From: Robin Murphy <robin.murphy@arm.com>
To: Dmitry Osipenko <digetx@gmail.com>,
Yuehaibing <yuehaibing@huawei.com>,
Sakari Ailus <sakari.ailus@linux.intel.com>
Cc: devel@driverdev.osuosl.org, linux-media@vger.kernel.org,
gregkh@linuxfoundation.org, linux-kernel@vger.kernel.org,
iommu@lists.linux-foundation.org, hverkuil-cisco@xs4all.nl,
mchehab@kernel.org, yong.zhi@intel.com
Subject: Re: [PATCH] media: staging: ipu3: Enable IOVA API only when IOMMU support is enabled
Date: Wed, 24 Jul 2019 15:23:55 +0100 [thread overview]
Message-ID: <eaf521ff-7dc6-70ae-0473-9c994def602b@arm.com> (raw)
In-Reply-To: <491dbca1-8a58-b26e-cf56-a1a419da288f@gmail.com>
On 24/07/2019 15:09, Dmitry Osipenko wrote:
> 24.07.2019 17:03, Yuehaibing пишет:
>> On 2019/7/24 21:49, Robin Murphy wrote:
>>> On 24/07/2019 11:30, Sakari Ailus wrote:
>>>> Hi Yue,
>>>>
>>>> On Mon, Jul 22, 2019 at 09:47:49PM +0800, YueHaibing wrote:
>>>>> If IOMMU_SUPPORT is not set, ipu3 driver may select IOMMU_IOVA to m.
>>>>> But for many drivers, they use "select IOMMU_IOVA if IOMMU_SUPPORT"
>>>>> in the Kconfig, for example, CONFIG_TEGRA_VDE is set to y but IOMMU_IOVA
>>>>> is m, then the building fails like this:
>>>>>
>>>>> drivers/staging/media/tegra-vde/iommu.o: In function `tegra_vde_iommu_map':
>>>>> iommu.c:(.text+0x41): undefined reference to `alloc_iova'
>>>>> iommu.c:(.text+0x56): undefined reference to `__free_iova'
>>>>>
>>>>> Reported-by: Hulk Robot <hulkci@huawei.com>
>>>>> Fixes: 7fc7af649ca7 ("media: staging/intel-ipu3: Add imgu top level pci device driver")
>>>>> Signed-off-by: YueHaibing <yuehaibing@huawei.com>
>>>>> ---
>>>>> drivers/staging/media/ipu3/Kconfig | 2 +-
>>>>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>>>>
>>>>> diff --git a/drivers/staging/media/ipu3/Kconfig b/drivers/staging/media/ipu3/Kconfig
>>>>> index 4b51c67..b7df18f 100644
>>>>> --- a/drivers/staging/media/ipu3/Kconfig
>>>>> +++ b/drivers/staging/media/ipu3/Kconfig
>>>>> @@ -4,7 +4,7 @@ config VIDEO_IPU3_IMGU
>>>>> depends on PCI && VIDEO_V4L2
>>>>> depends on MEDIA_CONTROLLER && VIDEO_V4L2_SUBDEV_API
>>>>> depends on X86
>>>>> - select IOMMU_IOVA
>>>>> + select IOMMU_IOVA if IOMMU_SUPPORT
>>>>
>>>> This doesn't seem right: the ipu3-cio2 driver needs IOMMU_IOVA
>>>> independently of IOMMU_SUPPORT.
>>>>
>>>> Looking at tegra-vde, it seems to depend on IOMMU_SUPPORT but that's not
>>>> declared in its Kconfig entry. I wonder if adding that would be the right
>>>> way to fix this.
>>>>
>>>> Cc'ing the IOMMU list.
> IOMMU_SUPPORT is optional for the Tegra-VDE driver.
>
>>> Right, I also had the impression that we'd made the IOVA library completely standalone. And what does the IPU3 driver's Kconfig have to do with some *other* driver failing to link anyway?
>
> I can see it failing if IPU3 is compiled as a loadable module, while
> Tegra-VDE is a built-in driver. Hence IOVA lib should be also a kernel
> module and thus the IOVA symbols will be missing during of linkage of
> the VDE driver.
>
>> Oh, I misunderstand that IOMMU_IOVA is depend on IOMMU_SUPPORT, thank you for clarification.
>>
>> I will try to fix this in tegra-vde.
>
> Probably IOVA could be selected independently of IOMMU_SUPPORT, but IOVA
> library isn't needed for the VDE driver if IOMMU_SUPPORT is disabled.
Oh, I think I get the problem now - tegra-vde/iommu.c is built
unconditionally and relies on the static inline stubs for IOMMU and IOVA
calls if !IOMMU_SUPPORT, but in a compile-test config where IOVA=m for
other reasons, it then picks up the real declarations from linux/iova.h
instead of the stubs, and things go downhill from there. So there is a
real issue, but indeed it's Tegra-VDE which needs to be restructured to
cope with such configurations, and not IPU3's (or anyone else who may
select IOVA=m in future) job to work around it.
Robin.
next prev parent reply other threads:[~2019-07-24 14:24 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-07-22 13:47 [PATCH] media: staging: ipu3: Enable IOVA API only when IOMMU support is enabled YueHaibing
2019-07-24 10:30 ` Sakari Ailus
2019-07-24 13:49 ` Robin Murphy
2019-07-24 14:03 ` Yuehaibing
2019-07-24 14:09 ` Dmitry Osipenko
2019-07-24 14:23 ` Robin Murphy [this message]
2019-07-24 14:28 ` Dmitry Osipenko
2019-07-25 2:41 ` [PATCH] media: staging: tegra-vde: Fix build error YueHaibing
2019-07-25 7:19 ` Dmitry Osipenko
2019-09-20 19:32 ` Arnd Bergmann
2019-09-20 22:26 ` Dmitry Osipenko
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=eaf521ff-7dc6-70ae-0473-9c994def602b@arm.com \
--to=robin.murphy@arm.com \
--cc=devel@driverdev.osuosl.org \
--cc=digetx@gmail.com \
--cc=gregkh@linuxfoundation.org \
--cc=hverkuil-cisco@xs4all.nl \
--cc=iommu@lists.linux-foundation.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-media@vger.kernel.org \
--cc=mchehab@kernel.org \
--cc=sakari.ailus@linux.intel.com \
--cc=yong.zhi@intel.com \
--cc=yuehaibing@huawei.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).