Skip to contentSkip to conversation

sfc: reduce the number of requested xdp ev queues

PATCH Applied Ivan Babrou 6 participants 6 messages Jan 21, 2021 – Feb 6, 2021

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 sfc: reduce the number of requested xdp ev queues

Without this change the driver tries to allocate too many queues, breaching the number of available msi-x interrupts on machines with many logical cpus and default adapter settings:

Insufficient resources for 12 XDP event queues (24 other channels, max 32)

Which in turn triggers EINVAL on XDP processing:

sfc 0000:86:00.0 ext0: XDP TX failed (-22)

---
 drivers/net/ethernet/sfc/efx_channels.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
drivers/net/ethernet/sfc/efx_channels.c +4 -2
index a4a626e9cd9a..1bfeee283ea9 100644
@@ -17,6 +17,7 @@
 #include "rx_common.h"
 #include "nic.h"
 #include "sriov.h"
+#include "workarounds.h"
 
 /* This is the first interrupt mode to try out of:
  * 0 => MSI-X
@@ -137,6 +138,7 @@ static int efx_allocate_msix_channels(struct efx_nic *efx,
 {
 	unsigned int n_channels = parallelism;
 	int vec_count;
+	int tx_per_ev;
 	int n_xdp_tx;
 	int n_xdp_ev;
 
@@ -149,9 +151,9 @@ static int efx_allocate_msix_channels(struct efx_nic *efx,
 	 * multiple tx queues, assuming tx and ev queues are both
 	 * maximum size.
 	 */
-
+	tx_per_ev = EFX_MAX_EVQ_SIZE / EFX_TXQ_MAX_ENT(efx);
 	n_xdp_tx = num_possible_cpus();
-	n_xdp_ev = DIV_ROUND_UP(n_xdp_tx, EFX_MAX_TXQ_PER_CHANNEL);
+	n_xdp_ev = DIV_ROUND_UP(n_xdp_tx, tx_per_ev);
 
 	vec_count = pci_msix_vec_count(efx->pci_dev);
 	if (vec_count < 0)
Signature · 1 line
2.29.2
maintainer Acked-by on v1 1/1
#1 lore raw
Quoted 11 lines · Ivan Babrou

Without this change the driver tries to allocate too many queues, breaching the number of available msi-x interrupts on machines with many logical cpus and default adapter settings:

Insufficient resources for 12 XDP event queues (24 other channels, max 32)

Which in turn triggers EINVAL on XDP processing:

sfc 0000:86:00.0 ext0: XDP TX failed (-22)

#2 lore raw
In reply to Ivan Babrou
Quoted 4 lines · On Wed, 20 Jan 2021 13:27:59 -0800 Ivan Babrou <[email protected]> wrote:

Without this change the driver tries to allocate too many queues, breaching the number of available msi-x interrupts on machines with many logical cpus and default adapter settings:

Quoted 8 lines

Insufficient resources for 12 XDP event queues (24 other channels, max 32)

Which in turn triggers EINVAL on XDP processing:

sfc 0000:86:00.0 ext0: XDP TX failed (-22)

---

I guess the patch is good in itself due to available msi-x interrupts.

Per earlier discussion: What will happen if a CPU with an ID higher than available XDP TX-queues redirect a packet out this driver?

Quoted 35 lines
 drivers/net/ethernet/sfc/efx_channels.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
drivers/net/ethernet/sfc/efx_channels.c +4 -2
index a4a626e9cd9a..1bfeee283ea9 100644
@@ -17,6 +17,7 @@
 #include "rx_common.h"
 #include "nic.h"
 #include "sriov.h"
+#include "workarounds.h"
 
 /* This is the first interrupt mode to try out of:
  * 0 => MSI-X
@@ -137,6 +138,7 @@ static int efx_allocate_msix_channels(struct efx_nic *efx,
 {
 	unsigned int n_channels = parallelism;
 	int vec_count;
+	int tx_per_ev;
 	int n_xdp_tx;
 	int n_xdp_ev;
 
@@ -149,9 +151,9 @@ static int efx_allocate_msix_channels(struct efx_nic *efx,
 	 * multiple tx queues, assuming tx and ev queues are both
 	 * maximum size.
 	 */
-
+	tx_per_ev = EFX_MAX_EVQ_SIZE / EFX_TXQ_MAX_ENT(efx);
 	n_xdp_tx = num_possible_cpus();
-	n_xdp_ev = DIV_ROUND_UP(n_xdp_tx, EFX_MAX_TXQ_PER_CHANNEL);
+	n_xdp_ev = DIV_ROUND_UP(n_xdp_tx, tx_per_ev);
 
 	vec_count = pci_msix_vec_count(efx->pci_dev);
 	if (vec_count < 0)
Signature · 4 lines
Best regards,
  Jesper Dangaard Brouer
  MSc.CS, Principal Kernel Engineer at Red Hat
  LinkedIn: http://www.linkedin.com/in/brouer
#3 lore raw
Quoted 7 lines · Jesper Dangaard Brouer
Quoted 4 lines · On Wed, 20 Jan 2021 13:27:59 -0800 Ivan Babrou <[email protected]> wrote:

Without this change the driver tries to allocate too many queues, breaching the number of available msi-x interrupts on machines with many logical cpus and default adapter settings:

Quoted 5 lines
Quoted 5 lines

Insufficient resources for 12 XDP event queues (24 other channels, max 32)

Which in turn triggers EINVAL on XDP processing:

sfc 0000:86:00.0 ext0: XDP TX failed (-22)

Please mention in commit message *how* you are addressing/fixing this issue.

Quoted 8 lines
Quoted 3 lines

---

I guess the patch is good in itself due to available msi-x interrupts.

Per earlier discussion: What will happen if a CPU with an ID higher than available XDP TX-queues redirect a packet out this driver?

+1 on that question

Quoted 46 lines
Quoted 37 lines

 drivers/net/ethernet/sfc/efx_channels.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
drivers/net/ethernet/sfc/efx_channels.c +4 -2
index a4a626e9cd9a..1bfeee283ea9 100644
@@ -17,6 +17,7 @@
 #include "rx_common.h"
 #include "nic.h"
 #include "sriov.h"
+#include "workarounds.h"
 
 /* This is the first interrupt mode to try out of:
  * 0 => MSI-X
@@ -137,6 +138,7 @@ static int efx_allocate_msix_channels(struct efx_nic *efx,
 {
 	unsigned int n_channels = parallelism;
 	int vec_count;
+	int tx_per_ev;
 	int n_xdp_tx;
 	int n_xdp_ev;
 
@@ -149,9 +151,9 @@ static int efx_allocate_msix_channels(struct efx_nic *efx,
 	 * multiple tx queues, assuming tx and ev queues are both
 	 * maximum size.
 	 */
-
+	tx_per_ev = EFX_MAX_EVQ_SIZE / EFX_TXQ_MAX_ENT(efx);
 	n_xdp_tx = num_possible_cpus();
-	n_xdp_ev = DIV_ROUND_UP(n_xdp_tx, EFX_MAX_TXQ_PER_CHANNEL);
+	n_xdp_ev = DIV_ROUND_UP(n_xdp_tx, tx_per_ev);
 
 	vec_count = pci_msix_vec_count(efx->pci_dev);
 	if (vec_count < 0)
Signature · 4 lines
Best regards,
  Jesper Dangaard Brouer
  MSc.CS, Principal Kernel Engineer at Red Hat
  LinkedIn: http://www.linkedin.com/in/brouer
#4 lore raw
In reply to Ivan Babrou
This patch was applied to netdev/net-next.git (refs/heads/master):

Hello:

This patch was applied to netdev/net-next.git (refs/heads/master):

Quoted 9 lines · On Wed, 20 Jan 2021 13:27:59 -0800 you wrote:

Without this change the driver tries to allocate too many queues, breaching the number of available msi-x interrupts on machines with many logical cpus and default adapter settings:

Insufficient resources for 12 XDP event queues (24 other channels, max 32)

Which in turn triggers EINVAL on XDP processing:

[...]

Here is the summary with links:
- [net-next] sfc: reduce the number of requested xdp ev queues

    https://git.kernel.org/netdev/net-next/c/e26ca4b53582

You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html

Acked-by on v1 1/1
#5 lore raw
In reply to Ivan Babrou
Quoted 11 lines · Ivan Babrou

Without this change the driver tries to allocate too many queues, breaching the number of available msi-x interrupts on machines with many logical cpus and default adapter settings:

Insufficient resources for 12 XDP event queues (24 other channels, max 32)

Which in turn triggers EINVAL on XDP processing:

sfc 0000:86:00.0 ext0: XDP TX failed (-22)

Quoted 38 lines
---
 drivers/net/ethernet/sfc/efx_channels.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
drivers/net/ethernet/sfc/efx_channels.c +4 -2
index a4a626e9cd9a..1bfeee283ea9 100644
@@ -17,6 +17,7 @@
 #include "rx_common.h"
 #include "nic.h"
 #include "sriov.h"
+#include "workarounds.h"
 
 /* This is the first interrupt mode to try out of:
  * 0 => MSI-X
@@ -137,6 +138,7 @@ static int efx_allocate_msix_channels(struct efx_nic *efx,
 {
 	unsigned int n_channels = parallelism;
 	int vec_count;
+	int tx_per_ev;
 	int n_xdp_tx;
 	int n_xdp_ev;
 
@@ -149,9 +151,9 @@ static int efx_allocate_msix_channels(struct efx_nic *efx,
 	 * multiple tx queues, assuming tx and ev queues are both
 	 * maximum size.
 	 */
-
+	tx_per_ev = EFX_MAX_EVQ_SIZE / EFX_TXQ_MAX_ENT(efx);
 	n_xdp_tx = num_possible_cpus();
-	n_xdp_ev = DIV_ROUND_UP(n_xdp_tx, EFX_MAX_TXQ_PER_CHANNEL);
+	n_xdp_ev = DIV_ROUND_UP(n_xdp_tx, tx_per_ev);
 
 	vec_count = pci_msix_vec_count(efx->pci_dev);
 	if (vec_count < 0)
Signature · 1 line
2.29.2