blob: 149208aa69da250449037fbe454dc53c261c63d6 (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
|
From d0055b03d9c8d48ad2b971821989b09ba95c39f8 Mon Sep 17 00:00:00 2001
From: Christian Marangi <ansuelsmth@gmail.com>
Date: Sun, 17 Sep 2023 20:18:31 +0200
Subject: [PATCH] net: qualcomm: ipqess: fix TX timeout errors
Currently logic to handle napi tx completion is flawed and on the long
run on loaded condition cause TX timeout error with the queue not being
able to handle any new packet.
There are 2 main cause of this:
- incrementing the packet done value wrongly
- handling 2 times the tx_ring tail
ipqess_tx_unmap_and_free may return 2 kind values:
- 0: we are handling first and middle descriptor for the packet
- packet len: we are at the last descriptor for the packet
Done value was wrongly incremented also for first and intermediate
descriptor for the packet resulting causing panic and TX timeouts by
comunicating to the kernel an inconsistent value of packet handling not
matching the expected ones.
Tx_ring tail was handled twice for ipqess_tx_complete run resulting in
again done value incremented wrongly and also problem with idx handling
by actually skipping descriptor for some packets.
Rework the loop logic to fix these 2 problem and also add some comments
to make sure ipqess_tx_unmap_and_free ret value is better
understandable.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
---
drivers/net/ethernet/qualcomm/ipqess/ipqess.c | 13 ++++++++++---
1 file changed, 10 insertions(+), 3 deletions(-)
--- a/drivers/net/ethernet/qualcomm/ipqess/ipqess.c
+++ b/drivers/net/ethernet/qualcomm/ipqess/ipqess.c
@@ -453,13 +453,22 @@ static int ipqess_tx_complete(struct ipq
tail >>= IPQESS_TPD_CONS_IDX_SHIFT;
tail &= IPQESS_TPD_CONS_IDX_MASK;
- do {
+ while ((tx_ring->tail != tail) && (done < budget)) {
ret = ipqess_tx_unmap_and_free(&tx_ring->ess->pdev->dev,
&tx_ring->buf[tx_ring->tail]);
- tx_ring->tail = IPQESS_NEXT_IDX(tx_ring->tail, tx_ring->count);
+ /* ipqess_tx_unmap_and_free may return 2 kind values:
+ * - 0: we are handling first and middle descriptor for the packet
+ * - packet len: we are at the last descriptor for the packet
+ * Increment total bytes handled and packet done only if we are
+ * handling the last descriptor for the packet.
+ */
+ if (ret) {
+ total += ret;
+ done++;
+ }
- total += ret;
- } while ((++done < budget) && (tx_ring->tail != tail));
+ tx_ring->tail = IPQESS_NEXT_IDX(tx_ring->tail, tx_ring->count);
+ };
ipqess_w32(tx_ring->ess, IPQESS_REG_TX_SW_CONS_IDX_Q(tx_ring->idx),
tx_ring->tail);
|
