en-sc has quit [Remote host closed the connection]
Hawk777 has joined #openocd
PlasmaHH has joined #openocd
Haohmaru has joined #openocd
<PlasmaHH>
Hi, trying to figure out the features of openocd regarding stm32 quadspi support... the docs for the stmqspi command say that the chip and registers have to be properly initialized for memory map access. does that mean I already need to have some code running there that does it? also does it only support reading or is there also some kind of write support?
lxsameer has joined #openocd
<Haohmaru>
PlasmaHH, well, the quadspi peripheral isn't pre-configured after reset, right?
<Haohmaru>
at least on H723 there was this peripheral that i think supported quadspi and also hyperbus, which basically has a lot of different modes of operation and different adjustable things, so it is probably not too likely for it to come with "at reset" settings that exactly match your setup, i'd rather think the peripheral is disabled on Reset
<Haohmaru>
there would be clocks and other stuff involved in it
<PlasmaHH>
Haohmaru: I am still trying to figure out how things work exactly... with segger tools for example you specify the pinout in the flash/gdbserver config... and they seem to support writing too...
<Haohmaru>
$$$segger is expensive and thus probably more fancy
<gamiee>
Yeah, In segger's price is also the luxury of more polished drivers.
<Haohmaru>
i'm ghetto, i so i use daplink ;P~
<gamiee>
PlasmaHH: You have two options, 1) run quadspi related functions only after your app initializes it, or you need to configure all required peripherals in TCL script by yourself.
<Haohmaru>
x_x
<gamiee>
"after your app initializes it" it -> Quad SPI Flash
<PlasmaHH>
Haohmaru: yeah just mentioning it because in principlpe it seems possible, but they are known to do funny hacks so its not clear how all this works... and since our pinout isnt supported by segger anyways I want to find another way
<Haohmaru>
well, i'd guess one way is to write a simple firmware that just initializes the needed stuff for the quadspi
<Haohmaru>
then flash it, run it, pause it and from there it should be "operational"
<PlasmaHH>
until the next resset... I think if its possible to do it in the openocd config that would be better..
<Haohmaru>
the alternative way is like mr gamiee said, use openocd commands for reading/writing random data into the chip, and carefully write the whole initialization procedure that way
<Haohmaru>
that would involve a lot of reading and writing funny values into funny addresses
<PlasmaHH>
probably not that muchm just a dozen registers Iguess
<Haohmaru>
H7xx?
<PlasmaHH>
yep
<Haohmaru>
well, in general my impression is that the more complex chips kinda have more adjustments needed to set them up
<Haohmaru>
you'll probably need to adjust clocks, pins, dma?
en-sc has joined #openocd
<Haohmaru>
in any case, it should be doable
<Haohmaru>
unless there's some tricky aspects that i'm not aware of
<Hawk777>
Don’t think you need DMA.
<Hawk777>
Just RCC, GPIO, and a couple of the QUADSPI regs.
<PlasmaHH>
Hawk777: I have no idea how the writing there is setup, might need mdma
<Hawk777>
It doesn’t.
<Hawk777>
Set the QUADSPI and GPIOs enabled in the RCC, set the GPIO AFRs, MODEs, and SPEEDs, and set the QUADSPI CR, DCR, and CCR.
<Hawk777>
I do that in the “reset-init” handler.
<Hawk777>
That’s all that’s needed for the way OpenOCD uses it.
<Haohmaru>
if it's "memory mapped" to some area of the address space, yes i'd think you wouldn't have to set up DMA as it should just work magically under the hood
<PlasmaHH>
Haohmaru: memory mapped is read only though
<Hawk777>
No, OpenOCD just doesn’t use DMA, even when writing.
<Haohmaru>
shouldn't matter whether you only wanna read from it
<PlasmaHH>
Haohmaru: well if I want to write to it and memory mapped doesnt support write then it kind of matters what I want ;)
<PaulFertser>
PlasmaHH: the Tcl config file can init the peripheral properly.
<Haohmaru>
why wouldn't the quadspi support writing to the memory? ;P~
<Hawk777>
It doesn’t support writing *through the memory mapped region*.
<PlasmaHH>
Haohmaru: it does, just not the memory map access
<Hawk777>
It supports writing in general.
<Hawk777>
But to use OpenOCD commands to read and write, you don’t need DMA and you don’t need memory mapped mode.
<PlasmaHH>
which kind of makes sense if you recall that writing means erasing whole pages
<PlasmaHH>
ah it seems there is quite some configs available in the board support configs I can adapt for our layout
<Haohmaru>
okay, i'm not too familiar, i know i supposedly was gonna use a similar peripheral on H723 to have a hyperbus PSRAM which i was gonna use as circular buffer, and was told that it'll work, that involves also writing, but i don't know how the code would look like for it and whether it'll be memory mapped
<PaulFertser>
PlasmaHH: see e.g. board/stm32h750b-disco.cfg reset-init handler
<Haohmaru>
PlasmaHH, yeah but the memory chip could be RAM, afaik
<PaulFertser>
The qspi code itself is in board/stm32h7x_dual_qspi.cfg
<Haohmaru>
but anyway, listen to what mr PaulFertser says ^
<PlasmaHH>
PaulFertser: yeah found taht qual qspi one, I think I have enough to start that stuff tomorrow then
<PaulFertser>
It makes sense it's in board config because different setup is needed for different flash chips and the way they're connected on a specific board.
Hawk777 has quit [Quit: Leaving.]
<PlasmaHH>
I wish it was easier to get the company to donate a fraction of the money I save them to the projects involved...
* karlp
grins
<karlp>
I got work to pay for shipping some beer once, otherhtan than, normally the best I've managed is being allowed to make contributions back upstream
<PlasmaHH>
heresy!
<Haohmaru>
beer o_O
<PlasmaHH>
here anything that is not paying an invoice is hard work
<Haohmaru>
pay me in beercoinz
<Haohmaru>
actually, i know people who would work for alcohol
Foxyloxy has quit [Read error: Connection reset by peer]
en-sc has quit [Remote host closed the connection]
en-sc has joined #openocd
en-sc has quit [Remote host closed the connection]
en-sc has joined #openocd
Pokey has quit [Quit: No Ping reply in 180 seconds.]