On Tue, Mar 26, 2024 at 12:47:34PM +0100, Stefan Wahren wrote:
> [add Dave since he's working on DMA for Raspberry Pi 4 and maybe have a
> opinion about this]
>
> [drop Emma Anholt old address since she is not involved anymore]
>
> Am 26.03.24 um 08:06 schrieb Krzysztof Kozlowski:
> > On 26/03/2024 01:49, Laurent Pinchart wrote:
> >> The raspberrypi,bcm2835-firmware devices requires a dma-ranges property,
> >> and, as a result, also needs to specify #address-cells and #size-cells.
> >> Those properties have been added to thebcm2835-rpi.dtsi in commits
> >> be08d278eb09 ("ARM: dts: bcm283x: Add cells encoding format to firmware
> >> bus") and 55c7c0621078 ("ARM: dts: bcm283x: Fix vc4's firmware bus DMA
> >> limitations"), but the DT bindings haven't been updated, resulting in
> >> validation errors:
> >>
> >> arch/arm64/boot/dts/broadcom/bcm2711-rpi-4-b.dtb: firmware: '#address-cells', '#size-cells', 'dma-ranges', 'gpio' do not match any of the regexes: 'pinctrl-[0-9]+'
> >> from schema $id: http://devicetree.org/schemas/arm/bcm/raspberrypi,bcm2835-firmware.yaml#
> >>
> >> Fix this by adding the properties to the bindings.
> >>
> >> Signed-off-by: Laurent Pinchart <laurent.pinchart@xxxxxxxxxxxxxxxx>
> >
> > Children do not perform any IO on their own, because everything is
> > handled by parent. It is really odd to see dma-ranges without ranges.
> > Referenced commits might be also wrong.
Comunication with the firmware goes through a mailbox interface, which
uses DMA transfers. See for instance
rpi_firmware_transaction(struct rpi_firmware *fw, u32 chan, u32 data)
{
u32 message = MBOX_MSG(chan, data);
int ret;
WARN_ON(data & 0xf);
mutex_lock(&transaction_lock);
reinit_completion(&fw->c);
ret = mbox_send_message(fw->chan, &message);
if (ret >= 0) {
if (wait_for_completion_timeout(&fw->c, HZ)) {
ret = 0;
} else {
ret = -ETIMEDOUT;
WARN_ONCE(1, "Firmware transaction timeout");
}
} else {
dev_err(fw->cl.dev, "mbox_send_message returned %d\n", ret);
}
mutex_unlock(&transaction_lock);
return ret;
}
int rpi_firmware_property_list(struct rpi_firmware *fw,
void *data, size_t tag_size)
{
size_t size = tag_size + 12;
u32 *buf;
dma_addr_t bus_addr;
int ret;
/* Packets are processed a dword at a time. */
if (size & 3)
return -EINVAL;
buf = dma_alloc_coherent(fw->cl.dev, PAGE_ALIGN(size), &bus_addr,
GFP_ATOMIC);
if (!buf)
return -ENOMEM;
/* The firmware will error out without parsing in this case. */
WARN_ON(size >= 1024 * 1024);
buf[0] = size;
buf[1] = RPI_FIRMWARE_STATUS_REQUEST;
memcpy(&buf[2], data, tag_size);
buf[size / 4 - 1] = RPI_FIRMWARE_PROPERTY_END;
wmb();
ret = rpi_firmware_transaction(fw, MBOX_CHAN_PROPERTY, bus_addr);
rmb();
memcpy(data, &buf[2], tag_size);
if (ret == 0 && buf[1] != RPI_FIRMWARE_STATUS_SUCCESS) {
/*
* The tag name here might not be the one causing the
* error, if there were multiple tags in the request.
* But single-tag is the most common, so go with it.
*/
dev_err(fw->cl.dev, "Request 0x%08x returned status 0x%08x\n",
buf[2], buf[1]);
ret = -EINVAL;
}
dma_free_coherent(fw->cl.dev, PAGE_ALIGN(size), buf, bus_addr);
return ret;
}
fw->cl.dev is the device for the firmware child node. That may be where
the problem comes from, shouldn't we use the mailbox device for DMA
mapping ?
--
Regards,
Laurent Pinchart
