[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-ID: <20240624175015.4520-1-shannon.nelson@amd.com>
Date: Mon, 24 Jun 2024 10:50:15 -0700
From: Shannon Nelson <shannon.nelson@....com>
To: <netdev@...r.kernel.org>, <davem@...emloft.net>, <kuba@...nel.org>,
<edumazet@...gle.com>, <pabeni@...hat.com>
CC: <brett.creeley@....com>, Shannon Nelson <shannon.nelson@....com>
Subject: [PATCH v3 net] ionic: use dev_consume_skb_any outside of napi
If we're not in a NAPI softirq context, we need to be careful
about how we call napi_consume_skb(), specifically we need to
call it with budget==0 to signal to it that we're not in a
safe context.
This was found while running some configuration stress testing
of traffic and a change queue config loop running, and this
curious note popped out:
[ 4371.402645] BUG: using smp_processor_id() in preemptible [00000000] code: ethtool/20545
[ 4371.402897] caller is napi_skb_cache_put+0x16/0x80
[ 4371.403120] CPU: 25 PID: 20545 Comm: ethtool Kdump: loaded Tainted: G OE 6.10.0-rc3-netnext+ #8
[ 4371.403302] Hardware name: HPE ProLiant DL360 Gen10/ProLiant DL360 Gen10, BIOS U32 01/23/2021
[ 4371.403460] Call Trace:
[ 4371.403613] <TASK>
[ 4371.403758] dump_stack_lvl+0x4f/0x70
[ 4371.403904] check_preemption_disabled+0xc1/0xe0
[ 4371.404051] napi_skb_cache_put+0x16/0x80
[ 4371.404199] ionic_tx_clean+0x18a/0x240 [ionic]
[ 4371.404354] ionic_tx_cq_service+0xc4/0x200 [ionic]
[ 4371.404505] ionic_tx_flush+0x15/0x70 [ionic]
[ 4371.404653] ? ionic_lif_qcq_deinit.isra.23+0x5b/0x70 [ionic]
[ 4371.404805] ionic_txrx_deinit+0x71/0x190 [ionic]
[ 4371.404956] ionic_reconfigure_queues+0x5f5/0xff0 [ionic]
[ 4371.405111] ionic_set_ringparam+0x2e8/0x3e0 [ionic]
[ 4371.405265] ethnl_set_rings+0x1f1/0x300
[ 4371.405418] ethnl_default_set_doit+0xbb/0x160
[ 4371.405571] genl_family_rcv_msg_doit+0xff/0x130
[...]
I found that ionic_tx_clean() calls napi_consume_skb() which calls
napi_skb_cache_put(), but before that last call is the note
/* Zero budget indicate non-NAPI context called us, like netpoll */
and
DEBUG_NET_WARN_ON_ONCE(!in_softirq());
Those are pretty big hints that we're doing it wrong. We can pass a
context hint down through the calls to let ionic_tx_clean() know what
we're doing so it can call napi_consume_skb() correctly.
Fixes: 386e69865311 ("ionic: Make use napi_consume_skb")
Signed-off-by: Shannon Nelson <shannon.nelson@....com>
---
.../net/ethernet/pensando/ionic/ionic_dev.h | 4 ++-
.../net/ethernet/pensando/ionic/ionic_lif.c | 2 +-
.../net/ethernet/pensando/ionic/ionic_txrx.c | 28 +++++++++++--------
3 files changed, 21 insertions(+), 13 deletions(-)
v3: use budget for the hint where we can (Jakub)
v2: replace softirq_count() with a napi context hint (Jakub)
https://lore.kernel.org/netdev/20240620192519.35395-1-shannon.nelson@amd.com/
v1: https://lore.kernel.org/netdev/20240619212022.30700-1-shannon.nelson@amd.com/
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_dev.h b/drivers/net/ethernet/pensando/ionic/ionic_dev.h
index f30eee4a5a80..b6c01a88098d 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_dev.h
+++ b/drivers/net/ethernet/pensando/ionic/ionic_dev.h
@@ -375,7 +375,9 @@ typedef void (*ionic_cq_done_cb)(void *done_arg);
unsigned int ionic_cq_service(struct ionic_cq *cq, unsigned int work_to_do,
ionic_cq_cb cb, ionic_cq_done_cb done_cb,
void *done_arg);
-unsigned int ionic_tx_cq_service(struct ionic_cq *cq, unsigned int work_to_do);
+unsigned int ionic_tx_cq_service(struct ionic_cq *cq,
+ unsigned int work_to_do,
+ bool in_napi);
int ionic_q_init(struct ionic_lif *lif, struct ionic_dev *idev,
struct ionic_queue *q, unsigned int index, const char *name,
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_lif.c b/drivers/net/ethernet/pensando/ionic/ionic_lif.c
index 1934e9d6d9e4..1837a30ba08a 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_lif.c
+++ b/drivers/net/ethernet/pensando/ionic/ionic_lif.c
@@ -1189,7 +1189,7 @@ static int ionic_adminq_napi(struct napi_struct *napi, int budget)
ionic_rx_service, NULL, NULL);
if (lif->hwstamp_txq)
- tx_work = ionic_tx_cq_service(&lif->hwstamp_txq->cq, budget);
+ tx_work = ionic_tx_cq_service(&lif->hwstamp_txq->cq, budget, !!budget);
work_done = max(max(n_work, a_work), max(rx_work, tx_work));
if (work_done < budget && napi_complete_done(napi, work_done)) {
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_txrx.c b/drivers/net/ethernet/pensando/ionic/ionic_txrx.c
index 2427610f4306..995d7f7d96d0 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_txrx.c
+++ b/drivers/net/ethernet/pensando/ionic/ionic_txrx.c
@@ -23,7 +23,8 @@ static void ionic_tx_desc_unmap_bufs(struct ionic_queue *q,
static void ionic_tx_clean(struct ionic_queue *q,
struct ionic_tx_desc_info *desc_info,
- struct ionic_txq_comp *comp);
+ struct ionic_txq_comp *comp,
+ bool in_napi);
static inline void ionic_txq_post(struct ionic_queue *q, bool ring_dbell)
{
@@ -935,7 +936,7 @@ int ionic_tx_napi(struct napi_struct *napi, int budget)
u32 work_done = 0;
u32 flags = 0;
- work_done = ionic_tx_cq_service(cq, budget);
+ work_done = ionic_tx_cq_service(cq, budget, !!budget);
if (unlikely(!budget))
return budget;
@@ -1019,7 +1020,7 @@ int ionic_txrx_napi(struct napi_struct *napi, int budget)
txqcq = lif->txqcqs[qi];
txcq = &lif->txqcqs[qi]->cq;
- tx_work_done = ionic_tx_cq_service(txcq, IONIC_TX_BUDGET_DEFAULT);
+ tx_work_done = ionic_tx_cq_service(txcq, IONIC_TX_BUDGET_DEFAULT, !!budget);
if (unlikely(!budget))
return budget;
@@ -1152,7 +1153,8 @@ static void ionic_tx_desc_unmap_bufs(struct ionic_queue *q,
static void ionic_tx_clean(struct ionic_queue *q,
struct ionic_tx_desc_info *desc_info,
- struct ionic_txq_comp *comp)
+ struct ionic_txq_comp *comp,
+ bool in_napi)
{
struct ionic_tx_stats *stats = q_to_tx_stats(q);
struct ionic_qcq *qcq = q_to_qcq(q);
@@ -1204,11 +1206,13 @@ static void ionic_tx_clean(struct ionic_queue *q,
desc_info->bytes = skb->len;
stats->clean++;
- napi_consume_skb(skb, 1);
+ napi_consume_skb(skb, likely(in_napi) ? 1 : 0);
}
static bool ionic_tx_service(struct ionic_cq *cq,
- unsigned int *total_pkts, unsigned int *total_bytes)
+ unsigned int *total_pkts,
+ unsigned int *total_bytes,
+ bool in_napi)
{
struct ionic_tx_desc_info *desc_info;
struct ionic_queue *q = cq->bound_q;
@@ -1230,7 +1234,7 @@ static bool ionic_tx_service(struct ionic_cq *cq,
desc_info->bytes = 0;
index = q->tail_idx;
q->tail_idx = (q->tail_idx + 1) & (q->num_descs - 1);
- ionic_tx_clean(q, desc_info, comp);
+ ionic_tx_clean(q, desc_info, comp, in_napi);
if (desc_info->skb) {
pkts++;
bytes += desc_info->bytes;
@@ -1244,7 +1248,9 @@ static bool ionic_tx_service(struct ionic_cq *cq,
return true;
}
-unsigned int ionic_tx_cq_service(struct ionic_cq *cq, unsigned int work_to_do)
+unsigned int ionic_tx_cq_service(struct ionic_cq *cq,
+ unsigned int work_to_do,
+ bool in_napi)
{
unsigned int work_done = 0;
unsigned int bytes = 0;
@@ -1253,7 +1259,7 @@ unsigned int ionic_tx_cq_service(struct ionic_cq *cq, unsigned int work_to_do)
if (work_to_do == 0)
return 0;
- while (ionic_tx_service(cq, &pkts, &bytes)) {
+ while (ionic_tx_service(cq, &pkts, &bytes, in_napi)) {
if (cq->tail_idx == cq->num_descs - 1)
cq->done_color = !cq->done_color;
cq->tail_idx = (cq->tail_idx + 1) & (cq->num_descs - 1);
@@ -1279,7 +1285,7 @@ void ionic_tx_flush(struct ionic_cq *cq)
{
u32 work_done;
- work_done = ionic_tx_cq_service(cq, cq->num_descs);
+ work_done = ionic_tx_cq_service(cq, cq->num_descs, false);
if (work_done)
ionic_intr_credits(cq->idev->intr_ctrl, cq->bound_intr->index,
work_done, IONIC_INTR_CRED_RESET_COALESCE);
@@ -1296,7 +1302,7 @@ void ionic_tx_empty(struct ionic_queue *q)
desc_info = &q->tx_info[q->tail_idx];
desc_info->bytes = 0;
q->tail_idx = (q->tail_idx + 1) & (q->num_descs - 1);
- ionic_tx_clean(q, desc_info, NULL);
+ ionic_tx_clean(q, desc_info, NULL, false);
if (desc_info->skb) {
pkts++;
bytes += desc_info->bytes;
--
2.17.1
Powered by blists - more mailing lists