public inbox for linux-nfs@vger.kernel.org
 help / color / mirror / Atom feed
From: Boaz Harrosh <bharrosh@panasas.com>
To: Christoph Hellwig <hch@infradead.org>
Cc: Trond Myklebust <Trond.Myklebust@netapp.com>,
	Fred Isaman <iisaman@netapp.com>,
	linux-nfs@vger.kernel.org
Subject: Re: [PATCH 07/13] RFC: pnfs: full mount/umount infrastructure
Date: Mon, 13 Sep 2010 17:14:47 +0200	[thread overview]
Message-ID: <4C8E3FE7.8000404@panasas.com> (raw)
In-Reply-To: <20100913144444.GA31105@infradead.org>

On 09/13/2010 04:44 PM, Christoph Hellwig wrote:
> On Mon, Sep 13, 2010 at 01:06:53PM +0200, Boaz Harrosh wrote:
>> Me two BTW.
>> I will later need a per super-block resources, mainly slabs and work_q.
> 
> Slab caches are always globals, as should workqueues be.  With Tejun's
> recent work most work queues should go away completely.
> 

Right, sorry wrong terminology. I have pages pre-allocation for raid calculation
I always keep a minimum of full stripe raid pages per SB and scale up to the amount
of concurrent writes, with some lazy deallocation. It's all very preliminary.

The thread is so I can issue async writes/reads from io completion. So you
are right this should be global now.

>> I think we should just pass nfss to the LD and let it take care of accessing
>> nfss->nfs_client for the device cache. (It is done so later in the tree, right?)
> 
> You can add this later once a laypout driver actually needing it is
> introduced.
> 

You are right in principle, but this is one place that should be considered
from higher view. What happens is that we can only make the call after an
actual mount when we know the version/pnfs-usage/layout-type to be used.
Then when we know that, we have a per-nfs_client initialization to do.
Later in both blocks and objects we have other things to do peculiar to
each. The one call vector could serve them all. And the overall name of
that vector is therefore init_mountpoint(). The original code was passing
nfs-server as proper for it's name. Later attempts to hide the code future
and comply with files-only needs changed that to be an nfs-client pointer
since that was the only thing used. Now the name is wrong and the place
it is called from is confusing.

I think it is just simple, more clear, to just call it what it is and
pass the proper pointer. For the binary code it is all the same. For
code readability and review, an attempt to clarify just made it more
complicated. So why not just drop it.

Boaz

  reply	other threads:[~2010-09-13 15:14 UTC|newest]

Thread overview: 55+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2010-09-02 18:00 [PATCH 00/13] RFC: pnfs: LAYOUTGET/DEVINFO submission Fred Isaman
2010-09-02 18:00 ` [PATCH 01/13] NFSD: remove duplicate NFS4_STATEID_SIZE Fred Isaman
2010-09-02 18:00 ` [PATCH 02/13] SUNRPC: define xdr_decode_opaque_fixed Fred Isaman
2010-09-02 18:00 ` [PATCH 03/13] RFC: pnfsd, pnfs: protocol level pnfs constants Fred Isaman
2010-09-02 18:00 ` [PATCH 04/13] RFC: nfs: change stateid to be a union Fred Isaman
2010-09-02 18:00 ` [PATCH 05/13] RFC: nfs: ask for layouttypes during fsinfo call Fred Isaman
2010-09-02 18:00 ` [PATCH 06/13] RFC: nfs: set layout driver Fred Isaman
2010-09-02 18:00 ` [PATCH 07/13] RFC: pnfs: full mount/umount infrastructure Fred Isaman
2010-09-10 19:23   ` Trond Myklebust
     [not found]     ` <1284146604.10062.68.camel-rJ7iovZKK19ZJLDQqaL3InhyD016LWXt@public.gmane.org>
2010-09-10 20:53       ` Fred Isaman
2010-09-13 11:06     ` Boaz Harrosh
2010-09-13 14:44       ` Christoph Hellwig
2010-09-13 15:14         ` Boaz Harrosh [this message]
2010-09-13 11:20     ` Benny Halevy
2010-09-10 23:58   ` Christoph Hellwig
2010-09-11  0:07     ` Trond Myklebust
2010-09-13 11:24       ` Benny Halevy
2010-09-13 12:29         ` Trond Myklebust
2010-09-13 14:37           ` Benny Halevy
2010-09-13 16:55             ` Trond Myklebust
2010-09-13 14:28         ` Christoph Hellwig
2010-09-13 14:39           ` Benny Halevy
2010-09-13 15:07   ` Christoph Hellwig
2010-09-13 15:27     ` Fred Isaman
2010-09-02 18:00 ` [PATCH 08/13] RFC: pnfs: filelayout: introduce minimal file layout driver Fred Isaman
2010-09-10 19:31   ` Trond Myklebust
2010-09-10 21:11     ` Fred Isaman
2010-09-10 22:37       ` Trond Myklebust
2010-09-13 10:32         ` Benny Halevy
2010-09-13 13:01           ` Fred Isaman
     [not found]             ` <AANLkTimONZfA6ZX4xtzbmy0NdfEtbwMAi+__PhFYznTn-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
2010-09-13 14:23               ` Benny Halevy
2010-09-13 14:48         ` Christoph Hellwig
2010-09-13 10:16       ` Benny Halevy
2010-09-10 23:56     ` Christoph Hellwig
2010-09-11  0:03       ` Trond Myklebust
2010-09-11  0:07         ` Christoph Hellwig
2010-09-11  0:13           ` Trond Myklebust
2010-09-13 11:28             ` Benny Halevy
2010-09-13 15:08   ` Christoph Hellwig
2010-09-13 15:16     ` Fred Isaman
2010-09-02 18:00 ` [PATCH 09/13] RFC: nfs: create and destroy inode's layout cache Fred Isaman
2010-09-10 19:43   ` Trond Myklebust
     [not found]     ` <1284147785.10062.80.camel-rJ7iovZKK19ZJLDQqaL3InhyD016LWXt@public.gmane.org>
2010-09-10 21:13       ` Fred Isaman
2010-09-13 11:32     ` Benny Halevy
2010-09-02 18:00 ` [PATCH 10/13] RFC: nfs: client needs to maintain list of inodes with active layouts Fred Isaman
2010-09-10 19:59   ` Trond Myklebust
     [not found]     ` <1284148768.10062.94.camel-rJ7iovZKK19ZJLDQqaL3InhyD016LWXt@public.gmane.org>
2010-09-10 21:18       ` Fred Isaman
2010-09-02 18:00 ` [PATCH 11/13] RFC: nfs: retry on certain pnfs errors Fred Isaman
2010-09-02 18:00 ` [PATCH 12/13] RFC: pnfs: add LAYOUTGET and GETDEVICEINFO infrastructure Fred Isaman
2010-09-10 20:11   ` Trond Myklebust
2010-09-10 21:47     ` Fred Isaman
2010-09-10 22:43       ` Trond Myklebust
2010-09-13 14:16       ` Benny Halevy
2010-09-02 18:00 ` [PATCH 13/13] RFC: pnfs: filelayout: add driver's " Fred Isaman
2010-09-10 20:33   ` Trond Myklebust

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=4C8E3FE7.8000404@panasas.com \
    --to=bharrosh@panasas.com \
    --cc=Trond.Myklebust@netapp.com \
    --cc=hch@infradead.org \
    --cc=iisaman@netapp.com \
    --cc=linux-nfs@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox