Commit Graph

2974 Commits

Author SHA1 Message Date
David Bauer
46d0ce19f1 iwinfo: update to latest Git HEAD
9f5a7c4 iwinfo: add missing HT modename for HT-None
06a03c9 Revert "iwinfo: add BSS load element to scan result"
9a4bae8 iwinfo: add device id for Qualcomm Atheros QCA9990
eba5a20 iwinfo: add device id for BCM43602
a6914dc iwinfo: add BSS load element to scan result
bb21698 iwinfo: add device id for Atheros AR9287
7483398 iwinfo: add device id for MediaTek MT7615E

Signed-off-by: David Bauer <mail@david-bauer.net>
2020-03-22 02:08:02 +01:00
Rafał Miłecki
8c33debb52 samba36: log error if getting device info failed
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
2020-03-21 22:35:45 +01:00
Jason A. Donenfeld
2bd56595a6 wireguard: bump to 0.0.20200318
WireGuard had a brief professional security audit. The auditors didn't find
any vulnerabilities, but they did suggest one defense-in-depth suggestion to
protect against potential API misuse down the road, mentioned below. This
compat snapshot corresponds with the patches I just pushed to Dave for
5.6-rc7.

* curve25519-x86_64: avoid use of r12

This buys us 100 extra cycles, which isn't much, but it winds up being even
faster on PaX kernels, which use r12 as a RAP register.

* wireguard: queueing: account for skb->protocol==0

This is the defense-in-depth change. We deal with skb->protocol==0 just fine,
but the advice to deal explicitly with it seems like a good idea.

* receive: remove dead code from default packet type case

A default case of a particular switch statement should never be hit, so
instead of printing a pretty debug message there, we full-on WARN(), so that
we get bug reports.

* noise: error out precomputed DH during handshake rather than config

All peer keys will now be addable, even if they're low order. However, no
handshake messages will be produced successfully. This is a more consistent
behavior with other low order keys, where the handshake just won't complete if
they're being used anywhere.

* send: use normaler alignment formula from upstream

We're trying to keep a minimal delta with upstream for the compat backport.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
2020-03-21 09:42:07 +01:00
Jason A. Donenfeld
858c6b17c8 wireguard-tools: bump to 1.0.20200319
* netlink: initialize mostly unused field
* curve25519: squelch warnings on clang

Code quality improvements.

* man: fix grammar in wg(8) and wg-quick(8)
* man: backlink wg-quick(8) in wg(8)
* man: add a warning to the SaveConfig description

Man page improvements. We hope to rewrite our man pages in mdocml at some
point soon.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
2020-03-21 09:41:52 +01:00
Daniel Golle
50a59b3a39 hostapd: fix segfault in wpa_supplicant ubus
When introducing ubus reload support, ubus initialization was moved
to the service level instead of being carried out when adding a BSS
configuration. While this works when using wpa_supplicant in that way,
it breaks the ability to run wpa_supplicant on the command line, eg.
for debugging purposes.
Fix that by re-introducing ubus context intialization when adding
configuration.

Reported-by: @PolynomialDivision https://github.com/openwrt/openwrt/pull/2417
Fixes: 60fb4c92b6 ("hostapd: add ubus reload")
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
2020-03-18 19:05:22 +01:00
Leon M. George
b78f61c336 hostapd: fix pointer cast warnings
Signed-off-by: Leon M. George <leon@georgemail.eu>
2020-03-17 10:23:28 +01:00
Leon M. George
a8a993e64c hostapd: remove trailing whitespace
Signed-off-by: Leon M. George <leon@georgemail.eu>
2020-03-17 10:23:28 +01:00
Hans Dedecker
3db9b83f16 curl: bump to 7.69.1
For changes in 7.69.1; see https://curl.haxx.se/changes.html#7_69_1

Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
2020-03-16 21:31:01 +01:00
Rozhuk Ivan
d890f85e59 wwan: fix hotplug event handling
Hotplug manager send: "remove" -> "add" -> "bind" events,
script interpret bind as "not add" = "remove" and mark device
as unavailable.

Signed-off-by: Rozhuk Ivan <rozhuk.im@gmail.com>
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com> [PKG_RELEASE increase]
2020-03-16 21:03:25 +01:00
Rozhuk Ivan
4821ff064b comgt: fix hotplug event handling
Hotplug manager send: "remove" -> "add" -> "bind" events,
script interpret bind as "not add" = "remove" and mark device
as unavailable.

Signed-off-by: Rozhuk Ivan <rozhuk.im@gmail.com>
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com> [PKG_RELEASE increase]
2020-03-16 21:02:33 +01:00
Hans Dedecker
d21f5aaa99 netifd: update to latest git HEAD
dbdef93 interface-ip: transfer prefix route ownership for deprecated ipv6addr to kernel

Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
2020-03-16 20:59:12 +01:00
Mathias Kresin
1a9408281b iproute2: revert add libcap support, enabled in ip-full
This reverts commit a6da3f9ef7.

The libcap isn't as optional as the commit messages suggests. A hard
dependency to the libcap package is added, which is only available in
the external packages feed. Therefore it is impossible to package
ip-full without having the external packages feed up and running, which
is a regression to the former behaviour.

Signed-off-by: Mathias Kresin <dev@kresin.me>
Acked-by: Hans Dedecker <dedeckeh@gmail.com>
2020-03-15 21:38:05 +01:00
Hans Dedecker
a5c30efeb1 odhcpd: update to latest git HEAD
6594c6b ubus: use dhcpv6 ia assignment flag
a90cc2e dhcpv6-ia: avoid setting lifetime to infinite for static assignments
bb07fa4 dhcpv4: avoid setting lifetime to infinite for static assignments

Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
2020-03-15 20:09:19 +01:00
Kevin Darbyshire-Bryant
d7613bd02f iptables: update to 1.8.4
Bump to iptable 1.8.4 and address packaging issue as mentioned in the
original bump/revert cycle.

"This reverts commit 10cbc896c0.
The updated iptables package does not build due to the following error
encountered on the buildbots:
    cp: cannot stat '.../iptables-1.8.4/ipkg-install/usr/lib/libiptc.so.*': No such file or directory

The changelog mentions "build: remove -Wl,--no-as-needed and libiptc.so" so
it appears as if further packaging changes are needed beyond a simple
version bump."

Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
2020-03-15 15:55:56 +00:00
Hans Dedecker
659ae99e9b curl: bump to 7.69.0
For changes in 7.69.0; see https://curl.haxx.se/changes.html#7_69_0

Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
2020-03-10 22:15:18 +01:00
Kevin Darbyshire-Bryant
04a21c26a0 dnsmasq: bump to v2.81rc3
Bump to latest release candidate and drop 2 local patches that have been
upstreamed.

Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
2020-03-10 12:49:07 +00:00
Kevin Darbyshire-Bryant
0b84b89251 dnsmasq: bump to 2.81rc2 + 2 local
Bump to dnsmasq 2.81rc2.  In the process discovered several compiler
warnings one with a logical error.

2 relevant patches sent upstream, added as 2 local patches for OpenWrt

Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
2020-03-06 15:47:56 +00:00
Kevin Darbyshire-Bryant
3251ac8f2d dnsmasq: bump to v2.81rc1
1st release candidate for v2.81 after 18 months.

Refresh patches & remove all upstreamed leaving:

110-ipset-remove-old-kernel-support.patch

Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
2020-03-04 20:44:23 +00:00
Alan Swanson
25cb5685c1 netifd: rename 20-smp-tune to 20-smp-packet-steering
Rename the script to be more obvious that this is for
packet steering only.

Signed-off-by: Alan Swanson <reiver@improbability.net>
2020-03-03 22:43:09 +01:00
Alan Swanson
d3868f15f8 netifd: change RPS/XPS handling to all CPUs and disable by default
The current implementation is significantly lowering lantiq
performace [1][2] by using RPS with non-irq CPUs and XPS
with alternating CPUs.

The previous netifd implementation (by default but could be
configured) simply used all CPUs and this patch essentially
reverts to this behaviour.

The only document suggesting using non-interrupt CPUs is Red
Hat [3] where if the network interrupt rate is extremely high
excluding the CPU that handles network interrupts *may* also
improve performance.

The original packet steering patches [4] advise that optimal
settings for the CPU mask seems to depend on architectures
and cache hierarcy so one size does not fit all. It also
advises that the overhead in processing for a lightly loaded
server can cause performance degradation.

Ideally, proper IRQ balancing is a better option with
the irqbalance daemon or manually.

The kernel does not enable packet steering by default, so
also disable in OpenWRT by default. (Though mvebu with its
hardware scheduling issues [5] might want to enable packet
steering by default.)

Change undocumented "default_ps" parameter to clearer
"packet_steering" parameter. The old parameter was only ever
set in target/linux/mediatek/base-files/etc/uci-defaults/99-net-ps
and matched the default.

[1] https://forum.openwrt.org/t/18-06-4-speed-fix-for-bt-homehub-5a
[2] https://openwrt.ebilan.co.uk/viewtopic.php?f=7&t=1105
[3] https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/6/html/performance_tuning_guide/network-rps
[4] https://marc.info/?l=linux-netdev&m=125792239522685&w=2
[5] https://git.openwrt.org/?p=openwrt/openwrt.git;a=commitdiff;h=2e1f6f1682d3974d8ea52310e460f1bbe470390f

Fixes: #1852
Fixes: #2573

Signed-off-by: Alan Swanson <reiver@improbability.net>
2020-03-03 22:43:08 +01:00
Petr Štetiar
2c3c83e40b ppp: activate PIE ASLR by default
This activates PIE ASLR support by default when the regular option is
selected.

Size increase on imx6:

 112681 ppp_2.4.8-2_arm_cortex-a9_neon.ipk
 121879 ppp_2.4.8-2_arm_cortex-a9_neon.ipk
 = 9198 diff

Acked-by: Alexander Couzens <lynxis@fe80.eu>
Signed-off-by: Petr Štetiar <ynezz@true.cz>
2020-03-01 21:35:59 +01:00
Petr Štetiar
35890514bb ppp: backport security fixes
8d45443bb5c9 pppd: Ignore received EAP messages when not doing EAP
8d7970b8f3db pppd: Fix bounds check in EAP code
858976b1fc31 radius: Prevent buffer overflow in rc_mksid()

Signed-off-by: Petr Štetiar <ynezz@true.cz>
Fixes: CVE-2020-8597
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
2020-02-26 16:38:43 +01:00
Jo-Philipp Wich
817e775319 Revert "ppp: backport security fixes"
This reverts commit 215598fd03 since it
didn't contain a reference to the CVE it addresses. The next commit
will re-add the commit including a CVE reference in its commit message.

Signed-off-by: Jo-Philipp Wich <jo@mein.io>
2020-02-26 16:37:27 +01:00
John Crispin
d3b7838ebe hostapd: enhance wifi reload
Add a radio_config_id property. If the radio config changes return an error
upon receiving the reconf call.

Signed-off-by: John Crispin <john@phrozen.org>
2020-02-25 17:01:55 +01:00
Piotr Dymacz
2d113f89d2 hostapd: start hostapd/wpa_supplicant for all wiphy devices
c888e17e06 ("hostapd: manage instances via procd instead of pidfile")
added procd support for managing hostapd and wpa_supplicant daemons
but at the same time limited wiphy names to 'phy*'.

This brings back initial behaviour (introduced in 60fb4c92b6 ("hostapd:
add ubus reload") and makes procd manage daemons for any wiphy device
found in '/sys/class/ieee80211'.

CC: Felix Fietkau <nbd@nbd.name>
CC: Daniel Golle <daniel@makrotopia.org>
Signed-off-by: Piotr Dymacz <pepe2k@gmail.com>
2020-02-24 23:27:50 +01:00
Piotr Dymacz
82679ca0b9 umbim: move package to 'WWAN' submenu
'uqmi' was moved to 'WWAN' submenu in 9abdeee0b7.
Let's be consistent and do the same with 'umbim'.

Signed-off-by: Piotr Dymacz <pepe2k@gmail.com>
2020-02-24 23:27:50 +01:00
Hauke Mehrtens
806354ab53 linux-atm: Fix compile warning
The function trace_on_exit() is given to atexit() as a parameter, but
atexit() only takes a function pointer to a function with a void
parameter.

This problem was introduced when the on_exit() function was incompletely
replaced by atexit().

Fixes: ba6c8bd614 ("linux-atm: add portability fixes")
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
2020-02-24 23:25:28 +01:00
DENG Qingfang
b9d29b78c8 iw: update to 5.4
Update iw to 5.4
This increases the ipk size of iw-tiny/full by about 400 bytes

Signed-off-by: DENG Qingfang <dengqf6@mail2.sysu.edu.cn>
2020-02-22 16:38:41 +01:00
Adrian Schmutzler
a5b2c6f5ed rssileds: add dependencies based on LDFLAGS
This adds the direct dependencies introduced by TARGET_LDFLAGS
to the package's DEPENDS variable.

This was found by accidentally building rssileds on octeon, which
resulted in:

"Package rssileds is missing dependencies for the following libraries:
libnl-tiny.so"

Though the dependencies are provided when building for the
relevant targets ar71xx, ath79 and ramips, it seems more tidy to
specify them explicitly.

Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
2020-02-22 14:26:01 +01:00
Stijn Tintel
a9b5473c92 lldpd: bump to 1.0.5
Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
2020-02-22 10:31:28 +02:00
Petr Štetiar
215598fd03 ppp: backport security fixes
8d45443bb5c9 pppd: Ignore received EAP messages when not doing EAP
8d7970b8f3db pppd: Fix bounds check in EAP code
858976b1fc31 radius: Prevent buffer overflow in rc_mksid()

Signed-off-by: Petr Štetiar <ynezz@true.cz>
2020-02-20 09:12:12 +01:00
Russell Senior
731f7ea48a dnsmasq: fix uci-defaults script to exit 0 so it is cleaned up
A file, package/network/services/dnsmasq/files/50-dnsmasq-migrate-resolv-conf-auto.sh,
was added in commit 6a28552120, but it
does not exit in a way that tells the uci-defaults mechanism that it
succeeded, and so it is not cleaned up after running successfully. Add
an exit 0 to the end to correct that.

Signed-off-by: Russell Senior <russell@personaltelco.net>
2020-02-19 22:02:59 +01:00
Jason A. Donenfeld
49caf9f98a wireguard: bump to 0.0.20200215
* send: cleanup skb padding calculation
* socket: remove useless synchronize_net

Sorry for the back-to-back releases. This fixes a regression spotted by Eric
Dumazet.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
2020-02-15 08:57:49 +01:00
DENG Qingfang
5715b21f80 iproute2: update to 5.5.0, enable LTO
Update iproute2 to 5.5.0
Enable LTO to save several KB of size

Signed-off-by: DENG Qingfang <dengqf6@mail2.sysu.edu.cn>
2020-02-13 21:35:13 +01:00
Jo-Philipp Wich
04069fde19 uhttpd: update to latest Git HEAD
2ee323c file: poke ustream after starting deferred program

Signed-off-by: Jo-Philipp Wich <jo@mein.io>
2020-02-12 18:01:13 +01:00
Jason A. Donenfeld
cb17d7aed7 wireguard-tools: bump to 1.0.20200206
* wg-quick: android: split uids into multiple commands

Newer android's ndc implementations have limits on uid size, so we have to
break these into several lists.

* man: document dynamic debug trick for Linux

This comes up occasionally, so it may be useful to mention its
possibility in the man page. At least the Arch Linux and Ubuntu kernels
support dynamic debugging, so this advice will at least help somebody. So that
you don't have to go digging into the commit, this adds this helpful tidbit
to the man page for getting debug logs on Linux:

 # modprobe wireguard && echo module wireguard +p > /sys/kernel/debug/dynamic_debug/control

* extract-{handshakes,keys}: rework for upstream kernel

These tools will now use the source code from the running kernel instead of
from the old monolithic repo. Essential for the functioning of Wireshark.

* netlink: remove libmnl requirement

We no longer require libmnl. It turns out that inlining the small subset of
libmnl that we actually use results in a smaller binary than the overhead of
linking to the external library. And we intend to gradually morph this code
into something domain specific as a libwg emerges. Performance has also
increased, thanks to the inliner. On all platforms, wg(8) only needs a normal
libc. Compile time on my system is still less than one second. So all in all
we have: smaller binary, zero dependencies, faster performance.

Packagers should no longer have their wireguard-tools package depend on
libmnl.

* embeddable-wg-library: use newer string_list
* netlink: don't pretend that sysconf isn't a function

Small cleanups.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
2020-02-09 21:25:51 +01:00
Jo-Philipp Wich
766e778226 hostapd: remove erroneous $(space) redefinition
The $(space) definition in the hostapd Makefile ceased to work with
GNU Make 4.3 and later, leading to syntax errors in the generated
Kconfig files.

Drop the superfluous redefinition and reuse the working $(space)
declaration from rules.mk to fix this issue.

Fixes: GH#2713
Ref: https://github.com/openwrt/openwrt/pull/2713#issuecomment-583722469
Reported-by: Karel Kočí <cynerd@email.cz>
Suggested-by: Jonas Gorski <jonas.gorski@gmail.com>
Tested-by: Shaleen Jain <shaleen@jain.sh>
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
2020-02-08 11:45:33 +01:00
Jason A. Donenfeld
71de48bd37 wireguard: bump to 0.0.20200205
* compat: support building for RHEL-8.2
* compat: remove RHEL-7.6 workaround

Bleeding edge RHEL users should be content now (which includes the actual
RedHat employees I've been talking to about getting this into the RHEL kernel
itself). Also, we remove old hacks for versions we no longer support anyway.

* allowedips: remove previously added list item when OOM fail
* noise: reject peers with low order public keys

With this now being upstream, we benefit from increased fuzzing coverage of
the code, uncovering these two bugs.

* netns: ensure non-addition of peers with failed precomputation
* netns: tie socket waiting to target pid

An added test to our test suite for the above and a small fix for high-load CI
scenarios.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
2020-02-05 21:56:02 +01:00
Jo-Philipp Wich
5f5ec7660c Revert "iwinfo: update to latest Git HEAD"
This reverts commit 96424c143d.

The commit changed libiwinfo's internal ABI which breaks a number of
downstream projects, including LuCI and rpcd-mod-iwinfo.

Signed-off-by: Jo-Philipp Wich <jo@mein.io>
2020-02-05 15:31:39 +01:00
David Bauer
96424c143d iwinfo: update to latest Git HEAD
eba5a20 iwinfo: add device id for BCM43602
a6914dc iwinfo: add BSS load element to scan result
bb21698 iwinfo: add device id for Atheros AR9287
7483398 iwinfo: add device id for MediaTek MT7615E

Signed-off-by: David Bauer <mail@david-bauer.net>
2020-02-04 20:14:47 +01:00
John Crispin
df773ead9a bcm4xxx: fix iwinfo behaviour
Signed-off-by: John Crispin <john@phrozen.org>
2020-02-04 07:48:09 +01:00
Kevin Darbyshire-Bryant
e481df07fa iptables: set-dscpmark follow upstreamimg attempt
I'm having another attempt at trying to getting the 'store dscp into
conntrack connmark' functionality into upstream kernel, since the
restore function (act_ctinfo) has been accepted.

The syntax has changed from 'savedscp' to 'set-dscpmark' since that
conforms more closely with existing functionality.

Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
2020-01-31 20:21:43 +00:00
Felix Fietkau
b3e86cbb4f hostapd: add back support for passing CSA events from sta/mesh to AP interfaces
Fixes handling CSA when using AP+STA or AP+Mesh
This change was accidentally dropped in commit 167028b75
("hostapd: Update to version 2.9 (2019-08-08)")

Signed-off-by: Felix Fietkau <nbd@nbd.name>
2020-01-29 12:25:10 +01:00
Jason A. Donenfeld
c2859bf126 wireguard: bump to 0.0.20200128
This fixes a few small oversights for the 5.5 compat layer.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
2020-01-28 22:33:40 +01:00
Felix Fietkau
03e9e4ba9e hostapd: unconditionally enable ap/mesh for wpa-cli
Without this change, wpa-cli features depend on which wpad build variant was
used to build the wpa-cli package

Signed-off-by: Felix Fietkau <nbd@nbd.name>
2020-01-28 14:38:43 +01:00
Sven Roederer
3519bf4976 hostapd: remove some bashisms
"[[" is a bash extension for test. As the ash-implementation is not
fully compatible we drop its usage.

Signed-off-by: Sven Roederer <devel-sven@geroedel.de>
[remove shebang, slightly facelift commit title/message]
Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
2020-01-26 22:03:00 +01:00
Sven Roederer
bad59fd51b 6in4/6in4.sh: remove some bashism (usage of [[)
"[[" is a bash extension for test. As the ash-implementation is not
fully compatible we drop its usage.

Signed-off-by: Sven Roederer <devel-sven@geroedel.de>
2020-01-26 22:02:51 +01:00
Sven Roederer
bc357aaa2b netifd/config.sh: remove some bashism (usage of [[)
"[[" is a bash extension for test. As the ash-implementation is not
fully compatible we drop its usage.

Signed-off-by: Sven Roederer <devel-sven@geroedel.de>
2020-01-26 22:02:39 +01:00
Jason A. Donenfeld
4576a753f2 wireguard-tools: bump to 1.0.20200121
* Makefile: remove pwd from compile output
* Makefile: add standard 'all' target
* Makefile: evaluate git version lazily

Quality of life improvements for packagers.

* ipc: simplify inflatable buffer and add fuzzer
* fuzz: add generic command argument fuzzer
* fuzz: add set and setconf fuzzers

More fuzzers and a slicker string list implementation. These fuzzers now find
themselves configuring wireguard interfaces from scratch after several million
mutations, which is fun to watch.

* netlink: make sure to clear return value when trying again

Prior, if a dump was interrupted by a concurrent set operation, we'd try
again, but forget to reset an error flag, so we'd keep trying again forever.
Now we do the right thing and succeed when we succeed.

* Makefile: sort inputs to linker so that build is reproducible

Earlier versions of make(1) passed GLOB_NOSORT to glob(3), resulting in the
linker receiving its inputs in a filesystem-dependent order. This screwed up
reproducible builds.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
2020-01-24 08:21:04 +01:00
Jason A. Donenfeld
ec13b34118 wireguard: bump to 0.0.20200121
* Makefile: strip prefixed v from version.h

This fixes a mistake in dmesg output and when parsing the sysfs entry in the
filesystem.

* device: skb_list_walk_safe moved upstream

This is a 5.6 change, which we won't support here, but it does make the code
cleaner, so we make this change to keep things in sync.

* curve25519: x86_64: replace with formally verified implementation

This comes from INRIA's HACL*/Vale. It implements the same algorithm and
implementation strategy as the code it replaces, only this code has been
formally verified, sans the base point multiplication, which uses code
similar to prior, only it uses the formally verified field arithmetic
alongside reproducable ladder generation steps. This doesn't have a
pure-bmi2 version, which means haswell no longer benefits, but the
increased (doubled) code complexity is not worth it for a single
generation of chips that's already old.

Performance-wise, this is around 1% slower on older microarchitectures,
and slightly faster on newer microarchitectures, mainly 10nm ones or
backports of 10nm to 14nm. This implementation is "everest" below:

Xeon E5-2680 v4 (Broadwell)

armfazh: 133340 cycles per call
everest: 133436 cycles per call

Xeon Gold 5120 (Sky Lake Server)

armfazh: 112636 cycles per call
everest: 113906 cycles per call

Core i5-6300U (Sky Lake Client)

armfazh: 116810 cycles per call
everest: 117916 cycles per call

Core i7-7600U (Kaby Lake)

armfazh: 119523 cycles per call
everest: 119040 cycles per call

Core i7-8750H (Coffee Lake)

armfazh: 113914 cycles per call
everest: 113650 cycles per call

Core i9-9880H (Coffee Lake Refresh)

armfazh: 112616 cycles per call
everest: 114082 cycles per call

Core i3-8121U (Cannon Lake)

armfazh: 113202 cycles per call
everest: 111382 cycles per call

Core i7-8265U (Whiskey Lake)

armfazh: 127307 cycles per call
everest: 127697 cycles per call

Core i7-8550U (Kaby Lake Refresh)

armfazh: 127522 cycles per call
everest: 127083 cycles per call

Xeon Platinum 8275CL (Cascade Lake)

armfazh: 114380 cycles per call
everest: 114656 cycles per call

Achieving these kind of results with formally verified code is quite
remarkable, especialy considering that performance is favorable for
newer chips.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
2020-01-24 08:21:04 +01:00