12:22 <Nemo_bis> montjoie: thank for trying :)

12:43 <montjoie> I just fixed it

12:43 <montjoie> but I hit a strange behavour I cannot rproduce now

12:43 <montjoie> dma finish copy all data, but never call callback

12:44 <montjoie> now time to bench it

22:03 <macromorgan> is there a reliable way to trigger a jump to FEL mode via software? Specifically, I'm thinking from U-Boot (like jumping to a memory address or something).

22:03 <macromorgan> tried something to the effect of ((void(*)(void))0xffff0020)(); but it didn't quite get me there...

22:04 <macromorgan> I'm trying to allow users to reboot into FEL from Linux. My thought is to place a value in nvram and if that specific value is present when U-Boot (SPL stage or U-Boot proper) loads it clears the value then jumps to FEL somehow.

22:07 <apritzel> macromorgan: that depends on the SoC you are aiming at

22:08 <apritzel> we had this idea for a while: a "felreset" command in U-Boot

22:08 <apritzel> but implementing this is tricky, especially for 64-bit SoCs

22:09 <macromorgan> the R8

22:09 <apritzel> for 32-bit SoCs it sounds possible from U-Boot, by carefully resetting all peripherals, bringing the core in a pristine state, then jumping to the FEL entry point

22:10 <macromorgan> would still be fine for me to have it in the "post SPL" phase, kind of like after you upload the SPL but before you do anything else

22:11 <apritzel> I once implement a GPIO FEL button feature in the SPL

22:11 <apritzel> it would configure and check a GPIO very early in the SPL, then jump to the FEL entry point in BROM

22:12 <macromorgan> how did you jump to the entry point?

22:12 <apritzel> so this could be used by checking some NVRAM instead (but I don't know the details of the A13 here)

22:12 <macromorgan> I tried it (just after the PMIC loaded so I could use the PMIC routines to check NVRAM) but it didn't work

22:13 <apritzel> mov r1, =0xffff0020; br r1

22:13 <macromorgan> okay, I'll try that

22:13 <apritzel> but you have to do this as early as possible

22:14 <macromorgan> hopefully I'm doing it early enough (before the DRAM is ready even...)

22:14 <apritzel> as the FEL code makes some assumption about being run in a certain state (system registers, peripherals, clocks, ...)

22:15 <apritzel> macromorgan: https://gist.github.com/apritzel/2d7668db6470a6d033b791a08029943a

22:25 <macromorgan> okay, I think that gets me started at least

22:26 <macromorgan> I'll need to do the BARE MINIMUM hardware init to read the value over i2c, clear the value, de-init the hardware, then jump to the FEL address

22:26 <macromorgan> does that sound about right?

22:28 <macromorgan> too bad I can't just write a magic value in SRAM and read it at boot up

22:34 <apritzel> yeah, that is one way, I guess (at least that part is easier with later SoCs integrating an RTC and NVRAM)

22:36 <apritzel> macromorgan: you could also try to avoid resetting the core, and manually set the peripherals, clocks and registers into a reset state

22:36 <apritzel> but that might be more involved than a simple series of MMIO accesses to the I2C controller

22:37 <apritzel> as you don't want to be very quick or generic, you might be able to cut some corners here

22:38 <apritzel> is it just one single register you need to read via I2C?

22:39 <macromorgan> yep

22:39 <macromorgan> I just want to read a persistent value from Linux; I'm thinking the fastest way is to use the nvram on the PMIC

22:42 <apritzel> yeah, it's a shame that this simple thing (actually just one bit of information) is so involved

22:48 <apritzel> macromorgan: why do you want FEL reset from *Linux*?

22:49 <macromorgan> as a courtesy, just seems like a nice thing if possible

22:49 <macromorgan> might decide against it and instead do a fastboot from Linux (which is way easier since U-Boot initializes most everything by then)

22:55 <apritzel> macromorgan: yeah, if you don't need any particular FEL killer feature (like loading low level firmware), then fastboot or a UMS export sound more sane

22:57 <macromorgan> well for the R8 (the CHIP) Fastboot doesn't support the NAND sadly :-(

22:57 <macromorgan> it supports NAND, just not this NAND the way I'm trying to run it.

22:58 <macromorgan> either way though, no biggie, this sounds like it's not worth the effort when a paperclip is handy...