Recently, several link-time issues were spotted in the ethernet/intel/
folder thanks to Kbuild bots and linux-next.
The fixes are pretty straightforward, just some stubs and CONFIG_*
guards, so resolve all of them in one shot and unbreak randconfig
builds.
Alexander Lobakin (3):
ice: fix undefined references to ice_is_*() when
!CONFIG_PTP_1588_CLOCK
ice: fix undefined references from DPLL code when
!CONFIG_PTP_1588_CLOCK
idpf: fix undefined reference to tcp_gro_complete() when !CONFIG_INET
drivers/net/ethernet/intel/ice/Makefile | 5 ++---
drivers/net/ethernet/intel/ice/ice_main.c | 8 ++++---
drivers/net/ethernet/intel/ice/ice_ptp_hw.h | 25 ++++++++++++++++++++-
drivers/net/ethernet/intel/idpf/idpf_txrx.c | 3 +++
4 files changed, 34 insertions(+), 7 deletions(-)
---
Directly to netdev/net-next, build bots are not happy and the next
linux-next is approaching :s
--
2.41.0
DPLL code in ice unconditionally calls several PTP functions which are
only built when CONFIG_PTP_1588_CLOCK is set. This throws a good bunch
of link errors:
ERROR: modpost: "ice_cgu_get_pin_name"
[drivers/net/ethernet/intel/ice/ice.ko] undefined!
ERROR: modpost: "ice_get_cgu_state"
[drivers/net/ethernet/intel/ice/ice.ko] undefined!
OR: modpost: "ice_is_cgu_present"
[drivers/net/ethernet/intel/ice/ice.ko] undefined!
ERROR: modpost: "ice_get_cgu_rclk_pin_info"
[drivers/net/ethernet/intel/ice/ice.ko] undefined!
ERROR: modpost: "ice_cgu_get_pin_type"
[drivers/net/ethernet/intel/ice/ice.ko] undefined!
ERROR: modpost: "ice_cgu_get_pin_freq_supp"
[drivers/net/ethernet/intel/ice/ice.ko] undefined!
ice_dpll_{,de}init() can be only called at runtime when the
corresponding feature flags are set, which is not the case when PTP
support is not compiled. However, the linker has no clue about this.
Compile DPLL code only when CONFIG_PTP_1588_CLOCK is enabled and guard
the mentioned init/deinit function calls, so that ice_dpll.o is only
referred when it gets compiled.
Note that ideally ice_is_feature_supported() needs to check for
compile-time flags first to be able to handle this without any
additional call guards, and we may want to do that in the future.
Fixes: d7999f5ea64b ("ice: implement dpll interface to control cgu")
Reported-by: kernel test robot <[email protected]>
Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]
Signed-off-by: Alexander Lobakin <[email protected]>
---
drivers/net/ethernet/intel/ice/Makefile | 5 ++---
drivers/net/ethernet/intel/ice/ice_main.c | 8 +++++---
2 files changed, 7 insertions(+), 6 deletions(-)
diff --git a/drivers/net/ethernet/intel/ice/Makefile b/drivers/net/ethernet/intel/ice/Makefile
index 00806ddf5bf0..16f96d5210d8 100644
--- a/drivers/net/ethernet/intel/ice/Makefile
+++ b/drivers/net/ethernet/intel/ice/Makefile
@@ -34,8 +34,7 @@ ice-y := ice_main.o \
ice_lag.o \
ice_ethtool.o \
ice_repr.o \
- ice_tc_lib.o \
- ice_dpll.o
+ ice_tc_lib.o
ice-$(CONFIG_PCI_IOV) += \
ice_sriov.o \
ice_virtchnl.o \
@@ -44,7 +43,7 @@ ice-$(CONFIG_PCI_IOV) += \
ice_vf_mbx.o \
ice_vf_vsi_vlan_ops.o \
ice_vf_lib.o
-ice-$(CONFIG_PTP_1588_CLOCK) += ice_ptp.o ice_ptp_hw.o
+ice-$(CONFIG_PTP_1588_CLOCK) += ice_dpll.o ice_ptp.o ice_ptp_hw.o
ice-$(CONFIG_DCB) += ice_dcb.o ice_dcb_nl.o ice_dcb_lib.o
ice-$(CONFIG_RFS_ACCEL) += ice_arfs.o
ice-$(CONFIG_XDP_SOCKETS) += ice_xsk.o
diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c
index e22f41fea8db..9b48918dcdb7 100644
--- a/drivers/net/ethernet/intel/ice/ice_main.c
+++ b/drivers/net/ethernet/intel/ice/ice_main.c
@@ -4665,8 +4665,9 @@ static void ice_init_features(struct ice_pf *pf)
if (ice_is_feature_supported(pf, ICE_F_GNSS))
ice_gnss_init(pf);
- if (ice_is_feature_supported(pf, ICE_F_CGU) ||
- ice_is_feature_supported(pf, ICE_F_PHY_RCLK))
+ if (IS_ENABLED(CONFIG_PTP_1588_CLOCK) &&
+ (ice_is_feature_supported(pf, ICE_F_CGU) ||
+ ice_is_feature_supported(pf, ICE_F_PHY_RCLK)))
ice_dpll_init(pf);
/* Note: Flow director init failure is non-fatal to load */
@@ -4695,7 +4696,8 @@ static void ice_deinit_features(struct ice_pf *pf)
ice_gnss_exit(pf);
if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
ice_ptp_release(pf);
- if (test_bit(ICE_FLAG_DPLL, pf->flags))
+ if (IS_ENABLED(CONFIG_PTP_1588_CLOCK) &&
+ test_bit(ICE_FLAG_DPLL, pf->flags))
ice_dpll_deinit(pf);
}
--
2.41.0
On 20/09/2023 19:07, Alexander Lobakin wrote:
> DPLL code in ice unconditionally calls several PTP functions which are
> only built when CONFIG_PTP_1588_CLOCK is set. This throws a good bunch
> of link errors:
>
> ERROR: modpost: "ice_cgu_get_pin_name"
> [drivers/net/ethernet/intel/ice/ice.ko] undefined!
> ERROR: modpost: "ice_get_cgu_state"
> [drivers/net/ethernet/intel/ice/ice.ko] undefined!
> OR: modpost: "ice_is_cgu_present"
> [drivers/net/ethernet/intel/ice/ice.ko] undefined!
> ERROR: modpost: "ice_get_cgu_rclk_pin_info"
> [drivers/net/ethernet/intel/ice/ice.ko] undefined!
> ERROR: modpost: "ice_cgu_get_pin_type"
> [drivers/net/ethernet/intel/ice/ice.ko] undefined!
> ERROR: modpost: "ice_cgu_get_pin_freq_supp"
> [drivers/net/ethernet/intel/ice/ice.ko] undefined!
>
> ice_dpll_{,de}init() can be only called at runtime when the
> corresponding feature flags are set, which is not the case when PTP
> support is not compiled. However, the linker has no clue about this.
> Compile DPLL code only when CONFIG_PTP_1588_CLOCK is enabled and guard
> the mentioned init/deinit function calls, so that ice_dpll.o is only
> referred when it gets compiled.
>
> Note that ideally ice_is_feature_supported() needs to check for
> compile-time flags first to be able to handle this without any
> additional call guards, and we may want to do that in the future.
>
There is another fix under review [1], which came from Jacob.
It converts the code a bit more, and will create conflicts.
I would suggest to drop this patch until another series is fully
reviewed.
[1]
https://lore.kernel.org/netdev/[email protected]/
On 9/20/23 20:07, Alexander Lobakin wrote:
> Recently, several link-time issues were spotted in the ethernet/intel/
> folder thanks to Kbuild bots and linux-next.
> The fixes are pretty straightforward, just some stubs and CONFIG_*
> guards, so resolve all of them in one shot and unbreak randconfig
> builds.
>
> Alexander Lobakin (3):
> ice: fix undefined references to ice_is_*() when
> !CONFIG_PTP_1588_CLOCK
> ice: fix undefined references from DPLL code when
> !CONFIG_PTP_1588_CLOCK
> idpf: fix undefined reference to tcp_gro_complete() when !CONFIG_INET
>
> drivers/net/ethernet/intel/ice/Makefile | 5 ++---
> drivers/net/ethernet/intel/ice/ice_main.c | 8 ++++---
> drivers/net/ethernet/intel/ice/ice_ptp_hw.h | 25 ++++++++++++++++++++-
> drivers/net/ethernet/intel/idpf/idpf_txrx.c | 3 +++
> 4 files changed, 34 insertions(+), 7 deletions(-)
>
> ---
> Directly to netdev/net-next, build bots are not happy and the next
> linux-next is approaching :s
Reviewed-by: Przemek Kitszel <[email protected]>
I like the `if (IS_ENABLED(xxx))` approach more than stubs :)
From: Vadim Fedorenko <[email protected]>
Date: Fri, 22 Sep 2023 00:58:04 +0100
> On 20/09/2023 19:07, Alexander Lobakin wrote:
>> DPLL code in ice unconditionally calls several PTP functions which are
>> only built when CONFIG_PTP_1588_CLOCK is set. This throws a good bunch
>> of link errors:
>>
>> ERROR: modpost: "ice_cgu_get_pin_name"
>> [drivers/net/ethernet/intel/ice/ice.ko] undefined!
>> ERROR: modpost: "ice_get_cgu_state"
>> [drivers/net/ethernet/intel/ice/ice.ko] undefined!
>> OR: modpost: "ice_is_cgu_present"
>> [drivers/net/ethernet/intel/ice/ice.ko] undefined!
>> ERROR: modpost: "ice_get_cgu_rclk_pin_info"
>> [drivers/net/ethernet/intel/ice/ice.ko] undefined!
>> ERROR: modpost: "ice_cgu_get_pin_type"
>> [drivers/net/ethernet/intel/ice/ice.ko] undefined!
>> ERROR: modpost: "ice_cgu_get_pin_freq_supp"
>> [drivers/net/ethernet/intel/ice/ice.ko] undefined!
>>
>> ice_dpll_{,de}init() can be only called at runtime when the
>> corresponding feature flags are set, which is not the case when PTP
>> support is not compiled. However, the linker has no clue about this.
>> Compile DPLL code only when CONFIG_PTP_1588_CLOCK is enabled and guard
>> the mentioned init/deinit function calls, so that ice_dpll.o is only
>> referred when it gets compiled.
>>
>> Note that ideally ice_is_feature_supported() needs to check for
>> compile-time flags first to be able to handle this without any
>> additional call guards, and we may want to do that in the future.
>>
>
> There is another fix under review [1], which came from Jacob.
> It converts the code a bit more, and will create conflicts.
> I would suggest to drop this patch until another series is fully
> reviewed.
I know, I already explained earlier in this topic that I sent the IDPF
fix separately and that series can be dropped.
>
> [1]
> https://lore.kernel.org/netdev/[email protected]/
Thanks,
Olek