On 05/15, Jiri Olsa wrote:
Changes from v1:
- rebased on top of tip/master + mm/mm-stable + 1 extra change [1]
- keep the refcrf offset update inside write_insn and enabling it
via function argument
- fixed locking comment for uprobe_write_opcode, but skiped suggested
comment on register_for_each_vma, need more thinking on that [Oleg]
- added acks
- removed refctr from uprobe_trampoline object [Oleg]
- change find_nearest_page to use vm_unmapped_area [Oleg]
- re-structured x86 set_swbp [Andrii]
- use -EINVAL in __arch_uprobe_optimize [Andrii]
- added usdt.h from libbpf/usdt project [Andrii]
- several minor test code changes [Andrii]
- man page updates [Alejandro]
I forgot to send this email a week ago, sorry...
This version seems to address my concerns. I don't see anything obviously
wrong. So if this optimization is important for users... FWIW,
For 1/22 - 10/22
Acked-by: Oleg Nesterov <oleg@redhat.com>