Date: Fri, 20 Apr 2007 08:54:36 +1000 From: Glen <glen.leeder@nokia.com> To: ext Howard Su <howard0su@gmail.com> Cc: fs@freebsd.org Subject: Re: handle special file type in tmpfs Message-ID: <4627F32C.4030301@nokia.com> In-Reply-To: <f126fae00704190102v3ca39538we34160e589cf3d54@mail.gmail.com> References: <4626EA38.6010703@nokia.com> <4626F339.5040602@nokia.com> <f126fae00704190102v3ca39538we34160e589cf3d54@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
ext Howard Su wrote: >> Glen wrote: >> > Hi Howard, >> > >> > May I ask if you are using the tmpfs port by Rohit Jalan? A link to >> > his work can be found on this page: >> > http://www.freebsd.org/projects/ideas/ > Yes. The ideas page bring me to this working. > >> > >> > I am currently using the BETA3 tmpfs port and have experienced an >> > issue with the FIFO support for tmpfs. I believe that this is because >> > the tmpfs_fifoops.c specifies an incomplete vnode operations vector >> > (vop_vector) for tmpfs. The vop_vector allows you to specify all the >> > vnode operations for a particular filesystem. > I already fix that in my WIP. you can check the code in the perforce. > http://perforce.freebsd.org/depotTreeBrowser.cgi?FSPC=//depot/user/howardsu/truss/sys/fs/tmpfs > > My working currently based on Beta3. >> > >> > If you check out ufs/ufs_vnops.c it specifies a global vop vector for >> > most file types and then a different vop_vector for FIFO file types. >> > Rohit's tmpfs port does a similar thing. I have populated the tmpfs >> > vop_vector in this port and FIFO seems to work a better although I >> > haven't fully tested it yet. Maybe this information helps you get to >> > a solution. > You are right. > >> > >> > I am keen to get tmpfs into the Freebsd tree and have been >> > communicating with Rohit regarding this. He currently doesn't have >> > time to do further improvements and I have offered to help get this >> > tmpfs port into Freebsd. I am awaiting a response to this (we only >> > started talking last week). >> > >> > This port has a few outstanding items: >> > >> > * MP safeness (some locking may be required). I have started testing >> > tmpfs on an MP system. > Can you try my WIP?(I can send you the patch against -Current) I add > some basic locking stuffs. I am glad to work with you together. Now I > don't have a MP system to test. I'll be very happy to try it out and should get a chance to try it today. I think I may need to look for/write some test software to exercise the code in an MP scenario unless you already know of a regression test that does this. >> > * Security audit (not sure what's required for this) >> > * Quota support, ACL work is pending > Do we really need this? What's user scenerio to use Quota or ACL for a > TMPFS? I can be conveienced. I guess not for the first release. I just mentioned this list as I know these things were not yet done. I see that Brook has expressed an interest in some of this stuff but also mentions it is not critical. > >> > * Two data I/O mechanisms are benchmarked, deciding which one to use >> > is still pending > You can help. > > I really help to see you have interesting. It will be better if we can > work together. Besides the locking & fifo fixing, I also working on > the following tasks: > 1. Catch up the main tree changes related FS after Rohit post his patch. > 2. Port regression test cases > 3. Remove un-named union so that tmpfs can build in a kernel I will be glad to help in any way I can. I will start with the MP system testing. Please let me know if you are aware of any other problems or tasks that I can help with. > >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4627F32C.4030301>