On martedì 26 aprile 2022 13:04:42 CEST Sebastian Andrzej Siewior wrote: > On 2022-04-26 11:43:03 [+0200], Fabio M. De Francesco wrote: > > Either way, according to the code kmap_atomic() does not always disable > preemption Hi Sebastian, In my last email (for patch 4/4) I wrote that I would add a deprecation notice to kunmap_local() and would talk about thread locality in kmap_local_page(). I want to confirm that I'll do these changes in v3. Then I closed that email asking if I was still overlooking something. Consider that I'm relatively new to kernel development and that it's just something I do in my spare time. So I was pretty sure I was still missing something :) After reading again the code of kmap_atomic() (as you suggested - thanks!) I noted that you correctly say that kmap_atomic() does not always disables preemption (i.e., preemption is disabled only for !PREEMPT_RT kernel's configurations). Therefore I'll also change the first sentence of kunmap_local() to the following: "[It] Unmaps an address previously mapped by kmap_atomic() and re-enables pagefaults and preemption (the latter disabled only for !PREEMP_RT configurations).". I will also be making this change in v3. Can you please say if I'm still missing something? Thanks, Fabio
