Re: Build errors in v3.18-stable-queue

[Guenter Roeck <linux@roeck-us.net>]

On 02/29/2016 05:14 PM, Gavin Shan wrote:
> On Mon, Feb 29, 2016 at 07:24:33PM -0500, Sasha Levin wrote:
>> On 02/28/2016 05:45 AM, Guenter Roeck wrote:
>>> Current build results:
>>>      total: 137 pass: 121 fail: 16
>>> Failed builds:
>>>      arm:allmodconfig
>>>      arm:omap2plus_defconfig
>>>      arm:exynos_defconfig
>>>      arm:s5pv210_defconfig
>>>      powerpc:defconfig
>>>      powerpc:allmodconfig
>>>      powerpc:allnoconfig
>>>      powerpc:ppc64e_defconfig
>>>      powerpc:cell_defconfig
>>>      powerpc:maple_defconfig
>>>      powerpc:ppc6xx_defconfig
>>>      powerpc:mpc83xx_defconfig
>>>      powerpc:tqm8xx_defconfig
>>>      powerpc:85xx/sbc8548_defconfig
>>>      powerpc:83xx/mpc834x_mds_defconfig
>>>      powerpc:86xx/sbc8641d_defconfig
>>>
>>> Culprits are listed below. Copying the patch authors for advice.
>>>
>
> Thanks for reporting it, Guenter. some questions as below:
>
> - The subject says you were building 3.18 stable kernel, but the PowerPC
>    related commit was merged to 4.1 kernel. So I assume you were building
>    4.1 stable kernel instead.

In this case it was 3.18. There were failures in 4.1 as well, but they are
different. All builds in 3.18 fail, but only two out of 12 builds in 4.1.
For 3.18, this includes both 32 bit and 64 bit builds.

> - All above PowerPC configuration failed to build? At least, those
>    configuration that has (CONFIG_PPC64=y) shouldn't fail because of
>    the error you pointed.

Possibly, but they still fail. Sorry if my report missed the 64 bit error(s).

> - Those configuration without (CONFIG_PPC64=y) needs below fix. Could you
>    please have a try? If it works, I'll post a patch as soon as possible.
>
> diff --git a/arch/powerpc/kernel/of_platform.c b/arch/powerpc/kernel/of_platform.c
> index b60a67d..25dec0b 100644
> --- a/arch/powerpc/kernel/of_platform.c
> +++ b/arch/powerpc/kernel/of_platform.c
> @@ -71,8 +71,10 @@ static int of_pci_phb_probe(struct platform_device *dev)
>          eeh_dev_phb_init_dynamic(phb);
>
>          /* Register devices with EEH */
> +#ifdef CONFIG_EEH
>          if (dev->dev.of_node->child)
>                  eeh_add_device_tree_early(PCI_DN(dev->dev.of_node));
> +#endif
>

Looking into a ppc 64 bit log (from a qemu test attempt), I see

In file included from arch/powerpc/mm/init_64.c:59:0:
./arch/powerpc/include/asm/eeh.h:344:48: error: 'struct pci_dn' declared inside parameter list [-Werror]
  static inline void eeh_add_device_early(struct pci_dn *pdn) { }

which won't be fixed by the above patch.

The build logs are public at http://kerneltests.org/builders.
They are typically available for at least 30 days, so feel free
to have a look yourself.

Thanks,
Guenter

> Thanks,
> Gavin
>
>>> Guenter
>>>
>>> ---
>>> powerpc:
>>>
>>> In file included from arch/powerpc/kernel/of_platform.c:30:0:
>>> ./arch/powerpc/include/asm/eeh.h:344:48: error: 'struct pci_dn' declared inside parameter list
>>>
>>> Culprit: "powerpc/eeh: Do probe on pci_dn"
>>>
>>> ---
>>> arm:
>>>
>>> drivers/pinctrl/samsung/pinctrl-samsung.c: In function 'samsung_gpio_set':
>>> drivers/pinctrl/samsung/pinctrl-samsung.c:551:9: error: implicit declaration of function 'gpiochip_get_data'
>>>
>>> Culprit: "pinctrl: samsung: fix SMP race condition"
>>>
>>> drivers/tty/serial/omap-serial.c: In function 'serial_omap_stop_tx':
>>> drivers/tty/serial/omap-serial.c:318:10: error: 'struct uart_port' has no member named 'rs485'
>>>
>>> drivers/tty/serial/omap-serial.c: In function 'serial_omap_probe':
>>> drivers/tty/serial/omap-serial.c:1683:10: error: 'struct uart_port' has no member named 'rs485_config'
>>>
>>> Culprit: "serial/omap: Use the rs485 functions on serial_core"
>>
>>
>> Thanks Guenter. I'll pull both of these out of 3.18 and 4.1 until we can get a proper
>> backport since it doesn't seem trivial.
>>
>>
>> Thanks,
>> Sasha
>>
>
>