Date: Tue, 12 Sep 2017 17:11:39 +0300 From: Andriy Gapon <avg@FreeBSD.org> To: Warner Losh <imp@bsdimp.com> Cc: freebsd-geom@FreeBSD.org Subject: Re: "unspoilable" geom labels Message-ID: <62cda94e-751d-3d59-914c-c5242a8f1d81@FreeBSD.org> In-Reply-To: <CANCZdfrVVfxC655v0fhT=h0fciVxyS8xmt5vOVWpjkp-fPGPqg@mail.gmail.com> References: <648bcd86-5ef8-b58e-ed04-48880f867fc0@FreeBSD.org> <CANCZdfrVVfxC655v0fhT=h0fciVxyS8xmt5vOVWpjkp-fPGPqg@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 12/09/2017 16:28, Warner Losh wrote: > I recently added aliases, and the thought was that once the kinks are worked out > they could also be used for labels and we could eliminate the extra geom nodes. I noticed that change, thanks! There is one thing that, in my opinion, is not perfect about the aliases unless I misunderstand the code. It's that unlike the current approach the aliasing is not automatically propagated up the GEOM graph and each geom class needs to be made aware of the aliases. I mean that currently we create a new geom that carries the label name and all the tasting, etc automatically happens for. It has its own problems as we know, but there is a lot of auto-magic too. With the aliases there are no additional geoms, so any classes attaching on top of geoms with aliases need to handle the aliases. E.g. the partition classes need to create an alias for each partition, etc. Also, the aliases seem to be rather adequate for the userland consumers, but they seem to be not as convenient for the in-kernel consumers that expect the names to be names of GEOM providers. But maybe I am extrapolating too much from the existing code and the bigger design for the GEOM aliases handles the mentioned issues. -- Andriy Gapon
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?62cda94e-751d-3d59-914c-c5242a8f1d81>