[PATCH 0/2] net/sfc: fix and extend secondary process support

[PATCH 0/2] net/sfc: fix and extend secondary process support

[Ivan Malov]

The series addresses two secondary process deficiencies.

Patch 1 adds a missing shared adapter pointer assignment to
the secondary process init path, preventing a NULL pointer
dereference at device close time.

Patch 2 pre-populates a primary-process device info cache
to satisfy the 'dev_infos_get' requirement now imposed by
the 'test-pmd' application on secondary processes.


Ivan Malov (2):
  net/sfc: fix shared adapter pointer set in secondary process
  net/sfc: provide cached dev info to use in secondary process

 drivers/net/sfc/sfc.h        |  7 +++++++
 drivers/net/sfc/sfc_ethdev.c | 30 ++++++++++++++++++++++++++++++
 2 files changed, 37 insertions(+)

-- 
2.47.3

[PATCH 1/2] net/sfc: fix shared adapter pointer set in secondary process

[Ivan Malov]

Add the missing assignment to preclude NULL pointer dereference that has
happened in the context of the secondary process at the dev close stage.

Fixes: 5313b441d8ae ("net/sfc: separate adapter primary process and shared data")
Cc: stable@dpdk.org

Signed-off-by: Ivan Malov <ivan.malov@arknetworks.am>
Reviewed-by: Viacheslav Galaktionov <viacheslav.galaktionov@arknetworks.am>
---
 drivers/net/sfc/sfc_ethdev.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/net/sfc/sfc_ethdev.c b/drivers/net/sfc/sfc_ethdev.c
index 6be91789cf..13619b4b9a 100644
--- a/drivers/net/sfc/sfc_ethdev.c
+++ b/drivers/net/sfc/sfc_ethdev.c
@@ -3224,6 +3224,7 @@ sfc_eth_dev_secondary_init(struct rte_eth_dev *dev, uint32_t logtype_main)
 
 	sap->dp_rx = dp_rx;
 	sap->dp_tx = dp_tx;
+	sap->shared = sas;
 
 	dev->process_private = sap;
 	dev->rx_pkt_burst = dp_rx->pkt_burst;
-- 
2.47.3

Re: [PATCH 1/2] net/sfc: fix shared adapter pointer set in secondary process

[Stephen Hemminger]

On Thu, 20 Aug 2026 17:03:13 +0400
Ivan Malov <ivan.malov@arknetworks.am> wrote:

> Add the missing assignment to preclude NULL pointer dereference that has
> happened in the context of the secondary process at the dev close stage.
> 
> Fixes: 5313b441d8ae ("net/sfc: separate adapter primary process and shared data")
> Cc: stable@dpdk.org
> 
> Signed-off-by: Ivan Malov <ivan.malov@arknetworks.am>
> Reviewed-by: Viacheslav Galaktionov <viacheslav.galaktionov@arknetworks.am>
> ---

This is correct.

[PATCH 2/2] net/sfc: provide cached dev info to use in secondary process

[Ivan Malov]

Secondary process support in the 'test-pmd' application now requires that
the driver expose the 'dev_infos_get' method within that context. Use the
cached dev info from the primary process in order to meet the requirement.

Signed-off-by: Ivan Malov <ivan.malov@arknetworks.am>
Reviewed-by: Viacheslav Galaktionov <viacheslav.galaktionov@arknetworks.am>
---
 drivers/net/sfc/sfc.h        |  7 +++++++
 drivers/net/sfc/sfc_ethdev.c | 29 +++++++++++++++++++++++++++++
 2 files changed, 36 insertions(+)

diff --git a/drivers/net/sfc/sfc.h b/drivers/net/sfc/sfc.h
index 629578549f..9fcc8c9044 100644
--- a/drivers/net/sfc/sfc.h
+++ b/drivers/net/sfc/sfc.h
@@ -156,6 +156,13 @@ struct sfc_adapter_shared {
 	unsigned int			nb_repr_txq;
 
 	struct sfc_nic_dma_info		nic_dma_info;
+
+	/*
+	 * Snapshot of the 'sfc_dev_infos_get' output created by the primary
+	 * process attach path for the secondary process to use in its own
+	 * implementation of the 'dev_infos_get' method.
+	 */
+	struct rte_eth_dev_info		dev_info_cache;
 };
 
 /* Adapter process private data */
diff --git a/drivers/net/sfc/sfc_ethdev.c b/drivers/net/sfc/sfc_ethdev.c
index 13619b4b9a..aa0e92b004 100644
--- a/drivers/net/sfc/sfc_ethdev.c
+++ b/drivers/net/sfc/sfc_ethdev.c
@@ -3163,7 +3163,22 @@ sfc_eth_dev_clear_ops(struct rte_eth_dev *dev)
 	sa->priv.dp_rx = NULL;
 }
 
+static int
+sfc_dev_infos_get_secondary(struct rte_eth_dev *dev,
+			    struct rte_eth_dev_info *dev_info)
+{
+	*dev_info = sfc_adapter_shared_by_eth_dev(dev)->dev_info_cache;
+
+	/*
+	 * The cache holds a stale primary-process device pointer;
+	 * restore the process-local one passed in by the caller.
+	 */
+	dev_info->device = dev->device;
+	return 0;
+}
+
 static const struct eth_dev_ops sfc_eth_dev_secondary_ops = {
+	.dev_infos_get			= sfc_dev_infos_get_secondary,
 	.dev_supported_ptypes_get	= sfc_dev_supported_ptypes_get,
 	.reta_query			= sfc_dev_rss_reta_query,
 	.rss_hash_conf_get		= sfc_dev_rss_hash_conf_get,
@@ -3443,6 +3458,20 @@ sfc_eth_dev_init(struct rte_eth_dev *dev, void *init_params)
 
 	sa->link_ev_need_poll = encp->enc_link_ev_need_poll;
 
+	/*
+	 * Pre-populate the dev info cache for the secondary process.
+	 * All prerequisites (probe, attach) are met at this point.
+	 * The 'sfc_dev_infos_get' helper always returns 0.
+	 *
+	 * Care to initialise the switch info and reset the device
+	 * pointer as it is going to be stale in the context of
+	 * the secondary process and it will have to fix it.
+	 */
+	sas->dev_info_cache.switch_info.domain_id =
+		RTE_ETH_DEV_SWITCH_DOMAIN_ID_INVALID;
+	(void)sfc_dev_infos_get(dev, &sas->dev_info_cache);
+	sas->dev_info_cache.device = NULL;
+
 	sfc_adapter_unlock(sa);
 
 	sfc_log_init(sa, "done");
-- 
2.47.3

Re: [PATCH 2/2] net/sfc: provide cached dev info to use in secondary process

[Stephen Hemminger]

On Thu, 20 Aug 2026 17:03:14 +0400
Ivan Malov <ivan.malov@arknetworks.am> wrote:

> Secondary process support in the 'test-pmd' application now requires that
> the driver expose the 'dev_infos_get' method within that context. Use the
> cached dev info from the primary process in order to meet the requirement.
> 
> Signed-off-by: Ivan Malov <ivan.malov@arknetworks.am>
> Reviewed-by: Viacheslav Galaktionov <viacheslav.galaktionov@arknetworks.am>
> ---

This patch has issues.



Patch 2/2: net/sfc: provide cached dev info to use in secondary process

Error: the cached snapshot is missing the defaults that
rte_eth_dev_info_get() fills in before it calls the driver callback,
so the secondary process reports zero for several fields.

rte_eth_dev_info_get() pre-populates the struct and then calls
.dev_infos_get(), so a PMD callback only has to set the fields it
actually knows about.  Besides switch_info.domain_id and device
(both handled by this patch) it pre-sets:

  rx_desc_lim.nb_seg_max      = UINT16_MAX
  rx_desc_lim.nb_mtu_seg_max  = UINT16_MAX
  tx_desc_lim.nb_seg_max      = UINT16_MAX
  tx_desc_lim.nb_mtu_seg_max  = UINT16_MAX
  rss_algo_capa               = RTE_ETH_HASH_ALGO_CAPA_MASK(DEFAULT)
  max_rx_bufsize              = UINT32_MAX

sfc_dev_infos_get() never writes any of these, and neither do the
datapath get_dev_info() helpers (sfc_ef100_rx_get_dev_info() and
sfc_ef100_get_dev_info() only touch nb_min and nb_align).  In the
primary process that is fine because the ethdev layer supplied the
values.  Here the cache is filled by calling sfc_dev_infos_get()
directly on a zeroed structure, so those fields stay zero, and
sfc_dev_infos_get_secondary() then overwrites the ethdev pre-fill
wholesale with

	*dev_info = sfc_adapter_shared_by_eth_dev(dev)->dev_info_cache;

A secondary process therefore sees nb_seg_max = 0,
nb_mtu_seg_max = 0, max_rx_bufsize = 0 and rss_algo_capa = 0, which
differs from what the same call returns in the primary.  Applications
that validate multi-segment Tx against tx_desc_lim.nb_seg_max, or
that check the RSS hash algorithm capability mask, will get wrong
answers.

Suggested fix: seed the cache with the same defaults before the
snapshot is taken, e.g.

	static const struct rte_eth_desc_lim lim = {
		.nb_max = UINT16_MAX,
		.nb_min = 0,
		.nb_align = 1,
		.nb_seg_max = UINT16_MAX,
		.nb_mtu_seg_max = UINT16_MAX,
	};

	sas->dev_info_cache.rx_desc_lim = lim;
	sas->dev_info_cache.tx_desc_lim = lim;
	sas->dev_info_cache.max_rx_bufsize = UINT32_MAX;
	sas->dev_info_cache.rss_algo_capa =
		RTE_ETH_HASH_ALGO_CAPA_MASK(DEFAULT);
	sas->dev_info_cache.switch_info.domain_id =
		RTE_ETH_DEV_SWITCH_DOMAIN_ID_INVALID;
	(void)sfc_dev_infos_get(dev, &sas->dev_info_cache);

This duplicates ethdev knowledge in the driver and will drift when
new pre-filled fields are added.  An alternative that avoids the
duplication is to have sfc_dev_infos_get_secondary() copy only the
fields the PMD owns, or to keep the caller's pre-filled struct and
merge the cached values into it.

Info: switch_info.name is left pointing at the primary process copy
of dev->device->driver->name.  The comment in
sfc_dev_infos_get_secondary() only mentions the device pointer, but
this is the same class of problem; the string lives in the driver
image rather than in per-process heap, so it happens to work under
the usual multi-process assumptions, but it would be more consistent
to re-derive it next to the device pointer:

	if (dev_info->switch_info.name != NULL)
		dev_info->switch_info.name = dev->device->driver->name;

Info: the cache is a snapshot taken at the end of sfc_eth_dev_init().
Everything sfc_dev_infos_get() reports is fixed at attach time today
(NIC config, rxq_max/txq_max, offload capabilities, MAE status), so
the snapshot is accurate.  Worth a note in the sfc.h comment that any
future dev_info field derived from post-attach state must not be
served from this cache.

[PATCH v2 0/2] net/sfc: fix and extend secondary process support

[Ivan Malov]

The series addresses two secondary process deficiencies.

Patch 1 adds a missing shared adapter pointer assignment to
the secondary process init path, preventing a NULL pointer
dereference at device close time.

Patch 2 pre-populates a primary-process device info cache
to satisfy the 'dev_infos_get' requirement now imposed by
the 'test-pmd' application on secondary processes.


v2:

- addressed https://mails.dpdk.org/archives/dev/2026-August/343941.html


Ivan Malov (2):
  net/sfc: fix shared adapter pointer set in secondary process
  net/sfc: provide cached dev info to use in secondary process

 drivers/net/sfc/sfc.h        | 19 ++++++++++++++
 drivers/net/sfc/sfc_ethdev.c | 50 ++++++++++++++++++++++++++++++++++++
 2 files changed, 69 insertions(+)

-- 
2.47.3

[PATCH v2 1/2] net/sfc: fix shared adapter pointer set in secondary process

[Ivan Malov]

Add the missing assignment to preclude NULL pointer dereference that has
happened in the context of the secondary process at the dev close stage.

Fixes: 5313b441d8ae ("net/sfc: separate adapter primary process and shared data")
Cc: stable@dpdk.org

Signed-off-by: Ivan Malov <ivan.malov@arknetworks.am>
Reviewed-by: Viacheslav Galaktionov <viacheslav.galaktionov@arknetworks.am>
---
 drivers/net/sfc/sfc_ethdev.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/net/sfc/sfc_ethdev.c b/drivers/net/sfc/sfc_ethdev.c
index 6be91789cf..13619b4b9a 100644
--- a/drivers/net/sfc/sfc_ethdev.c
+++ b/drivers/net/sfc/sfc_ethdev.c
@@ -3224,6 +3224,7 @@ sfc_eth_dev_secondary_init(struct rte_eth_dev *dev, uint32_t logtype_main)
 
 	sap->dp_rx = dp_rx;
 	sap->dp_tx = dp_tx;
+	sap->shared = sas;
 
 	dev->process_private = sap;
 	dev->rx_pkt_burst = dp_rx->pkt_burst;
-- 
2.47.3

[PATCH v2 2/2] net/sfc: provide cached dev info to use in secondary process

[Ivan Malov]

Secondary process support in the 'test-pmd' application now requires that
the driver expose the 'dev_infos_get' method within that context. Use the
cached dev info from the primary process in order to meet the requirement.

Signed-off-by: Ivan Malov <ivan.malov@arknetworks.am>
Reviewed-by: Viacheslav Galaktionov <viacheslav.galaktionov@arknetworks.am>
---
 drivers/net/sfc/sfc.h        | 19 ++++++++++++++
 drivers/net/sfc/sfc_ethdev.c | 49 ++++++++++++++++++++++++++++++++++++
 2 files changed, 68 insertions(+)

diff --git a/drivers/net/sfc/sfc.h b/drivers/net/sfc/sfc.h
index 629578549f..809ad59148 100644
--- a/drivers/net/sfc/sfc.h
+++ b/drivers/net/sfc/sfc.h
@@ -156,6 +156,25 @@ struct sfc_adapter_shared {
 	unsigned int			nb_repr_txq;
 
 	struct sfc_nic_dma_info		nic_dma_info;
+
+	/*
+	 * Snapshot of the 'rte_eth_dev_info_get' output created by the primary
+	 * process attach path for the secondary process to use in its own
+	 * implementation of the 'dev_infos_get' method.
+	 *
+	 * Some driver-computed fields derived from mutable post-attach state
+	 * are knowingly stale, which is acceptable for the secondary process.
+	 *
+	 * This also contains a handful of stale fields which are normally
+	 * set by the ethdev layer upon invocation of the 'dev_infos_get',
+	 * so they will be overridden anyway in the secondary process.
+	 */
+	struct rte_eth_dev_info		dev_info_cache;
+	/*
+	 * Set to 'true' by the probe function from the primary process. The
+	 * secondary 'dev_infos_get' returns '-EAGAIN' when this is 'false'.
+	 */
+	RTE_ATOMIC(bool)		dev_info_cache_is_valid;
 };
 
 /* Adapter process private data */
diff --git a/drivers/net/sfc/sfc_ethdev.c b/drivers/net/sfc/sfc_ethdev.c
index 13619b4b9a..2739985bab 100644
--- a/drivers/net/sfc/sfc_ethdev.c
+++ b/drivers/net/sfc/sfc_ethdev.c
@@ -16,7 +16,9 @@
 #include <bus_pci_driver.h>
 #include <rte_errno.h>
 #include <rte_string_fns.h>
+#include <rte_stdatomic.h>
 #include <rte_bitops.h>
+#include <rte_ethdev.h>
 #include <rte_ether.h>
 
 #include "efx.h"
@@ -3163,7 +3165,32 @@ sfc_eth_dev_clear_ops(struct rte_eth_dev *dev)
 	sa->priv.dp_rx = NULL;
 }
 
+static int
+sfc_dev_infos_get_secondary(struct rte_eth_dev *dev,
+			    struct rte_eth_dev_info *dev_info)
+{
+	const struct sfc_adapter_shared *sas =
+		sfc_adapter_shared_by_eth_dev(dev);
+	bool valid = rte_atomic_load_explicit(&sas->dev_info_cache_is_valid,
+					      rte_memory_order_acquire);
+
+	if (!valid)
+		return -EAGAIN;
+
+	*dev_info = sas->dev_info_cache;
+
+	/*
+	 * The cache holds stale primary-process pointers; restore
+	 * the process-local values from the caller-supplied 'dev'.
+	 */
+	if (dev_info->switch_info.name != NULL)
+		dev_info->switch_info.name = dev->device->driver->name;
+	dev_info->device = dev->device;
+	return 0;
+}
+
 static const struct eth_dev_ops sfc_eth_dev_secondary_ops = {
+	.dev_infos_get			= sfc_dev_infos_get_secondary,
 	.dev_supported_ptypes_get	= sfc_dev_supported_ptypes_get,
 	.reta_query			= sfc_dev_rss_reta_query,
 	.rss_hash_conf_get		= sfc_dev_rss_hash_conf_get,
@@ -3749,6 +3776,7 @@ static int sfc_eth_dev_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
 	struct rte_pci_device *pci_dev)
 {
 	struct sfc_ethdev_init_data init_data;
+	struct sfc_adapter_shared *sas;
 	struct rte_eth_devargs eth_da;
 	struct rte_eth_dev *dev;
 	bool dev_created;
@@ -3787,6 +3815,27 @@ static int sfc_eth_dev_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
 	if (rc != 0)
 		return rc;
 
+	sas = sfc_adapter_shared_by_eth_dev(dev);
+
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
+		struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev);
+
+		/*
+		 * Pre-fill the dev info cache for the secondary
+		 * process. The port has been registered at this
+		 * point, allowing use of the public API.
+		 */
+		rc = rte_eth_dev_info_get(dev->data->port_id,
+					  &sas->dev_info_cache);
+		if (rc == 0) {
+			sas->dev_info_cache.device = NULL;
+			rte_atomic_store_explicit(&sas->dev_info_cache_is_valid,
+				true, rte_memory_order_release);
+		} else {
+			sfc_warn(sa, "failed to cache dev info for the secondary process");
+		}
+	}
+
 	rc = sfc_eth_dev_create_representors(dev, &eth_da);
 	if (rc != 0) {
 		if (dev_created)
-- 
2.47.3

[PATCH v3 0/2] net/sfc: fix and extend secondary process support

[Ivan Malov]

The series addresses two secondary process deficiencies.

Patch 1 adds a missing shared adapter pointer assignment to
the secondary process init path, preventing a NULL pointer
dereference at device close time.

Patch 2 pre-populates a primary-process device info cache
to satisfy the 'dev_infos_get' requirement now imposed by
the 'test-pmd' application on secondary processes.


v3:

- cosmetic fix


v2:

- addressed https://mails.dpdk.org/archives/dev/2026-August/343941.html


Ivan Malov (2):
  net/sfc: fix shared adapter pointer set in secondary process
  net/sfc: provide cached dev info to use in secondary process

 drivers/net/sfc/sfc.h        | 19 ++++++++++++++
 drivers/net/sfc/sfc_ethdev.c | 49 ++++++++++++++++++++++++++++++++++++
 2 files changed, 68 insertions(+)

-- 
2.47.3

[PATCH v3 1/2] net/sfc: fix shared adapter pointer set in secondary process

[Ivan Malov]

Add the missing assignment to preclude NULL pointer dereference that has
happened in the context of the secondary process at the dev close stage.

Fixes: 5313b441d8ae ("net/sfc: separate adapter primary process and shared data")
Cc: stable@dpdk.org

Signed-off-by: Ivan Malov <ivan.malov@arknetworks.am>
Reviewed-by: Viacheslav Galaktionov <viacheslav.galaktionov@arknetworks.am>
---
 drivers/net/sfc/sfc_ethdev.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/net/sfc/sfc_ethdev.c b/drivers/net/sfc/sfc_ethdev.c
index 6be91789cf..13619b4b9a 100644
--- a/drivers/net/sfc/sfc_ethdev.c
+++ b/drivers/net/sfc/sfc_ethdev.c
@@ -3224,6 +3224,7 @@ sfc_eth_dev_secondary_init(struct rte_eth_dev *dev, uint32_t logtype_main)
 
 	sap->dp_rx = dp_rx;
 	sap->dp_tx = dp_tx;
+	sap->shared = sas;
 
 	dev->process_private = sap;
 	dev->rx_pkt_burst = dp_rx->pkt_burst;
-- 
2.47.3

[PATCH v3 2/2] net/sfc: provide cached dev info to use in secondary process

[Ivan Malov]

Secondary process support in the 'test-pmd' application now requires that
the driver expose the 'dev_infos_get' method within that context. Use the
cached dev info from the primary process in order to meet the requirement.

Signed-off-by: Ivan Malov <ivan.malov@arknetworks.am>
Reviewed-by: Viacheslav Galaktionov <viacheslav.galaktionov@arknetworks.am>
---
 drivers/net/sfc/sfc.h        | 19 ++++++++++++++
 drivers/net/sfc/sfc_ethdev.c | 48 ++++++++++++++++++++++++++++++++++++
 2 files changed, 67 insertions(+)

diff --git a/drivers/net/sfc/sfc.h b/drivers/net/sfc/sfc.h
index 629578549f..809ad59148 100644
--- a/drivers/net/sfc/sfc.h
+++ b/drivers/net/sfc/sfc.h
@@ -156,6 +156,25 @@ struct sfc_adapter_shared {
 	unsigned int			nb_repr_txq;
 
 	struct sfc_nic_dma_info		nic_dma_info;
+
+	/*
+	 * Snapshot of the 'rte_eth_dev_info_get' output created by the primary
+	 * process attach path for the secondary process to use in its own
+	 * implementation of the 'dev_infos_get' method.
+	 *
+	 * Some driver-computed fields derived from mutable post-attach state
+	 * are knowingly stale, which is acceptable for the secondary process.
+	 *
+	 * This also contains a handful of stale fields which are normally
+	 * set by the ethdev layer upon invocation of the 'dev_infos_get',
+	 * so they will be overridden anyway in the secondary process.
+	 */
+	struct rte_eth_dev_info		dev_info_cache;
+	/*
+	 * Set to 'true' by the probe function from the primary process. The
+	 * secondary 'dev_infos_get' returns '-EAGAIN' when this is 'false'.
+	 */
+	RTE_ATOMIC(bool)		dev_info_cache_is_valid;
 };
 
 /* Adapter process private data */
diff --git a/drivers/net/sfc/sfc_ethdev.c b/drivers/net/sfc/sfc_ethdev.c
index 13619b4b9a..a02697a7cd 100644
--- a/drivers/net/sfc/sfc_ethdev.c
+++ b/drivers/net/sfc/sfc_ethdev.c
@@ -16,7 +16,9 @@
 #include <bus_pci_driver.h>
 #include <rte_errno.h>
 #include <rte_string_fns.h>
+#include <rte_stdatomic.h>
 #include <rte_bitops.h>
+#include <rte_ethdev.h>
 #include <rte_ether.h>
 
 #include "efx.h"
@@ -3163,7 +3165,32 @@ sfc_eth_dev_clear_ops(struct rte_eth_dev *dev)
 	sa->priv.dp_rx = NULL;
 }
 
+static int
+sfc_dev_infos_get_secondary(struct rte_eth_dev *dev,
+			    struct rte_eth_dev_info *dev_info)
+{
+	const struct sfc_adapter_shared *sas =
+		sfc_adapter_shared_by_eth_dev(dev);
+	bool valid = rte_atomic_load_explicit(&sas->dev_info_cache_is_valid,
+					      rte_memory_order_acquire);
+
+	if (!valid)
+		return -EAGAIN;
+
+	*dev_info = sas->dev_info_cache;
+
+	/*
+	 * The cache holds stale primary-process pointers; restore
+	 * the process-local values from the caller-supplied 'dev'.
+	 */
+	if (dev_info->switch_info.name != NULL)
+		dev_info->switch_info.name = dev->device->driver->name;
+	dev_info->device = dev->device;
+	return 0;
+}
+
 static const struct eth_dev_ops sfc_eth_dev_secondary_ops = {
+	.dev_infos_get			= sfc_dev_infos_get_secondary,
 	.dev_supported_ptypes_get	= sfc_dev_supported_ptypes_get,
 	.reta_query			= sfc_dev_rss_reta_query,
 	.rss_hash_conf_get		= sfc_dev_rss_hash_conf_get,
@@ -3787,6 +3814,27 @@ static int sfc_eth_dev_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
 	if (rc != 0)
 		return rc;
 
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
+		struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev);
+		struct sfc_adapter_shared *sas =
+			sfc_adapter_shared_by_eth_dev(dev);
+
+		/*
+		 * Pre-fill the dev info cache for the secondary
+		 * process. The port has been registered at this
+		 * point, allowing use of the public API.
+		 */
+		rc = rte_eth_dev_info_get(dev->data->port_id,
+					  &sas->dev_info_cache);
+		if (rc == 0) {
+			sas->dev_info_cache.device = NULL;
+			rte_atomic_store_explicit(&sas->dev_info_cache_is_valid,
+				true, rte_memory_order_release);
+		} else {
+			sfc_warn(sa, "failed to cache dev info for the secondary process");
+		}
+	}
+
 	rc = sfc_eth_dev_create_representors(dev, &eth_da);
 	if (rc != 0) {
 		if (dev_created)
-- 
2.47.3

Re: [PATCH v3 2/2] net/sfc: provide cached dev info to use in secondary process

[Stephen Hemminger]

On Sat, 22 Aug 2026 04:06:46 +0400
Ivan Malov <ivan.malov@arknetworks.am> wrote:

> Secondary process support in the 'test-pmd' application now requires that
> the driver expose the 'dev_infos_get' method within that context. Use the
> cached dev info from the primary process in order to meet the requirement.
> 
> Signed-off-by: Ivan Malov <ivan.malov@arknetworks.am>
> Reviewed-by: Viacheslav Galaktionov <viacheslav.galaktionov@arknetworks.am>
> ---

Looks like AI does see one valid race in the use of the cache.
Sorry for the word salad. It is hard to get it be concise.

Review of [PATCH v3 2/2] net/sfc: provide cached dev info to use in
secondary process

Error: the driver sets RTE_PCI_DRV_PROBE_AGAIN, and re-probe of an
existing device (to add representors) runs this code again when the
ethdev already exists, dev_info_cache_is_valid is already true, and
secondary processes are attached and may call dev_infos_get at any
time.  rte_eth_dev_info_get() begins by memset()ing the structure it
is given, which here is the shared cache, so it is zeroed and then
refilled while the flag stays true; a secondary reading during that
window gets zeroed or partially refilled dev info reported as valid.
This is not the initial-fill race raised on v2 -- there the release
store after the fill does order things correctly.

Fix: only fill the cache when the ethdev was just created, since the
values do not change between probes of the same device.  dev_created
is already in scope:

	if (dev_created && rte_eal_process_type() == RTE_PROC_PRIMARY) {

[PATCH v4 0/2] net/sfc: fix and extend secondary process support

[Ivan Malov]

The series addresses two secondary process deficiencies.

Patch 1 adds a missing shared adapter pointer assignment to
the secondary process init path, preventing a NULL pointer
dereference at device close time.

Patch 2 pre-populates a primary-process device info cache
to satisfy the 'dev_infos_get' requirement now imposed by
the 'test-pmd' application on secondary processes.


v4:

- addressed https://mails.dpdk.org/archives/dev/2026-August/344068.html


v3:

- cosmetic fix


v2:

- addressed https://mails.dpdk.org/archives/dev/2026-August/343941.html


Ivan Malov (2):
  net/sfc: fix shared adapter pointer set in secondary process
  net/sfc: provide cached dev info to use in secondary process

 drivers/net/sfc/sfc.h        | 19 ++++++++++++++
 drivers/net/sfc/sfc_ethdev.c | 49 ++++++++++++++++++++++++++++++++++++
 2 files changed, 68 insertions(+)

-- 
2.47.3

[PATCH v4 1/2] net/sfc: fix shared adapter pointer set in secondary process

[Ivan Malov]

Add the missing assignment to preclude NULL pointer dereference that has
happened in the context of the secondary process at the dev close stage.

Fixes: 5313b441d8ae ("net/sfc: separate adapter primary process and shared data")
Cc: stable@dpdk.org

Signed-off-by: Ivan Malov <ivan.malov@arknetworks.am>
Reviewed-by: Viacheslav Galaktionov <viacheslav.galaktionov@arknetworks.am>
---
 drivers/net/sfc/sfc_ethdev.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/net/sfc/sfc_ethdev.c b/drivers/net/sfc/sfc_ethdev.c
index 6be91789cf..13619b4b9a 100644
--- a/drivers/net/sfc/sfc_ethdev.c
+++ b/drivers/net/sfc/sfc_ethdev.c
@@ -3224,6 +3224,7 @@ sfc_eth_dev_secondary_init(struct rte_eth_dev *dev, uint32_t logtype_main)
 
 	sap->dp_rx = dp_rx;
 	sap->dp_tx = dp_tx;
+	sap->shared = sas;
 
 	dev->process_private = sap;
 	dev->rx_pkt_burst = dp_rx->pkt_burst;
-- 
2.47.3

[PATCH v4 2/2] net/sfc: provide cached dev info to use in secondary process

[Ivan Malov]

Secondary process support in the 'test-pmd' application now requires that
the driver expose the 'dev_infos_get' method within that context. Use the
cached dev info from the primary process in order to meet the requirement.

Signed-off-by: Ivan Malov <ivan.malov@arknetworks.am>
Reviewed-by: Viacheslav Galaktionov <viacheslav.galaktionov@arknetworks.am>
---
 drivers/net/sfc/sfc.h        | 19 ++++++++++++++
 drivers/net/sfc/sfc_ethdev.c | 48 ++++++++++++++++++++++++++++++++++++
 2 files changed, 67 insertions(+)

diff --git a/drivers/net/sfc/sfc.h b/drivers/net/sfc/sfc.h
index 629578549f..809ad59148 100644
--- a/drivers/net/sfc/sfc.h
+++ b/drivers/net/sfc/sfc.h
@@ -156,6 +156,25 @@ struct sfc_adapter_shared {
 	unsigned int			nb_repr_txq;
 
 	struct sfc_nic_dma_info		nic_dma_info;
+
+	/*
+	 * Snapshot of the 'rte_eth_dev_info_get' output created by the primary
+	 * process attach path for the secondary process to use in its own
+	 * implementation of the 'dev_infos_get' method.
+	 *
+	 * Some driver-computed fields derived from mutable post-attach state
+	 * are knowingly stale, which is acceptable for the secondary process.
+	 *
+	 * This also contains a handful of stale fields which are normally
+	 * set by the ethdev layer upon invocation of the 'dev_infos_get',
+	 * so they will be overridden anyway in the secondary process.
+	 */
+	struct rte_eth_dev_info		dev_info_cache;
+	/*
+	 * Set to 'true' by the probe function from the primary process. The
+	 * secondary 'dev_infos_get' returns '-EAGAIN' when this is 'false'.
+	 */
+	RTE_ATOMIC(bool)		dev_info_cache_is_valid;
 };
 
 /* Adapter process private data */
diff --git a/drivers/net/sfc/sfc_ethdev.c b/drivers/net/sfc/sfc_ethdev.c
index 13619b4b9a..02231e4bce 100644
--- a/drivers/net/sfc/sfc_ethdev.c
+++ b/drivers/net/sfc/sfc_ethdev.c
@@ -16,7 +16,9 @@
 #include <bus_pci_driver.h>
 #include <rte_errno.h>
 #include <rte_string_fns.h>
+#include <rte_stdatomic.h>
 #include <rte_bitops.h>
+#include <rte_ethdev.h>
 #include <rte_ether.h>
 
 #include "efx.h"
@@ -3163,7 +3165,32 @@ sfc_eth_dev_clear_ops(struct rte_eth_dev *dev)
 	sa->priv.dp_rx = NULL;
 }
 
+static int
+sfc_dev_infos_get_secondary(struct rte_eth_dev *dev,
+			    struct rte_eth_dev_info *dev_info)
+{
+	const struct sfc_adapter_shared *sas =
+		sfc_adapter_shared_by_eth_dev(dev);
+	bool valid = rte_atomic_load_explicit(&sas->dev_info_cache_is_valid,
+					      rte_memory_order_acquire);
+
+	if (!valid)
+		return -EAGAIN;
+
+	*dev_info = sas->dev_info_cache;
+
+	/*
+	 * The cache holds stale primary-process pointers; restore
+	 * the process-local values from the caller-supplied 'dev'.
+	 */
+	if (dev_info->switch_info.name != NULL)
+		dev_info->switch_info.name = dev->device->driver->name;
+	dev_info->device = dev->device;
+	return 0;
+}
+
 static const struct eth_dev_ops sfc_eth_dev_secondary_ops = {
+	.dev_infos_get			= sfc_dev_infos_get_secondary,
 	.dev_supported_ptypes_get	= sfc_dev_supported_ptypes_get,
 	.reta_query			= sfc_dev_rss_reta_query,
 	.rss_hash_conf_get		= sfc_dev_rss_hash_conf_get,
@@ -3787,6 +3814,27 @@ static int sfc_eth_dev_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
 	if (rc != 0)
 		return rc;
 
+	if (dev_created && rte_eal_process_type() == RTE_PROC_PRIMARY) {
+		struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev);
+		struct sfc_adapter_shared *sas =
+			sfc_adapter_shared_by_eth_dev(dev);
+
+		/*
+		 * Pre-fill the dev info cache for the secondary
+		 * process. The port has been registered at this
+		 * point, allowing use of the public API.
+		 */
+		rc = rte_eth_dev_info_get(dev->data->port_id,
+					  &sas->dev_info_cache);
+		if (rc == 0) {
+			sas->dev_info_cache.device = NULL;
+			rte_atomic_store_explicit(&sas->dev_info_cache_is_valid,
+				true, rte_memory_order_release);
+		} else {
+			sfc_warn(sa, "failed to cache dev info for the secondary process");
+		}
+	}
+
 	rc = sfc_eth_dev_create_representors(dev, &eth_da);
 	if (rc != 0) {
 		if (dev_created)
-- 
2.47.3

Re: [PATCH v4 0/2] net/sfc: fix and extend secondary process support

[Stephen Hemminger]

On Mon, 24 Aug 2026 20:30:31 +0400
Ivan Malov <ivan.malov@arknetworks.am> wrote:

> The series addresses two secondary process deficiencies.
> 
> Patch 1 adds a missing shared adapter pointer assignment to
> the secondary process init path, preventing a NULL pointer
> dereference at device close time.
> 
> Patch 2 pre-populates a primary-process device info cache
> to satisfy the 'dev_infos_get' requirement now imposed by
> the 'test-pmd' application on secondary processes.

Applied to next-net