On Mon, 28 Dec 2020 at 04:24, Cong Wang <xiyou.wangc...@gmail.com> wrote: > > On Sun, Dec 27, 2020 at 6:40 AM Taehee Yoo <ap420...@gmail.com> wrote: > > But I'm so sorry I didn't understand some points. > > > > 1. you said "both side" and I understand these as follows: > > a) failure of allocation because of a high order and it is fixed > > by 72e09ad107e7 > > b) kernel panic because of 72e09ad107e7 > > Are these two issues right? > > Yes, we can't fix one by reverting the fix for the other. > > > > > 2. So, as far as I understand your mention, these timers are > > good to be changed to the delayed works And these timers are mca_timer, > > mc_gq_timer, mc_ifc_timer, mc_dad_timer. > > Do I understand your mention correctly? > > If so, what is the benefit of it? > > I, unfortunately, couldn't understand the relationship between changing > > timers to the delayed works and these issues. > > Because a work has process context so we can use GFP_KERNEL > allocation rather than GFP_ATOMIC, which is what commit 72e09ad107e7 > addresses. >
Thank you for explaining! I now understand why you suggested it. I will send a v2 patch which will change timers to delay works. Thanks a lot! Taehee Yoo
