Ira,
On Mon, Nov 16 2020 at 10:49, Ira Weiny wrote:
On Sun, Nov 15, 2020 at 07:58:52PM +0100, Thomas Gleixner wrote:
Currently struct irqentry_state_t only contains a single bool value which makes passing it by value is reasonable. However, future patches propose to add information to this struct, for example the PKRS register/thread state.
Adding information to irqentry_state_t makes passing by value less efficient. Therefore, change the entry/exit calls to pass irq_state by reference.
The PKRS muck needs to add an u32 to that struct. So how is that a problem?
There are more fields to be added for the kmap/pmem support. So this will be needed eventually. Even though it is not strictly necessary in the next patch.
if you folks could start to write coherent changelogs with proper explanations why something is needed and why it's going beyond the immediate use case in the next patch, then getting stuff sorted would be way easier.
Sorry my crystalball got lost years ago and I'm just not able to keep up with the flurry of patch sets which have dependencies in one way or the other.
The resulting struct still fits into 64bit which is by far more efficiently passed by value than by reference. So which problem are you solving here?
I'm getting ahead of myself a bit. I will be adding more fields for the kmap/pmem tracking.
Would you accept just a clean up for the variable names in this patch? I could then add the pass by reference when I add the new fields later. Or would an update to the commit message be ok to land this now?
Can you provide a coherent explanation for the full set of things which needs to be added here first please?
Thanks,
tglx