This marks __cxa_allocate_exception, __cxa_free_exception and
__cxa_init_primary_exception noexcept, to ensure compatibility with
libc++'s declarations.
This updates llvm, clang, compiler-rt, libc++, libunwind, lld, lldb and
openmp to llvm-project main llvmorg-18-init-18359-g93248729cfae, the
last commit before the upstream release/18.x branch was created.
Dimitry Andric [Sun, 14 Jan 2024 13:20:42 +0000 (14:20 +0100)]
Redo libc++ customizations
* Remove osreldate include because _LIBCPP_HAS_NO_GETS has disappeared
* Instead, add direct major __FreeBSD__ check for using ::gets declaration
* Mark EINTEGRITY values as FreeBSD customization
* Reformat _LIBCPP_TYPE_VISIBILITY_DEFAULT customization
Dimitry Andric [Sun, 7 Jan 2024 11:32:29 +0000 (12:32 +0100)]
Add two FreeBSD customizations to libc++
This is intended to better support ports usage of older clang versions:
* Add _LIBCPP_ENABLE_COMPILER_VERSION_CHECKS block around compiler
version checks, to avoid any warnings about support. This makes some
ports that use -Werror fall over.
* When using clang < 15.0, avoid using a type visibility attribute on
the std namespace, as older versions of clang do not support this.
[libcxx] Re-include <osreldate.h> in __config for FreeBSD
In 0a97720d0197 some changes were made to `__config` for assuming that
`__BYTE_ORDER__` is always present.
However, this deleted a `<osreldate.h>` include for FreeBSD, which is
required to get the value of `__FreeBSD_version`, and that is used later
in the file to determine whether `_LIBCPP_C_HAS_NO_GETS` needs to be
enabled.
Include `<osreldate.h>` just after the other includes used for feature
detection, to fix this.
Note that when FreeBSD 13 is EOLed, this can be removed, as then all
supported FreeBSD versions will no longer have `gets()`.
Warner Losh [Wed, 3 Apr 2024 17:26:17 +0000 (11:26 -0600)]
nvme: Add LPA bits
Add all the bits from the NVMe 2.0 base specification: CMD_EFFECTS to
indicate the commands and effects log page is supported, TELEMETRY to
indicate that the telemetry log pages and protocols are supported,
PERSISTENT_EVENTS to indicate the persistent event log is supported,
LOG_PAGES_PAGE to indicate that various log pages related to log page
and command support are supported: L0, L5, L12, and L13. and
DA4_TELEMETRY to indicate that the DA4 area is supported for telemetry
data.
arm64: Add a CPU reset hook instead of expecting PSCI
Some SoCs do not include a PSCI for power management and defer it to
something else instead. Add a CPU reset hook to account for this, and
use it in the psci driver.
Reviewed by: andrew
Obtained from: Juniper Networks, Inc.
Differential Revision: https://reviews.freebsd.org/D44535
Collin Funk [Sun, 11 Feb 2024 04:26:38 +0000 (20:26 -0800)]
which: Use size_t instead of ssize_t for pathlen
The "pathlen" variable is the return value of strlen(3) and is then
passed as an argument to malloc(3) and memcpy(3). The size_t type
matches the prototype for these functions. The size_t type is unsigned
so it can fit larger $PATH values than ssize_t. However, in practice
ssize_t should be larger enough so this change is just for clarity.
LinuxKPI: Remove the temporary variable fileid from the macro request_module
The variable fileid stores the result from kern_kldload() but never gets
used. Since the third parameter `*fileid` of kern_kldload() can be NULL,
this unused variable can be safely removed.
Michael Tuexen [Fri, 5 Apr 2024 15:47:03 +0000 (17:47 +0200)]
tcp rack: fix sending
In rack_output(), idle is used as a boolean variable. So don't use it
as an int and don't clear it afterwards.
This avoids setting idle to false, when it is not intended.
Reported by: olivier
Reviewed by: rrs, rscheff
MFC after: 3 days
Sponsored by: Netflix, Inc.
Differential Revision: https://reviews.freebsd.org/D44610
Mark Johnston [Fri, 5 Apr 2024 15:14:36 +0000 (11:14 -0400)]
tarfs: Implement VOP_BMAP
This lets tarfs provide readahead/behind hints to the VFS, which helps
memory-mapped I/O performance, important when running faulting in
executables out of a tarfs mount as one might if tarfs is used to back
the root filesystem, for example. The improvement is particularly
noticeable when the backing tarball is zstd-compressed.
The implementation simply returns the extent of the virtual block
containing the target offset, clamped by the maximum I/O size. This is
perhaps simplistic; it effectively just chooses values that would
correspond to a single VOP_READ call in tarfs_read_file().
The revert was not directly due to the attack (CVE-2024-3094):
our import process have removed the test cases and build scripts
that would have enabled the attack. However, reverting would
help to reduce potential confusion and false positives from
security scanners that assess risk based solely on version
numbers.
Another commit will follow to restore binary compatibility with
the liblzma 5.6.0 library by making the previously private
symbol (lzma_mt_block_size) public.
* Give link(1) its own usage message.
* Use getprogname(3) instead of rolling our own.
* Verify that the target file does not already exist.
* Add tests specific to link(1).
MFC after: 3 days
Sponsored by: Klara, Inc.
Reviewed by: allanjude
Differential Revision: https://reviews.freebsd.org/D44635
* The empty test case no longer fails because 89f1dcb3eb46 causes empty
files to bypass the bug.
* The bug still exists, so add a test case which exercises it.
* While here, tighten up some of the checks.
A similar patch has been submitted upstream.
PR: 274615
X-MFC-With: 89f1dcb3eb46
Sponsored by: Juniper Networks, Inc.
Sponsored by: Klara, Inc.
Reviewed by: markj
Differential Revision: https://reviews.freebsd.org/D44609
It's possible for the capture buffer to be smaller than indicated by the
header length. However, pfsync_print() only took the header length into
account. As a result we could read outside of the buffer.
Check that we have at least the expected amount of data before we start
parsing.
There is a conflict between bsm/audit.h and security/audit/audit.h due
to the way that staging is being set up using .PATH to point to the
full directory and the leaf files being specified in the list. Due to
this, the bsm/audit.h was getting staged as both bsm/audit.h and
security/audit/audit.h since the sys/bsm directory is listed first in
the .PATH list.
Use sys/security in the .PATH instead of sys/security/audit and specify
the audit header files as audit/<name>.h. This ensures that we get the
correct audit.h stanged for security/audit/audit.h.
Reviewed by: sjg
Obtained from: Juniper Networks, Inc.
Alan Cox [Wed, 3 Apr 2024 05:21:08 +0000 (00:21 -0500)]
arm64: correctly handle a failed BTI check in pmap_enter_l2()
If pmap_enter_l2() does not create a mapping because the BTI check
fails, then we should release the reference on the page table page
acquired from pmap_alloc_l2(). Otherwise, the page table page will
never be reclaimed.
The ddb pretty-printer currently does not print out enum values that
are not labeled (e.g. X | Y).
The enum printer was reworked to print non-labeled values.
Mark Johnston [Wed, 3 Apr 2024 15:29:25 +0000 (11:29 -0400)]
nextboot: Write nextboot.conf safely
As in the old nextboot.sh script:
- First write everything to a tempfile instead of /boot/nextboot.conf.
- fsync() the tempfile before renaming it to nextboot.conf.
Fixes: fd6d47375a78 ("rescue,nextboot: Install nextboot as a link to reboot, rm nextboot.sh")
Reviewed by: imp
Differential Revision: https://reviews.freebsd.org/D44572
Previously, we would error out if we encountered a global extended
header, because we don't know what it means. This doesn't really
matter though, and traditionally, tar implementations have either
ignored them or treated them as plain files, so just ignore them.
This allows tarfs to mount tar files created by `git archive`.
MFC after: 3 days
Sponsored by: Juniper Networks, Inc.
Sponsored by: Klara, Inc.
Reviewed by: kevans
Differential Revision: https://reviews.freebsd.org/D44600
Reinstate returning EOVERFLOW from stats_v1_blob_clone()
a0993376ec5f (from D43179) subtly changed stats_v1_blob_clone() to stop returning EOVERFLOW in the case where the user buffer is not large enough to receive the entire statsblob. This results in any consumers which are implemented to retry on receiving EOVERFLOW to instead give up after receiving an empty statsblob header.
Fix by latching any errors recorded prior to copyout.
Reviewed by: markj
Obtained from: Netflix, Inc.
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D44585
Fixes: a0993376ec5f ("stats: Check for errors from copyout()")
sound: Move sndstat_prepare_pcm() to pcm/sndstat.c and remove sndstat_entry->handler
Since all sndstat_entry->handler fields point to sndstat_prepare_pcm(),
we can just call the function directly, without assigning it to a
function pointer and calling it indirectly.
While here, move sndstat_prepare_pcm() to pcm/sndstat.c, as it is more
suitable there.
No functional change intended.
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
Reviewed by: dev_submerge.ch, markj
Differential Revision: https://reviews.freebsd.org/D44571
Fix the logic which determines if the destination Q variable can represent the source Q variable's value with full accuracy.
The new logic is mostly self explanatory except for the value fit checks.
If b has fewer integer bits than a, 0 == (Q_GIABSVAL(a) & (~Q_TC(a, 0) << Q_NIBITS(b))) is checking that a's integer value does not have high-order bits set above what b is capable of storing.
If b has fewer fractional bits than a, 0 == (Q_GFABSVAL(a) & ~(~Q_TC(a, 0) << (Q_NFBITS(a) - Q_NFBITS(b))))) is checking that a's fractional value does not have low-order bits set below what b is capable of storing.
Michael Tuexen [Mon, 1 Apr 2024 19:51:59 +0000 (21:51 +0200)]
tcp hpts: improve consistency
The target_slot argument of max_slots_available() can be NULL.
Therefore, check for this in all places.
Right now, all callers provide non-NULL pointer.
Reported by: Coverity Scan
CID: 1527732
Reviewed by: rrs
MFC after: 3 days
Sponsored by: Netflix, Inc.
Differential Revision: https://reviews.freebsd.org/D44527
cp: Improved conformance when copying directories.
* When copying a directory, if the destination exists and is not a
directory, we would previously emit an error message and exit. The
correct behavior according to POSIX is to emit an error message and
continue without descending further into the source directory.
* When copying a directory, if the destination does not exist and we
fail to create it, we would previously emit an error message and
exit. The correct behavior according to POSIX is to emit an error
message and continue. Whether to descend further into the source
directory is explicitly left unspecified; GNU cp does not, which
seems to me to be the safer and less surprising option, so let's not
either.
Mark Johnston [Mon, 1 Apr 2024 17:20:55 +0000 (13:20 -0400)]
wg: Use ENETUNREACH when transmitting to a non-existent peer
The old errno value used is specifically for Capsicum and shouldn't be
co-opted in this way. It has special handling in the generic syscall
layer (see syscallret()). OpenBSD returns ENETUNREACH in this case;
let's do the same thing.
Florian Walpen [Sun, 31 Mar 2024 19:14:16 +0000 (20:14 +0100)]
snd_hdspe(4): Only buffer_copy() audio data once.
Instead of blindly copying two periods of audio data to and from DMA
buffers, keep track of the writing position and derive the actual
part of audio data that needs to be copied.
This approximately halves the number of samples copied in total.
Rick Macklem [Sun, 31 Mar 2024 19:00:08 +0000 (12:00 -0700)]
mountd.c: Add warning messages for administrative controls
When "administrative controls" (which are exports of subdirectories
within a NFS server's local file system) are used, they export the
entire local server file system. (The subdirectory only applies to
the Mount protocol used for NFSv3 mounts.)
To minimize the risk that this causes confusion w.r.t. what is exported
to NFS client(s), this patch generates warning messages for these.
Only one message is generated for each server local file system.
The messages can be silenced via a new "-A" command line option.
The mountd.8 man page will be patched via a separate commit.
Mark Johnston [Sun, 31 Mar 2024 18:14:02 +0000 (14:14 -0400)]
kern linker: Don't invoke dtors without having invoked ctors
I have a kernel module which fails to load because of an unrecognized
relocation type. link_elf_load_file() fails before the module's ctors
are invoked and it calls linker_file_unload(), which causes the module's
dtors to be executed, resulting in a kernel panic.
Add a flag to the linker file to ensure that dtors are not invoked if
unloading due to an error prior to ctors being invoked.
At the moment I only implemented this for link_elf_obj.c since
link_elf.c doesn't invoke dtors, but I refactored link_elf.c to make
them more similar.