scx_rusty: arena library backports #2017
Conversation
|
Yes, the arena code in rusty has been copied off of the main library, so this is essentially a backport. This includes recomputing the padding. Do you mind if I rename the pull request as "scx_rusty: arena library backports" for clarity? |
Yes, renaming is clearer. |
EricccTaiwan
changed the title
EricccTaiwan
force-pushed
the
rusty-lib-fix
branch
from
4836036 to
7df7724
Compare
|
Oh I guess we had not added the padding feature to the allocator when we forked off the code, in that cases the padding is always 0 so it is not necessary. |
EricccTaiwan
force-pushed
the
rusty-lib-fix
branch
from
7df7724 to
a6032ca
Compare
OK, I've cleaned up the commit message, PTAL thanks! |
Add a check for failed allocation when calling sdt_alloc(). Ref: 9649028 Co-authored-by: Po-Ying Chiu <charlie910417@gmail.com> Signed-off-by: Cheng-Yang Chou <yphbchou0911@gmail.com>
Check `memory` instead of `scx_static.memory` after arena allocation. Ref: 99b2c79 Co-authored-by: Po-Ying Chiu <charlie910417@gmail.com> Signed-off-by: Cheng-Yang Chou <yphbchou0911@gmail.com>
Ref: ece44e4 Signed-off-by: Cheng-Yang Chou <yphbchou0911@gmail.com>
Add a missing `bpf_spin_unlock(&sdt_lock)` before returning the allocated pointer to ensure the lock is always released on the normal path. Co-authored-by: Po-Ying Chiu <charlie910417@gmail.com> Signed-off-by: Cheng-Yang Chou <yphbchou0911@gmail.com>
When the static allocator runs out of space it now allocates a fresh arena block, resets the offset to zero. Ref: 8d8a85d Co-authored-by: Po-Ying Chiu <charlie910417@gmail.com> Signed-off-by: Cheng-Yang Chou <yphbchou0911@gmail.com>
EricccTaiwan
force-pushed
the
rusty-lib-fix
branch
from
a6032ca to
0006e57
Compare
|
Thank you! I'll take another look. |
2 participants
Here are the changes I made:
[RFC] Recompute padding on allocation$\to$ should this align with *scx_static_alloc(size_t bytes, size_t alignment)?Edit: