[v2] alloc_tag: fix a leak and a deadlock around shutdown_mem_profiling()

[PATCH v2 0/2] alloc_tag: fix a leak and a deadlock around shutdown_mem_profiling()

Hao Ge posted 2 patches 1 month, 1 week ago

Diff against v2 v2

Only 0 patches received!

There is a newer version of this series

lib/codetag.c  | 10 ++++++++--
mm/alloc_tag.c | 14 ++++++++++++-
2 files changed, 21 insertions(+), 3 deletions(-)

Expand all Fold all

[PATCH v2 0/2] alloc_tag: fix a leak and a deadlock around shutdown_mem_profiling()

Posted by Hao Ge 1 month, 1 week ago

Two fixes for issues reported by sashiko:

  1. percpu counter leak on modules loaded after profiling is disabled.
  2. AB-BA deadlock between module load and /proc/allocinfo readers.

Changes since the RFC [2]:

  - The "move codetag section placement decision to layout_sections()"
    patch is dropped from this series. Its retry path depends on [1],
    so it will be sent as part of that patchset instead.
  - Patch 1 now returns -EOPNOTSUPP instead of introducing
    CODETAG_MODULE_LOAD/CODETAG_MODULE_EXCLUDED defines (suggested by
    Suren).
  - Reported-by tags added.

Patch 1 skips percpu counter allocation when profiling is disabled,
so the module loads without its tags instead of leaking counters.
Patch 2 defers remove_proc_entry() to a workqueue.

[1]: https://lore.kernel.org/all/20260812054105.102637-3-hao.ge@linux.dev/
[2]: https://lore.kernel.org/all/20260813093421.135230-1-hao.ge@linux.dev/

Hao Ge (2):
  alloc_tag: skip percpu counter allocation when profiling is disabled
  alloc_tag: remove /proc/allocinfo outside of mod_lock

 lib/codetag.c  | 10 ++++++++--
 mm/alloc_tag.c | 14 ++++++++++++-
 2 files changed, 21 insertions(+), 3 deletions(-)

--
2.25.1