Date: Sat, 20 Feb 2016 13:49:20 +0200 From: Konstantin Belousov <kostikbel@gmail.com> To: Stefan Esser <se@freebsd.org> Cc: Eric van Gyzen <vangyzen@FreeBSD.org>, Bryan Drewery <bdrewery@FreeBSD.org>, src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org, dwmalone@FreeBSD.org Subject: Re: svn commit: r295800 - head/usr.bin/cap_mkdb Message-ID: <20160220114919.GZ91220@kib.kiev.ua> In-Reply-To: <56C833CD.4040309@freebsd.org> References: <201602190842.u1J8gDOc015177@repo.freebsd.org> <56C7B60E.8080002@FreeBSD.org> <56C7D019.3020807@FreeBSD.org> <56C833CD.4040309@freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, Feb 20, 2016 at 10:37:17AM +0100, Stefan Esser wrote: > I had assumed that one of the guarantees soft-updates makes is that > data and meta-data operations are ordered relative to each other. This is not true, SU never provided such guarantee. SU only ensures the correctness of the metadata for any moment in time, or innocent inconsistencies, where operation cannot be made atomic and consistent simultaneously. In other words, in this case, SU guarantees that the data blocks which are allocated for writes, are not used by other inode, as recorded by the metadata on disk, _if_ blocks are recorded in the on-disk inode. It makes absolutely no guarantees about the flushing of the data buffers.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20160220114919.GZ91220>