RPi4B's DMA11 (DMA4 engine example) vs. xHCI/pcie

[Mark Millard]

On 2020-Sep-30, at 15:40, Mark Millard <marklmi at yahoo.com> wrote:

> On 2020-Sep-30, at 14:15, Robert Crowston <crowston at protonmail.com> wrote:
> 
>> Very interesting analysis. Certainly uncovered a few things I wasn't aware of.
>> 
>> By default sc->sc_bus.dma_bits in xhci_init is 64 bits; I toggle it back to 32 bits in the xhci shim I wrote for the Pi 4. You can see that output in a verbose dmesg.
>> 
> 
> My biggest worry from all that material, for things as they currently
> are, is that it appears that FreeBSD could try to use a RPi4B DMA LITE
> engine but not follow its limitations when doing so. For example,
> ending up using a smaller size DMA transfer than intended (just 16 bits
> for size). (Other RPi's might have the same issue?)

So I looked at BCM2835-ARM-Peripherals.pdf, comparing to
rpi_DATA_2711_1p0.pdf and:

BCM2835-ARM-Peripherals.pdf indicated DMA engines 7-14 are all
DMA LITE engines for the context it was covering. DMA 0-6 are
normal DMA engines.

So the RPi4B reduced the number of DMA LITE engines (now only 7-10)
and filled the removed slots (11-14) with the new type DMA4 engines.

So I learn from this some about how to interpret "brcm,bcm2835-dma"
as shown by what the RPi4B is doing:

                dma at 7e007000 {
                        compatible = "brcm,bcm2835-dma";
                        reg = <0x7e007000 0x00000b00>;
                        interrupts = * 0x0000000007ef645c [0x00000084];
                        interrupt-names = "dma0", "dma1", "dma2", "dma3", "dma4", "dma5", "dma6", "dma7", "dma8", "dma9", "dma10";
                        #dma-cells = <0x00000001>;
                        brcm,dma-channel-mask = <0x000001f5>;
                        phandle = <0x0000000b>;
                };

The brcm,dma-channel-mask value and name list is different than in prior
examples of "brcm,bcm2835-dma" but the subset that is present follows the
original protocol. For example, how to tell DMA (normal) from DMA LITE
works the same [a debug register bit to check] --but only within the subset
present.

Seeing "brcm,bcm2835-dma" should not lead to presuming some fixed number of
interrupts or dma-channels to process.

The protocol for DMA (normal) vs. DMA LITE does not even apply to
the new RPi4B DMA4 engines (11-14):

                dma at 7e007b00 {
                        compatible = "brcm,bcm2711-dma";
                        reg = <0x00000000 0x7e007b00 0x00000000 0x00000400>;
                        interrupts = <0x00000000 0x00000059 0x00000004 0x00000000 0x0000005a 0x00000004 0x00000000 0x0000005b 0x00000004 0x00000000 0x0000005c 0x00000004>;
                        interrupt-names = "dma11", "dma12", "dma13", "dma14";
                        #dma-cells = <0x00000001>;
                        brcm,dma-channel-mask = <0x00007000>;
                        phandle = <0x0000003d>;
                };

11-14 have a new, distinct protocol in various respects.


===
Mark Millard
marklmi at yahoo.com
( dsl-only.net went
away in early 2018-Mar)