On 2025-10-28 11:56, Thomas Gleixner wrote:
On Tue, Oct 28 2025 at 10:24, Mathieu Desnoyers wrote:
On 2025-10-27 04:44, Thomas Gleixner wrote:
From: Thomas Gleixner <[email protected]>
Replace the open coded implementation with the new get/put_user_inline()
helpers. This might be replaced by a regular get/put_user(), but that needs
a proper performance evaluation.
I understand that this is aiming to keep the same underlying code,
but I find it surprising that the first user of the "inline" get/put
user puts the burden of the proof on moving this to regular
get/put_user() rather than on using the inlined version.
The comment above the inline API clearly states that performance
numbers are needed to justify the use of inline, not the opposite.
I am concerned that this creates a precedent that may be used by future
users of the inline API to use it without performance numbers
justification.
There was not justification for the open coded inline either and
converting it to get/put must be a completely seperate change.
AFAIU there was justification from Linus in
commit 43a43faf5376 ("futex: improve user space accesses").
Perhaps it's good enough to refer to that prior justification
in the commit message.
Thanks,
Mathieu
--
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com
