From: NeilBrown <neilb@suse.com>
To: Trond Myklebust <trondmy@primarydata.com>,
"anna.schumaker\@netapp.com" <anna.schumaker@netapp.com>
Cc: "linux-nfs\@vger.kernel.org" <linux-nfs@vger.kernel.org>
Subject: Re: [PATCH 2/8] NFSv4: don't let hanging mounts block other mounts
Date: Sat, 19 Aug 2017 10:50:31 +1000 [thread overview]
Message-ID: <87378omld4.fsf@notabene.neil.brown.name> (raw)
In-Reply-To: <1503083150.11511.4.camel@primarydata.com>
[-- Attachment #1: Type: text/plain, Size: 2481 bytes --]
On Fri, Aug 18 2017, Trond Myklebust wrote:
> On Fri, 2017-08-18 at 17:12 +1000, NeilBrown wrote:
>> If you try an NFSv4 mount from an inaccessible server, it will hang
>> as
>> you would expect.
>> If you then try an NFSv4 mount from a different accessible server,
>> it will also hang. This is not expected.
>>
>> The second mount is blocked in
>> nfs4_init_client()
>> -> nfs4_discover_server_trunking()
>> -> nfs40_discover_server_trunking()
>> -> nfs40_walk_client_list()
>> -> nfs4_match_client()
>> -> nfs_wait_client_init_complete()
>> It is waiting for the first mount to complete so that it can then
>> see if the two servers are really one and the same.
>>
>> It is not necessary to wait here when an nfs_client cl_cons_state is
>> NFS_CS_INITING. Such a client will, after changing cl_cons_state,
>> call
>> nfs4_discover_server_trunking() itself. So if the current client
>> just
>> skips those clients, trunking will still be discovered if necessary.
>>
>> I am unsure of situation with NFS_CS_SESSION_INITING, but I suspect
>> that the comment "Wait for CREATE_SESSION to finish" implies that
>> it is only clients in NFS_CS_SESSION_INITING that need to be waited
>> for.
>>
>> Signed-off-by: NeilBrown <neilb@suse.com>
>> ---
>> fs/nfs/nfs4client.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/fs/nfs/nfs4client.c b/fs/nfs/nfs4client.c
>> index e9bea90dc017..d8b9b7ff19a9 100644
>> --- a/fs/nfs/nfs4client.c
>> +++ b/fs/nfs/nfs4client.c
>> @@ -482,7 +482,7 @@ static int nfs4_match_client(struct
>> nfs_client *pos, struct nfs_client *new,
>> * remaining fields in "pos", especially the client
>> * ID and serverowner fields. Wait for CREATE_SESSION
>> * to finish. */
>> - if (pos->cl_cons_state > NFS_CS_READY) {
>> + if (pos->cl_cons_state == NFS_CS_SESSION_INITING) {
>> atomic_inc(&pos->cl_count);
>> spin_unlock(&nn->nfs_client_lock);
>
> This could cause us to declare a false positive match with a client
> that is uninitialised.
Thanks for the review.
A positive match is reported by returning zero.
If cl_cons_state is not NFS_CS_READY, nfs4_match_client() will not
return zero.
So I don't see how a false positive is possible.
A false negative might be possible with an uninitialized client, but
once that client is initialised, it will call walk_client_list and find
the match. Won't it?
Thanks,
NeilBrown
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 832 bytes --]
next prev parent reply other threads:[~2017-08-19 0:50 UTC|newest]
Thread overview: 17+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-08-18 7:12 [PATCH 0/8] Assorted NFS patches NeilBrown
2017-08-18 7:12 ` [PATCH 4/8] NFS: flush out dirty data on file fput() NeilBrown
2017-08-18 19:15 ` Trond Myklebust
2017-08-19 1:02 ` NeilBrown
2017-08-19 1:27 ` Trond Myklebust
2017-08-21 1:35 ` NeilBrown
2017-08-23 11:12 ` Jeff Layton
2017-08-18 7:12 ` [PATCH 1/8] SUNRPC: ECONNREFUSED should cause a rebind NeilBrown
2017-08-18 14:40 ` Chuck Lever
2017-08-18 7:12 ` [PATCH 2/8] NFSv4: don't let hanging mounts block other mounts NeilBrown
2017-08-18 19:05 ` Trond Myklebust
2017-08-19 0:50 ` NeilBrown [this message]
2017-08-18 7:12 ` [PATCH 3/8] NFS: don't expect errors from mempool_alloc() NeilBrown
2017-08-18 7:12 ` [PATCH 7/8] NFS: remove error handling for callers of rpc_new_task() NeilBrown
2017-08-18 7:12 ` [PATCH 6/8] NFS: flush data when locking a file to ensure cache coherence for mmap NeilBrown
2017-08-18 7:12 ` [PATCH 8/8] NFSv4.1: don't use machine credentials for CLOSE when using 'sec=sys' NeilBrown
2017-08-18 7:12 ` [PATCH 5/8] SUNRPC: remove some dead code NeilBrown
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=87378omld4.fsf@notabene.neil.brown.name \
--to=neilb@suse.com \
--cc=anna.schumaker@netapp.com \
--cc=linux-nfs@vger.kernel.org \
--cc=trondmy@primarydata.com \
/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;
as well as URLs for NNTP newsgroup(s).