Attachment 295447 Details for Bug 432393 – patch -- fix several problems with callback thread shutdown

[patch] patch -- fix several problems with callback thread shutdown

0025-BZ-432393-NFSv4-fix-problems-with-callback-thread.patch (text/plain), 2.69 KB, created by Jeff Layton on 2008-02-20 20:50:07 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Jeff Layton

Created: 2008-02-20 20:50:07 UTC

Size: 2.69 KB

patch

obsolete

>From 0aedfc62fee16aa471b0a92b607399484180cdd9 Mon Sep 17 00:00:00 2001
>From: Jeff Layton <jlayton@redhat.com>
>Date: Tue, 19 Feb 2008 13:26:57 -0500
>Subject: [PATCH] BZ#432393: NFSv4: fix problems with callback thread shutdown
>
>This patch fixes several problems when the NFSv4 callback thread is
>shut down:
>
>1) The callback thread doesn't call svc_exit_thread() on shutdown and
>   so doesn't attempt to clean up the svc_rqst and associated info,
>   leading to a memory leak. The patch adds this call.
>
>2) The reference count of the thread stays artificially high and so
>   even when svc_exit_thread() is called, it doesn't actually clean
>   up anything. The fix there is to make sure that nfs_callback_up()
>   calls svc_destroy on both success and failure.
>
>3) The callback thread is signalled by nfs_callback_down() but the
>   thread never flushes this signal. This means that all of the tasks
>   done at thread shutdown are done with a signal pending. This leads
>   to some spurious error messages and problems unregistering the
>   port with the portmapper.
>---
> fs/nfs/callback.c |   18 +++++++++++++-----
> 1 files changed, 13 insertions(+), 5 deletions(-)
>
>diff --git a/fs/nfs/callback.c b/fs/nfs/callback.c
>index 00dbab6..db86d2c 100644
>--- a/fs/nfs/callback.c
>+++ b/fs/nfs/callback.c
>@@ -73,6 +73,8 @@ static void nfs_callback_svc(struct svc_rqst *rqstp)
> 		svc_process(serv, rqstp);
> 	}
> 
>+	flush_signals(current);
>+	svc_exit_thread(rqstp);
> 	nfs_callback_info.pid = 0;
> 	complete(&nfs_callback_info.stopped);
> 	unlock_kernel();
>@@ -84,7 +86,7 @@ static void nfs_callback_svc(struct svc_rqst *rqstp)
>  */
> int nfs_callback_up(void)
> {
>-	struct svc_serv *serv;
>+	struct svc_serv *serv = NULL;
> 	struct svc_sock *svsk;
> 	int ret = 0;
> 
>@@ -101,7 +103,7 @@ int nfs_callback_up(void)
> 	/* FIXME: We don't want to register this socket with the portmapper */
> 	ret = svc_makesock(serv, IPPROTO_TCP, 0);
> 	if (ret < 0)
>-		goto out_destroy;
>+		goto out_err;
> 	if (!list_empty(&serv->sv_permsocks)) {
> 		svsk = list_entry(serv->sv_permsocks.next,
> 				struct svc_sock, sk_list);
>@@ -111,15 +113,21 @@ int nfs_callback_up(void)
> 		BUG();
> 	ret = svc_create_thread(nfs_callback_svc, serv);
> 	if (ret < 0)
>-		goto out_destroy;
>+		goto out_err;
> 	nfs_callback_info.serv = serv;
> 	wait_for_completion(&nfs_callback_info.started);
> out:
>+	/*
>+	 * svc_create creates the svc_serv with sv_nrthreads == 1, and then
>+	 * svc_create_thread increments that. So we need to call svc_destroy
>+	 * on both success and failure so that the refcount is 1 when the
>+	 * thread exits.
>+	 */
>+	if (serv)
>+		svc_destroy(serv);
> 	up(&nfs_callback_sema);
> 	unlock_kernel();
> 	return ret;
>-out_destroy:
>-	svc_destroy(serv);
> out_err:
> 	nfs_callback_info.users--;
> 	goto out;
>-- 
>1.5.3.6
>

Actions: View | Diff

Attachments on bug 432393: 295316 | 295447