[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <583aeb80-4ac8-06ac-67f7-d717415523dd@infradead.org>
Date: Thu, 6 May 2021 18:30:39 -0700
From: Randy Dunlap <rdunlap@...radead.org>
To: Shannon Nelson <snelson@...sando.io>,
Jakub Kicinski <kuba@...nel.org>
Cc: netdev@...r.kernel.org, davem@...emloft.net, drivers@...sando.io,
Allen Hubbe <allenbh@...sando.io>
Subject: Re: [PATCH v3 net] ionic: fix ptp support config breakage
On 5/6/21 6:16 PM, Randy Dunlap wrote:
> On 5/6/21 6:12 PM, Randy Dunlap wrote:
>> On 5/6/21 5:48 PM, Shannon Nelson wrote:
>>> On 5/6/21 5:21 PM, Randy Dunlap wrote:
>>>> On 5/6/21 5:15 PM, Jakub Kicinski wrote:
>>>>> On Wed, 5 May 2021 21:18:46 -0700 Shannon Nelson wrote:
>>>>>> Driver link failed with undefined references in some
>>>>>> kernel config variations.
>>>>> This is really vague and the patch is not very obvious.
When IONIC=y and PTP_1588_CLOCK=m...
>>>>>> ionic-y := ionic_main.o ionic_bus_pci.o ionic_devlink.o ionic_dev.o \
>>>>>> ionic_debugfs.o ionic_lif.o ionic_rx_filter.o ionic_ethtool.o \
>>>>>> - ionic_txrx.o ionic_stats.o ionic_fw.o
>>>>>> -ionic-$(CONFIG_PTP_1588_CLOCK) += ionic_phc.o
>>>>>> + ionic_txrx.o ionic_stats.o ionic_fw.o ionic_phc.o
>>>>> So we'd replace a build dependency..
>>>>>
>>>>>> diff --git a/drivers/net/ethernet/pensando/ionic/ionic_phc.c b/drivers/net/ethernet/pensando/ionic/ionic_phc.c
>>>>>> index a87c87e86aef..30c78808c45a 100644
>>>>>> --- a/drivers/net/ethernet/pensando/ionic/ionic_phc.c
>>>>>> +++ b/drivers/net/ethernet/pensando/ionic/ionic_phc.c
>>>>>> @@ -1,6 +1,8 @@
>>>>>> // SPDX-License-Identifier: GPL-2.0
>>>>>> /* Copyright(c) 2017 - 2021 Pensando Systems, Inc */
>>>>>> +#if IS_ENABLED(CONFIG_PTP_1588_CLOCK)
>>>>>> +
>>>>>> #include <linux/netdevice.h>
>>>>>> #include <linux/etherdevice.h>
>>>>>> @@ -613,3 +615,4 @@ void ionic_lif_free_phc(struct ionic_lif *lif)
>>>>>> devm_kfree(lif->ionic->dev, lif->phc);
>>>>>> lif->phc = NULL;
>>>>>> }
>>>>>> +#endif /* IS_ENABLED(CONFIG_PTP_1588_CLOCK) */
>>>>> .. with an ifdef around an entire file? Does not feel very clean.
>>>>>
>>>>> The construct of using:
>>>>>
>>>>> drv-$(CONFIG_PTP_1588_CLOCK) += ptp.o
>>>>>
>>>>> seems relatively common, why does it now work here?
>>>>>
>>>>> Maybe the config in question has PTP as a module and ionic built in?
>>>>> Then you should add depends on PTP_1588_CLOCK || !PTP_1588_CLOCK.
>>>>>
>>>>> Maybe somehow the "ionic-y" confuses kbuild and it should be ionic-objs?
>>>>>
>>>>> At the very least we need a better explanation in the commit message.
>>>>>
>>>> I'll take a look if someone can point me to the .config file.
>>>>
>>>
>>> These are the notes I got from kernel test robot:
>>> https://lore.kernel.org/lkml/202105041020.efEaBOYC-lkp@intel.com/
>>> https://lore.kernel.org/lkml/202105041154.GrLZmjGh-lkp@intel.com/
>>> https://lore.kernel.org/lkml/202105041634.paURyDp0-lkp@intel.com/
>>> https://lore.kernel.org/lkml/202105050636.UXXDl7m2-lkp@intel.com/
>>
>> At first glance it looks like Jakub's suggestion of
>>>>> Maybe the config in question has PTP as a module and ionic built in?
>>>>> Then you should add depends on PTP_1588_CLOCK || !PTP_1588_CLOCK.
>>
>> is what is needed, but I'm still doing build testing ATM.
>
> Nope, eat my words. These build issues are not about PTP.
> I'm still looking.
I have been trying to go to fast.. slow down, wait for the old computer.
Back to Jakub's suggestion -- that works for me. (copy-paste, whitespace damaged)
--- linux-next-20210506.orig/drivers/net/ethernet/pensando/Kconfig
+++ linux-next-20210506/drivers/net/ethernet/pensando/Kconfig
@@ -20,6 +20,7 @@ if NET_VENDOR_PENSANDO
config IONIC
tristate "Pensando Ethernet IONIC Support"
depends on 64BIT && PCI
+ depends on PTP_1588_CLOCK || !PTP_1588_CLOCK
select NET_DEVLINK
select DIMLIB
help
If PTP_1588_CLOCK=m, the depends limits IONIC to =m (or disabled).
If PTP_1588_CLOCK is disabled, IONIC can be any of y/m/n.
--
~Randy
Powered by blists - more mailing lists