On Tue, 2018-01-23 at 17:09 -0500, J. Bruce Fields wrote:
> 
> The object still has useful information in it so we can't just
> reinitalize it completely.  I guess we could make nlm_get_host do
> 
>       if (refcount_read(&host->h_count))
>               refcount_inc(&host->h_count);
>       else
>               refcount_set(&host->h_count, 1);
> 
> Or we could just change the code so the refcount is always 1 higher
> in
> the NFS server case, so "1" instead of "0" is used to mean "nobody's
> using this, you can garbage collect this", and then it won't go to 0
> until the garbage collector actually destroys it.
> 
> This isn't an unusual pattern, what have other subsystems been doing?
> 

Hi Bruce,

Sorry I forgot about the issues with the server garbage collector, and
I applied these patches to my linux-next a couple of weeks ago.

What say we fix the issue with something like the following?

8<------------------------------------------------------------
>From 83ce0f55ca54337a573f1d70038714815a9cd645 Mon Sep 17 00:00:00 2001
From: Trond Myklebust <[email protected]>
Date: Tue, 23 Jan 2018 19:39:04 -0500
Subject: [PATCH] lockd: Fix server refcounting

The server shouldn't actually delete the struct nlm_host until it hits
the garbage collector. In order to make that work correctly with the
refcount API, we can bump the refcount by one, and then use
refcount_dec_if_one() in the garbage collector.

Signed-off-by: Trond Myklebust <[email protected]>
---
 fs/lockd/host.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/fs/lockd/host.c b/fs/lockd/host.c
index 7d6ab72bbe65..d35cd6be0675 100644
--- a/fs/lockd/host.c
+++ b/fs/lockd/host.c
@@ -388,6 +388,8 @@ struct nlm_host *nlmsvc_lookup_host(const struct svc_rqst 
*rqstp,
        ln->nrhosts++;
        nrhosts++;
 
+       refcount_inc(&host->h_count);
+
        dprintk("lockd: %s created host %s (%s)\n",
                __func__, host->h_name, host->h_addrbuf);
 
@@ -662,8 +664,7 @@ nlm_gc_hosts(struct net *net)
        for_each_host_safe(host, next, chain, nlm_server_hosts) {
                if (net && host->net != net)
                        continue;
-               if (refcount_read(&host->h_count) || host->h_inuse
-                || time_before(jiffies, host->h_expires)) {
+               if (host->h_inuse || time_before(jiffies, host->h_expires)) {
                        dprintk("nlm_gc_hosts skipping %s "
                                "(cnt %d use %d exp %ld net %x)\n",
                                host->h_name, refcount_read(&host->h_count),
@@ -671,7 +672,8 @@ nlm_gc_hosts(struct net *net)
                                host->net->ns.inum);
                        continue;
                }
-               nlm_destroy_host_locked(host);
+               if (refcount_dec_if_one(&host->h_count))
+                       nlm_destroy_host_locked(host);
        }
 
        if (net) {
-- 
2.14.3

-- 
Trond Myklebust
Linux NFS client maintainer, PrimaryData
[email protected]

Reply via email to