On Mon, Nov 10, 2008 at 04:48:42PM -0500, Trond Myklebust wrote:
> On Mon, 2008-11-10 at 16:18 -0500, J. Bruce Fields wrote:
> > On Mon, Nov 10, 2008 at 03:37:57PM -0500, bfields wrote:
> > > On Mon, Nov 10, 2008 at 03:35:22PM -0500, Trond Myklebust wrote:
> > > > BTW: This is racy. You have to set first_open _after_ you take the
> > > > inode->i_mutex.
> > >
> > > Whoops! Good catch, thanks--fixed.
> >
> > (Updated version follows.)
>
> Err... This doesn't look right. This would be a mix of 4/9 and 5/9...
Erp, right. I folded the fix for the race into the wrong patch. New
5/9 looks like the below. (I'm also keeping the whole series in the
for-trond branch at
git://linux-nfs.org/~bfields/linux.git for-trond
But let me know if you want any patches resent at some point.)
--b.
commit 4e0278688f280ff815505c8cd5287e4e9d553990
Author: J. Bruce Fields <bfields@xxxxxxxxxxxxxx>
Date: Fri Nov 7 18:53:24 2008 -0500
rpc: call release_pipe only on last close
I can't see any reason we need to call this until either the kernel or
the last gssd closes the pipe.
Also, this allows to guarantee that open_pipe and release_pipe are
called strictly in pairs; open_pipe on gssd's first open, release_pipe
on gssd's last close (or on the close of the kernel side of the pipe, if
that comes first).
That will make it very easy for the gss code to keep track of which
pipes gssd is using.
Signed-off-by: J. Bruce Fields <bfields@xxxxxxxxxxxxxx>
diff --git a/net/sunrpc/rpc_pipe.c b/net/sunrpc/rpc_pipe.c
index 57342aa..a68aaac 100644
--- a/net/sunrpc/rpc_pipe.c
+++ b/net/sunrpc/rpc_pipe.c
@@ -126,13 +126,14 @@ rpc_close_pipes(struct inode *inode)
{
struct rpc_inode *rpci = RPC_I(inode);
struct rpc_pipe_ops *ops;
+ int need_release;
mutex_lock(&inode->i_mutex);
ops = rpci->ops;
if (ops != NULL) {
LIST_HEAD(free_list);
-
spin_lock(&inode->i_lock);
+ need_release = rpci->nreaders != 0 || rpci->nwriters != 0;
rpci->nreaders = 0;
list_splice_init(&rpci->in_upcall, &free_list);
list_splice_init(&rpci->pipe, &free_list);
@@ -141,7 +142,7 @@ rpc_close_pipes(struct inode *inode)
spin_unlock(&inode->i_lock);
rpc_purge_list(rpci, &free_list, ops->destroy_msg, -EPIPE);
rpci->nwriters = 0;
- if (ops->release_pipe)
+ if (need_release && ops->release_pipe)
ops->release_pipe(inode);
cancel_delayed_work_sync(&rpci->queue_timeout);
}
@@ -196,6 +197,7 @@ rpc_pipe_release(struct inode *inode, struct file *filp)
{
struct rpc_inode *rpci = RPC_I(inode);
struct rpc_pipe_msg *msg;
+ int last_close;
mutex_lock(&inode->i_mutex);
if (rpci->ops == NULL)
@@ -222,7 +224,8 @@ rpc_pipe_release(struct inode *inode, struct file *filp)
rpci->ops->destroy_msg, -EAGAIN);
}
}
- if (rpci->ops->release_pipe)
+ last_close = rpci->nwriters == 0 && rpci->nreaders == 0;
+ if (last_close && rpci->ops->release_pipe)
rpci->ops->release_pipe(inode);
out:
mutex_unlock(&inode->i_mutex);
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
