Skip to contentSkip to conversation

Creation of a readdirplus-like system call

LSF/MM TOPIC Steven Whitehouse 6 participants 6 messages Jan 23, 2013 – Jan 24, 2013

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

Hi,

As part of the work we've been doing in relation to integration between NFS/Samba and GFS2, Abhi has been working on a readdirplus system call in order to investigate the issues involved with creating such a call. It is still early days yet, but by April there should be some interesting results to present.

Please add Abhi to the attendee list as well as myself.

Also, it has been some time since we had a NFS/Samba meeting to discuss the other issues which are pending, such as locking, ACLs, etc. So we could perhaps also allow some time to do that face to face rather than over the phone as we've been doing up til now,

Steve.

#1 lore raw

Hello Steven,

Quoted 7 lines · Steven Whitehouse

Hi,

As part of the work we've been doing in relation to integration between NFS/Samba and GFS2, Abhi has been working on a readdirplus system call in order to investigate the issues involved with creating such a call. It is still early days yet, but by April there should be some interesting results to present.

I think having such a call would be good for other cluster file systems as well, e.g. Lustre and FhGFS.

Quoted 7 lines

Please add Abhi to the attendee list as well as myself.

Also, it has been some time since we had a NFS/Samba meeting to discuss the other issues which are pending, such as locking, ACLs, etc. So we could perhaps also allow some time to do that face to face rather than over the phone as we've been doing up til now,

Among others, locking is another issue for cluster file systems. For example Lustre and FhGFS currently disable cluster wide locks by default and add a configuration parameter for cluster wide locks. The reason is simply that one mostly only needs local locks and that cluster wide locks involve network messages and are therefore rather expensive. Are you doing any work into that direction?

I definitely need to create a list, but I think several other system calls could be optimized for cluster file systems.

Thanks,
Bernd

#2 lore raw
In reply to Steven Whitehouse
Quoted 1 line · Steven Whitehouse

Please add Abhi to the attendee list as well as myself.

Just as a point of clarification: no attendee request from the individual, as specified in the CFP, no invitation.

Also, the way I compile the list of requests is from thread heads ... that means don't send your attendee request as a reply to something else either otherwise it might get missed.

Thanks,

James

#3 lore raw
In reply to Steven Whitehouse
Quoted 8 lines · Steven Whitehouse

Hi,

As part of the work we've been doing in relation to integration between NFS/Samba and GFS2, Abhi has been working on a readdirplus system call in order to investigate the issues involved with creating such a call. It is still early days yet, but by April there should be some interesting results to present.

Quoted 6 lines

Please add Abhi to the attendee list as well as myself.

Also, it has been some time since we had a NFS/Samba meeting to discuss the other issues which are pending, such as locking, ACLs, etc. So we could perhaps also allow some time to do that face to face rather than over the phone as we've been doing up til now,

This would be an opportunity to discuss issues related to FedFS support, such as how junctions are represented for NFS and SMB/CIFS in physical file systems.

Signature · 2 lines
Chuck Lever
chuck[dot]lever[at]oracle[dot]com
#4 lore raw

[LSF/MM TOPIC] [ATTEND] readdirplus-like (xgetdents) system call

In reply to Steven Whitehouse

Like Steve mentioned in his email earlier, I'm working on a readdirplus-like (I'm calling it xgetdents for now) system call. This came up as a desired filesystem feature from GFS2 from a Clustered Samba perspective. I hope to have some results to present by April.

Could you please add me to the attendee list.

I also second Steve's thought of having a session to discuss various clustered samba/NFS issues we haven't talked about in a while.

Cheers!
--Abhi

----- Original Message -----

Quoted 35 lines

From: "Steven Whitehouse" <[email protected]>
To: [email protected], [email protected]
Cc: "Abhijith Das" <[email protected]>
Sent: Wednesday, January 23, 2013 3:43:49 AM
Subject: [LSF/MM TOPIC] [ATTEND] Creation of a readdirplus-like system call

Hi,

As part of the work we've been doing in relation to integration between
NFS/Samba and GFS2, Abhi has been working on a readdirplus system call
in order to investigate the issues involved with creating such a call.
It is still early days yet, but by April there should be some interesting results to present.

Please add Abhi to the attendee list as well as myself.

Also, it has been some time since we had a NFS/Samba meeting to discuss
the other issues which are pending, such as locking, ACLs, etc. So we could perhaps also allow some time to do that face to face rather than
over the phone as we've been doing up til now,

Steve.

--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#5 lore raw

[LSF/MM TOPIC] [ATTEND] New API's for better exporting of FSs from user-mode daemons, like readdirplus and more

In reply to Steven Whitehouse

This is of high interest also for the NFS-Ganesha project, a free software NFS user-mode daemon project. (CC: Ganesha ML)

I would (highly) second this topic, please, which is of cardinal importance for any such sharing servers from user-mode.

Also in Ganesha we do not use (and cannot use) any form of POSIX readdir, we must directly open the directory, and retrieve the undocumented dirents structures. There are lots of reasons why we do this, but mainly because of the directory-seek and the readdir continuation "cookie". Which is impossible with POSIX.

I can't, remember all the details right now, but Trond has showed us how POSIX readdir is totally broken in regard to the continuation "cookie", because POSIX assumes that the directory is a flat array.

So surly any implementation should also solve this issue.

Also in Ganesha we have portability layers for other OSs, and we have been porting some Linux APIs to FreeBSD, A new proposal should be well defined and have Portability and "generic-ness" in mind. For example it should not only map readdirplus call from an NFS daemon to a Local file system, it should also fit well with an application API to an NFS client, with maximum efficiency and minimum overhead.

Any implementation should also consider the in Kernel daemon. Any generic VFS code that could be shared should be considered. KNFSD should not be duplicating code already done by the above IOCTL layer.

I would also like to propose that we might want to broaden this topic to more API's that we might need from user-mode, like delegation locks, and other. (We had a list of them I'll try to collect and produce a list)

Thanks, and Steven, please forgive me if you feel I'm hijacking your topic, if you feel our topics should be kept separate, please tell me and I will send it as a separate topic.

Thanks
Boaz

Quoted 15 lines · Steven Whitehouse

Hi,

As part of the work we've been doing in relation to integration between NFS/Samba and GFS2, Abhi has been working on a readdirplus system call in order to investigate the issues involved with creating such a call. It is still early days yet, but by April there should be some interesting results to present.

Please add Abhi to the attendee list as well as myself.

Also, it has been some time since we had a NFS/Samba meeting to discuss the other issues which are pending, such as locking, ACLs, etc. So we could perhaps also allow some time to do that face to face rather than over the phone as we've been doing up til now,

Quoted 8 lines

Steve.

--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html

--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html