[PATCH] net: rds: fix per-cpu helper usage

[PATCH] net: rds: fix per-cpu helper usage

[Gerald Schaefer]

commit ae4b46e9d "net: rds: use this_cpu_* per-cpu helper" broke per-cpu
handling for rds. chpfirst is the result of __this_cpu_read(), so it is
an absolute pointer and not __percpu. Therefore, __this_cpu_write()
should not operate on chpfirst, but rather on cache->percpu->first, just
like __this_cpu_read() did before.

Cc: <stable@vger.kernel.org> # 3.8+
Signed-off-byd Gerald Schaefer <gerald.schaefer@de.ibm.com>
---
 net/rds/ib_recv.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/net/rds/ib_recv.c b/net/rds/ib_recv.c
index 8eb9501..b7ebe23 100644
--- a/net/rds/ib_recv.c
+++ b/net/rds/ib_recv.c
@@ -421,8 +421,7 @@ static void rds_ib_recv_cache_put(struct list_head *new_item,
 				 struct rds_ib_refill_cache *cache)
 {
 	unsigned long flags;
-	struct list_head *old;
-	struct list_head __percpu *chpfirst;
+	struct list_head *old, *chpfirst;
 
 	local_irq_save(flags);
 
@@ -432,7 +431,7 @@ static void rds_ib_recv_cache_put(struct list_head *new_item,
 	else /* put on front */
 		list_add_tail(new_item, chpfirst);
 
-	__this_cpu_write(chpfirst, new_item);
+	__this_cpu_write(cache->percpu->first, new_item);
 	__this_cpu_inc(cache->percpu->count);
 
 	if (__this_cpu_read(cache->percpu->count) < RDS_IB_RECYCLE_BATCH_COUNT)
@@ -452,7 +451,7 @@ static void rds_ib_recv_cache_put(struct list_head *new_item,
 	} while (old);
 
 
-	__this_cpu_write(chpfirst, NULL);
+	__this_cpu_write(cache->percpu->first, NULL);
 	__this_cpu_write(cache->percpu->count, 0);
 end:
 	local_irq_restore(flags);
-- 
1.8.4.5

Re: [PATCH] net: rds: fix per-cpu helper usage

[David Miller]

From: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Date: Thu, 16 Jan 2014 16:54:48 +0100

> commit ae4b46e9d "net: rds: use this_cpu_* per-cpu helper" broke per-cpu
> handling for rds. chpfirst is the result of __this_cpu_read(), so it is
> an absolute pointer and not __percpu. Therefore, __this_cpu_write()
> should not operate on chpfirst, but rather on cache->percpu->first, just
> like __this_cpu_read() did before.
> 
> Cc: <stable@vger.kernel.org> # 3.8+
> Signed-off-byd Gerald Schaefer <gerald.schaefer@de.ibm.com>

Applied and queued up for -stable, thanks.