Attachment 585310 Details for Bug 822189 – patch -- fix loss of task->tk_status after rpc_delay call in xprt_alloc_slot

[patch] patch -- fix loss of task->tk_status after rpc_delay call in xprt_alloc_slot

0001-sunrpc-fix-loss-of-task-tk_status-after-rpc_delay-ca.patch (text/plain), 3.08 KB, created by Jeff Layton on 2012-05-17 20:48:46 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Jeff Layton

Created: 2012-05-17 20:48:46 UTC

Size: 3.08 KB

patch

obsolete

>From f4a6ef69b0e524070c6785b4c320f82ae923b753 Mon Sep 17 00:00:00 2001
>From: Jeff Layton <jlayton@redhat.com>
>Date: Thu, 17 May 2012 16:45:10 -0400
>Subject: [RHEL6 PATCH] sunrpc: fix loss of task->tk_status after rpc_delay call in xprt_alloc_slot
>
>xprt_alloc_slot will call rpc_delay() to make the task wait a bit before
>retrying when it gets back an -ENOMEM error from xprt_dynamic_alloc_slot.
>
>The problem there is that rpc_delay does this:
>
>	rpc_sleep_on(&delay_queue, task, __rpc_atrun);
>
>...and __rpc_atrun will reset the task->tk_status back to 0 when the
>task wakes back up. This makes call_reserveresult abort the task with
>-EIO.
>
>Fix this by adding a new rpc_delay_action() variant that makes allows
>us to pass in a rpc_action to reset the status back to -EAGAIN instead.
>That should prevent the tk_status from being reset back to 0 after the
>task comes back to life.
>
>Signed-off-by: Jeff Layton <jlayton@redhat.com>
>---
> include/linux/sunrpc/sched.h |    1 +
> net/sunrpc/sched.c           |    9 +++++++--
> net/sunrpc/xprt.c            |    7 ++++++-
> 3 files changed, 14 insertions(+), 3 deletions(-)
>
>diff --git a/include/linux/sunrpc/sched.h b/include/linux/sunrpc/sched.h
>index c600035..8d254ca 100644
>--- a/include/linux/sunrpc/sched.h
>+++ b/include/linux/sunrpc/sched.h
>@@ -237,6 +237,7 @@ void		rpc_wake_up(struct rpc_wait_queue *);
> struct rpc_task *rpc_wake_up_next(struct rpc_wait_queue *);
> void		rpc_wake_up_status(struct rpc_wait_queue *, int);
> int		rpc_queue_empty(struct rpc_wait_queue *);
>+void		rpc_delay_action(struct rpc_task *, unsigned long, rpc_action);
> void		rpc_delay(struct rpc_task *, unsigned long);
> void *		rpc_malloc(struct rpc_task *, size_t);
> void		rpc_free(void *);
>diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c
>index 39c75ea..153fe95 100644
>--- a/net/sunrpc/sched.c
>+++ b/net/sunrpc/sched.c
>@@ -591,13 +591,18 @@ static void __rpc_atrun(struct rpc_task *task)
> 	task->tk_status = 0;
> }
> 
>+void rpc_delay_action(struct rpc_task *task, unsigned long delay, rpc_action action)
>+{
>+	task->tk_timeout = delay;
>+	rpc_sleep_on(&delay_queue, task, action);
>+}
>+
> /*
>  * Run a task at a later time
>  */
> void rpc_delay(struct rpc_task *task, unsigned long delay)
> {
>-	task->tk_timeout = delay;
>-	rpc_sleep_on(&delay_queue, task, __rpc_atrun);
>+	rpc_delay_action(task, delay, __rpc_atrun);
> }
> EXPORT_SYMBOL_GPL(rpc_delay);
> 
>diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c
>index 6dc2ec3..d9125ca 100644
>--- a/net/sunrpc/xprt.c
>+++ b/net/sunrpc/xprt.c
>@@ -964,6 +964,11 @@ static bool xprt_dynamic_free_slot(struct rpc_xprt *xprt, struct rpc_rqst *req)
> 	return false;
> }
> 
>+static void xprt_alloc_reset_status(struct rpc_task *task)
>+{
>+	task->tk_status = -EAGAIN;
>+}
>+
> static void xprt_alloc_slot(struct rpc_task *task)
> {
> 	struct rpc_xprt	*xprt = task->tk_xprt;
>@@ -979,7 +984,7 @@ static void xprt_alloc_slot(struct rpc_task *task)
> 		goto out_init_req;
> 	switch (PTR_ERR(req)) {
> 	case -ENOMEM:
>-		rpc_delay(task, HZ >> 2);
>+		rpc_delay_action(task, HZ >> 2, xprt_alloc_delay_reset_status);
> 		dprintk("RPC:       dynamic allocation of request slot "
> 				"failed! Retrying\n");
> 		break;
>-- 
>1.7.1
>

Actions: View | Diff

Attachments on bug 822189: 584998 | 585018 | 585310 | 585624