1272 Commits

This Branch

This Branch

All Branches

Author	SHA1	Message	Date
Akilan	22d9b65b4b	added morello build instruction	2025-11-26 12:23:41 +00:00
Akilan	97a73d7b7b	added ARM cheri variables	2025-11-26 12:11:14 +00:00
Akilan	465df659f5	added cheri permissions directly	2025-11-26 12:05:33 +00:00
Max	92ca918488	Remove interface linking for PRIVATE linkage (#799)	2025-10-20 10:27:12 +01:00
Matthew Parkinson	198f1fc464	Removing benchmarking from PR triggers (#800) ... The benchmarking workflow cannot run successfully on the PR as it requires access to secrets which are not available in PR contexts. This commit removes that trigger and adds a trigger on pushes to branches starting with "bench/" so that testing can be done before merging to main by those with write access.	2025-10-20 09:31:55 +01:00
Matthew Parkinson	fbedb31504	CMAKE: Libs should not require headers (#797) ... The API exposed by the various libraries in snmalloc all obey standard C std lib conventions, so no headers are required to consume them. This change updates the CMake configuration to not require headers for the libraries.	2025-10-17 16:42:50 +01:00
Matthew Parkinson	0b70ad97a6	Benchmarking automatically (#795) ... Co-authored-by: Amaury Chamayou <amaury@xargs.fr>	2025-08-18 09:14:54 +01:00
Matthew Parkinson	3385660fd7	Update minimum CMake version (#793) ... The use of CACHE in the SNMALLOC_STATIC_LIBRARY_PREFIX was not correctly handled by cmake before 3.21. > The “set(CACHE)” command no longer removes a normal variable of the same name, if any. See policy “CMP0126”. Changing the minimum version means that containing projects can do: ```CMake set(SNMALLOC_STATIC_LIBRARY_PREFIX "my_prefix") ``` Without this, there were weird behaviours that the first compile used "sn_" and then subsequent compiles used "my_prefix". Thanks to @achamayou for debugging this in CCF. https://github.com/microsoft/CCF/pull/7161	2025-08-05 11:19:42 +01:00
Matthew Parkinson	452fcc44b0	Add exception handling for new operator (#791) ... The new operator in the snmalloc did not throw exceptions in the case of allocation failure. Moreover, it was not possible to override the behaviour of the failure of the new operator using the std::set_new_handler function. This PR adds the necessary code to the snmalloc new operator to throw std::bad_alloc when the allocation fails. It also allows the std::set_new_handler function to be used to set a custom handler for allocation failures.	2025-07-15 09:57:34 +01:00
Matthew Parkinson	b49a42d9be	Code Quality - inlining (#792) ... * Inline lambdas for better codegen * Disable PreventFork in the freelist queue	2025-07-09 15:29:54 +01:00
Matthew Parkinson	71e205fe17	Fallible notify using (#790) ... On Windows, the most common way to out-of-memory is to fail to commit a page. The current design of the Pal assumes that notify_using always succeeds, thus we have to raise an error if it fails. This changes the specification of notify_using to return a bool, indicating whether the notify succeeded or not. This allows Windows to fail the notification, and then the surrounding code can handle the failure appropriately, such as by throwing an exception or returning the nullptr for the allocation. Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>	2025-07-09 07:37:58 +00:00
Matthew Parkinson	d4c7e01cd7	Pass continuations for success and failure cases (#788) ... This PR provides a templated parameter to the allocation routines. This can be used to add special behaviour in both the successful allocation behaviour, and in the failing to allocate cases. The intent of this is two enable two future features * set_new_handler - so that the failure case doesn't just set ENOMEM, and return nullptr. But can handle both the Windows and C++ versions of (_)set_new_handler. * The success handler can be used to add checking, zeroing and in the future storing precise size information in metadata for each allocation.	2025-07-03 20:05:52 +01:00
Matthew Parkinson	012138e29f	Small sizeclass lookup improvement, (#777) ... * Fix off by one in deciding where to switch from small to large. * Improve codegen on Windows by making it simpler for compiler to remove a redundant branch	2025-07-01 12:02:10 +01:00
Matthew Parkinson	2d33e4f33b	Adjust func-memory test to be shorter. (#780) ... Some CI pipelines were occasionally timing out due to the func-memory test taking too long. This change reduces the number of iterations from 100 to 50 for each test run, which should help avoid timeouts while still providing sufficient coverage. It also adds some debug output to indicate the time taken for each test.	2025-07-01 11:01:53 +01:00
Matthew Parkinson	9d29d828b8	Index cache by build run (#779)	2025-07-01 11:01:12 +01:00
Matthew Parkinson	76c67a3ee3	Drop Win2019 from CI (#778)	2025-07-01 09:18:07 +00:00
Matthew Parkinson	9a3d12a724	Make Secondary Allocator a template parameter (#774) ... This refactors the use of the Secondary Allocator, so that out of tree implementations can be used.	2025-06-30 21:23:07 +01:00
Neil Monday	e600bf1b38	Release all pagemap reservations at the very end of the program or DLL (#773) ... * Release all pagemap reservations at the very end of the program or DLL Co-authored-by: Neil Monday <neil.monday@amd.com> Co-authored-by: Matthew Parkinson <mjp41@users.noreply.github.com>	2025-06-30 13:15:49 +00:00
Mark Johnston	890bcf94a6	Remove tests for experimental jemalloc ABIs (#770) ... Commit 5680cf6dc6e2 ("jemalloc: don't expose 3.0 compat symbols") in FreeBSD src removes the default symbols for allocm() and friends, so this test no longer links. Compatibility is stil provided for programs linking against FBSD_1.3, but here it seems easier to simply stop referencing them.	2025-06-30 14:06:29 +01:00
Matthew Parkinson	16d96b9f8c	Fix CI (#776) ... This PR updates to aspects of the CI to fix blocked PRs: It moves the Bazel CI into its own set of actions, so failure can be isolated from other tasks. Updates the clang-cl version used in CI for self-vendored build as it was failing with the default.	2025-06-30 13:57:07 +01:00
Matthew Parkinson	0064c01b82	Improving handling first allocation being for TLS (#767) ... * handle reentrancy during initialization * use finialization list if possible * Add test for reentrancy of C++ destructors and allocation * Add test for reentrancy of setspecific * Add new mode for directly calling __cxa_thread_atexit_impl --------- Co-authored-by: Schrodinger ZHU Yifan <yifanzhu@rochester.edu>	2025-05-07 19:28:37 +01:00
Jaya Kasa	a63f0c1ac3	feat: add support for bazel (#759)	2025-05-07 09:09:44 +01:00
Matthew Parkinson	6325edefcf	Remove 20.04 from CI (#768) ... * Remove 20.04 as no longer supported. * Remove PowerPC due to existing issues with updated QEMU.	2025-04-28 15:11:21 +01:00
Matthew Parkinson	b8e28be14b	Fix bug in pool sort (#765) ... The pool sort routine is used by Verona's systematic testing. There was a bug introduced in #612 that could corrupt the list when a sort occured. This is not used by snmalloc and the test coverage was insufficient to expose the bug. This PR fixes the bug, and improves test coverage.	2025-04-10 10:25:24 +01:00
Will Brown	e6bf8570d0	[Rust] add support for optionally compiling libc functions (#763) ... * add SNMALLOC_RUST_LIBC_API option	2025-04-03 13:56:41 +01:00
Jeff Sooknarine	59987ca908	Header-only build fix for WIN32_LEAN_AND_MEAN redefinition (#764) ... When using snmalloc as header-only library in a project that globally defines `WIN32_LEAN_AND_MEAN`, there is a compile error for macro redefinition. Guard the pal_windows.h define.	2025-04-01 21:16:55 +00:00
Matthew Parkinson	80bdcd999f	CI: Cancel inflight checks on a new push. (#760)	2025-03-21 17:58:54 +00:00
Matthew Parkinson	dff1057db2	Refactor representation of thread local state. (#751) ... * Lift checking for init to ThreadAlloc The check init code was tightly integrated into LocalAllocator. This commit pull that code out into ThreadAlloc, and passes a template parameter into the remaining LocalAllocator to perform the relevant TLS manipulations. This removes some of the awkward layering around register_clean_up. * Reduce size of test due to failures. Fully disable lotsofthreads test Need to investigate if the test is unreliable, or we have actually regressed perf. A quick mimalloc-bench didn't show any regressions. * Simplify message queue initialisation This introduces one additional branch on when processing a batch of messages, but it is likely to only be hit when a lot of messages are processed. * Patch Domestication test. * Refactor CoreAlloc/LocalAlloc This combines the notion of CoreAlloc, LocalAlloc and LocalCache into a single class. Previously, these were separated so that a more complex structure would be stored directly in the TLS. This however, proved to be bad for compatibility if the allocator is part of the libc implementation. This commit collapses all the stages of the allocator into a single class. This simplifies the sequencing and overall is a nice reduction in complexity. * Re-enable lots of threads test. * Reenable concept using alternative lazy checking for concepts. * Self code review	2025-03-21 15:13:32 +00:00
Matthew Parkinson	3348bf9e56	Fix test when using GWP-Asan (#758)	2025-03-17 15:37:54 +00:00
Matthew Parkinson	a106a2e69d	Refactor check_bounds (#756) ... This changes the shape of check_bounds to take a continuation to call if the bounds check succeeds. This is designed to allow for easily wrapping existing code with a bounds check, e.g. ``` void* memcpy(void* dest, const void* src, size_t n) { return check_bounds(dest, n, [&] { return memcpy_impl(dest, src, n); }); } ```	2025-03-14 16:09:09 +00:00
Trithek	dfda7a8442	Fixed pagemap reservations for a fixed-region use-case using the default PAL (#757) ... * Made reserve_range do the correct thing when a fixed range configuration was used * Corrected the check for out-of-bounds	2025-03-13 10:20:43 +00:00
Matthew Parkinson	ccc03ce0fc	Protection against fork (#735) ... If a thread forks, while another thread is holding an snmalloc lock, then the allocator could stop working. This patch attempts to protect against the cases of this. There is one case that is not covered. If a fork occurs during the very first allocation. This can result in the installation of the fork handler racing with the fork, and all bets are off.	2025-03-04 13:44:59 +00:00
Matthew Parkinson	c2e22ccb79	Prevent internal errno setting escaping to the client. (#754)	2025-02-28 11:12:22 +00:00
Matthew Parkinson	06df9dd9e1	Update realloc(p,0) semantics. (#753) ... * Factor out small sizeclass check * Update realloc(p,0) semantics This commit changes the behaviour of realloc(p,0) to be free(p) if p!=nullptr, and malloc(0) if p== nullptr.	2025-02-27 15:20:31 +00:00
Matthew Parkinson	6622dc584e	Bug fixes (#752) ... * Fix overflow by alignment * Bug fix: Ensure bytes_free is the total The bytes freed was not added to the total, but overrode it. This meant it never fired. This commit fixes that.	2025-02-27 08:52:50 +00:00
Matthew Parkinson	5f7baef755	Refactor: Remove unused features and functions, and move most allocator operations to a global namespace. (#750) ... * Factor out explicit Config type Instead of using snmalloc::Alloc::Config, expose snmalloc::Config, which is then used to derive the allocator type. * Move globalalloc to front end. * Remove unneed template parameter from global snmalloc functions. * Remove SNMALLOC_PASS_THROUGH VeronaRT now has an abstraction layer which can easily replace the allocator. Having such a complex integration still in snmalloc does not make sense. * Take some global functions off of local alloc. * Drop comparison overloads on atomic Capptr. Performing a comparison on two atomic ptr is a complex operation, and should not be implicit. The memory model order and such things needs to be considered by the caller. * Remove function_ref and use templates The implementation prefers to use templates over the function_ref. This now only exists in the Pal for a currently unused feature. * Removing function_ref reduces stl needs. * Remove use of __is_convertible to support older g++ * Inline function that is only used once. * Remove unused function * Restrict ThreadAlloc usage to globalalloc This commit introduces various inline functions on snmalloc:: that perform allocation/deallocation using the thread local allocator. They remove all usage from a particular test. * Move cheri checks to own file. * Refactor is_owned checks. * Move alloc_size and check_size to globalalloc. * Minor simplification of dealloc path * Fix up is_owned to take a config * Improve usage of scoped allocator. * Handle Config_ in globalalloc.	2025-02-22 19:53:27 +00:00
Matthew Parkinson	6d50141e35	build: Fix nothrow include. (#749)	2025-02-14 12:50:18 +00:00
Matthew Parkinson	0cd9ccf170	CI: Refresh Targets (#748) ... Add newer platforms to CI. * ubuntu-24.04-arm * windows-2021 * windows-2025	2025-02-13 09:47:36 +00:00
Neil Monday	ef4748289d	Windows: Remove exception handler for pagemap on unloading (#746) ... Co-authored-by: Neil Monday <neil.monday@amd.com>	2025-02-12 15:38:54 +00:00
Matthew Parkinson	a3fe420af2	Fix HAS_EXCEPTION on Windows. (#743)	2025-02-05 16:54:51 +00:00
Matthew Parkinson	b3efd49c9a	Mistaken commit of message_once (#741)	2025-02-05 11:16:41 +00:00
Schrodinger ZHU Yifan	16b96245f6	Enable a seconary allocator support (e.g. GWP-Asan) (#737)	2025-02-05 09:03:11 +00:00
Matthew Parkinson	32495fd42d	Multiple Pals could return spuriously from wake on address (#739) ... * Add stress test benchmark Co-authored-by: Alexander Nadeau <wareya@gmail.com> * Add defensive code against spurious wakeup This commit checks that wait_on_address has not returned spuriously. * pal: spurious wake up. The code in the Pal for wake on address was incorrectly assuming the operation returning success meant it had actually changed. The specification allows for spurious wake ups. This change makes the Pals recheck for a change. --------- Co-authored-by: Alexander Nadeau <wareya@gmail.com>	2025-02-04 18:44:02 +00:00
Matthew Parkinson	e3e558472d	Update CMakeLists.txt (#734) ... * Handle platforms that have `_GLIBCXX_ASSERTIONS`, which require the stdlib++ to be included. * Stop override of memcpy with FORTIFY_SOURCE enabled * Add to .gitignore * Add data for graphs in release notes. * Minor tidy on CMake.	2025-01-23 13:36:08 +00:00
Jean-Michaël Celerier	aa07749f46	pal_windows.h: missing <cerrno> for ENOMEM on MinGW (#733) ... Co-authored-by: Matthew Parkinson <mjp41@users.noreply.github.com>	2025-01-21 09:25:11 +00:00
Schrodinger ZHU Yifan	229c6d04fe	remove extra puts definitions (#731) ... LGTM, thanks.	2025-01-16 10:31:31 +00:00
Matthew Parkinson	0fbc325a0a	Fix merge issue for out of data CI run. (#730)	2025-01-10 08:37:33 +00:00
Matthew Parkinson	9e632047de	Add Windows Support for Lazy Commit of page map (#727) ... The Pagemap can potentially be accessed where there is no direct knowledge of the location. On platforms with lazy commit this can be handled by the OS faulting in a zero page. Here we modify the Windows PAL to add an exception handler that will commit pages in the read only range, i.e. the page map. We still require explicit notify_using to be able to modify the pagemap.	2025-01-09 17:57:52 +00:00
Schrodinger ZHU Yifan	0111a410a2	[proxy](7/n) self vendor array and algorithm headers (#722) ... * [proxy](7/n) proxy array and algorithm headers * [wip] address CRs * undo comment * format * fix build * simplify macro dispatching	2025-01-09 13:49:22 +00:00
Schrodinger ZHU Yifan	0027f02396	[proxy](6/n) self-vendor utility headers (#721)	2025-01-09 10:25:35 +00:00