TLSF and the Temple of Corrupted Memory #3653
Conversation
* add tlsf as submodule * libs: tlsf * Furi: tlsf as allocator * Furi: heap walker * shmal fixshesh * f18: tlsf * PVS: ignore tlsf * I like to moving * merge upcoming changes * memmgr: alloc aligned, realloc * Furi: distinct name for auxiliary memory pool * Furi: put idle and timer thread to mem2 * Furi: fix smal things in allocator * Furi: remove aligned_free. Use free instead. * aligned_malloc -> aligned_alloc * aligned_alloc, parameters order * aligned_alloc: check that alignment is correct * unit test: malloc * unit tests: realloc and test with memory fragmentation * unit tests: aligned_alloc * update api * updater: properly read large update file Co-authored-by: Aleksandr Kutuzov <[email protected]>
DrZlo13
requested review from
skotopes,
hedger,
gsurkov,
nminaylov,
gornekich,
Astrrra and
Skorpionm
as code owners
|
Compiled f7 firmware for commit |
hedger
added
Sub-GHz
|
Regarding the 8-byte alignment, GCC indeed appears to make that assumption, with or without This code: int32_t align_test_app(void* p) {
UNUSED(p);
FURI_LOG_I("TEST", "Hello world");
FURI_LOG_I("TEST", "I'm align_test!");
void* b = malloc(4);
uintptr_t valB = (uintptr_t)b;
return (valB & 7) == 0;
}compiled into: So it's a very good idea to guarantee proper alignment, at least for |
|
PVS-Studio report for commit |
|
I disabled all dolphin animations and ran some tests: tlsf
heap4
It looks like in the current version tlsf has no meaning except for the correct implementation of realloc. |
What's new
Verification
Checklist (For Reviewer)