Date: Sun, 28 Aug 2022 21:30:39 +0000 From: bugzilla-noreply@freebsd.org To: bugs@FreeBSD.org Subject: [Bug 265974] SMR has several missing barriers Message-ID: <bug-265974-227-TfyZYucReo@https.bugs.freebsd.org/bugzilla/> In-Reply-To: <bug-265974-227@https.bugs.freebsd.org/bugzilla/> References: <bug-265974-227@https.bugs.freebsd.org/bugzilla/>
next in thread | previous in thread | raw e-mail | index | archive | help
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D265974 --- Comment #9 from Pierre Habouzit <pierre@habouzit.net> --- `atomic_thread_fence_rel()` is a `dmb sy` ? I would have expected that to b= e a `dmb st` .. but indeed it shows I'm not familiar with with FreeBSD: ```C static __inline void atomic_thread_fence_rel(void) { dmb(sy); } ``` and indeed it makes the deferred advance correct then. It also means smr_synchronize() has a stronger semantics than I expected already and makes the smr_poll barrier on entry useless because the `atomic_load_acq_int()` in smr_poll() already give the semantics I was afte= r on entry. I think you might be right that the one on exit of smr_poll() isn't quite needed. I think I just didn't think it through and just took `ck_epoch_synchronize_wait` as an example and just was like "it needs one so smr_scan does too" but trying to come up with a reason why it breaks yields nothing.. --=20 You are receiving this mail because: You are the assignee for the bug.=
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?bug-265974-227-TfyZYucReo>