* [PATCH net-next v03 1/4] hinic3: Solve an stall issue in hinic3_rx_poll()
2026-11-12 7:06 [PATCH net-next v03 0/4] net: hinic3: Bugfix for later patch-set Fan Gong
@ 2026-11-12 7:06 ` Fan Gong
2026-11-16 13:51 ` netdev-bot+sashiko
2026-11-12 7:06 ` [PATCH net-next v03 2/4] hinic3: Avoid concurrent updates on 32-bit systems Fan Gong
` (2 subsequent siblings)
3 siblings, 1 reply; 8+ messages in thread
From: Fan Gong @ 2026-11-12 7:06 UTC (permalink / raw)
To: Fan Gong, Wu Di, Teng Peisen, netdev, David S. Miller,
Eric Dumazet, Jakub Kicinski, Paolo Abeni, Simon Horman,
Andrew Lunn, Ioana Ciornei, Maxime Chevallier, Mohsin Bashir,
Dimitri Daskalakis, Harshitha Ramamurthy, Larysa Zaremba,
Vadim Fedorenko
Cc: linux-kernel, linux-doc, Zhou Shuai, Wu Like
When Large Receive Offload (LRO) packets are large, such as 32KB, each
packet consumes 16 Work Queue Elements (WQEs). If multiple large LRO
packets occur at this time, it will cause rapid consumption of WQEs in
the RQ and WQEs cannot be replenished in time. The chip will drop
packets due to lack of WQEs, causing the RQ to be unable to receive
packets. Furthermore, the NAPI instance will not be polled again,
resulting in residual packets in the RQ.
When napi_alloc_skb() fails inside recv_one_pkt(), hinic3_rx_poll()
breaks out of the loop without advancing rxq->cons_idx or clearing
rx_cqe->status, and returns nr_pkts (< budget). hinic3_poll() then
calls napi_complete_done() and re-enables the edge-triggered MSI-X
interrupt while the unprocessed CQE remains at the head of the ring,
stalling the receive queue until a new packet arrives to trigger
another interrupt.
Fixes: 17fcb3dc12bb ("hinic3: module initialization and tx/rx logic")
Co-developed-by: Wu Di <wudi234@huawei.com>
Signed-off-by: Wu Di <wudi234@huawei.com>
Co-developed-by: Teng Peisen <tengpeisen@huawei.com>
Signed-off-by: Teng Peisen <tengpeisen@huawei.com>
Signed-off-by: Fan Gong <gongfan1@huawei.com>
---
drivers/net/ethernet/huawei/hinic3/hinic3_rx.c | 11 ++++++++---
1 file changed, 8 insertions(+), 3 deletions(-)
diff --git a/drivers/net/ethernet/huawei/hinic3/hinic3_rx.c b/drivers/net/ethernet/huawei/hinic3/hinic3_rx.c
index 309ab5901379..f2262cdd6628 100644
--- a/drivers/net/ethernet/huawei/hinic3/hinic3_rx.c
+++ b/drivers/net/ethernet/huawei/hinic3/hinic3_rx.c
@@ -555,6 +555,7 @@ int hinic3_rx_poll(struct hinic3_rxq *rxq, int budget)
struct hinic3_nic_dev *nic_dev = netdev_priv(rxq->netdev);
u32 sw_ci, status, pkt_len, vlan_len;
struct hinic3_rq_cqe *rx_cqe;
+ bool poll_again = false;
u32 num_wqe = 0;
int nr_pkts = 0;
u16 num_lro;
@@ -571,8 +572,10 @@ int hinic3_rx_poll(struct hinic3_rxq *rxq, int budget)
vlan_len = le32_to_cpu(rx_cqe->vlan_len);
pkt_len = RQ_CQE_SGE_GET(vlan_len, LEN);
- if (recv_one_pkt(rxq, rx_cqe, pkt_len, vlan_len, status))
+ if (recv_one_pkt(rxq, rx_cqe, pkt_len, vlan_len, status)) {
+ poll_again = true;
break;
+ }
nr_pkts++;
num_lro = RQ_CQE_STATUS_GET(status, NUM_LRO);
@@ -581,12 +584,14 @@ int hinic3_rx_poll(struct hinic3_rxq *rxq, int budget)
rx_cqe->status = 0;
- if (num_wqe >= nic_dev->lro_replenish_thld)
+ if (num_wqe >= nic_dev->lro_replenish_thld) {
+ poll_again = true;
break;
+ }
}
if (rxq->delta >= HINIC3_RX_BUFFER_WRITE)
hinic3_rx_fill_buffers(rxq);
- return nr_pkts;
+ return poll_again ? budget : nr_pkts;
}
--
2.43.0
^ permalink raw reply related [flat|nested] 8+ messages in thread* Re: [PATCH net-next v03 1/4] hinic3: Solve an stall issue in hinic3_rx_poll()
2026-11-12 7:06 ` [PATCH net-next v03 1/4] hinic3: Solve an stall issue in hinic3_rx_poll() Fan Gong
@ 2026-11-16 13:51 ` netdev-bot+sashiko
0 siblings, 0 replies; 8+ messages in thread
From: netdev-bot+sashiko @ 2026-11-16 13:51 UTC (permalink / raw)
To: gongfan1
Cc: wudi234, tengpeisen, netdev, davem, edumazet, kuba, pabeni, horms,
andrew+netdev, ioana.ciornei, maxime.chevallier, mohsin.bashr,
dimitri.daskalakis1, hramamurthy, larysa.zaremba, vadim.fedorenko,
linux-kernel, linux-doc, zhoushuai28, wulike1
Thank you for your contribution! Sashiko AI review found 1 potential
issue(s) to consider.
Critical: 0 · High: 1 · Medium: 0 · Low: 0
If you believe none of the issues are legitimate, please reply to each
issue of Medium severity or higher and explain why the AI is wrong.
If any of them are legitimate and you plan to respin, please let patchwork
know by including "pw-bot: cr" as a separate line at the end of your reply
(one such reply per series is enough).
- [High] The fix does not cover the case where the RX refill fails, so
the RQ can still stall.
--
hinic3: Solve an stall issue in hinic3_rx_poll()
This patch makes hinic3_rx_poll() return the full budget in two cases:
recv_one_pkt() fails, or the LRO replenish threshold is reached. In both
cases NAPI keeps polling instead of completing and re-enabling the
interrupt while unprocessed CQEs or depleted WQEs remain.
> diff --git a/drivers/net/ethernet/huawei/hinic3/hinic3_rx.c b/drivers/net/ethernet/huawei/hinic3/hinic3_rx.c
> index 309ab59013797..f2262cdd66287 100644
> --- a/drivers/net/ethernet/huawei/hinic3/hinic3_rx.c
> +++ b/drivers/net/ethernet/huawei/hinic3/hinic3_rx.c
[ ... ]
> @@ -581,12 +584,14 @@ int hinic3_rx_poll(struct hinic3_rxq *rxq, int budget)
>
> rx_cqe->status = 0;
>
> - if (num_wqe >= nic_dev->lro_replenish_thld)
> + if (num_wqe >= nic_dev->lro_replenish_thld) {
> + poll_again = true;
> break;
> + }
> }
>
> if (rxq->delta >= HINIC3_RX_BUFFER_WRITE)
> hinic3_rx_fill_buffers(rxq);
>
> - return nr_pkts;
> + return poll_again ? budget : nr_pkts;
[Severity: High]
Can the RQ still stall when the refill itself fails? The return value of
hinic3_rx_fill_buffers() is ignored here, and poll_again is not set when
fewer WQEs than needed get posted.
hinic3_rx_fill_buffers() stops at the first page_pool allocation failure.
If that happens on the very first WQE, it rings no doorbell at all:
drivers/net/ethernet/huawei/hinic3/hinic3_rx.c:hinic3_rx_fill_buffers() {
...
for (i = 0; i < free_wqebbs; i++) {
...
err = rx_alloc_mapped_page(rxq->page_pool, rx_info,
rxq->buf_len);
if (unlikely(err))
break;
...
}
if (likely(i)) {
hinic3_write_db(rxq->rq, rxq->q_id & 3, DB_CFLAG_DP_RQ,
rxq->next_to_update << HINIC3_NORMAL_RQ_WQE);
...
}
Under memory pressure, refills can fail across several polls while
hinic3_fetch_rx_buffer() keeps adding sge_num to rxq->delta for each
consumed packet. Large LRO frames, which use up to 16 WQEs each, drain the
posted WQEs quickly.
On the last such poll the loop drains all CQEs and exits on !RXDONE. The
refill fails again, and hinic3_rx_poll() returns nr_pkts < budget. Then
hinic3_poll() does this:
drivers/net/ethernet/huawei/hinic3/hinic3_irq.c:hinic3_poll() {
...
if (likely(napi_complete_done(napi, work_done))) {
hinic3_net_dim(nic_dev, irq_cfg);
hinic3_set_msix_state(nic_dev->hwdev, irq_cfg->msix_entry_idx,
HINIC3_MSIX_ENABLE);
...
}
As the commit message says, the chip drops packets when it has no WQEs, so
no CQE is written and no RX interrupt is raised. qp_irq() is the only
napi_schedule() caller in the driver. hinic3_rx_fill_buffers() is called
only from hinic3_configure_rxqs() and hinic3_rx_poll(), and the periodic
work does not refill RX buffers.
Wouldn't the queue then stay dead even after memory frees up, until the
interface is reconfigured? That looks like the same "WQEs cannot be
replenished in time ... the NAPI instance will not be polled again"
failure, reached through the refill allocation failure path.
Would it make sense to also set poll_again when hinic3_rx_fill_buffers()
posts fewer WQEs than requested? Several other drivers do this; i40e, for
example, does "return failure ? budget : total_rx_packets". A deferred
refill would be another option. The later patches in this series don't
seem to touch hinic3_rx.c, so this case still looks open at the end of
the series.
> }
--
Sashiko AI review · https://netdev-ai.bots.linux.dev/sashiko/#/patchset/cover.1794467919.git.gongfan1%40huawei.com
^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH net-next v03 2/4] hinic3: Avoid concurrent updates on 32-bit systems.
2026-11-12 7:06 [PATCH net-next v03 0/4] net: hinic3: Bugfix for later patch-set Fan Gong
2026-11-12 7:06 ` [PATCH net-next v03 1/4] hinic3: Solve an stall issue in hinic3_rx_poll() Fan Gong
@ 2026-11-12 7:06 ` Fan Gong
2026-11-16 13:51 ` netdev-bot+sashiko
2026-11-12 7:06 ` [PATCH net-next v03 3/4] hinic3: fix some checkpatchs Fan Gong
2026-11-12 7:06 ` [PATCH net-next v03 4/4] hinic3: Ensure that the value of max_qps is not changed Fan Gong
3 siblings, 1 reply; 8+ messages in thread
From: Fan Gong @ 2026-11-12 7:06 UTC (permalink / raw)
To: Fan Gong, Wu Di, Teng Peisen, netdev, David S. Miller,
Eric Dumazet, Jakub Kicinski, Paolo Abeni, Simon Horman,
Andrew Lunn, Ioana Ciornei, Maxime Chevallier, Mohsin Bashir,
Dimitri Daskalakis, Harshitha Ramamurthy, Larysa Zaremba,
Vadim Fedorenko
Cc: linux-kernel, linux-doc, Zhou Shuai, Wu Like
Within the while loop of hinic3_get_stats64(), the counters between
u64_stats_fetch_retry and u64_stats_fetch_begin are incremented by +=.
This operation will cause repeated counting of the counters on a 32-bit
system. Therefore, change += to + and the counters are incremented
out side the wile loop.
csum_errors and other_errors are not the same things. csum_errors will
go to csum_bad in netdev netlink qstats.
Remove unnecessary checks for the nic_dev->txqs and nic_dev->rxqs.
Remove dead stores before the loop.
Fixes: f47872bed40f ("hinic3: Add .ndo_tx_timeout and .ndo_get_stats64")
Co-developed-by: Wu Di <wudi234@huawei.com>
Signed-off-by: Wu Di <wudi234@huawei.com>
Co-developed-by: Teng Peisen <tengpeisen@huawei.com>
Signed-off-by: Teng Peisen <tengpeisen@huawei.com>
Signed-off-by: Fan Gong <gongfan1@huawei.com>
---
.../huawei/hinic3/hinic3_netdev_ops.c | 42 +++++++------------
1 file changed, 14 insertions(+), 28 deletions(-)
diff --git a/drivers/net/ethernet/huawei/hinic3/hinic3_netdev_ops.c b/drivers/net/ethernet/huawei/hinic3/hinic3_netdev_ops.c
index da73811641a9..d1fd91d97ccf 100644
--- a/drivers/net/ethernet/huawei/hinic3/hinic3_netdev_ops.c
+++ b/drivers/net/ethernet/huawei/hinic3/hinic3_netdev_ops.c
@@ -791,49 +791,35 @@ static void hinic3_get_stats64(struct net_device *netdev,
unsigned int start;
int i;
- bytes = 0;
- packets = 0;
- dropped = 0;
for (i = 0; i < nic_dev->max_qps; i++) {
- if (!nic_dev->txqs)
- break;
-
txq = &nic_dev->txqs[i];
txq_stats = &txq->txq_stats;
do {
start = u64_stats_fetch_begin(&txq_stats->syncp);
- bytes += txq_stats->bytes;
- packets += txq_stats->packets;
- dropped += txq_stats->dropped;
+ bytes = txq_stats->bytes;
+ packets = txq_stats->packets;
+ dropped = txq_stats->dropped;
} while (u64_stats_fetch_retry(&txq_stats->syncp, start));
+ stats->tx_packets += packets;
+ stats->tx_bytes += bytes;
+ stats->tx_dropped += dropped;
}
- stats->tx_packets = packets;
- stats->tx_bytes = bytes;
- stats->tx_dropped = dropped;
- bytes = 0;
- packets = 0;
- errors = 0;
- dropped = 0;
for (i = 0; i < nic_dev->max_qps; i++) {
- if (!nic_dev->rxqs)
- break;
-
rxq = &nic_dev->rxqs[i];
rxq_stats = &rxq->rxq_stats;
do {
start = u64_stats_fetch_begin(&rxq_stats->syncp);
- bytes += rxq_stats->bytes;
- packets += rxq_stats->packets;
- errors += rxq_stats->csum_errors +
- rxq_stats->other_errors;
- dropped += rxq_stats->dropped;
+ bytes = rxq_stats->bytes;
+ packets = rxq_stats->packets;
+ errors = rxq_stats->other_errors;
+ dropped = rxq_stats->dropped;
} while (u64_stats_fetch_retry(&rxq_stats->syncp, start));
+ stats->rx_packets += packets;
+ stats->rx_bytes += bytes;
+ stats->rx_errors += errors;
+ stats->rx_dropped += dropped;
}
- stats->rx_packets = packets;
- stats->rx_bytes = bytes;
- stats->rx_errors = errors;
- stats->rx_dropped = dropped;
}
static void hinic3_nic_set_rx_mode(struct net_device *netdev)
--
2.43.0
^ permalink raw reply related [flat|nested] 8+ messages in thread* Re: [PATCH net-next v03 2/4] hinic3: Avoid concurrent updates on 32-bit systems.
2026-11-12 7:06 ` [PATCH net-next v03 2/4] hinic3: Avoid concurrent updates on 32-bit systems Fan Gong
@ 2026-11-16 13:51 ` netdev-bot+sashiko
0 siblings, 0 replies; 8+ messages in thread
From: netdev-bot+sashiko @ 2026-11-16 13:51 UTC (permalink / raw)
To: gongfan1
Cc: wudi234, tengpeisen, netdev, davem, edumazet, kuba, pabeni, horms,
andrew+netdev, ioana.ciornei, maxime.chevallier, mohsin.bashr,
dimitri.daskalakis1, hramamurthy, larysa.zaremba, vadim.fedorenko,
linux-kernel, linux-doc, zhoushuai28, wulike1
Thank you for your contribution! Sashiko AI review found more severe
issue(s) in other patches of this series. This mail is only an FYI, in case
you want to address the nit picks below when you respin.
Low priority / nit picks for your consideration:
- [Low] The commit message and Fixes: tag describe a bug this driver can
never hit.
- [Low] The patch drops `csum_errors` from `rx_errors` in
hinic3_get_stats64() (`errors = rxq_stats->other_errors;`).
--
hinic3: Avoid concurrent updates on 32-bit systems.
This commit reworks the u64_stats reader loops in hinic3_get_stats64().
Per-queue values are now assigned inside the fetch/retry loop and added
into rtnl_link_stats64 outside it. The patch also drops csum_errors from
rx_errors, removes the NULL checks on nic_dev->txqs and nic_dev->rxqs, and
removes the zeroing of the local counters before each loop.
> hinic3: Avoid concurrent updates on 32-bit systems.
>
> Within the while loop of hinic3_get_stats64(), the counters between
> u64_stats_fetch_retry and u64_stats_fetch_begin are incremented by +=.
> This operation will cause repeated counting of the counters on a 32-bit
> system. Therefore, change += to + and the counters are incremented
> out side the wile loop.
[Severity: Low]
Can this driver be built on a 32-bit kernel? The Kconfig in
drivers/net/ethernet/huawei/hinic3/ has:
depends on PCI_MSI && 64BIT
When BITS_PER_LONG == 64, include/linux/u64_stats_sync.h defines:
static inline bool __u64_stats_fetch_retry(const struct u64_stats_sync *syncp,
unsigned int start)
{
return false;
}
So the old loop body ran exactly once per queue. The += into locals that
started at zero already gave correct totals in every configuration that
can be built.
The subject says "Avoid concurrent updates". However, the patch only
changes the reader side and does not change how updates are serialized.
This patch carries "Fixes: f47872bed40f" and will be picked up for
stable. Would it be better described as a cleanup of the reader idiom,
without the Fixes: tag?
> diff --git a/drivers/net/ethernet/huawei/hinic3/hinic3_netdev_ops.c b/drivers/net/ethernet/huawei/hinic3/hinic3_netdev_ops.c
> index da73811641a9d..d1fd91d97ccfc 100644
> --- a/drivers/net/ethernet/huawei/hinic3/hinic3_netdev_ops.c
> +++ b/drivers/net/ethernet/huawei/hinic3/hinic3_netdev_ops.c
> @@ -791,49 +791,35 @@ static void hinic3_get_stats64(struct net_device *netdev,
[ ... ]
> rxq = &nic_dev->rxqs[i];
> rxq_stats = &rxq->rxq_stats;
> do {
> start = u64_stats_fetch_begin(&rxq_stats->syncp);
> - bytes += rxq_stats->bytes;
> - packets += rxq_stats->packets;
> - errors += rxq_stats->csum_errors +
> - rxq_stats->other_errors;
> - dropped += rxq_stats->dropped;
> + bytes = rxq_stats->bytes;
> + packets = rxq_stats->packets;
> + errors = rxq_stats->other_errors;
[Severity: Low]
The commit message says:
csum_errors and other_errors are not the same things. csum_errors will
go to csum_bad in netdev netlink qstats.
Where is csum_errors reported as csum_bad? I could not find a netdev
queue stats implementation (stat_ops, get_queue_stats_rx or csum_bad)
anywhere under drivers/net/ethernet/huawei/hinic3/. It is not in this
patch, and it is not added later in the series.
After this change, the only remaining reference to csum_errors is the
reset in hinic3_rxq_clean_stats():
rxq_stats->csum_errors = 0;
rxq_stats->other_errors = 0;
Neither counter is incremented anywhere today, so rx_errors does not
change in practice. Once the datapath starts incrementing csum_errors,
would those counts then be missing from every uAPI?
This rx_errors change is unrelated to the retry loop change. Could it be
split into its own patch, or sent together with the qstats support it
refers to?
[ ... ]
--
Sashiko AI review · https://netdev-ai.bots.linux.dev/sashiko/#/patchset/cover.1794467919.git.gongfan1%40huawei.com
^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH net-next v03 3/4] hinic3: fix some checkpatchs.
2026-11-12 7:06 [PATCH net-next v03 0/4] net: hinic3: Bugfix for later patch-set Fan Gong
2026-11-12 7:06 ` [PATCH net-next v03 1/4] hinic3: Solve an stall issue in hinic3_rx_poll() Fan Gong
2026-11-12 7:06 ` [PATCH net-next v03 2/4] hinic3: Avoid concurrent updates on 32-bit systems Fan Gong
@ 2026-11-12 7:06 ` Fan Gong
2026-11-12 7:06 ` [PATCH net-next v03 4/4] hinic3: Ensure that the value of max_qps is not changed Fan Gong
3 siblings, 0 replies; 8+ messages in thread
From: Fan Gong @ 2026-11-12 7:06 UTC (permalink / raw)
To: Fan Gong, Wu Di, Teng Peisen, netdev, David S. Miller,
Eric Dumazet, Jakub Kicinski, Paolo Abeni, Simon Horman,
Andrew Lunn, Ioana Ciornei, Maxime Chevallier, Mohsin Bashir,
Dimitri Daskalakis, Harshitha Ramamurthy, Larysa Zaremba,
Vadim Fedorenko
Cc: linux-kernel, linux-doc, Zhou Shuai, Wu Like
Cleanup checkpatchs of lines exceeding 80 columns.
Co-developed-by: Wu Di <wudi234@huawei.com>
Signed-off-by: Wu Di <wudi234@huawei.com>
Co-developed-by: Teng Peisen <tengpeisen@huawei.com>
Signed-off-by: Teng Peisen <tengpeisen@huawei.com>
Signed-off-by: Fan Gong <gongfan1@huawei.com>
---
.../huawei/hinic3/hinic3_netdev_ops.c | 34 +++++++++++--------
1 file changed, 19 insertions(+), 15 deletions(-)
diff --git a/drivers/net/ethernet/huawei/hinic3/hinic3_netdev_ops.c b/drivers/net/ethernet/huawei/hinic3/hinic3_netdev_ops.c
index d1fd91d97ccf..621d8b12e247 100644
--- a/drivers/net/ethernet/huawei/hinic3/hinic3_netdev_ops.c
+++ b/drivers/net/ethernet/huawei/hinic3/hinic3_netdev_ops.c
@@ -121,8 +121,9 @@ static void hinic3_destroy_num_qps(struct net_device *netdev)
kfree(nic_dev->qps_msix_entries);
}
-static int hinic3_alloc_txrxq_resources(struct net_device *netdev,
- struct hinic3_dyna_txrxq_params *q_params)
+static int
+hinic3_alloc_txrxq_resources(struct net_device *netdev,
+ struct hinic3_dyna_txrxq_params *q_params)
{
int err;
@@ -176,8 +177,9 @@ static int hinic3_alloc_txrxq_resources(struct net_device *netdev,
return err;
}
-static void hinic3_free_txrxq_resources(struct net_device *netdev,
- struct hinic3_dyna_txrxq_params *q_params)
+static void
+hinic3_free_txrxq_resources(struct net_device *netdev,
+ struct hinic3_dyna_txrxq_params *q_params)
{
hinic3_free_rxqs_res(netdev, q_params->num_qps, q_params->rq_depth,
q_params->rxqs_res);
@@ -251,16 +253,17 @@ static void hinic3_remove_configure(struct net_device *netdev)
hinic3_rss_uninit(netdev);
}
-static int hinic3_alloc_channel_resources(struct net_device *netdev,
- struct hinic3_dyna_qp_params *qp_params,
- struct hinic3_dyna_txrxq_params *trxq_params)
+static int
+hinic3_alloc_channel_resources(struct net_device *netdev,
+ struct hinic3_dyna_qp_params *qp_params,
+ struct hinic3_dyna_txrxq_params *q_params)
{
struct hinic3_nic_dev *nic_dev = netdev_priv(netdev);
int err;
- qp_params->num_qps = trxq_params->num_qps;
- qp_params->sq_depth = trxq_params->sq_depth;
- qp_params->rq_depth = trxq_params->rq_depth;
+ qp_params->num_qps = q_params->num_qps;
+ qp_params->sq_depth = q_params->sq_depth;
+ qp_params->rq_depth = q_params->rq_depth;
err = hinic3_alloc_qps(nic_dev, qp_params);
if (err) {
@@ -268,7 +271,7 @@ static int hinic3_alloc_channel_resources(struct net_device *netdev,
return err;
}
- err = hinic3_alloc_txrxq_resources(netdev, trxq_params);
+ err = hinic3_alloc_txrxq_resources(netdev, q_params);
if (err) {
netdev_err(netdev, "Failed to alloc txrxq resources\n");
hinic3_free_qps(nic_dev, qp_params);
@@ -278,13 +281,14 @@ static int hinic3_alloc_channel_resources(struct net_device *netdev,
return 0;
}
-static void hinic3_free_channel_resources(struct net_device *netdev,
- struct hinic3_dyna_qp_params *qp_params,
- struct hinic3_dyna_txrxq_params *trxq_params)
+static void
+hinic3_free_channel_resources(struct net_device *netdev,
+ struct hinic3_dyna_qp_params *qp_params,
+ struct hinic3_dyna_txrxq_params *q_params)
{
struct hinic3_nic_dev *nic_dev = netdev_priv(netdev);
- hinic3_free_txrxq_resources(netdev, trxq_params);
+ hinic3_free_txrxq_resources(netdev, q_params);
hinic3_free_qps(nic_dev, qp_params);
}
--
2.43.0
^ permalink raw reply related [flat|nested] 8+ messages in thread* [PATCH net-next v03 4/4] hinic3: Ensure that the value of max_qps is not changed
2026-11-12 7:06 [PATCH net-next v03 0/4] net: hinic3: Bugfix for later patch-set Fan Gong
` (2 preceding siblings ...)
2026-11-12 7:06 ` [PATCH net-next v03 3/4] hinic3: fix some checkpatchs Fan Gong
@ 2026-11-12 7:06 ` Fan Gong
2026-11-16 13:51 ` netdev-bot+sashiko
3 siblings, 1 reply; 8+ messages in thread
From: Fan Gong @ 2026-11-12 7:06 UTC (permalink / raw)
To: Fan Gong, Wu Di, Teng Peisen, netdev, David S. Miller,
Eric Dumazet, Jakub Kicinski, Paolo Abeni, Simon Horman,
Andrew Lunn, Ioana Ciornei, Maxime Chevallier, Mohsin Bashir,
Dimitri Daskalakis, Harshitha Ramamurthy, Larysa Zaremba,
Vadim Fedorenko
Cc: linux-kernel, linux-doc, Zhou Shuai, Wu Like
nic_dev->max_qps is used to allocate hardware resources. Therefore,
it is assigned the value 1 only when its value is abnormal. In other
scenarios, the value cannot be changed. However, the value of
nic_dev->q_params.num_qps can be dynamically changed.
Fixes: 1f3838b84a63 ("hinic3: Add Rss function")
Co-developed-by: Wu Di <wudi234@huawei.com>
Signed-off-by: Wu Di <wudi234@huawei.com>
Co-developed-by: Teng Peisen <tengpeisen@huawei.com>
Signed-off-by: Teng Peisen <tengpeisen@huawei.com>
Signed-off-by: Fan Gong <gongfan1@huawei.com>
---
| 9 +++------
1 file changed, 3 insertions(+), 6 deletions(-)
--git a/drivers/net/ethernet/huawei/hinic3/hinic3_rss.c b/drivers/net/ethernet/huawei/hinic3/hinic3_rss.c
index 25db74d8c7dd..75dc0c044071 100644
--- a/drivers/net/ethernet/huawei/hinic3/hinic3_rss.c
+++ b/drivers/net/ethernet/huawei/hinic3/hinic3_rss.c
@@ -307,15 +307,13 @@ void hinic3_try_to_enable_rss(struct net_device *netdev)
int err;
nic_dev->max_qps = hinic3_func_max_qnum(hwdev);
- if (nic_dev->max_qps <= 1 ||
+ if (nic_dev->max_qps == 1 ||
!hinic3_test_support(nic_dev, HINIC3_NIC_F_RSS))
goto err_reset_q_params;
err = alloc_rss_resource(netdev);
- if (err) {
- nic_dev->max_qps = 1;
+ if (err)
goto err_reset_q_params;
- }
set_bit(HINIC3_RSS_ENABLE, &nic_dev->flags);
decide_num_qps(netdev);
@@ -324,7 +322,6 @@ void hinic3_try_to_enable_rss(struct net_device *netdev)
if (err) {
dev_err(hwdev->dev, "Failed to set hardware rss parameters\n");
hinic3_clear_rss_config(netdev);
- nic_dev->max_qps = 1;
goto err_reset_q_params;
}
@@ -332,5 +329,5 @@ void hinic3_try_to_enable_rss(struct net_device *netdev)
err_reset_q_params:
clear_bit(HINIC3_RSS_ENABLE, &nic_dev->flags);
- nic_dev->q_params.num_qps = nic_dev->max_qps;
+ nic_dev->q_params.num_qps = 1;
}
--
2.43.0
^ permalink raw reply related [flat|nested] 8+ messages in thread* Re: [PATCH net-next v03 4/4] hinic3: Ensure that the value of max_qps is not changed
2026-11-12 7:06 ` [PATCH net-next v03 4/4] hinic3: Ensure that the value of max_qps is not changed Fan Gong
@ 2026-11-16 13:51 ` netdev-bot+sashiko
0 siblings, 0 replies; 8+ messages in thread
From: netdev-bot+sashiko @ 2026-11-16 13:51 UTC (permalink / raw)
To: gongfan1
Cc: wudi234, tengpeisen, netdev, davem, edumazet, kuba, pabeni, horms,
andrew+netdev, ioana.ciornei, maxime.chevallier, mohsin.bashr,
dimitri.daskalakis1, hramamurthy, larysa.zaremba, vadim.fedorenko,
linux-kernel, linux-doc, zhoushuai28, wulike1
Thank you for your contribution! Sashiko AI review found 1 potential
issue(s) to consider.
Critical: 0 · High: 0 · Medium: 1 · Low: 0
If you believe none of the issues are legitimate, please reply to each
issue of Medium severity or higher and explain why the AI is wrong.
If any of them are legitimate and you plan to respin, please let patchwork
know by including "pw-bot: cr" as a separate line at the end of your reply
(one such reply per series is enough).
- [Medium] The patch is presented as a bug fix (it carries `Fixes:
1f3838b84a63 ("hinic3: Add Rss function")`), but I could not find any…
--
hinic3: Ensure that the value of max_qps is not changed
This patch stops hinic3_try_to_enable_rss() from setting nic_dev->max_qps
to 1 when RSS is unsupported or its setup fails. On those paths it now
sets nic_dev->q_params.num_qps to 1 directly. It also changes the check
on the function's queue count from "<= 1" to "== 1".
> diff --git a/drivers/net/ethernet/huawei/hinic3/hinic3_rss.c b/drivers/net/ethernet/huawei/hinic3/hinic3_rss.c
> index 25db74d8c7ddb..75dc0c044071c 100644
> --- a/drivers/net/ethernet/huawei/hinic3/hinic3_rss.c
> +++ b/drivers/net/ethernet/huawei/hinic3/hinic3_rss.c
> @@ -307,15 +307,13 @@ void hinic3_try_to_enable_rss(struct net_device *netdev)
> int err;
>
> nic_dev->max_qps = hinic3_func_max_qnum(hwdev);
> - if (nic_dev->max_qps <= 1 ||
> + if (nic_dev->max_qps == 1 ||
> !hinic3_test_support(nic_dev, HINIC3_NIC_F_RSS))
> goto err_reset_q_params;
>
> err = alloc_rss_resource(netdev);
> - if (err) {
> - nic_dev->max_qps = 1;
> + if (err)
> goto err_reset_q_params;
> - }
[ ... ]
> @@ -332,5 +329,5 @@ void hinic3_try_to_enable_rss(struct net_device *netdev)
>
> err_reset_q_params:
> clear_bit(HINIC3_RSS_ENABLE, &nic_dev->flags);
> - nic_dev->q_params.num_qps = nic_dev->max_qps;
> + nic_dev->q_params.num_qps = 1;
> }
[Severity: Medium]
Which bug does this fix? The patch has a Fixes: tag for 1f3838b84a63
("hinic3: Add Rss function"), but I couldn't find a reachable problem
in the old fallback.
The only caller of hinic3_try_to_enable_rss() is hinic3_sw_init(). Every
allocation sized from nic_dev->max_qps happens after that call, so the
value is already final when it is used:
hinic3_sw_init()
hinic3_try_to_enable_rss() <- max_qps settled here
hinic3_alloc_txrxqs()
hinic3_alloc_txqs() <- num_txqs = nic_dev->max_qps
hinic3_alloc_rxqs() <- num_rxqs = nic_dev->max_qps
hinic3_init_intr_coalesce()
ndo_open
hinic3_setup_num_qps() <- qps_msix_entries sized by max_qps
With the old code, an RSS failure gave a consistent single-queue setup:
max_qps = 1, one txq and one rxq, one coalesce entry, one msix entry and
num_qps = 1.
The commit message says:
"However, the value of nic_dev->q_params.num_qps can be dynamically
changed."
Does that hold at this commit? hinic3_config_num_qps() already forces a
single queue when RSS is off:
if (!test_bit(HINIC3_RSS_ENABLE, &nic_dev->flags))
q_params->num_qps = 1;
The driver also has no ethtool set_channels or set_rxfh callback, so
nothing can turn RSS on or raise the queue count later.
The commit message also says:
"nic_dev->max_qps is used to allocate hardware resources. Therefore,
it is assigned the value 1 only when its value is abnormal."
The diff removes both "nic_dev->max_qps = 1" assignments and adds none,
so after this patch nothing ever sets max_qps to 1.
On the hardware side, the limit is nic_io->max_qps, which
hinic3_init_nic_io() sets separately and this function never changed.
nic_dev->max_qps only sizes software arrays. alloc_etherdev_mq() in
hinic3_nic_probe() uses its own local max_qps.
The commit message also doesn't mention the "<= 1" to "== 1" change. It
looks harmless, because max_sqs is computed as
min(nic_max_sq_id + 1, HINIC3_CFG_MAX_QP) and is always at least 1.
As far as I can tell, the only change in behaviour is on the RSS failure
path. There, txqs, rxqs, intr_coalesce and qps_msix_entries are now
sized for the full hardware maximum while the device still runs on one
queue.
Is this meant as a design change rather than a fix? If so, could the
commit message say so, and could the Fixes: tag be dropped so this
doesn't get backported to stable?
--
Sashiko AI review · https://netdev-ai.bots.linux.dev/sashiko/#/patchset/cover.1794467919.git.gongfan1%40huawei.com
^ permalink raw reply [flat|nested] 8+ messages in thread