[PATCH 1/2] mac80211: Fix ieee80211_amsdu_aggregate frag_tail bug

[PATCH 1/2] mac80211: Fix ieee80211_amsdu_aggregate frag_tail bug

[Ping-Ke Shih]

From: Chih-Kang Chang <gary.chang@realtek.com>

In ieee80211_amsdu_aggregate() set a pointer frag_tail point to the
end of skb_shinfo(head)->frag_list, and use it to bind other skb in
the end of this function. But when execute ieee80211_amsdu_aggregate()
->ieee80211_amsdu_realloc_pad()->pskb_expand_head(), the address of
skb_shinfo(head)->frag_list will be changed. However, the
ieee80211_amsdu_aggregate() not update frag_tail after call
pskb_expand_head(). That will cause the second skb can't bind to the
head skb appropriately.So we update the address of frag_tail to fix it.

Fixes: 6e0456b54545 ("mac80211: add A-MSDU tx support")
Signed-off-by: Chih-Kang Chang <gary.chang@realtek.com>
Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
---
 net/mac80211/tx.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/net/mac80211/tx.c b/net/mac80211/tx.c
index 8509778ff31f..9e0d4fca1c76 100644
--- a/net/mac80211/tx.c
+++ b/net/mac80211/tx.c
@@ -3406,6 +3406,12 @@ static bool ieee80211_amsdu_aggregate(struct ieee80211_sub_if_data *sdata,
 
 	head->len += skb->len;
 	head->data_len += skb->len;
+
+	/* frag_list should be updated after pskb_expand_head() */
+	frag_tail = &skb_shinfo(head)->frag_list;
+	while (*frag_tail)
+		frag_tail = &(*frag_tail)->next;
+
 	*frag_tail = skb;
 
 out_recalc:
-- 
2.25.1

[PATCH 2/2] mac80211: Fix insufficient headroom issue for AMSDU

[Ping-Ke Shih]

From: Chih-Kang Chang <gary.chang@realtek.com>

ieee80211_amsdu_realloc_pad() lacks some check for extra_tx_headroom,
the original reserved headroom might be eaten. So we add the
extra_tx_headroom check.

Fixes: 6e0456b54545 ("mac80211: add A-MSDU tx support")
Signed-off-by: Chih-Kang Chang <gary.chang@realtek.com>
Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
---
 net/mac80211/tx.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/net/mac80211/tx.c b/net/mac80211/tx.c
index 9e0d4fca1c76..a69dc58d9768 100644
--- a/net/mac80211/tx.c
+++ b/net/mac80211/tx.c
@@ -3242,7 +3242,8 @@ static bool ieee80211_amsdu_prepare_head(struct ieee80211_sub_if_data *sdata,
 	if (info->control.flags & IEEE80211_TX_CTRL_AMSDU)
 		return true;
 
-	if (!ieee80211_amsdu_realloc_pad(local, skb, sizeof(*amsdu_hdr)))
+	if (!ieee80211_amsdu_realloc_pad(local, skb, sizeof(*amsdu_hdr) +
+					 local->hw.extra_tx_headroom))
 		return false;
 
 	data = skb_push(skb, sizeof(*amsdu_hdr));
-- 
2.25.1

Re: [PATCH 1/2] mac80211: Fix ieee80211_amsdu_aggregate frag_tail bug

[Felix Fietkau]

On 2021-08-16 10:51, Ping-Ke Shih wrote:
> From: Chih-Kang Chang <gary.chang@realtek.com>
> 
> In ieee80211_amsdu_aggregate() set a pointer frag_tail point to the
> end of skb_shinfo(head)->frag_list, and use it to bind other skb in
> the end of this function. But when execute ieee80211_amsdu_aggregate()
> ->ieee80211_amsdu_realloc_pad()->pskb_expand_head(), the address of
> skb_shinfo(head)->frag_list will be changed. However, the
> ieee80211_amsdu_aggregate() not update frag_tail after call
> pskb_expand_head(). That will cause the second skb can't bind to the
> head skb appropriately.So we update the address of frag_tail to fix it.
I think instead of iterating over fragments a second time, we should
move the ieee80211_amsdu_prepare_head call further up.

- Felix

RE: [PATCH 1/2] mac80211: Fix ieee80211_amsdu_aggregate frag_tail bug

[Pkshih]

> -----Original Message-----
> From: Felix Fietkau [mailto:nbd@nbd.name]
> Sent: Monday, August 16, 2021 8:23 PM
> To: Pkshih; johannes@sipsolutions.net
> Cc: linux-wireless@vger.kernel.org; Gary Chang
> Subject: Re: [PATCH 1/2] mac80211: Fix ieee80211_amsdu_aggregate frag_tail bug
> 
> On 2021-08-16 10:51, Ping-Ke Shih wrote:
> > From: Chih-Kang Chang <gary.chang@realtek.com>
> >
> > In ieee80211_amsdu_aggregate() set a pointer frag_tail point to the
> > end of skb_shinfo(head)->frag_list, and use it to bind other skb in
> > the end of this function. But when execute ieee80211_amsdu_aggregate()
> > ->ieee80211_amsdu_realloc_pad()->pskb_expand_head(), the address of
> > skb_shinfo(head)->frag_list will be changed. However, the
> > ieee80211_amsdu_aggregate() not update frag_tail after call
> > pskb_expand_head(). That will cause the second skb can't bind to the
> > head skb appropriately.So we update the address of frag_tail to fix it.
> I think instead of iterating over fragments a second time, we should
> move the ieee80211_amsdu_prepare_head call further up.
> 

The first iteration is used to calculate 'n' and 'nfrags' to decide if
we need to do AMSDU as well as ieee80211_amsdu_prepare_head(), so we
don't have a good idea to move the function upward. However, we have
an alternative idea [1] to fix frag_tail. 

[1] https://lore.kernel.org/linux-wireless/20210830061728.10332-1-pkshih@realtek.com/T/#u

--
Ping-Ke