The vn_start_write() call there is already interruptible. Check for
user signals before restarting due to ERELOOKUP, or after failed
vn_start_write(). Note that vn_start_write(V_XSLEEP | V_PCATCH)
does not check for signals if not sleeping.
PR: 295826
Reviewed by: markj
Tested by: pho
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Differential revision: https://reviews.freebsd.org/D57453
Prior to this commit, we'd compute the page tables and have the last
entries point to the staging area. We'd then add some more metadata to
the image and boot. This assumed the staging area didn't need to move
for this last bit of data.
However, if we go over the staging limit, when we copyin new data, we
grow the staging area, usually be moving it to a lower address. This
overage usually happens when we're loading modules and so things work
out nicely. Sometimes we're close to the limit, and we need to do this
growing inside bi_load, after we've computed the page table, making the
page table wrong, and the code we jump to random rather than the btext
routine we normally start at.
To fix this, move computation of the table (but not its allocation) to
after bi_load, but before we call the trampoline.
This problem was most observed when loading microcode for many peole,
but Gleb reproduced the error with a set of modules that didn't include
ucode.
This bug hunt was greatly assisted by Claude who looked at the crash
from the EFI boot loader and surmised that we weren't jumping to the
code we thought we were jumping to. After inspecting the code, I asked
claude how corruption could happen (I thought overwriting the page
table), but claude notice the possibility that staging might change
after we computed the page table, and this fix is the result. Claude
didn't suggest a diff, but did provide many helpful clues that lead me
to this fix.
PR: 294630
Reviewed by: kib (prior version)
Sponsored by: Netflix
MFC After: insta per re@ request
Differential Revision: https://reviews.freebsd.org/D57462
This change refactors fatal error delivery via APEI and prints more info:
- Makes the NMI handler call into the ge handler to establish a common
code flow, no matter how the error is delivered
- Adds the FRU to the panic string so as to provide more information than
just "APEI Fatal Hardware Error!" such as
"APEI Fatal Hardware Error: PcieError"
- Prints more details about fatal pcie errors. Note that we skip acquiring
Giant on fatal errors
- Hexdumps the full GED data on fatal errors, so as to facilitate
offline data analysis
Reviewed by: imp
Sponsored by: Netflix
Differential Revision: https://reviews.freebsd.org/D57417
This reduces the size of a trigger entry from 24 bytes to 16 (or from
12 bytes to 8 on 32-bit) with no additional complexity.
MFC after: 1 week
Sponsored by: Klara, Inc.
Reviewed by: kevans, emaste
Differential Revision: https://reviews.freebsd.org/D57464
It was previously not possible to poll() or select() on the trigger
device, which made implementing proper signal handling in auditd
difficult.
MFC after: 1 week
Sponsored by: Klara, Inc.
Reviewed by: kevans, markj
Differential Revision: https://reviews.freebsd.org/D57457
powerpc is still relevant for lib32. powerpcspe cannot use lib32 so
remains removed.
Reported by: Minsoo Choo <minsoo@minsoo.io>
Reviewed by: Minsoo Choo <minsoo@minsoo.io>
Sponsored by: The FreeBSD Foundation
Fixes: 2ea49bb595 ("arch.7: Remove 32-bit powerpc from tables")
Differential Revision: https://reviews.freebsd.org/D57473
Add ATF regression tests for previously-fixed fts(3) bugs:
- PR 45723: directory with read but no execute is traversed via
FTS_DONTCHDIR fallback, not silently skipped
(commit 1e03bff7f2)
- PR 196724: FTS_SLNONE must not be returned for a non-symlink;
time-bounded race test runs for 1 second with concurrent
file creation/deletion
(commit bf4374c54589)
- PR 262038: readdir(2) errors produce FTS_DNR with fts_errno
set, not silently treated as end-of-directory
(commit 0cff70ca66)
- SVN r246641: normal traversal works correctly with O_DIRECTORY
fix in fts_safe_changedir()
(commit f9928f1705ee)
- SVN r261589: no crash when tree modified during traversal;
time-bounded race test runs for 1 second with concurrent
file creation/deletion
(commit c6d38f088e5c)
Sponsored by: Google LLC (GSoC 2026)
Reviewed by: asomers
MFC after: 2 weeks
Pull Request: https://github.com/freebsd/freebsd-src/pull/2257
There's no point in acquiring a hold if the proc lock isn't going to be
dropped.
Reviewed by: kib
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D57468
Most 32-bit architecture support has been deprecated for FreeBSD 16.
Reviewed by: des
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D57472
Most architectures have the same signedness for char and wchar_t, but
powerpc has the unusual combination of unsigned char and signed wchar_t.
Reviewed by: des, jhb, markj
Sponsored by: The FreeBSD Foundation
At present, the exception syndrome register (ESR) is not accessible via
the user context on FreeBSD AArch64, which makes it difficult to
determine the cause of an exception. For example, a signal handler might
get a `SIGSEGV` with `SEGV_ACCERR`, but it can't know whether that
occurred due to a bad read or write. The change implemented here
includes ESR in `struct __mcontext`, thereby allowing access via
`ucontext_t` for use by signal handlers, sanitizers, debuggers, etc.
Note that this addition consumes one of the seven spare 64-bit slots in
`struct __mcontext`.
Signed-off-by: Alex Arslan <ararslan@comcast.net>
Reviewed by: andrew
Pull Request: https://github.com/freebsd/freebsd-src/pull/2053
The `ksiginfo_t` flag `KSI_TRAP` is set both for exceptions and when
copying between userspace and the kernel fails. In the latter case, the
exception syndrome register as captured in `struct trapframe` won't be
valid. That means we can't use `KSI_TRAP` to determine whether `tf_esr`
is valid. This motivates the addition of a new flag, here called
`KSI_EXCEPT`, for specifically identifying signals caused by exceptions.
It is added to `ksi_flags` via `trapsignal`.
Signed-off-by: Alex Arslan <ararslan@comcast.net>
Reported by: andrew
Pull Request: https://github.com/freebsd/freebsd-src/pull/2053
bhyve's -p allows to pin guest's virtual CPU vcpu to hostcpu, however
this becomes very tedious work when you have to pin more than a single
CPU.
This allows to pass a range to -p, e.g. -p 0-3:4-7 which will pin the
cpus 0:4, 1:5, 2:6, 3:7. The ranges must be equal and the CPU numbers
must be ascending.
Sponsored by: Armenian Bioinformatics Institute
Reviewed by: corvink, markj
Tested by: bnovkov
MFC after: 3 weeks
Differential Revision: https://reviews.freebsd.org/D54937
On startup, check that /dev/null exists and is a character device.
Otherwise, one of two things will happen: either /dev is a writable
directory and we will immediately create /dev/null as a regular file
and dump garbage into it, or it does not and we will spit out a stream
of error messages about failing to create /dev/null.
PR: 295782
MFC after: 1 week
Reviewed by: jhb, emaste
Differential Revision: https://reviews.freebsd.org/D57447
The common case for etcupdate is to run it after building and installing
the world, in which case we already have an object directory to draw on.
Add a -b option to turn nobuild off (opposite of -B), and turn nobuild
on by default.
MFC after: 1 week
Reviewed by: jhb, emaste
Differential Revision: https://reviews.freebsd.org/D57306
- Remove fts_check_debug() which is only needed by test cases
that use the fts_test() helper; fts_children_test.c does not
call fts_test()
- Remove fts_lexical_compar where traversal order does not matter
for the test result
- Drop fts_test.h and copy fts_lexical_compar locally
- Pull up NULL argument onto same line as fts_open
Fixes: e624417db8 ("lib/libc/tests/gen: add fts_children() tests")
Sponsored by: Google LLC (GSoC 2026)
Reviewed by: des, asomers
MFC after: 2 weeks
Pull Request: https://github.com/freebsd/freebsd-src/pull/2260
Per review by des@:
- Remove fts_check_debug() which has no effect in fts_set_test.c, and
remove unused #include "fts_test.h"
- Add errno = 0 before fts_read() so errno check is meaningful
- Remove fts_lexical_compar where traversal order does not matter
- Change int boolean flags to bool throughout
- Use !saw_inside instead of saw_inside == false
- Move fts_set_clientptr/fts_get_clientptr test inside fts_read()
loop so the pointer is exercised during active traversal
Fixes: 940142d610 ("lib/libc/tests/gen: add fts_set() tests")
Sponsored by: Google LLC (GSoC 2026)
Reviewed by: des, asomers
MFC after: 2 weeks
Pull Request: https://github.com/freebsd/freebsd-src/pull/2258
Commit 72e57bc264 added support for striping to the pNFS
server when using the Flexible File layout.
It is configured globally via the following sysctls:
vfs.nfsd.pnfsstripeunit - Size (in bytes) of a stripe
vfs.nfsd.pnfsstripecnt - # of DSs to stripe across
This patch allows the above settings to be overridden
on a per-MDS exported file system basis. With this
patch, a stripeunit can optionally be specified for
the MDS file system listed when one is listed after
a '#' in the "-p" nfsd argument. This is done by
appending "@NNN", where NNN is the stripeunit, in bytes.
The current syntax implies that neither a '#' nor '@'
can be in the MDS mount point's directory path.
This patch does not affect current pNFS server
configurations (of which there appears to be very
few, anyhow).
- Update /etc/protocols with IANA list updated 2026-03-09.
- Document that 240 (pfsync) is not assigned by IANA.
- Document deprecated protocols.
PR: 295739
Reviewed by: des
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D57445
In the output path where we are sending a frame to the driver mark it
if it should be part of an A-MPDU based on its tid, type, and whether
net80211 thinks that we are in the right state for this.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
Fill in more details for lkpi_ic_addba_request(), lkpi_ic_addba_response(),
and lkpi_ic_addba_response_timeout().
Migrate the ltxq flags seen_dequeue and stopped to a bitfield and add %b
support to log messages. This seemed the better approach after needing
an additional stop field for BA while we have to hold packets from being
transmitted.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
Adjust the pure debug tracing to no longer be under HT and add more
to the other functions we are interested in, so we can follow the
calls more easily.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
We currently do not support rate control in LinuxKPI. As more drivers
and chipsets gain suppport for higher throughput add a note under
bootverbose if we hit one of these cases (currently only older iwlwifi
chipsets).
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
Implement ieee80211_start_tx_ba_session() as a start for rtw8x (and
select mt76 chipsets) to support more throughput.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
Add (*link_sta_rc_update), (*set_bitrate_mask), and
(*sta_set_decap_offload) mac80211 driver downcalls in preparation
for further work.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
Add support for the 'mounts' cloud-config key which configures
mount points by appending entries to /etc/fstab and creating
the corresponding directories.
Add support for adding the instance hostname to /etc/hosts on the
127.0.0.1 and ::1 localhost lines, matching cloud-init's default
behaviour (manage_etc_hosts: true).
create a revolve_hostname helper to avoid code duplucation.
Add support for the 'bootcmd' cloud-config directive, which allows
running commands very early in the boot process, before the hostname
is set and before the network is configured.
- nuageinit: bootcmd() function follows the same pattern as runcmd(),
writing commands to /var/cache/nuageinit/bootcmds instead of runcmds.
It is the first entry in the pre_network_calls table.
- rc.d/nuageinit: execute /var/cache/nuageinit/bootcmds immediately
after /usr/libexec/nuageinit completes, before unmounting the config
drive. This ensures bootcmd runs before NETWORKING per cloud-init spec.
The unquote() function took a const char * parameter but modified the
string in-place (removing quote characters). Change the parameter to
char * and update callers that passed const char * to cast explicitly.
The 'name' variable could be left uninitialized if neither the
positional argument nor -n is supplied, leading to undefined
behavior when passed to getgroup().
Some upstream result files introduced in file 5.47 (e.g., bgcode.result)
lack a trailing newline, causing the contrib_file_tests ATF test to
fail with "cmp: EOF on bgcode.result". Generate normalized copies
of the expected results and install those instead.
MFC after: 3 days
Fixes: e949ce9dc0
* pipebuf is a size but is listed as a count
PR: 295623
MFC after: 1 week
Fixes: f54f41403d ("usr.bin/limits: support RLIMIT_PIPEBUF")
Reviewed by: kib
Differential Revision: https://reviews.freebsd.org/D57456
Commit 4d80d4913e fixed a long standing bug in the recovery
code. However. glebius@ reported seeing multiple
recovery cycles with this patch during an NFSv4.1/4.2
server reboot.
This commit should minimize the risk of multiple
recovery cycles.
PR: 294925
Reported by: Jov <amutu@amutu.com>
MFC after: 2 weeks
Fixes: 4d80d4913e ("nfs: Fix argument typo to avoid a crash")
Konstantin Belousov