tcp: avoid call to soisconnected() on transition to ESTABLISHED
This call existed since pre-FreeBSD times, and it is hard to understand
why it was there in the first place. After 6f3caa6d815 it definitely
became necessary always and commit message from f1ee30ccd60 confirms that.
Now that 6f3caa6d815 is effectively backed out by 07285bb4c22, the call
appears to be useful only for sockets that landed on the incomplete queue,
e.g. sockets that have accept_filter(9) enabled on them.
Provide a new TCP flag to mark connections that are known to be on the
incomplete queue, and call soisconnected() only for those connections.
stand: i386: take into account disk sector size for blk calculation
disk_blocks assumes BIOSDISK_SECSIZE, but the media may not be using
it. In particular, bioscd on Parallels presents a 2K sector size, so
we end up with a short disk_blocks and subsequent validation fails when
trying to read /boot/lua.
Andrew Turner [Wed, 7 Sep 2022 11:43:16 +0000 (12:43 +0100)]
Use the cached dc zva length in arm64 memset
On boot we cache the length the 'dc zva' instruction will zero. Use
this in the memset function to decide when to use it. As the cached
value is in .bss it will be zero on boot so memset is safe to use
before the value has been read.
Andrew Turner [Wed, 7 Sep 2022 11:41:13 +0000 (12:41 +0100)]
Import the arm64 Cortex Strings memset into the kernel
Bring in the last version of the optimized memset from the Cortex
Strings library that didn't use the VFP registers. While here clean up
to use the ENTRY/EDIT macros.
Andrew Turner [Wed, 7 Sep 2022 09:51:21 +0000 (10:51 +0100)]
Update the arm64 kernel memcpy/memmove
Bring in the latest Arm Optimized Routines memcpy/memmove into the
arm64 kernel. As these functions have been merged in the current
version remove the now unneeded memmove.S.
routing: add abitity to set the protocol that installed route/nexthop.
Routing daemons such as bird need to know if they install certain route
so they can clean it up on startup, as a form of achieving consistent
state during the crash recovery.
Currently they use combination of routing flags (RTF_PROTO1) to detect
these routes when interacting via route(4) rtsock protocol.
Netlink protocol has a special "rtm_protocol" field that is filled and
checked by the route originator. To prepare for the upcoming netlink
introduction, add ability to record origing to both nexthops and
nexthop groups via <nhop|nhgrp>_<get|set>_origin() KPI. The actual
calls will be used in the followup commits.
In 8db2e8fd16c4 ("Remove the secondary_stacks array in arm64 [...]"),
bootstacks was setup to be allocated dynamically. While this is
generally how x86 does it, it inadvertently shrunk each boot stack from
KSTACK_PAGES pages to a single page.
Resize these back up to the expected size using the kstack_pages
tunable, as we'll need larger stacks with upcoming sanitizer work.
Reviewed by: andrew, imp, markj
Fixes: 8db2e8fd16c4 ("Remove the secondary_stacks array [...]")
Sponsored by: Juniper Networks, Inc.
Sponsored by: Klara, Inc.
Differential Revision: https://reviews.freebsd.org/D36475
Doug Moore [Thu, 8 Sep 2022 04:46:19 +0000 (23:46 -0500)]
rb_tree: reduce duplication in balancing code
Change RB_INSERT_COLOR and RB_REMOVE_COLOR so that the blocks of code
that are identical except for left and right being exchanged are made
only one block with a variable to indicate left- or right-handedness.
Rename RB macros so that those not intended for external use begin
with an underscore.
Add comments to the balancing code so that another might understand it.
Doug Moore [Thu, 8 Sep 2022 02:40:05 +0000 (21:40 -0500)]
rb_tree: test rank balance
With _RB_DIAGNOSTIC defined, provide an RB_RANK method to compute the
rank of a node in an rb-tree, if the subtree rooted at that node is
rank-balanced, and -1 otherwise.
In rb_test, rewrite a bit to avoid malloc/free and nondeterministic
running times because of randomness. Allocate all the nodes on the
stack, and shuffle a set of keys to get randomness for the testing.
Michael Gmelin [Wed, 7 Sep 2022 16:56:49 +0000 (18:56 +0200)]
stand: Unbreak FAT32 in loader
This corrects an issue introduced in b4cb3fe0e39a3, where a freshly
allocated `DOS_FS` structure would not be initialized properly before
use in `dos_open`.
In case of FAT32 file systems, this would leave `fs->dirents`
uninitialized and - depending on its content and due to checks in
`parsebs` - prevent mounting the file system successfully.
This particularily impacted the EFI loader, as it was sometimes not
able to read files from a FAT32-formatted EFI partition, including
LoaderEnv (`/efi/freebsd/loader.env`).
USB: add 2 (3) more device IDs for rtw88-usb devices and a quirk
Add DLink DWA-182 rev D1 and generic Realtek RTW8821CU entry found on
a Tenda U10 USB WLAN Stick, AC 650 Mbps (and possibly more devices).
The latter first presents itself as a CD device with Windows drivers
(useless on FreeBSD) first so add a quirk for that we get the wireless
device right away.
MFC after: 2 weeks
While here sort some other Realtek entries by DeviceID.
Update the mac80211 ops KPI for (*conf_tx), (*assign_vif_chanctx),
(*unassign_vif_chanctx), (*start_ap), and (*stop_ap), as well as
ieee80211_beacon_get_tim() and ieee80211_beacon_get_template().
Update in-tree drivers iwlwifi and rtw88 accordingly based on upstream
changes (as well as out-of-tree ones). This was triggered by trying to
synchronize more drivers to a common state.
There are drivers directly accessing napi->state testing for bits
(NAPI_STATE_SCHED encountered so far). Rename the internal _flags
struct field to state and expose our internal state flag bits along
with the one official aliased.
As I left in a comment, I wished Linux would hide these accesses
behind inline functions or by other means and not public expose
the implementation details.
A (so far out-of-tree) driver update needs
request_partial_firmware_into_buf(). Given we load the full .ko file
using firmware(9) just do that and copy the requeste data into the
buffer (rather than poissibly only reading portions of the firmware
file).
net80211: update (parts of) Action field categories
Update to the full list of action field category values based on
802.11-2020 with 11ax extensions and remove one value Reserved nowadays.
While here annotate them with descriptions.
Also add Spectrum Management and Radio Measurement action field values
needed in LinuxKPI.
No functional changes.
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Ed Maste [Wed, 7 Sep 2022 14:12:37 +0000 (10:12 -0400)]
Cirrus-CI: move QEMU pkg installation to test script
Occasionally the QEMU package fails to build and isn't available on
-CURRENT using the Latest package set. Move the package installation
to the test script, so that if the package isn't available we at least
still perform a build test.
Code setting it was removed in:
commit 325fab802e1f40c992141f945d0788c0edfdb1a4
Author: Eric van Gyzen <vangyzen@FreeBSD.org>
Date: Tue Dec 4 23:46:43 2018 +0000
Albert Jakieła [Tue, 6 Sep 2022 08:33:34 +0000 (10:33 +0200)]
mvneta: Use clkgen API on armv7
With addition of clkgen devices to the Armada38x we no longer
need to rely on get_tclk() to get the device frequency.
Leave it as a fallback, just in case.
Reviewed by: manu
Obtained from: Semihalf
Differential Revision: https://reviews.freebsd.org/D36465
Albert Jakieła [Tue, 5 Jul 2022 15:01:02 +0000 (17:01 +0200)]
Add ARMADA38X clkgen driver
This patch introduces clkgen driver for Armada38x SoCs.
Clock topology consists of single coreclk which supplies
clock signal to CPU cores and peripherials.
Reviewed by: manu
Obtained from: Semihalf
Differential Revision: https://reviews.freebsd.org/D36453
It is stored in the clock-frequency property.
In case of failure, fallback to the harcoded value stored in the
compat data.
Also, while here improve style.
Tested on LS1046ARDB and x86 PC.
Add VM_EXITCODE_IPI to permit returning unhandled IPIs to userland.
INIT and Startup IPIs are now returned to userland. Due to backward
compatibility reasons, a new capability is added for enabling
VM_EXITCODE_IPI.
vcpus could be restarted by the guest by sending an INIT SIPI SIPI
sequence to a vcpu. That's not supported by bhyve yet but it will be
supported in a future commit. So, create the vcpu threads only once on
startup to make restarting a vcpu easier.
Although originally socket was intended to use with ipfw(4) only, now
it also can be used with pf(4). On a kernel without packet filters,
it still can be used to inject traffic.
Python's socketmodule isn't able to construct sockaddr when it doesn't
recognize the address family. Until this is solved in python let the
tests use the compatibility way to create divert(4) socket.
The SIOCSTAT1 ioctl is only used in ip_auth and is unused in ip_state.
The ip_state version was likely added to support a new statistic yet
to be developed in ipfstat(8) or for some sample userspace application
(similar in fashion to the sample provided for authentication rules).
There is no need to report individual state hash table bucket lengths
to any future userspace application.
If needed for any future debugging purposes a DTrace probe would be a
better vehicle.
This unused ioctl in ip_stat results in a panic.
PR: 266124
Reported by: Robert Morris <rtm@lcs.mit.edu>
MFC after: 3 days
When something was found wrong with an inode the error message
was always "UNKNOWN FILE TYPE". This error is now used only when
the file type field is wrong. Other errors have their own messages:
"BAD FILE SIZE", "NEGATIVE FILE SIZE", "BAD SPECIAL-FILE RDEV",
"INVALID DIRECT BLOCK", and "INVALID INDIRECT BLOCK".
More complete information about the inode is also provided.
Andrew Turner [Tue, 6 Sep 2022 16:20:29 +0000 (17:20 +0100)]
Import the updated Arm Optimized Routines
The main changes this brings in are:
- Improves the performance of memcmp
- Adds SVE implementation of memcpy
- Uses the MTE version of some str* functions as they are faster
loader.efi(8): document slop control, amd64 nocopy, and amd64 fault commands
Reviewed by: imp
Discussed with: gbe (man pages)
English wording help by: rpokala
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
Differential revision: https://reviews.freebsd.org/D36435
ifconfig: print interface name on SIOCIFCREATE2 error
We have repeatedly gotten reports of unclassified SIOCIFCREATE2 errors
(usually "Device not configured"). This can happen if there is
configuration for interfaces in rc.conf which do not (yet) exist and
we try to configure. I can, e.g., provoke this by configuring wlan
interfaces with their physical interface not installed.
In order to cut support (guesswork) down print the name of the
interface to be configured with the error message.
Hopefully this will help us in the future to improve other configuration
or driver problems.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
Fix a couple of problems with printing of anchors, in particular recursive
printing, both of inline anchors and when requested explicitly with a '*'
in the anchor.
- Correct recursive printing of wildcard anchors (recurse into child anchors
rather than rules, which don't exist)
- Print multi-part anchor paths correctly (pr6065)
- Fix comments and prevent users from specifying multi-component names for
inline anchors.
tested by phessler
ok henning
Also fix the relevant pfctl test case to reflect the new (and now
correct) behaviour).
most programs in ports are looking for .pc files in order to get the
necessary information on how to compile and link against openssl.
The ports now also has a way to hide or force a path for pkgconf.
Providing .pc files along with openssl in base will allow (once all
the supported version of FreeBSD has it) so improve the framework to
deal with openssl in base vs openssl in ports (and libressl)
This will also greatly reduce the number of patches necessary to
workaround the build systems which only knows how to detect where
openssl is installed via pkgconf.
When calling shutdown, shutdown sends a signal to init and exits. This
causes a race condition for the waitpid function. If the signal wins the
race, wpid will be set to -1 and init calls death_single. If shutdown
wins the race, wpid will be set to the pid of the shutdown process and
the requested_transition will be ignored.
For one initialise vif->hw_queues later set in lkpi_80211_txq_tx_one()
for drivers using them. This may still need slightly more cleanup once
we get QUEUE_CONTROL-supporting drivers into the tree.
For the other rework and improve how we deal with tid/ac/txq in
lkpi_80211_txq_tx_one() and cleanup old comments and unused code.
This seems to reduce (remove) "Invalid TXQ id" reports from iwlwifi.
(The assumption is that the frame(s) triggering this WARN_ONCE or the
added FreeBSD specific logging were sent with lkpi_80211_mo_tx()).
Adjust the one logging from e674ddec0b4138274539587fe9336b577ff1242a
to IWL_DEBUG_TX so that now this is fixed it is also not always logged
anymore but leave it in case we need to further debug queues in the
future.
Tested by: pstef, Kevin Oberman (rkoberman gmail.com)
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
Continue what was started in 26a36948331bd08d9afaddfc0da724eacdb953dd
in iwlwifi and extend out internal implementation of
linuxkpi_ieee80211_get_tid() by an argument as to whether "no-QoS"
answers are acceptable. For the LinuxKPI ieee80211_get_tid() set
this to false as the Linux derived drivers seem to do extra checks
for the QoS-Data frame before acquiring the tid.
Add KASSERTs to enforce the extra argument.
This allows us to use the net80211 variant in LinuxKPI for other
means explicitly documenting that we do accept a IEEE80211_NONQOS_TID.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
Cy Schubert [Sun, 28 Aug 2022 12:48:25 +0000 (05:48 -0700)]
libexec/rc: Add var_run rc script
Users with a tmpfs /var/run will lose the directory tree state of
/var/run at reboot. This rc script will optionally (by default)
capture the state of the directory structure in /var/run prior to
shutdown and recreate it at system boot.
Alternatively a user can save the state of the /var/run directories
manually using service var_run save and disable the autosaving of
/var/run state using the var_run_autosave variable, for those
paranoid SSD users.
It is only there to check if the packet was reassembled,
relevant if we are forwarding. But if the packet originated
locally it could not have been reassembled, thus the flag is
redundant.
netinet6: call lle_event eventhandler after updating state
Fix nd6_na_input() eventhandler call: run eventhandler after lle
state transition.
Old behaviour (as seen by event handler):
* fe80::5054:ff:fe8c:63e9 dev vtnet0 lladdr 52:54:00:8c:63:e9 INCOMPLETE
New behaviour:
* fe80::5054:ff:fe8c:63e9 dev vtnet0 lladdr 52:54:00:8c:63:e9 REACHABLE
Ed Maste [Fri, 19 Aug 2022 14:43:19 +0000 (10:43 -0400)]
ssh: describe deprecated options in general in update doc
Rename "HPN" to more general "Retired patches." We handle two now-
removed patches the same way: to avoid breaking existing configurations
we accept, but ignore, the option.
Rick Macklem [Sun, 4 Sep 2022 20:09:33 +0000 (13:09 -0700)]
nfscl: Allow "nolockd" to work for NFSv4 mounts
Commit 40ada74ee1da modified the NFSv4.1/4.2 client so
that it would issue a DestroySession to the server when
all session slots are marked bad. This handles the
case where session slots get broken when "intr" or "soft"
NFSv4 fairly well.1/4.2 mounts are done.
There are two other cases where having an NFSv4.1/4.2
RPC attempt terminate without completion can leave
state in a non-determinate condition.
One is file locking RPCs. If the "nolockd" option is
used, this avoids file locking RPCs by doing locking
locally within the client.
The other is Open locks, but since all FreeBSD Open
locks are done with OPEN_SHARE_DENY_NONE, the locking
state for these should not be critical.
This patch enables use of "nolockd" for NFSv4 mounts,
so that it can be combined with "intr" and/or "soft",
making the latter more usable.
Use of "intr" or "soft" NFSv4 mounts are still not
recommended, but when combined with "nolockd" should
now work fairly well.
A man page update will be done as a separate commit.
Warner Losh [Sun, 4 Sep 2022 15:31:51 +0000 (09:31 -0600)]
stand/kboot: Add note about why we use MACHINE_ARCH here
Normally in the boot loader, we key off of MACHINE since that specifies
the kernel and the loader is very tuned to each type of MACHINE in
general. In this case, however, we're producing a Linux binary, with
Linux system calls encoded in it. These align better along the
MACHINE_ARCH axis of FreeBSD. For PowerPC the system calls are radically
different for each of our MACHINE_ARCHes, with only powerpc64 and
powerpc64le sharing the same numbers and memory layout. The same was
true about mips when it was in the tree. 32-bit arm uses the same
layout, however, for both armv6 and armv7 ports: that can be easily
shared in the unlikely event we support that in the future.