't yet work for IPv6 (The PCBGROUP code in netinet6/ doesn't
yet know about this); nor does it yet fully work for UDP.
ifa_ifwithnet() and ifa_ifwithdstaddr() The legacy functions will call the
_fib() versions with RT_ALL_FIBS, preserving legacy behavior.
sys/net/if_var.h
sys/net/if.c
Add legacy-compatible functions as described above. Ensure legacy
behavior when RT_ALL_FIBS is passed as fibnum.
sys/netinet/in_pcb.c
sys/netinet/ip_output.c
sys/netinet/ip_options.c
sys/net/route.c
sys/net/rtsock.c
sys/netinet6/nd6.c
Call with _fib() functions if we must use a specific fib, or the
legacy functions otherwise.
tests/sys/netinet/fibs_test.sh
tests/sys/netinet/udp_dontroute.c
Improve the udp_dontroute test. The bug that this test exercises is
that ifa_ifwithnet() will return the wrong address, if multiple
interfaces have addresses on the same subnet but with different
fibs. The previous version of the test only considered one possible
failure mode: that ifa_ifwithnet_fib() might fail to find any
suitable address at all. The new version also checks whether
ifa_ifwithnet_fib() finds the correct address by checking where the
ARP request goes.
Reported by: bz, hrs
Reviewed by: hrs
MFC after: 1 week
X-MFC-with: 264905
Sponsored by: Spectra Logic
These two bugs are closely related. The root cause is that ifa_ifwithnet
does not consider FIBs when searching for an interface address.
sys/net/if_var.h
sys/net/if.c
Add a fib argument to ifa_ifwithnet and ifa_ifwithdstadddr. Those
functions will only return an address whose interface fib equals the
argument.
sys/net/route.c
Update calls to ifa_ifwithnet and ifa_ifwithdstaddr with fib
arguments.
sys/netinet/in.c
Update in_addprefix to consider the interface fib when adding
prefixes. This will prevent it from not adding a subnet route when
one already exists on a different fib.
sys/net/rtsock.c
sys/netinet/in_pcb.c
sys/netinet/ip_output.c
sys/netinet/ip_options.c
sys/netinet6/nd6.c
Add RT_DEFAULT_FIB arguments to ifa_ifwithdstaddr and ifa_ifwithnet.
In some cases it there wasn't a clear specific fib number to use.
In others, I was unable to test those functions so I chose
RT_DEFAULT_FIB to minimize divergence from current behavior. I will
fix some of the latter changes along with PR kern/187553.
tests/sys/netinet/fibs_test.sh
tests/sys/netinet/udp_dontroute.c
tests/sys/netinet/Makefile
Revert r263738. The udp_dontroute test was right all along.
However, bugs kern/187550 and kern/187553 cancelled each other out
when it came to this test. Because of kern/187553, ifa_ifwithnet
searched the default fib instead of the requested one, but because
of kern/187550, there was an applicable subnet route on the default
fib. The new test added in r263738 doesn't work right, however. I
can verify with dtrace that ifa_ifwithnet returned the wrong address
before I applied this commit, but route(8) miraculously found the
correct interface to use anyway. I don't know how.
Clear expected failure messages for kern/187550 and kern/187552.
PR: kern/187550
PR: kern/187552
Reviewed by: melifaro
MFC after: 3 weeks
Sponsored by: Spectra Logic
linking NIC Receive Side Scaling (RSS) to the network stack's
connection-group implementation. This prototype (and derived patches)
are in use at Juniper and several other FreeBSD-using companies, so
despite some reservations about its maturity, merge the patch to the
base tree so that it can be iteratively refined in collaboration rather
than maintained as a set of gradually diverging patch sets.
(1) Merge a software implementation of the Toeplitz hash specified in
RSS implemented by David Malone. This is used to allow suitable
pcbgroup placement of connections before the first packet is
received from the NIC. Software hashing is generally avoided,
however, due to high cost of the hash on general-purpose CPUs.
(2) In in_rss.c, maintain authoritative versions of RSS state intended
to be pushed to each NIC, including keying material, hash
algorithm/ configuration, and buckets. Provide software-facing
interfaces to hash 2- and 4-tuples for IPv4 and IPv6 using both
the RSS standardised Toeplitz and a 'naive' variation with a hash
efficient in software but with poor distribution properties.
Implement rss_m2cpuid()to be used by netisr and other load
balancing code to look up the CPU on which an mbuf should be
processed.
(3) In the Ethernet link layer, allow netisr distribution using RSS as
a source of policy as an alternative to source ordering; continue
to default to direct dispatch (i.e., don't try and requeue packets
for processing on the 'right' CPU if they arrive in a directly
dispatchable context).
(4) Allow RSS to control tuning of connection groups in order to align
groups with RSS buckets. If a packet arrives on a protocol using
connection groups, and contains a suitable hardware-generated
hash, use that hash value to select the connection group for pcb
lookup for both IPv4 and IPv6. If no hardware-generated Toeplitz
hash is available, we fall back on regular PCB lookup risking
contention rather than pay the cost of Toeplitz in software --
this is a less scalable but, at my last measurement, faster
approach. As core counts go up, we may want to revise this
strategy despite CPU overhead.
Where device drivers suitably configure NICs, and connection groups /
RSS are enabled, this should avoid both lock and line contention during
connection lookup for TCP. This commit does not modify any device
drivers to tune device RSS configuration to the global RSS
configuration; patches are in circulation to do this for at least
Chelsio T3 and Intel 1G/10G drivers. Currently, the KPI for device
drivers is not particularly robust, nor aware of more advanced features
such as runtime reconfiguration/rebalancing. This will hopefully prove
a useful starting point for refinement.
No MFC is scheduled as we will first want to nail down a more mature
and maintainable KPI/KBI for device drivers.
Sponsored by: Juniper Networks (original work)
Sponsored by: EMC/Isilon (patch update and merge)
