[ASan][Windows] Fix false positive for zero sized rtl allocations (#181015)
This is a follow up to #155943
On Windows, ASan's allocator internally upgrades zero-size allocation
requests to size 1 (since malloc(0) must return a unique non-NULL
pointer). However, when the user queries the allocation size through
Windows heap APIs (RtlSizeHeap, HeapSize, \_msize, GlobalSize,
LocalSize), ASan reports the internal size (1) instead of the originally
requested size (0).
This causes false positive heap-buffer-overflow errors in a common
pattern:
```c++
void *buf = HeapAlloc(GetProcessHeap(), 0, 0);
SIZE_T size = HeapSize(GetProcessHeap(), 0, buf); // Returns 1, should be 0
if(size > 0) // could remove this and still be correct
memset(buf, 0, size); // ASan reports heap-buffer-overflow
[8 lines not shown]
[Support] Support 5-component VersionTuples (#181275)
LLDB parses compiler versions out of DW_AT_producer DWARF attributes
into a VersionTuple. The Swift compiler recently switched to 5-component
version numbers. In order to support this version scheme without growing
the size of VersionTuple, this patch dedicates the last 10 bits of the
build version to a 5th "sub-build" component. The Swift compiler
currently uses 1 digit for this and promises to never use more than 3
digits for the last 3 components.
This patch still leaves 6 decimal digits for the build component for
other version schemes.
rdar://170181060
Implement vector version of memchr, and dispatch to same (#177711)
As in the description.
This implementation shares quite a bit of code with the wide-read
versions of string_length.
[RFC][Docs][Clang][AMDGPU] Add AMDGPU builtins documentation
Add comprehensive documentation for AMDGPU target-specific builtins
(`AMDGPUBuiltins.rst`) covering argument semantics, restrictions, and
lowering notes for all builtin families.
This documentation was generated by AI (Claude) by cross-referencing:
- `clang/include/clang/Basic/BuiltinsAMDGPU.td` (builtin definitions)
- `llvm/include/llvm/IR/IntrinsicsAMDGPU.td` (intrinsic definitions)
- `clang/lib/Sema/SemaAMDGPU.cpp` (argument validation/constraints)
- `clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp` (lowering logic)
I did my best to proofread the parts I'm familiar with, but it would be greatly
appreciated if more people could help review it as well.
Reapply "[clang] Fix sema on ObjCLifetime conversion (#178524)" (#180817)
Clang can't handle objc lifetime correctly when casting We reuse the
approach similar to lifetime: First remove it before the conversion,
then add it back.
Add a test
Fixes https://github.com/llvm/llvm-project/issues/177478
NAS-139300 / 26.0.0-BETA.1 / Add dataset and relative_path fields to share entries (#18021)
This represents the first change for
[NAS-137337](https://ixsystems.atlassian.net/browse/NAS-137337). We may
choose to do something similar for other services that use dataset
paths.
* Add "dataset" field which is calculated on create and update based on
"path"
* Add "relative_path" field which is calculated on create and update
based on "path"
* Keep "path" field to avoid concatenating the fields and prepending
"/mnt/" when the full path is required
* No extra mess in `extend` and `compress`
1. Alembic migration adds columns to all `SharingServiceTask` services.
2. migration/0018 attempts to populate the columns based on the existing
path fields.
3. `pool.import` and `pool.dataset.unlock` call hooks to query
[2 lines not shown]
timeout: Clean up
* Annotate logv() and fix format string bug.
* Don't reinvent str2sig(3).
* Reorganize kill_self() so we unblock signals as late as possible, and
use raise(2) instead of kill(2).
* Explicitly close unused pipe descriptors.
* Use correct type to collect result of read(2) and write(2).
* Compare return values to 0, not -1.
* Sort local variables according to style(9).
* Reduce unnecessary nesting.
[8 lines not shown]
diff: Tweak recursion tests
The -r flag is not required to compare two directories; it is only
required to compare them recursively, i.e. descend into their common
subdirectories. Adjust tests that use -r needlessly, and adjust the
dirloop test to verify that these two cases remain distinct.
MFC after: 1 week
Sponsored by: Klara, Inc.
Reviewed by: kevans
Differential Revision: https://reviews.freebsd.org/D55262
diff: Tweak range of -C and -U arguments
POSIX uses the terms “positive decimal integer” for -C and “non-negative
decimal integer” for -U, which translates into lower bounds of 1 for -C
and 0 for -U.
POSIX does not specify a minimum upper bound for either mode, but as of
5fc739eb5949 both our backends support context sizes up to and including
INT_MAX, so use that.
Having had the opportunity to consult the Unix System Test Suite, the
diff test cases found therein happen to precisely match these bounds.
While here, switch to using strtonum() to parse numerical arguments, and
try to be more consistent in how we report usage errors.
MFC after: 1 week
Sponsored by: Klara, Inc.
Reviewed by: kevans
Differential Revision: https://reviews.freebsd.org/D55261
Merge tag 'mm-hotfixes-stable-2026-02-13-07-14' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
Pull MM fixes from Andrew Morton:
"Three MM hotfixes, all three are cc:stable"
* tag 'mm-hotfixes-stable-2026-02-13-07-14' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm:
procfs: fix possible double mmput() in do_procmap_query()
mm/page_alloc: skip debug_check_no_{obj,locks}_freed with FPI_TRYLOCK
mm/hugetlb: restore failed global reservations to subpool
[lldb/Breakpoint] Fix condition hash after updating condition text (#181409)
StopCondition::SetText was computing the hash from the moved-from text
parameter instead of the stored m_text member. After std::move(text),
the source parameter becomes empty, causing the hash to always be
computed from an empty string.
This caused breakpoint condition updates to fail silently. When a user
modified a condition (e.g., from "x == y" to "x > y"), the hash remained
unchanged. Breakpoint locations use this hash to detect when conditions
need re-evaluation, so with a stale hash they would continue using
cached state for the old condition, triggering at incorrect locations.
The patch fixes this issue by computing the hash from m_text after the
move operation, ensuring it reflects the actual stored condition text.
It also adds API test that updates a breakpoint condition and verifies
the new condition is properly evaluated.
[2 lines not shown]
NAS-139796 / 26.0.0-BETA.1 / Add developer documentation related to PAM (#18210)
* Internal truenas developers should be using our vetted / approved PAM
configuration files rather than rolling their own.
* Relationship between PAM / NSS, and how we use PAM files should be
documented.
Merge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost
Pull virtio updates from Michael Tsirkin:
- in-order support in virtio core
- multiple address space support in vduse
- fixes, cleanups all over the place, notably dma alignment fixes for
non-cache-coherent systems
* tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost: (59 commits)
vduse: avoid adding implicit padding
vhost: fix caching attributes of MMIO regions by setting them explicitly
vdpa/mlx5: update MAC address handling in mlx5_vdpa_set_attr()
vdpa/mlx5: reuse common function for MAC address updates
vdpa/mlx5: update mlx_features with driver state check
crypto: virtio: Replace package id with numa node id
crypto: virtio: Remove duplicated virtqueue_kick in virtio_crypto_skcipher_crypt_req
[14 lines not shown]
[GISel][RISCV] Simplify the generated code for narrowScalarCTLS. (#180827)
Instead of trying to make CTLS work for the Lo part, conditionally
invert Lo using the Hi sign bit, then do a CTLZ. If the CTLZ is
zero, then the Lo sign bit differs from the Hi sign bit. Otherwise,
each leading zero represents an additional sign bit.
This generates better code when CTLS and CTLZ are both supported.
I've added Zbb to the P extension command line for RISC-V since
P is likely to imply Zbb when it is ratified, but that isn't written
in the spec yet. If that doesn't happen, I expect CLZ would get
added back to the P extension.
[RISCV] In tryWideningMulAcc, check multiple users before checking operand 1. (#181321)
If both operands are WMUL, we should still consider the second one if
the first one has multiple users.