netdev - [PATCH 3/9] net: davinci_cpdma: requeue on early end-of-queue

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Mon, 13 Sep 2010 14:07:25 -0400
From:	Cyril Chemparathy <cyril@...com>
To:	netdev@...r.kernel.org,
	davinci-linux-open-source@...ux.davincidsp.com,
	linux-omap@...r.kernel.org
Cc:	michael.williamson@...ticallink.com, caglarakyuz@...il.com,
	bparrot@...com, Cyril Chemparathy <cyril@...com>
Subject: [PATCH 3/9] net: davinci_cpdma: requeue on early end-of-queue

When chaining descriptors to the end of a cpdma queue, there is a chance that
the cpdma engine has already traversed the last descriptor and signalled an
end-of-queue.  The original cpdma code detected this condition and requeued
descriptors on submission.

With this patch, an early end-of-queue condition is detected on descriptor
submission as well.  This is necessary on davinci style emac controllers, and
prevents transmit timeouts when sending out a burst of packets.

Signed-off-by: Cyril Chemparathy <cyril@...com>
Signed-off-by: Michael Williamson <michael.williamson@...ticallink.com>
Signed-off-by: Caglar Akyuz <caglarakyuz@...il.com>
---
 drivers/net/davinci_cpdma.c |   12 ++++++++++--
 drivers/net/davinci_cpdma.h |    1 +
 2 files changed, 11 insertions(+), 2 deletions(-)

diff --git a/drivers/net/davinci_cpdma.c b/drivers/net/davinci_cpdma.c
index ab7ecd7..abc2b14 100644
--- a/drivers/net/davinci_cpdma.c
+++ b/drivers/net/davinci_cpdma.c
@@ -488,6 +488,7 @@ static void __cpdma_chan_submit(struct cpdma_chan *chan,
 	struct cpdma_desc __iomem	*prev = chan->tail;
 	struct cpdma_desc_pool		*pool = ctlr->pool;
 	dma_addr_t			desc_dma;
+	u32				mode;
 
 	desc_dma = desc_phys(pool, desc);
 
@@ -507,8 +508,10 @@ static void __cpdma_chan_submit(struct cpdma_chan *chan,
 	chan->stats.tail_enqueue++;
 
 	/* next check if EOQ has been triggered already */
-	if (desc_read(prev, hw_mode) & CPDMA_DESC_EOQ &&
-	    chan->state == CPDMA_STATE_ACTIVE) {
+	mode = desc_read(prev, hw_mode);
+	if (((mode & (CPDMA_DESC_EOQ | CPDMA_DESC_OWNER)) == CPDMA_DESC_EOQ) &&
+	    (chan->state == CPDMA_STATE_ACTIVE)) {
+		desc_write(prev, hw_mode, mode & ~CPDMA_DESC_EOQ);
 		chan_write(chan, hdp, desc_dma);
 		chan->stats.misqueued++;
 	}
@@ -618,6 +621,11 @@ static int __cpdma_chan_process(struct cpdma_chan *chan)
 	chan->count--;
 	chan->stats.good_dequeue++;
 
+	if (status & CPDMA_DESC_EOQ) {
+		chan->stats.requeue++;
+		chan_write(chan, hdp, desc_phys(pool, chan->head));
+	}
+
 	spin_unlock_irqrestore(&chan->lock, flags);
 
 	__cpdma_chan_free(chan, desc, outlen, status);
diff --git a/drivers/net/davinci_cpdma.h b/drivers/net/davinci_cpdma.h
index d50ee35..9baacca 100644
--- a/drivers/net/davinci_cpdma.h
+++ b/drivers/net/davinci_cpdma.h
@@ -56,6 +56,7 @@ struct cpdma_chan_stats {
 	u32			empty_dequeue;
 	u32			busy_dequeue;
 	u32			good_dequeue;
+	u32			requeue;
 	u32			teardown_dequeue;
 };
 
-- 
1.7.0.4

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html