(not a formal patch, take this as a bug report for now, I can clean
up depending on the feedback I get here)
There's at least 3 issues here, and the patch fixes 2/3 only, I'm not sure
how/if 1 should be handled.
1. ath10k_sdio_mbox_rx_alloc allocating skb of a incorrect size (too
small)
2. ath10k_sdio_mbox_rx_packet calling skb_put with that incorrect size.
3. ath10k_sdio_mbox_rx_process_packet attempts to fixup the size, but
does not use proper skb_put commands to do so, so we end up with
a mismatch between skb->head + skb->tail and skb->data + skb->len.
Let's start with 3, this is quite serious as this and causes corruptions
in the TCP stack, as the stack tries to coalesce packets, and relies on
skb->tail being correct (that is, skb_tail_pointer must point to the
first byte _after_ the data): one must never manipulate skb->len
directly.
Instead, we need to use skb_put to allocate more space (which updates
skb->len and skb->tail). But it seems odd to do that in
ath10k_sdio_mbox_rx_process_packet, so I move the code to
ath10k_sdio_mbox_rx_packet (point 2 above).
However, there is still something strange (point 1 above), why is
ath10k_sdio_mbox_rx_alloc allocating packets of the incorrect
(too small?) size? What happens if the packet is bigger than alloc_len?
Does this lead to corruption/lost data?
Fixes: 8530b4e7b22bc3b ("ath10k: sdio: set skb len for all rx packets")
Signed-off-by: Nicolas Boichat <drinkcat@xxxxxxxxxxxx>
---
One simple way to test this is this scriplet, that sends a lot of
small packets over SSH:
(for i in `seq 1 300`; do echo $i; sleep 0.1; done) | ssh $IP cat
In my testing it rarely ever reach 300 without failure.
drivers/net/wireless/ath/ath10k/sdio.c | 18 ++++++++++++------
1 file changed, 12 insertions(+), 6 deletions(-)
diff --git a/drivers/net/wireless/ath/ath10k/sdio.c b/drivers/net/wireless/ath/ath10k/sdio.c
index 8ed4fbd8d6c3888..a9f5002863ee7bb 100644
--- a/drivers/net/wireless/ath/ath10k/sdio.c
+++ b/drivers/net/wireless/ath/ath10k/sdio.c
@@ -381,16 +381,14 @@ static int ath10k_sdio_mbox_rx_process_packet(struct ath10k *ar,
struct ath10k_htc_hdr *htc_hdr = (struct ath10k_htc_hdr *)skb->data;
bool trailer_present = htc_hdr->flags & ATH10K_HTC_FLAG_TRAILER_PRESENT;
enum ath10k_htc_ep_id eid;
- u16 payload_len;
u8 *trailer;
int ret;
- payload_len = le16_to_cpu(htc_hdr->len);
- skb->len = payload_len + sizeof(struct ath10k_htc_hdr);
+ /* TODO: Remove this? */
+ WARN_ON(skb->len != le16_to_cpu(htc_hdr->len) + sizeof(*htc_hdr));
if (trailer_present) {
- trailer = skb->data + sizeof(*htc_hdr) +
- payload_len - htc_hdr->trailer_len;
+ trailer = skb->data + skb->len - htc_hdr->trailer_len;
eid = pipe_id_to_eid(htc_hdr->eid);
@@ -637,8 +635,16 @@ static int ath10k_sdio_mbox_rx_packet(struct ath10k *ar,
ret = ath10k_sdio_readsb(ar, ar_sdio->mbox_info.htc_addr,
skb->data, pkt->alloc_len);
pkt->status = ret;
- if (!ret)
+ if (!ret) {
+ /* Update actual length. */
+ /* FIXME: This looks quite wrong, why is pkt->act_len not
+ * correct in the first place?
+ */
+ struct ath10k_htc_hdr *htc_hdr =
+ (struct ath10k_htc_hdr *)skb->data;
+ pkt->act_len = le16_to_cpu(htc_hdr->len) + sizeof(*htc_hdr);
skb_put(skb, pkt->act_len);
+ }
return ret;
}
--
2.23.0.187.g17f5b7556c-goog
