From: John Feeney <jfeeney@redhat.com>
Date: Thu, 30 Sep 2010 21:10:52 -0400
Subject: [net] tg3: fix 5717/57765/5719 memory leak
Message-id: <4CA4FCDC.7040608@redhat.com>
Patchwork-id: 28528
O-Subject: [RHEL5.6 PATCH] tg3: Fix 5717/57765/5719 memory leak
Bugzilla: 631963
RH-Acked-by: Stanislaw Gruszka <sgruszka@redhat.com>
RH-Acked-by: David S. Miller <davem@redhat.com>

bz631963
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=631963
tg3: 5717 / 57765 / 5719 devices leak memory

Description of problem:
The rx producer ring does not get cleaned up properly.

Solution:
Move the call to tg3_rx_prodring_free so it is called for each
vector.

>From the commit:
"The rx resources for MSI-X interrupt vector 0 were not being freed
correctly.  This happens because the teardown loop continue's to the
next loop iteration if it detects the tx ring for that vector is not
setup, thus bypassing the rx teardown code.  This patch moves the
call to tg3_rx_prodring_free() earlier in the loop."

Upstream commit:
tg3: Fix a memory leak on 5717+ devices
 b28f6428af279ffb9e97ee00486a29498b7fcfdc

Brew:
Successfully built in Brew for all architectures (task_2785135).

Testing:
Sanity tested with Connectathon on a number of rhts/beaker systems
with tg3 NICs.

Note: We don't have any 5719, 5717, or 57765 NICs so I have asked
Broadcom to test but so far have not heard back. Will report to
list when they do.

Acks would be appreciated. Thanks.

Signed-off-by: Jarod Wilson <jarod@redhat.com>

diff --git a/drivers/net/tg3.c b/drivers/net/tg3.c
index 66bf556..9a2e452 100644
--- a/drivers/net/tg3.c
+++ b/drivers/net/tg3.c
@@ -5758,6 +5758,8 @@ static void tg3_free_rings(struct tg3 *tp)
 	for (j = 0; j < tp->irq_cnt; j++) {
 		struct tg3_napi *tnapi = &tp->napi[j];
 
+		tg3_rx_prodring_free(tp, &tp->prodring[j]);
+
 		if (!tnapi->tx_buffers)
 			continue;
 
@@ -5779,8 +5781,6 @@ static void tg3_free_rings(struct tg3 *tp)
 
 			dev_kfree_skb_any(skb);
 		}
-
-		tg3_rx_prodring_free(tp, &tp->prodring[j]);
 	}
 }