Bug 1178
Summary: | Deadlock with heap lock | ||
---|---|---|---|
Product: | DPDK | Reporter: | Maxime Coquelin (maxime.coquelin) |
Component: | core | Assignee: | Anatoly Burakov (anatoly.burakov) |
Status: | UNCONFIRMED --- | ||
Severity: | normal | CC: | david.marchand |
Priority: | Normal | ||
Version: | unspecified | ||
Target Milestone: | --- | ||
Hardware: | All | ||
OS: | All |
Description
Maxime Coquelin
2023-03-10 18:31:54 CET
Looking more closely at rte_malloc_heap_destroy() with David, the function is not buggy, the lock is implicitly unlocked in malloc_heap_destroy() when doing the memset to 0. David's patch [0] moving the lock handling in malloc_heap_destroy(), and using spinlock API to unlock it makes it clearer. Furthermore, the API is not used with test-pmd, so it is not related to the issue I faced. Since other heap->lock usage look sane, the issue faced may be due to a memory corruption. I let the Bz open for now. In case someone manages to reproduce. [0]: https://patchwork.dpdk.org/project/dpdk/patch/20230224151143.3274897-2-david.marchand@redhat.com/ |