Skip to contentSkip to conversation

bnxt: fix building without CONFIG_RFS_ACCEL

PATCH In review Arnd Bergmann 4 participants 4 messages Jan 3, 2024

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 bnxt: fix building without CONFIG_RFS_ACCEL

From: Arnd Bergmann <[email protected]>

A recent patch series generalized the filter logic in bnxt to no longer just be used for RFS, but it now fails to build when RFS_ACCEL is disabled:

drivers/net/ethernet/broadcom/bnxt/bnxt.c: In function 'bnxt_cfg_ntp_filters':
drivers/net/ethernet/broadcom/bnxt/bnxt.c:14077:37: error: implicit declaration of function 'rps_may_expire_flow' [-Werror=implicit-function-declaration]
14077 | if (rps_may_expire_flow(bp->dev, fltr->base.rxq,

      |                                     ^~~~~~~~~~~~~~~~~~~

Add back one #ifdef check around a call to the missing rps_may_expire_flow() function.

Fixes 59cde76f33fa ("bnxt_en: Refactor filter insertion logic in bnxt_rx_flow_steer().")
Cc Michael Chan <[email protected]>

---
I don't know if this is a correct fix, only checked that it is plausible and that it does address the build failure. If a different fix is needed, please just treat this as a bug report.

---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/net/ethernet/broadcom/bnxt/bnxt.c +2 -0
index 827821e89c40..83a97c65b728 100644
@@ -14074,6 +14074,7 @@ static void bnxt_cfg_ntp_filters(struct bnxt *bp)
 			if (test_bit(BNXT_FLTR_VALID, &fltr->base.state)) {
 				if (fltr->base.flags & BNXT_ACT_NO_AGING)
 					continue;
+#if IS_ENABLED(CONFIG_RFS_ACCEL)
 				if (rps_may_expire_flow(bp->dev, fltr->base.rxq,
 							fltr->flow_id,
 							fltr->base.sw_id)) {
@@ -14081,6 +14082,7 @@ static void bnxt_cfg_ntp_filters(struct bnxt *bp)
 									 fltr);
 					del = true;
 				}
+#endif
 			} else {
 				rc = bnxt_hwrm_cfa_ntuple_filter_alloc(bp,
 								       fltr);
Signature · 1 line
2.39.2
#1 lore raw
Quoted 14 lines · Arnd Bergmann

From: Arnd Bergmann <[email protected]>

A recent patch series generalized the filter logic in bnxt to no longer just be used for RFS, but it now fails to build when RFS_ACCEL is disabled:

drivers/net/ethernet/broadcom/bnxt/bnxt.c: In function 'bnxt_cfg_ntp_filters':
drivers/net/ethernet/broadcom/bnxt/bnxt.c:14077:37: error: implicit declaration of function 'rps_may_expire_flow' [-Werror=implicit-function-declaration]
14077 | if (rps_may_expire_flow(bp->dev, fltr->base.rxq,

      |                                     ^~~~~~~~~~~~~~~~~~~

Add back one #ifdef check around a call to the missing rps_may_expire_flow() function.

Quoted 7 lines
Fixes 59cde76f33fa ("bnxt_en: Refactor filter insertion logic in bnxt_rx_flow_steer().")
Cc Michael Chan <[email protected]>

---
I don't know if this is a correct fix, only checked that it is plausible and that it does address the build failure. If a different fix is needed, please just treat this as a bug report.

Are you using a kernel config with CONFIG_SMP=n ? That was how I was able to reproduce this.

There is a good oppportunity to clean this up a little better. When CONFIG_RFS_ACCEL is not set there is no reason to even have bnxt_cfg_ntp_filters included in the driver build.

I'll talk to Michael and we will post a fix for this.

Quoted 27 lines
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/net/ethernet/broadcom/bnxt/bnxt.c +2 -0
index 827821e89c40..83a97c65b728 100644
@@ -14074,6 +14074,7 @@ static void bnxt_cfg_ntp_filters(struct bnxt *bp)
 			if (test_bit(BNXT_FLTR_VALID, &fltr->base.state)) {
 				if (fltr->base.flags & BNXT_ACT_NO_AGING)
 					continue;
+#if IS_ENABLED(CONFIG_RFS_ACCEL)
 				if (rps_may_expire_flow(bp->dev, fltr->base.rxq,
 							fltr->flow_id,
 							fltr->base.sw_id)) {
@@ -14081,6 +14082,7 @@ static void bnxt_cfg_ntp_filters(struct bnxt *bp)
 									 fltr);
 					del = true;
 				}
+#endif
 			} else {
 				rc = bnxt_hwrm_cfa_ntuple_filter_alloc(bp,
 								       fltr);
Signature · 1 line
2.39.2
#2 lore raw
Quoted 23 lines · Andy Gospodarek
Quoted 14 lines · Arnd Bergmann

From: Arnd Bergmann <[email protected]>

A recent patch series generalized the filter logic in bnxt to no longer just be used for RFS, but it now fails to build when RFS_ACCEL is disabled:

drivers/net/ethernet/broadcom/bnxt/bnxt.c: In function 'bnxt_cfg_ntp_filters':
drivers/net/ethernet/broadcom/bnxt/bnxt.c:14077:37: error: implicit declaration of function 'rps_may_expire_flow' [-Werror=implicit-function-declaration]
14077 | if (rps_may_expire_flow(bp->dev, fltr->base.rxq,

      |                                     ^~~~~~~~~~~~~~~~~~~

Add back one #ifdef check around a call to the missing rps_may_expire_flow() function.

Quoted 7 lines
Fixes 59cde76f33fa ("bnxt_en: Refactor filter insertion logic in bnxt_rx_flow_steer().")
Cc Michael Chan <[email protected]>

---
I don't know if this is a correct fix, only checked that it is plausible and that it does address the build failure. If a different fix is needed, please just treat this as a bug report.

Quoted 2 lines

Are you using a kernel config with CONFIG_SMP=n ? That was how I was able to reproduce this.

I saw this in two randconfig test builds on 32-bit arm, both with SMP disabled, though I did not expect that to make a difference.

In case it helps, this is one of the two config files:
https://pastebin.com/raw/uxHEXzG2

Quoted 5 lines

There is a good oppportunity to clean this up a little better. When CONFIG_RFS_ACCEL is not set there is no reason to even have bnxt_cfg_ntp_filters included in the driver build.

I'll talk to Michael and we will post a fix for this.

Ok, thanks!

Arnd

maintainer
#3 lore raw
Quoted 9 lines · Arnd Bergmann
Quoted 5 lines · Andy Gospodarek

There is a good oppportunity to clean this up a little better. When CONFIG_RFS_ACCEL is not set there is no reason to even have bnxt_cfg_ntp_filters included in the driver build.

I'll talk to Michael and we will post a fix for this.

Ok, thanks!

Yes, we can clean this up better. User configured ntuple filters are directly added and don't go through bnxt_cfg_ntp_filters() by way of the workqueue. I'll post the fix later today. Thanks.