On Tue, 2024-03-05 at 16:43 +0100, Jan Beulich wrote: > On 05.03.2024 16:15, Oleksii wrote: > > I agree that upon examining the current state of the code around > > these > > functions, it appears safe to provide stubs. However, the reason my > > patch was rejected is that it may not be entirely safe, as Julien > > pointed out that even with Arm, some functions shouldn't be empty. > > > > What I would like to propose is that it might be beneficial, at > > least > > in CONFIG_DEBUG=y, to have a warning message. Does that make sense? > > A warning message to what effect? And are you thinking of a build- > time > warning, or a runtime one? Plus wouldn't different aspects quickly > lead > to proliferation of warnings? A warning message about that an empty definition is provided for evaluate_nospec/block_speculation functions, so a developer will know that it can be an issue.
Personally, I am OK with having this function empty by default as it is done in the patch with opportunity to being redefined in arch specific code if it is necessary, but I remembered that I had the similar questions in my patch series which probably should be covered in this patch series. Runtime message can also be an option, but I thought about build-time, but I agree that it can lead to proliferation of warning, so not the best one idea. ~ Oleksii
