[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <ea5433f6-ef8d-3cd0-0645-dd89c4806dca@xenosoft.de>
Date: Thu, 19 Dec 2019 14:54:27 +0100
From: Christian Zigotzky <chzigotzky@...osoft.de>
To: Michael Ellerman <mpe@...erman.id.au>,
Christoph Hellwig <hch@....de>,
Benjamin Herrenschmidt <benh@...nel.crashing.org>,
Paul Mackerras <paulus@...ba.org>
Cc: linux-arch@...r.kernel.org, linux-mm@...ck.org,
iommu@...ts.linux-foundation.org, linuxppc-dev@...ts.ozlabs.org,
linux-kernel@...r.kernel.org, "R.T.Dickinson" <rtd2@...a.co.nz>,
"contact@...on.com" <contact@...on.com>,
mad skateman <madskateman@...il.com>,
Darren Stevens <darren@...vens-zone.net>
Subject: Re: use generic DMA mapping code in powerpc V4
Hi All,
We still have some issues with PCI cards in our FSL P5020 and P5040
systems since the DMA mapping updates. [1, 2]
We have to limit the RAM to 3500MB for some problematic PCI cards.
(kernel boot argument 'mem=3500M')
The problematic DMA mapping code was added with the PowerPC updates
4.21-1 to the official kernel source code last year. [3]
We have created a bug report. [4]
The old 4.x kernels aren't affected because they use the old DMA code.
Please check the new DMA code again.
Thanks,
Christian
[1]
http://forum.hyperion-entertainment.com/viewtopic.php?f=58&p=49486#p49486
[2]
http://forum.hyperion-entertainment.com/viewtopic.php?f=58&t=4349&start=50#p49099
[3]
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=8d6973327ee84c2f40dd9efd8928d4a1186c96e2
[4] https://bugzilla.kernel.org/show_bug.cgi?id=205201
On 28 November 2018 at 4:55 pm, Christian Zigotzky wrote:
> On 28 November 2018 at 12:05PM, Michael Ellerman wrote:
>> Nothing specific yet.
>>
>> I'm a bit worried it might break one of the many old obscure platforms
>> we have that aren't well tested.
>>
> Please don't apply the new DMA mapping code if you don't be sure if it
> works on all supported PowerPC machines. Is the new DMA mapping code
> really necessary? It's not really nice, to rewrote code if the old
> code works perfect. We must not forget, that we work for the end
> users. Does the end user have advantages with this new code? Is it
> faster? The old code works without any problems. I am also worried
> about this code. How can I test this new DMA mapping code?
>
> Thanks
>
>
Powered by blists - more mailing lists