lagg: Avoid dropping locks when starting the interface
The init routine of a lagg(4) interface will not change during the whole
lifecycle. So we can call lagg_init() directly instead of through the
function pointer. Well, that requires a drop and pickup lock, which
unnecessarily expose a small race window. Refactor lagg_init() into
lagg_init_locked() and call the later one to avoid that.
Meanwhile, delay updating the driver managed status until after the
interface is really ready.
Reviewed by: markj
MFC after: 5 days
Differential Revision: https://reviews.freebsd.org/D55198
diff: Improve directory loop detection
When we're done processing a directory, remove its entry from the tree
of visited inodes, ensuring that we only report a loop when we encounter
a descendant-to-ancestor link, not when we encounter a cousin-to-cousin
or sibling-to-sibling link.
MFC after: 1 week
Reported by: Bakul Shah <bakul at iitbombay.org>
Sponsored by: Klara, Inc.
Reviewed by: kevans
Differential Revision: https://reviews.freebsd.org/D55248
install: Expect EINTR while copying
Both copy_file_range() and read() / write() in our fallback loop can be
interrupted before copying anything at all, in which case it returns -1
and sets errno to EINTR. If that happens, we should retry, not fail.
While here, drop the size argument from copy() (we always want to copy
the entire file anyway) and add test cases which exercise the metalog
and digest functionality.
PR: 293028
MFC after: 1 week
Reviewed by: kevans
Differential Revision: https://reviews.freebsd.org/D55168
cp: Expect EINTR while copying
Both copy_file_range() and copy_fallback() can be interrupted before
they have read anything at all, in which case they return -1 and set
errno to EINTR. If that happens, we should retry, not fail.
PR: 293028
MFC after: 1 week
Reviewed by: kevans
Differential Revision: https://reviews.freebsd.org/D55167
amd64: add LASS support
In short, LASS enforces all kernel memory accesses to have bit 63 set to
1, and all userspace accesses have bit 63 set to 0. Violations of these
rules cause #GP. There are natural loopholes, like SMAP with rflags.AC=1
allows kernel to access userspace.
Enablement is simple, we need to set CR4.LASS bit on all CPUs. There
are complications when kernel has to execute code at low addresses, e.g.
for la57 trampoline, or calling into EFI RT. The patch turns CR4.LASS
off around these regions.
LASS is officially documented in SDM, since at least rev. 085, October
2024. Tested in simics.
Reviewed by: markj
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Differential revision: https://reviews.freebsd.org/D55218
freebsd-update: Error for -b basedir without UNAME_r set
freebsd-update sets the currently running release from UNAME -r, which
can be overridden via the --currently-running commandline option (or by
setting UNAME_r in the environment). This may be invalid if -b is used
to specify a basedir other than /, so error out if -b is specified
without setting the currently running version.
PR: 283229
Reviewed by: cperciva
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D48016
lagg: Remove the member pr_num from struct lagg_proto
It is set but never used. Remove it to avoid confusion and save a
little space.
While here, use designated initializers to initialize the LAGG protocol
table. That improves readability, and it will be safer to initialize the
table if we introduce new protocols in the future.
No functional change intended.
Reviewed by: glebius
MFC after: 5 days
Differential Revision: https://reviews.freebsd.org/D55124
(cherry picked from commit 5ba503fc2cabc1a614997f102ace671d996bcc53)
qlnxe: Overhaul setting the multicast MAC filters
When operating the multicast MAC filters, the current usage of
ECORE_FILTER_ADD and ECORE_FILTER_REMOVE are rather misleading.
ECORE_FILTER_ADD reads "adding new filter", but it actually removes
any existing filters and then addes a new one. ECORE_FILTER_REMOVE
reads "removing a filter", but it actually removes all filters.
Let's use ECORE_FILTER_REPLACE and ECORE_FILTER_FLUSH instead to
avoid confusion.
In the current implementation, only one MAC address is passed to
ecore_sp_eth_filter_mcast() and any previously installed filters are
removed, hence it breaks the multicast function. That can be observed
via either assigning new IPv6 addresses to the interface or putting
the interface as a member of lagg(4) interface with LACP aggregation
protocol. Fix that by calculating the multicast filter bins directly
from multicast MAC addresses and replace the filters every time
the bins changes.
[20 lines not shown]
lagg: Make the none protocol a first-class citizen
All the other protocols have corresponding start and input routines,
which are used in the fast path. Currently the none protocol is
treated specially. In the fast path it is checked to indicate whether
a working protocol is configured. There are two issues raised by this
design:
1. In production, other protocols are commonly used, but not the
none protocol. It smells like an overkill to always check it in the
fast path. It is unfair to other commonly used protocols.
2. PR 289017 reveals that there's a small window between checking the
protocol and calling lagg_proto_start(). lagg_proto_start() is possible
to see the none protocol and do NULL deferencing.
Fix them by making the none protocol a first-class citizen so that it
has start and input routines just the same as other protocols. Then we
can stop checking it in the fast path, since lagg_proto_start() and
[15 lines not shown]
qlnxe: Allow tapping the TX packets
Currently only the packets in the RX path can be captured by tcpdump
as the ETHER_BPF_MTAP call in the TX path is missing. Add it so that
packets in both directions can be captured.
PR: 290973
Reviewed by: kbowling
MFC after: 5 days
Differential Revision: https://reviews.freebsd.org/D54891
(cherry picked from commit 968647502ec21464ad3aecc7577ff0e8dfd41693)
qlnxe: Refactor setting the promiscuous and allmulti mode
There are two entry points to set the promiscuous and allmulti mode.
One is ioctl, and another is the init routine. Given they share almost
the identical logic, refactor a little to make the code more clear.
While here, for the ioctl, translate the error to EINVAL to avoid
confusing the net stack.
Reviewed by: kbowling
MFC after: 5 days
Differential Revision: https://reviews.freebsd.org/D54890
(cherry picked from commit 45b1718fadae7d56051ba04ef9d7a175a602a226)
qlnxe: Avoid reinitializing the interface when it is already initialized
qlnx_init_locked() unconditionally uninitialize the interface thus is
actually reinitializing the interface. Well the init routine qlnx_init()
is to initialize the interface by net stack when assigned with the first
inet or inet6 address. The ioctl SIOCSIFADDR for the first inet6 address
is handled by ether_ioctl() thus the interface is reinitialized no matter
it was initialized or not.
Add a driver status check for that to avoid reinitializing. Further plan
is removing SIOCSIFADDR ioctl from the driver and let ether_ioctl() handle
it.
Reviewed by: kbowling
MFC after: 5 days
Differential Revision: https://reviews.freebsd.org/D54887
(cherry picked from commit c10e6bc0f0079e90cb484323ad71d437f1882422)
qlnxe: Prevent potential concurrency between ioctls
The driver-managed status flags should be lock protected to be touched.
Also this can serialize ioctls those check the IFF_DRV_RUNNING status.
Reviewed by: kbowling
MFC after: 5 days
Differential Revision: https://reviews.freebsd.org/D54886
(cherry picked from commit 0df8a998a9fe28af659cb401c537c6d785e55f81)
qlnxe: Let ether_ioctl() handle SIOCSIFADDR ioctl
Since the change [1], the init routine qlnx_init() works as intended.
Let ether_ioctl() handle SIOCSIFADDR to simplify the code.
Combined with the change [1], this shall be a better fix for PR 287445.
[1] c10e6bc0f007 qlnxe: Avoid reinitializing the interface when it is already initialized
PR: 287445
Reviewed by: kbowling
MFC after: 5 days
Differential Revision: https://reviews.freebsd.org/D54888
(cherry picked from commit 4012b63889e40bb877bc0e4c8da1792bce472c08)
qlnxe: Remove a pointless copy back from the link-layer address
On ifnet attaching, ether_ifattach() makes the link-layer address by
shadow copying the ha->primary_mac. Well, the link-layer address will
not be altered during attaching, thus it is pointless to copy it back.
No functional change intended.
Reviewed by: kbowling
MFC after: 5 days
Differential Revision: https://reviews.freebsd.org/D54883
(cherry picked from commit 4ac3081b282800158df7abe93f307d76e1b5b808)
qlnxe: Fix setting the unicast MAC filter of RX path
When an Ethernet interface is added to lagg(4) as a child interface, its
type, aka if_type, is changed from IFT_ETHER to IFT_IEEE8023ADLAG. Well
changing the link-layer address of the lagg(4) interface will be
propagated to all child interfaces, hence the drivers of child interfaces
shall not presume the type of the interface will not be changed.
Meanwhile, on initializing, an ifnet has been fully attached and it is
guaranteed to have non-null link-layer address so stop NULL checking for
it.
Reviewed by: kbowling
Fixes: 792226e53023 qlnxe: Allow MAC address override
MFC after: 5 days
Differential Revision: https://reviews.freebsd.org/D54885
(cherry picked from commit f250852c9a0c1021c3be4b498e27cfc7b42a81db)
qlnxe: Avoid memcpy with same source and destination
In case the device is VF, qlnx_get_mac_addr() returns ha->primary_mac
hence it ends up memcpy with same source and destination. Refactor
slightly to avoid that.
Reviewed by: kbowling
MFC after: 5 days
Differential Revision: https://reviews.freebsd.org/D54884
(cherry picked from commit 3aeeedc7e0dc231c16406ff64f4a08a716964c40)
release: Turn off debugging in pkg(8)
Running `pkg -d` in pkg-stage.sh results in multiple GB of network
traffic being written into the log files, which is less than helpful
when it comes to tracking down build failures. Remove the -d flag.
MFC after: 5 days
X-MFC-note: The code in 15 has diverged from 14, but the flag is
there, just in a different place.
(cherry picked from commit 18721be356043f6749a6e2470bc9f7351c450c6e)
iwx: migrate to new net80211 encryption key API
Migrate to the new encryption key API rather than poking at the
key struct directly.
Locally tested:
* STA mode - iwx0: <Wi-Fi 6 AX210> mem 0x84c00000-0x84c03fff at device 0.0 on pci4
Differential Revision: https://reviews.freebsd.org/D54463
run: rum: migrate to new net80211 encryption key API
Migrate to the new encryption key API rather than poking at the
key struct directly.
Locally tested:
* STA mode - run0: MAC/BBP RT2860 (rev 0x0101), RF RT2820 (MIMO 2T2R), address 00:0e:2e:e5:ae:3a
Differential Revision: https://reviews.freebsd.org/D54462
net80211: clean up time comparison in HT code
There's at least one place where the time comparison is wrong, leading
to some bad beahaviour around deciding to try and establish/reestablish
AMPDU TX sessions.
Start addressing these!
Differential Revision: https://reviews.freebsd.org/D54390
