Date: Wed, 16 Oct 2019 16:04:13 +0300 From: Andriy Gapon <avg@FreeBSD.org> To: d@delphij.net, "freebsd-fs@freebsd.org" <freebsd-fs@freebsd.org> Subject: Re: solaris assert: (dn->dn_id_flags & DN_ID_NEW_EXIST) == 0 Message-ID: <57f9cce9-8190-4694-318b-3676952bf0b7@FreeBSD.org> In-Reply-To: <347bc8c9-5cf6-8cc9-129d-f1b1a07a3579@delphij.net> References: <ceed4f94-a079-74e3-92e4-8e0327109583@delphij.net> <347bc8c9-5cf6-8cc9-129d-f1b1a07a3579@delphij.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On 13/10/2019 12:38, Xin Li wrote: > On 10/12/19 12:36, Xin Li via freebsd-fs wrote: >> Not sure how this was triggered (yet; this is mostly an idle system) but >> will try to find out. > > It seems that this is related to the recent large dnode change (kernel > without r353176 r353304 would work just fine). It seems like I can > trigger it with some build-type workload (by doing portmaster -Bd icu > postfix ...), but it's not always triggerable and one has to build > multiple times to catch it. > > I think OpenZFS aa02ea01948372a32cbf08bfc31c72c32e3fc81e would fix this > (Note: 424fd7c3e080255935646d2beaa2655c116cc37a added zfs_ to all > refcount_* functions but it was not incorporated to FreeBSD, it should > either be applied, or you would need to change the additional > zfs_refcount_ to refcount_ when applying that revision). My system > seems to stand after applying it. > > Could you please import it or both? Thank you very much for the debugging and research! And sorry for the trouble. I haven't done enough of "look-ahead" through illumos commits when importing the change. Both requested commits are in FreeBSD now. -- Andriy Gapon
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?57f9cce9-8190-4694-318b-3676952bf0b7>