Hi Stefan

On Fri, 2022-01-14 at 17:24 +0100, Stefan Wahren wrote:

Hi Marcel,

Am 14.01.22 um 14:39 schrieb Marcel Ziswiler:

quoted

Hi Stefan

On Thu, 2022-01-13 at 19:42 +0100, Stefan Wahren wrote:

quoted

Hi Marcel,

Am 07.01.22 um 19:03 schrieb Marcel Ziswiler:

quoted

From: Marcel Ziswiler <redacted>

This avoids firmware load error and sysfs fallback reported as follows:

[    0.199448] imx-sdma 302c0000.dma-controller: Direct firmware load
 for imx/sdma/sdma-imx7d.bin failed with error -2
[    0.199487] imx-sdma 302c0000.dma-controller: Falling back to sysfs
 fallback for: imx/sdma/sdma-imx7d.bin

unfortunately this isn't an explanation why this solve the issue. It
looks like that the SDMA firmware is not available for the kernel. Build
the driver as a module hide the issue and just load it if the rootfs is
available.

Well, it seems to default to some built-in firmware which, unfortunately, might have some severe known
errata
anyway.

can you please explain what you mean with built-in firmware? AFAIK there
is a firmware in the SoC ROM, but there is also the option to append
firmware files to the kernel image (regardless of initramfs and so on).

Yes, sorry. I should have worded it differently. I really meant the one that comes with the SoC ROM.

And yes, it may be appended to the kernel image but that would require changes to how one goes about building
it which, at that time, one could simply also change that configuration.

However, for a generic case, like I assume this defconfig is supposed to be used, having it as a module seems
much more straight forward.

For instance the NXP vendor kernel of the i.MX6ULL (uses the same driver
with a different firmware file) appended the necessary firmware per
default.

Yes, we know all about those "great" vendor kernels ;-p.

But yes, the external firmware should be preferred.

quoted

quoted

What makes you sure that the driver is not critical for other boards
during boot?

That is a good question. I remember that there was some more discussion back when I proposed the same
change
for imx_v6_v7_defconfig in the 32-bit Arm world. While nobody was able to fully answer it all it did get
accepted by the i.MX maintainer back in the day [1]. I will mention that in v2.

Okay, in this case the change should be fine.

Thanks

Thank you.

quoted

[1] https://lore.kernel.org/lkml/20210920144938.314588-6-marcel@ziswiler.com/ (local)

Thanks for your review.

quoted

Best regards

quoted

Signed-off-by: Marcel Ziswiler <redacted>
---

 arch/arm64/configs/defconfig | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/arm64/configs/defconfig b/arch/arm64/configs/defconfig
index 3c8106c5776a..a18785a827dc 100644
--- a/arch/arm64/configs/defconfig
+++ b/arch/arm64/configs/defconfig

@@ -915,7 +915,7 @@ CONFIG_DMADEVICES=y

 CONFIG_DMA_BCM2835=y
 CONFIG_DMA_SUN6I=m
 CONFIG_FSL_EDMA=y
-CONFIG_IMX_SDMA=y
+CONFIG_IMX_SDMA=m
 CONFIG_K3_DMA=y
 CONFIG_MV_XOR=y
 CONFIG_MV_XOR_V2=y

Cheers

Marcel
_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Cheers

Marcel
_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel