Re: pull-request: wireless-drivers-2021-02-05

From: Lorenzo Bianconi <lorenzo.bianconi@redhat.com>
To: Jakub Kicinski <kuba@kernel.org>
Cc: Kalle Valo <kvalo@codeaurora.org>,
	Lorenzo Bianconi <lorenzo@kernel.org>,
	netdev@vger.kernel.org, linux-wireless@vger.kernel.org,
	Felix Fietkau <nbd@nbd.name>
Subject: Re: pull-request: wireless-drivers-2021-02-05
Date: Sat, 6 Feb 2021 20:43:25 +0100	[thread overview]
Message-ID: <20210206194325.GA134674@lore-desk> (raw)
In-Reply-To: <20210206093537.0bfaf0db@kicinski-fedora-pc1c0hjn.dhcp.thefacebook.com>

[-- Attachment #1: Type: text/plain, Size: 2071 bytes --]

> On Fri,  5 Feb 2021 16:34:34 +0000 (UTC) Kalle Valo wrote:
> > Hi,
> > 
> > here's a pull request to net tree, more info below. Please let me know if there
> > are any problems.
> 
> Pulled, thanks! One thing to confirm tho..
> 
> > ath9k
> > 
> > * fix build regression related to LEDS_CLASS
> > 
> > mt76
> > 
> > * fix a memory leak
> 
> Lorenzo, I'm just guessing what this code does, but you're dropping a
> frag without invalidating the rest of the SKB, which I presume is now
> truncated? Shouldn't the skb be dropped?
> 

Hi Jakub,

I agree. We can do something like:

diff --git a/drivers/net/wireless/mediatek/mt76/dma.c b/drivers/net/wireless/mediatek/mt76/dma.c
index e81dfaf99bcb..6d84533d1df2 100644
--- a/drivers/net/wireless/mediatek/mt76/dma.c
+++ b/drivers/net/wireless/mediatek/mt76/dma.c
@@ -511,8 +511,9 @@ mt76_add_fragment(struct mt76_dev *dev, struct mt76_queue *q, void *data,
 {
 	struct sk_buff *skb = q->rx_head;
 	struct skb_shared_info *shinfo = skb_shinfo(skb);
+	int nr_frags = shinfo->nr_frags;
 
-	if (shinfo->nr_frags < ARRAY_SIZE(shinfo->frags)) {
+	if (nr_frags < ARRAY_SIZE(shinfo->frags)) {
 		struct page *page = virt_to_head_page(data);
 		int offset = data - page_address(page) + q->buf_offset;
 
@@ -526,7 +527,10 @@ mt76_add_fragment(struct mt76_dev *dev, struct mt76_queue *q, void *data,
 		return;
 
 	q->rx_head = NULL;
-	dev->drv->rx_skb(dev, q - dev->q_rx, skb);
+	if (nr_frags < ARRAY_SIZE(shinfo->frags))
+		dev->drv->rx_skb(dev, q - dev->q_rx, skb);
+	else
+		dev_kfree_skb(skb);
 }
 

I do not know if it can occur, but I guess we should even check q->rx_head
pointer before overwriting it because if the hw does not report more set to
false for last fragment we will get a memory leak as well. Something like:

@@ -578,6 +582,8 @@ mt76_dma_rx_process(struct mt76_dev *dev, struct mt76_queue *q, int budget)
 		done++;
 
 		if (more) {
+			if (q->rx_head)
+				dev_kfree_skb(q->rx_head);
 			q->rx_head = skb;
 			continue;
 		}

Regards,
Lorenzo

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]
