[PATCH net] net/mlx5e: Skip NAPI polling when PCI channel is offline

[PATCH net] net/mlx5e: Skip NAPI polling when PCI channel is offline

[Breno Leitao]

When a PCI error (e.g. AER error or DPC containment) marks the PCI
channel as frozen or permanently failed, the IOMMU mappings for the
device may already be torn down. If mlx5e_napi_poll() continues
processing CQEs in this state, every call to dma_unmap_page() triggers
a WARN_ON in iommu_dma_unmap_phys().

In a real-world crash scenario on an NVIDIA Grace (ARM64) platform,
a DPC event froze the PCI channel and the mlx5 NAPI poll continued
processing error CQEs, calling dma_unmap for each pending WQE. Here is
an example:

The DPC event on port 0007:00:00.0 fires and eth1 (on 0017:01:00.0) starts
seeing error CQEs almost immediately:

    pcieport 0007:00:00.0: DPC: containment event, status:0x2009
    mlx5_core 0017:01:00.0 eth1: Error cqe on cqn 0x54e, ci 0xb06, ...

The WARN_ON storm begins ~0.4s later and repeats for every pending WQE:

    WARNING: CPU: 32 PID: 0 at drivers/iommu/dma-iommu.c:1237 iommu_dma_unmap_phys
    Call trace:
     iommu_dma_unmap_phys+0xd4/0xe0
     mlx5e_tx_wi_dma_unmap+0xb4/0xf0
     mlx5e_poll_tx_cq+0x14c/0x438
     mlx5e_napi_poll+0x6c/0x5e0
     net_rx_action+0x160/0x5c0
     handle_softirqs+0xe8/0x320
     run_ksoftirqd+0x30/0x58

After 23 seconds of WARN_ON() storm, the watchdog fires:

    watchdog: BUG: soft lockup - CPU#32 stuck for 23s! [ksoftirqd/32:179]
    Kernel panic - not syncing: softlockup: hung tasks

Each unmap hit the WARN_ON in the IOMMU layer, printing a full stack
trace. With dozens of pending WQEs, this created a storm of WARN_ON
dumps in softirq context that monopolized the CPU for over 23 seconds,
triggering a soft lockup panic.

Fix this by checking pci_channel_offline() at the top of
mlx5e_napi_poll() and bailing out immediately when the channel is
offline. napi_complete_done() is called before returning to clear the
NAPI_STATE_SCHED bit, ensuring that napi_disable() in the teardown path
does not spin forever waiting for it. No CQ interrupts are re-armed
since the explicit mlx5e_cq_arm() calls are skipped, so the NAPI
instance will not be re-scheduled. The pending DMA buffers are left for
device removal to clean up.

Fixes: e586b3b0baee ("net/mlx5: Ethernet Datapath files")
Signed-off-by: Breno Leitao <leitao@debian.org>
---
 drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c | 13 +++++++++++++
 1 file changed, 13 insertions(+)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
index 76108299ea57d..934ad7fafa801 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
@@ -138,6 +138,19 @@ int mlx5e_napi_poll(struct napi_struct *napi, int budget)
 	bool xsk_open;
 	int i;
 
+	/*
+	 * When the PCI channel is offline, IOMMU mappings may already be torn
+	 * down.  Processing CQEs would call dma_unmap for every pending WQE,
+	 * each hitting a WARN_ON in the IOMMU layer.  The resulting storm of
+	 * warnings in softirq context can monopolise the CPU long enough to
+	 * trigger a soft lockup and prevent any RCU grace period from
+	 * completing.
+	 */
+	if (unlikely(pci_channel_offline(c->mdev->pdev))) {
+		napi_complete_done(napi, 0);
+		return 0;
+	}
+
 	rcu_read_lock();
 
 	qos_sqs = rcu_dereference(c->qos_sqs);

---
base-commit: a956792a1543c2bf4a2266cb818dc7c4135006f0
change-id: 20260209-mlx5_iommu-c8b238b1bb14

Best regards,
--  
Breno Leitao <leitao@debian.org>

Re: [PATCH net] net/mlx5e: Skip NAPI polling when PCI channel is offline

[Jijie Shao]

on 2026/2/10 2:01, Breno Leitao wrote:
> When a PCI error (e.g. AER error or DPC containment) marks the PCI
> channel as frozen or permanently failed, the IOMMU mappings for the
> device may already be torn down. If mlx5e_napi_poll() continues
> processing CQEs in this state, every call to dma_unmap_page() triggers
> a WARN_ON in iommu_dma_unmap_phys().

Hi:
   My comment has nothing to do with the changes made in this patch itself.


I am more interested in this error itself.
1. If there is an issue with dma_unmp, does dma_map in tx have a similar problem?
2. Can this error be detected by mlx5_pci_err_detected()? If not, does this mean that all PCIe NIC drivers might have similar issues?
    Do other drivers need to do similar checks?

Thanks,
Jijie Shao

>
> In a real-world crash scenario on an NVIDIA Grace (ARM64) platform,
> a DPC event froze the PCI channel and the mlx5 NAPI poll continued
> processing error CQEs, calling dma_unmap for each pending WQE. Here is
> an example:
>
> The DPC event on port 0007:00:00.0 fires and eth1 (on 0017:01:00.0) starts
> seeing error CQEs almost immediately:
>
>      pcieport 0007:00:00.0: DPC: containment event, status:0x2009
>      mlx5_core 0017:01:00.0 eth1: Error cqe on cqn 0x54e, ci 0xb06, ...
>
> The WARN_ON storm begins ~0.4s later and repeats for every pending WQE:
>
>      WARNING: CPU: 32 PID: 0 at drivers/iommu/dma-iommu.c:1237 iommu_dma_unmap_phys
>      Call trace:
>       iommu_dma_unmap_phys+0xd4/0xe0
>       mlx5e_tx_wi_dma_unmap+0xb4/0xf0
>       mlx5e_poll_tx_cq+0x14c/0x438
>       mlx5e_napi_poll+0x6c/0x5e0
>       net_rx_action+0x160/0x5c0
>       handle_softirqs+0xe8/0x320
>       run_ksoftirqd+0x30/0x58
>
> After 23 seconds of WARN_ON() storm, the watchdog fires:
>
>      watchdog: BUG: soft lockup - CPU#32 stuck for 23s! [ksoftirqd/32:179]
>      Kernel panic - not syncing: softlockup: hung tasks
>
> Each unmap hit the WARN_ON in the IOMMU layer, printing a full stack
> trace. With dozens of pending WQEs, this created a storm of WARN_ON
> dumps in softirq context that monopolized the CPU for over 23 seconds,
> triggering a soft lockup panic.
>
> Fix this by checking pci_channel_offline() at the top of
> mlx5e_napi_poll() and bailing out immediately when the channel is
> offline. napi_complete_done() is called before returning to clear the
> NAPI_STATE_SCHED bit, ensuring that napi_disable() in the teardown path
> does not spin forever waiting for it. No CQ interrupts are re-armed
> since the explicit mlx5e_cq_arm() calls are skipped, so the NAPI
> instance will not be re-scheduled. The pending DMA buffers are left for
> device removal to clean up.
>
> Fixes: e586b3b0baee ("net/mlx5: Ethernet Datapath files")
> Signed-off-by: Breno Leitao <leitao@debian.org>
> ---
>   drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c | 13 +++++++++++++
>   1 file changed, 13 insertions(+)
>
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
> index 76108299ea57d..934ad7fafa801 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
> @@ -138,6 +138,19 @@ int mlx5e_napi_poll(struct napi_struct *napi, int budget)
>   	bool xsk_open;
>   	int i;
>   
> +	/*
> +	 * When the PCI channel is offline, IOMMU mappings may already be torn
> +	 * down.  Processing CQEs would call dma_unmap for every pending WQE,
> +	 * each hitting a WARN_ON in the IOMMU layer.  The resulting storm of
> +	 * warnings in softirq context can monopolise the CPU long enough to
> +	 * trigger a soft lockup and prevent any RCU grace period from
> +	 * completing.
> +	 */
> +	if (unlikely(pci_channel_offline(c->mdev->pdev))) {
> +		napi_complete_done(napi, 0);
> +		return 0;
> +	}
> +
>   	rcu_read_lock();
>   
>   	qos_sqs = rcu_dereference(c->qos_sqs);
>
> ---
> base-commit: a956792a1543c2bf4a2266cb818dc7c4135006f0
> change-id: 20260209-mlx5_iommu-c8b238b1bb14
>
> Best regards,
> --
> Breno Leitao <leitao@debian.org>
>
>

Re: [PATCH net] net/mlx5e: Skip NAPI polling when PCI channel is offline

[Breno Leitao]

On Tue, Feb 10, 2026 at 10:19:46AM +0800, Jijie Shao wrote:
> 
> on 2026/2/10 2:01, Breno Leitao wrote:
> > When a PCI error (e.g. AER error or DPC containment) marks the PCI
> > channel as frozen or permanently failed, the IOMMU mappings for the
> > device may already be torn down. If mlx5e_napi_poll() continues
> > processing CQEs in this state, every call to dma_unmap_page() triggers
> > a WARN_ON in iommu_dma_unmap_phys().
> 
> Hi:
>   My comment has nothing to do with the changes made in this patch itself.
> 
> 
> I am more interested in this error itself.
> 1. If there is an issue with dma_unmp, does dma_map in tx have a similar problem?

I suspect that dma_map will succeed in such a case (when the DMA maps are
gone). 

dma_map_single/dma_map_page creates new page table entries — it doesn't look up
existing ones. Even if existing mappings are gone, new mappings succeed !?

I haven't seen this instance on the TX path as well.

> 2. Can this error be detected by mlx5_pci_err_detected()? If not, does this mean that all PCIe NIC drivers might have similar issues?

mlx5_pci_err_detected() is called for the device under DPC — that's not the
issue.

From my naive view, the issue seems to be timing: there's a potential race
between DPC setting the PCI channel to frozen and the error handler completing
(which eventually calls napi_disable_locked).

During that window, NAPI poll can still fire and process CQEs, triggering the
dma_unmap WARN_ON storm, and crash.

>    Do other drivers need to do similar checks?

I really don't know, honestly. Are other drivers solving the problem
differently?! 

Thanks for the question,
--breno

Re: [PATCH net] net/mlx5e: Skip NAPI polling when PCI channel is offline

[Jijie Shao]

on 2026/2/10 23:18, Breno Leitao wrote:
> On Tue, Feb 10, 2026 at 10:19:46AM +0800, Jijie Shao wrote:
>> on 2026/2/10 2:01, Breno Leitao wrote:
>>> When a PCI error (e.g. AER error or DPC containment) marks the PCI
>>> channel as frozen or permanently failed, the IOMMU mappings for the
>>> device may already be torn down. If mlx5e_napi_poll() continues
>>> processing CQEs in this state, every call to dma_unmap_page() triggers
>>> a WARN_ON in iommu_dma_unmap_phys().
>> Hi:
>>    My comment has nothing to do with the changes made in this patch itself.
>>
>>
>> I am more interested in this error itself.
>> 1. If there is an issue with dma_unmp, does dma_map in tx have a similar problem?
> I suspect that dma_map will succeed in such a case (when the DMA maps are
> gone).
>
> dma_map_single/dma_map_page creates new page table entries — it doesn't look up
> existing ones. Even if existing mappings are gone, new mappings succeed !?
>
> I haven't seen this instance on the TX path as well.
>
>> 2. Can this error be detected by mlx5_pci_err_detected()? If not, does this mean that all PCIe NIC drivers might have similar issues?
> mlx5_pci_err_detected() is called for the device under DPC — that's not the
> issue.
>
> >From my naive view, the issue seems to be timing: there's a potential race
> between DPC setting the PCI channel to frozen and the error handler completing
> (which eventually calls napi_disable_locked).
>
> During that window, NAPI poll can still fire and process CQEs, triggering the
> dma_unmap WARN_ON storm, and crash.
>
>>     Do other drivers need to do similar checks?
> I really don't know, honestly. Are other drivers solving the problem
> differently?!
>
> Thanks for the question,
> --breno

Thanks for the reply, and I will continue to follow up on this issue.
This change may require further confirmation from the mlx driver's maintainer.

The code look good to me.
Reviewed-by: Jijie Shao <shaojijie@huawei.com>

Re: [PATCH net] net/mlx5e: Skip NAPI polling when PCI channel is offline

[Tariq Toukan]

On 09/02/2026 20:01, Breno Leitao wrote:
> When a PCI error (e.g. AER error or DPC containment) marks the PCI
> channel as frozen or permanently failed, the IOMMU mappings for the
> device may already be torn down. If mlx5e_napi_poll() continues
> processing CQEs in this state, every call to dma_unmap_page() triggers
> a WARN_ON in iommu_dma_unmap_phys().
> 
> In a real-world crash scenario on an NVIDIA Grace (ARM64) platform,
> a DPC event froze the PCI channel and the mlx5 NAPI poll continued
> processing error CQEs, calling dma_unmap for each pending WQE. Here is
> an example:
> 
> The DPC event on port 0007:00:00.0 fires and eth1 (on 0017:01:00.0) starts
> seeing error CQEs almost immediately:
> 
>      pcieport 0007:00:00.0: DPC: containment event, status:0x2009
>      mlx5_core 0017:01:00.0 eth1: Error cqe on cqn 0x54e, ci 0xb06, ...
> 
> The WARN_ON storm begins ~0.4s later and repeats for every pending WQE:
> 
>      WARNING: CPU: 32 PID: 0 at drivers/iommu/dma-iommu.c:1237 iommu_dma_unmap_phys
>      Call trace:
>       iommu_dma_unmap_phys+0xd4/0xe0
>       mlx5e_tx_wi_dma_unmap+0xb4/0xf0
>       mlx5e_poll_tx_cq+0x14c/0x438
>       mlx5e_napi_poll+0x6c/0x5e0
>       net_rx_action+0x160/0x5c0
>       handle_softirqs+0xe8/0x320
>       run_ksoftirqd+0x30/0x58
> 
> After 23 seconds of WARN_ON() storm, the watchdog fires:
> 
>      watchdog: BUG: soft lockup - CPU#32 stuck for 23s! [ksoftirqd/32:179]
>      Kernel panic - not syncing: softlockup: hung tasks
> 
> Each unmap hit the WARN_ON in the IOMMU layer, printing a full stack
> trace. With dozens of pending WQEs, this created a storm of WARN_ON
> dumps in softirq context that monopolized the CPU for over 23 seconds,
> triggering a soft lockup panic.
> 
> Fix this by checking pci_channel_offline() at the top of
> mlx5e_napi_poll() and bailing out immediately when the channel is
> offline. napi_complete_done() is called before returning to clear the
> NAPI_STATE_SCHED bit, ensuring that napi_disable() in the teardown path
> does not spin forever waiting for it. No CQ interrupts are re-armed
> since the explicit mlx5e_cq_arm() calls are skipped, so the NAPI
> instance will not be re-scheduled. The pending DMA buffers are left for
> device removal to clean up.
> 
> Fixes: e586b3b0baee ("net/mlx5: Ethernet Datapath files")
> Signed-off-by: Breno Leitao <leitao@debian.org>
> ---
>   drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c | 13 +++++++++++++
>   1 file changed, 13 insertions(+)
> 
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
> index 76108299ea57d..934ad7fafa801 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
> @@ -138,6 +138,19 @@ int mlx5e_napi_poll(struct napi_struct *napi, int budget)
>   	bool xsk_open;
>   	int i;
>   
> +	/*
> +	 * When the PCI channel is offline, IOMMU mappings may already be torn
> +	 * down.  Processing CQEs would call dma_unmap for every pending WQE,
> +	 * each hitting a WARN_ON in the IOMMU layer.  The resulting storm of
> +	 * warnings in softirq context can monopolise the CPU long enough to
> +	 * trigger a soft lockup and prevent any RCU grace period from
> +	 * completing.
> +	 */
> +	if (unlikely(pci_channel_offline(c->mdev->pdev))) {
> +		napi_complete_done(napi, 0);
> +		return 0;
> +	}
> +
>   	rcu_read_lock();
>   
>   	qos_sqs = rcu_dereference(c->qos_sqs);
> 
> ---
> base-commit: a956792a1543c2bf4a2266cb818dc7c4135006f0
> change-id: 20260209-mlx5_iommu-c8b238b1bb14
> 
> Best regards,
> --
> Breno Leitao <leitao@debian.org>
> 
> 

Hi,

Thanks for your patch.

You're introducing an interesting problem, but I am not convinced by 
this solution approach.

Why would the driver perform this check if it doesn't guarantee 
prevention of invalid access? It only "allows one napi cycle", which 
happen to be good enough to prevent the soft lockup in your case.

What if a napi cycle is configured with larger budget?

If the problem is that the WARN_ON is being called at a high rate, then 
it should be rate-limited.

Re: [PATCH net] net/mlx5e: Skip NAPI polling when PCI channel is offline

[Breno Leitao]

Hello Tariq,

On Wed, Feb 11, 2026 at 01:26:35PM +0200, Tariq Toukan wrote:
> On 09/02/2026 20:01, Breno Leitao wrote:
> > When a PCI error (e.g. AER error or DPC containment) marks the PCI
> > channel as frozen or permanently failed, the IOMMU mappings for the
> > device may already be torn down. If mlx5e_napi_poll() continues
> > processing CQEs in this state, every call to dma_unmap_page() triggers
> > a WARN_ON in iommu_dma_unmap_phys().
> > 
> > In a real-world crash scenario on an NVIDIA Grace (ARM64) platform,
> > a DPC event froze the PCI channel and the mlx5 NAPI poll continued
> > processing error CQEs, calling dma_unmap for each pending WQE. Here is
> > an example:
> > 
> > The DPC event on port 0007:00:00.0 fires and eth1 (on 0017:01:00.0) starts
> > seeing error CQEs almost immediately:
> > 
> >      pcieport 0007:00:00.0: DPC: containment event, status:0x2009
> >      mlx5_core 0017:01:00.0 eth1: Error cqe on cqn 0x54e, ci 0xb06, ...
> > 
> > The WARN_ON storm begins ~0.4s later and repeats for every pending WQE:
> > 
> >      WARNING: CPU: 32 PID: 0 at drivers/iommu/dma-iommu.c:1237 iommu_dma_unmap_phys
> >      Call trace:
> >       iommu_dma_unmap_phys+0xd4/0xe0
> >       mlx5e_tx_wi_dma_unmap+0xb4/0xf0
> >       mlx5e_poll_tx_cq+0x14c/0x438
> >       mlx5e_napi_poll+0x6c/0x5e0
> >       net_rx_action+0x160/0x5c0
> >       handle_softirqs+0xe8/0x320
> >       run_ksoftirqd+0x30/0x58
> > 
> > After 23 seconds of WARN_ON() storm, the watchdog fires:
> > 
> >      watchdog: BUG: soft lockup - CPU#32 stuck for 23s! [ksoftirqd/32:179]
> >      Kernel panic - not syncing: softlockup: hung tasks
> > 
> > Each unmap hit the WARN_ON in the IOMMU layer, printing a full stack
> > trace. With dozens of pending WQEs, this created a storm of WARN_ON
> > dumps in softirq context that monopolized the CPU for over 23 seconds,
> > triggering a soft lockup panic.
...
> You're introducing an interesting problem, but I am not convinced by this
> solution approach.
> 
> Why would the driver perform this check if it doesn't guarantee prevention
> of invalid access? It only "allows one napi cycle", which happen to be good
> enough to prevent the soft lockup in your case.
> 
> What if a napi cycle is configured with larger budget?

Very good point. In this case, we will still see some WARN_ON() in DMA, and the
patch might eventually not help much if the AER hits mid-NAPI and there is
still a long budget remaining.

> If the problem is that the WARN_ON is being called at a high rate, then it
> should be rate-limited.

That would be a solution as well, and I am happy to pursue it, if that one is
more appropriate

Thanks for reviewing it,
--breno

Re: [PATCH net] net/mlx5e: Skip NAPI polling when PCI channel is offline

[Breno Leitao]

On Wed, Feb 11, 2026 at 05:44:56AM -0800, Breno Leitao wrote:
> On Wed, Feb 11, 2026 at 01:26:35PM +0200, Tariq Toukan wrote:
> > On 09/02/2026 20:01, Breno Leitao wrote:

> > If the problem is that the WARN_ON is being called at a high rate, then it
> > should be rate-limited.
> 
> That would be a solution as well, and I am happy to pursue it, if that one is
> more appropriate

In fact, I got more inclined to try to fix it in iommu code, let's see how it
goes.

https://lore.kernel.org/all/20260211-dma_io_mmu-v1-1-cf89e24437af@debian.org/

--
pv-bot: closed

Re: [PATCH net] net/mlx5e: Skip NAPI polling when PCI channel is offline

[Breno Leitao]

On Wed, Feb 11, 2026 at 07:17:34AM -0800, Breno Leitao wrote:
> On Wed, Feb 11, 2026 at 05:44:56AM -0800, Breno Leitao wrote:
> > On Wed, Feb 11, 2026 at 01:26:35PM +0200, Tariq Toukan wrote:
> > > On 09/02/2026 20:01, Breno Leitao wrote:
> 
> > > If the problem is that the WARN_ON is being called at a high rate, then it
> > > should be rate-limited.
> > 
> > That would be a solution as well, and I am happy to pursue it, if that one is
> > more appropriate
> 
> In fact, I got more inclined to try to fix it in iommu code, let's see how it
> goes.
> 
> https://lore.kernel.org/all/20260211-dma_io_mmu-v1-1-cf89e24437af@debian.org/
> 
> --
> pv-bot: closed
--
pw-bot: closed