Re: arm-smmu 5000000.iommu: Cannot accommodate DMA offset for IOMMU page tables

From: Florian Fainelli
Date: Sun Oct 11 2020 - 17:59:51 EST


+Abhimanyu, Ioana, Fabio

On 10/11/2020 1:36 PM, Jim Quinlan wrote:
On Sat, Oct 10, 2020 at 2:53 PM Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx> wrote:

Hi Naresh,

Just adding Christoph and Jim to cc]

On Fri, 9 Oct 2020 19:26:24 +0530 Naresh Kamboju <naresh.kamboju@xxxxxxxxxx> wrote:

On Fri, 9 Oct 2020 at 19:24, Naresh Kamboju <naresh.kamboju@xxxxxxxxxx> wrote:



On Thu, 24 Sep 2020 at 15:26, Joerg Roedel <joro@xxxxxxxxxx> wrote:

On Thu, Sep 24, 2020 at 10:36:47AM +0100, Robin Murphy wrote:
Yes, the issue was introduced by one of the changes in "dma-mapping:
introduce DMA range map, supplanting dma_pfn_offset", so it only existed in
the dma-mapping/for-next branch anyway.


FYI,
The reported problem still exists on 5.9.0-rc8-next-20201009.

[ 1.843814] Driver must set ecc.strength when using hardware ECC
[ 1.849847] WARNING: CPU: 4 PID: 1 at
drivers/mtd/nand/raw/nand_base.c:5687 nand_scan_with_ids+0x1450/0x1470
[ 1.859676] Modules linked in:
[ 1.862730] CPU: 4 PID: 1 Comm: swapper/0 Not tainted
5.9.0-rc8-next-20201009 #1
[ 1.870125] Hardware name: Freescale Layerscape 2088A RDB Board (DT)
[ 1.876478] pstate: 40000005 (nZcv daif -PAN -UAO -TCO BTYPE=--)
[ 1.882483] pc : nand_scan_with_ids+0x1450/0x1470
[ 1.887183] lr : nand_scan_with_ids+0x1450/0x1470

Hi,

I'm having a hard time coming up with a theory regarding how a commit
concerning DMA offsets can affect the operation of a NAND driver that
appears not to use DMA or the dma-ranges property. Does anyone else
have some ideas, or is there perhaps someone familiar with this test
configuration that I can correspond with to get to the bottom of the
warning?

I believe you are given only a partial warning and just a glimpse of the thread here which is why understanding the connection to the dma ranges parsing is not clear. The start of the thread can be found here:

https://lore.kernel.org/lkml/CA+G9fYvuq58q+GsWnzni0sKSHbubuQz-UaK3TASX26V_a7yBVw@xxxxxxxxxxxxxx/

Robin indicated that the IOMMU probe failure was fixed with:

https://lore.kernel.org/linux-iommu/daedc9364a19dc07487e4d07b8768b1e5934abd4.1600700881.git.robin.murphy@xxxxxxx/T/#u

which is confirmed with the new log from Naresh:

https://qa-reports.linaro.org/lkft/linux-next-master/build/next-20201009/testrun/3284876/suite/linux-log-parser/test/check-kernel-warning-92014/log

however the NAND warning still remains. Someone else familiar with these NXP development boards should fix the DTS so as to provide the require ECC strength property.
--
Florian