This migrates ksud execution decision-making to bprm_check_security.
This requires passing proper argv and envp to a modified _ksud handler
aptly named 'ksu_handle_bprm_ksud'.
Introduces:
int ksu_handle_bprm_ksud(const char *filename, const char *argv1,
const char *envp, size_t envp_len)
which is adapted from:
int ksu_handle_execveat_ksud(int *fd, struct filename **filename_ptr,
struct user_arg_ptr *argv,
struct user_arg_ptr *envp,
int *flags)
ksu_handle_bprm_ksud handles all the decision making, it decides when it is
time to apply_kernelsu_rules depending if it sees "second_stage".
For LSM hook, turns out we can pull out argv and envp from mm_struct.
The code in here explains itself on how to do it.
whole blob exists on arg_start to arg_end, so we just pull it out and grab next
array after the first null terminator.
as for envp, we pass the pointer then hunt for it when needed
My reasoning on adding a fallback on usercopy is that on some devices a fault
happens, and it copies garbled data. On my creation of this, I actually had to lock
that _nofault copy on a spinlock as a way to mimic preempt_disable/enable without
actually doing it. As per user reports, no failed _nofault copies anyway but we
have-to-have a fallback for resilience.
References:
- old version1 6efcd8193e
- old version2 37d5938e66
- bad usercopy #21
This now provides a small helper function, ksu_copy_from_user_retry, which explains
itself. First we attempt a _nofault copy, if that fails, we try plain.
With that, It also provides an inlined copy_from_user_nofault for < 5.8.
While using strncpy_from_user_nofault was considered, this wont do, this will
only copy up to the first \0.
devlog:
16e5dce9e7...16c1f5f52128642e60d7...728de0c571
References:
https://elixir.bootlin.com/linux/v4.14.1/source/include/linux/mm_types.h#L429https://elixir.bootlin.com/linux/v4.14.1/source/include/linux/lsm_hooks.h
Stale: https://github.com/tiann/KernelSU/pull/2653
Signed-off-by: backslashxx <118538522+backslashxx@users.noreply.github.com>
Hi @tiann.
Thanks for the great project, I had great fun playing around with it.
This PR mainly tries to further minimize the possible delays caused by
KernelSU hooking.
There are 3 major changes:
- Processes with 0 < UID < 2000 are blocked straight-up before going
through the allow_list.
I don't see any need for such processes to be interested in root, and
this allows returning early before going through a more expensive
lookup.
If there's an expected breakage due to this change, I'll remove it. Let
me know.
- A page-sized (4K) bitmap is added.
This allows O(1) lookup for UID <= 32767.
This speeds up `ksu_is_allow_uid()` by about 4.8x by sacrificing a 4K
memory. IMHO, a good trade-off.
Most notably, this reduces the 99.999% result previously from worrying
milliseconds scale to microseconds scale.
For UID > 32767, another page-sized (4K) sequential array is used to
cache allow_list.
Compared to the previous PR #557, this new approach gives another nice
25% performance boost in average, 63-96% boost in worst cases.
Benchmark results are available at
https://docs.google.com/spreadsheets/d/1w_tO1zRLPNMFRer49pL1TQfL6ndEhilRrDU1XFIcWXY/edit?usp=sharing
Thanks!
---------
Signed-off-by: Juhyung Park <qkrwngud825@gmail.com>
