Hi Jiří.

On Fri, Mar 13, 2015 at 06:42:33PM +0100, Jiří Zárevúcky 
<[email protected]> wrote:
> At the moment, I would suggest that we (re)open a discussion about what
> each of us expects from the future file system services and libraries, and
> figure out steps to get there. I am thinking Michal Koutny who is currently
> working on service manager might provide some valuable input with respect
> to file system and block device management.
From my (current) point of view, main responsibility of VFS is brokering
connections from client to appropriate FS servers. I expect that VFS
would communicate with the service manager via general interface for
brokers. Effectively, it would mean that service manager would be aware
of the state of mount units (terminology I borrowed from systemd) whose
(un)mount operation was triggered manualy (and not by service manager).
Changes in current VFS will be minimal and could be IMO realized in any
future version of VFS too.

In the described model, block devices are plain servers that are brokered
via Location Service, i.e. current state.


Then there are some ideas that would allow early startup process to be
more streamlined and would lead to more convenient environment. Some of
them are in this ticket [2].

What I think may be benefitial is some sort of filesystem that won't
need an underlying block device. That could reduce number of tasks
necessary to be launched directly by the kernel in the initial phase
(locsrv and initrd), thus needing less tasks to operate in the
restricted initial environment (e.g. currently FS servers cannot use
logger). It's rather a brainstorm idea as I don't know whether it would
be better to have something like tmpfs crossbred with initrd or employ a
caching mechanism directly inside VFS.


Summarized -- current VFS implementation is conceptually sufficient
(regarding service management), later it would be nice to have "less
dependent" in-memory filesystem and finer control over placing mounted
filesystem into VFS namespace (chroot, pivot mount).


Michal


[1] http://trac.helenos.org/ticket/525#comment:7
[2] http://trac.helenos.org/ticket/447

_______________________________________________
HelenOS-devel mailing list
[email protected]
http://lists.modry.cz/listinfo/helenos-devel

Reply via email to