On Thu, Sep 28, 2023 at 8:07 AM Jim Quinlan [off-list ref] wrote:

On Wed, Sep 27, 2023 at 7:10 PM Linus Walleij [off-list ref] wrote:

quoted

Hi Jim,

thanks for your patch!

On Tue, Sep 26, 2023 at 7:52 PM Jim Quinlan [off-list ref] wrote:

quoted

Without this commit, the use of dma_alloc_coherent() while
using CONFIG_DMA_RESTRICTED_POOL=y breaks devices from working.
For example, the common Wifi 7260 chip (iwlwifi) works fine
on arm64 with restricted memory but not on arm, unless this
commit is applied.

Signed-off-by: Jim Quinlan <redacted>

(...)

quoted

+       select DMA_DIRECT_REMAP

Christoph invented that symbol so he can certainly
explain what is missing to use this on ARM.

This looks weird to me, because:

quoted

git grep atomic_pool_init

arch/arm/mm/dma-mapping.c:static int __init atomic_pool_init(void)
kernel/dma/pool.c:static int __init dma_atomic_pool_init(void)

Now you have two atomic DMA pools in the kernel,
and a lot more than that is duplicated. I'm amazed that it
compiles at all.

Ah, I did not communicate this well at all.  The patch compiles on our
ARM brcmstb_defconfig
for 5.15, 6.1, and upstream.  The kernel test-bot tells me it doesn't
compile on whatever
config it is using (looks like a missing header file).

My patch does not work on upstream; I only supplied it to show what
"fixes" 6.1 and 5.15.
For upstream on ARM, restricted-memory does not work w/ or w/o the patch.
For upstream on ARM64, restricted memory does not seem to be working either.

Regards,
Jim Quinlan
Broadcom STB/CM

quoted

Clearly if you want to do this, surely the ARM-specific
arch/arm/mm/dma-mapping.c and arch/arm/mm/dma-mapping-nommu.c
needs to be removed at the same time?

However I don't think it's that simple, because Christoph would surely
had done this a long time ago if it was that simple.

Hello Linus,

Yes, this is the reason I used "RFC" as the fix looked too easy to be viable :-)
I debugged it enough to see that the host driver's
writes to the dma_alloc_coherent() region  were not appearing in
memory, and that
led me to DMA_DIRECT_REMAP.

BTW, I tested "restricted-dma" on the master-tip the other day and it
failed for both arm64 and arm.
Please take this with a large grain of salt  as this was a quick test
and  I won't  have time to
confirm and bisect until  next week at the earliest.

Regards,
Jim Quinlan
Broadcom STB/CM

quoted

Yours,
Linus Walleij