Skip to contentSkip to conversation

net: mana: hardening: Reject zero max_num_queues from GDMA_QUERY_MAX_RESOURCES

PATCH Applied Erni Sri Satya Vennela 3 participants 3 messages Apr 30 – May 5

Subsystems (4)
Networking Network drivers Hyper-V/Azure core and drivers Microsoft MANA RDMA driver
lore
Share & tools
lore

Erni Sri Satya Vennela

author
#0 lore raw

PATCH v2 1/1 net: mana: hardening: Reject zero max_num_queues from GDMA_QUERY_MAX_RESOURCES

In a CVM environment, hardware responses cannot be trusted. The GDMA_QUERY_MAX_RESOURCES command returns resource limits used to determine the maximum number of queues.

In mana_gd_query_max_resources(), gc->max_num_queues is initialized from num_online_cpus() and successively clamped by the hardware-reported max_eq, max_cq, max_sq, max_rq, and num_msix_usable values. If any of these hardware values is zero, gc->max_num_queues becomes zero and the function returns success. This leads to a confusing failure later when alloc_etherdev_mq() is called with zero queues, returning NULL and producing a misleading -ENOMEM error.

Add an explicit zero check for gc->max_num_queues after all clamping steps and return -ENOSPC for a clear early failure, consistent with the existing gc->num_msix_usable <= 1 guard.

Signed-off-by Erni Sri Satya Vennela <[email protected]>

---
Changes in v2:
* Rebase to latest main.

---
 drivers/net/ethernet/microsoft/mana/gdma_main.c | 3 +++
 1 file changed, 3 insertions(+)
drivers/net/ethernet/microsoft/mana/gdma_main.c +3 -0
index 098fbda0d128..f3316e929175 100644
@@ -194,6 +194,9 @@ static int mana_gd_query_max_resources(struct pci_dev *pdev)
 	if (gc->max_num_queues > gc->num_msix_usable - 1)
 		gc->max_num_queues = gc->num_msix_usable - 1;
 
+	if (gc->max_num_queues == 0)
+		return -ENOSPC;
+
 	return 0;
 }
Signature · 1 line
2.34.1
Reviewed-by on v2 1/1
#1 lore raw
Quoted 29 lines · Erni Sri Satya Vennela

In a CVM environment, hardware responses cannot be trusted. The GDMA_QUERY_MAX_RESOURCES command returns resource limits used to determine the maximum number of queues.

In mana_gd_query_max_resources(), gc->max_num_queues is initialized from num_online_cpus() and successively clamped by the hardware-reported max_eq, max_cq, max_sq, max_rq, and num_msix_usable values. If any of these hardware values is zero, gc->max_num_queues becomes zero and the function returns success. This leads to a confusing failure later when alloc_etherdev_mq() is called with zero queues, returning NULL and producing a misleading -ENOMEM error.

Add an explicit zero check for gc->max_num_queues after all clamping steps and return -ENOSPC for a clear early failure, consistent with the existing gc->num_msix_usable <= 1 guard.

Signed-off-by Erni Sri Satya Vennela <[email protected]>

---
Changes in v2:
* Rebase to latest main.

---
 drivers/net/ethernet/microsoft/mana/gdma_main.c | 3 +++
 1 file changed, 3 insertions(+)
drivers/net/ethernet/microsoft/mana/gdma_main.c +0 -0
index 098fbda0d128..f3316e929175 100644

drivers/net/ethernet/microsoft/mana/gdma_main.c static int mana_gd_query_max_resources(struct pci_dev *pdev) · line 202

Quoted 9 lines
drivers/net/ethernet/microsoft/mana/gdma_main.c +3 -0
@@ -194,6 +194,9 @@ static int mana_gd_query_max_resources(struct pci_dev *pdev)
 	if (gc->max_num_queues > gc->num_msix_usable - 1)
 		gc->max_num_queues = gc->num_msix_usable - 1;
 
+	if (gc->max_num_queues == 0)
+		return -ENOSPC;
+
 	return 0;
 }
Reviewed-by Shradha Gupta <[email protected]>
Quoted 2 lines
Signature · 1 line
2.34.1
CI Applied
#2 lore raw
In reply to Erni Sri Satya Vennela
This patch was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:

Hello:

This patch was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:

Quoted 12 lines · On Thu, 30 Apr 2026 01:36:21 -0700 you wrote:

In a CVM environment, hardware responses cannot be trusted. The GDMA_QUERY_MAX_RESOURCES command returns resource limits used to determine the maximum number of queues.

In mana_gd_query_max_resources(), gc->max_num_queues is initialized from num_online_cpus() and successively clamped by the hardware-reported max_eq, max_cq, max_sq, max_rq, and num_msix_usable values. If any of these hardware values is zero, gc->max_num_queues becomes zero and the function returns success. This leads to a confusing failure later when alloc_etherdev_mq() is called with zero queues, returning NULL and producing a misleading -ENOMEM error.

Quoted 1 line

[...]

Here is the summary with links:
- [net-next,v2] net: mana: hardening: Reject zero max_num_queues from GDMA_QUERY_MAX_RESOURCES

    https://git.kernel.org/netdev/net-next/c/f7622e58e802

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html