13:30 <somlo> daveshah: with external 12V power, plugging both ethernet and microSD cards into the trellisboard no longer causes it to glitch

13:30 <daveshah> Great, that does seem like it was a power problem then

13:31 <somlo> yeah, and there's a deterministic solution (use ext. 12V - and a fuse!) that doesn't involve playing usb cable lottery :D

13:35 <_florent_> somlo: nice, just for info i'm looking at the LiteSDCard insert-init-eject issue

13:41 <somlo> _florent_: thanks! I'll hold off on creating a github issue then, unless it turns out to be a longer term project (hopefully it's just a quick FSM tweak though :)

13:47 <_florent_> somlo: the insert-init-eject issue is probably more related to software

13:49 <somlo> _florent_: the softer the better :D

13:59 <_florent_> somlo: i'm able to reproduce (even without ejecting the card)

14:02 <_florent_> somlo: it seems the "Switch driver strength" command is only accepted once: https://github.com/enjoy-digital/litex/blob/master/litex/soc/software/liblitesdcard/sdcard.c#L455-L457

14:03 <_florent_> it's passing on the first sdcard_init, but failing on the next ones

14:07 <_florent_> it's maybe not returning SD_OK because it's already configured correctly, i'm going to look at the spec

14:32 <somlo> _florent_: I just emailed you the sequence of commands being sent to the card by linux

14:32 <somlo> most of that is generic (i.e. not specifically done by the litex_mmc.c driver)

14:34 <somlo> so the question is whether sending "switch strength" repeatedly should be accepted/ignored rather than allow the card to get "legitimately upset" about it?

14:47 <_florent_> somlo: i just removed the driver strength configuration since that's only supported in 1.8V signaling in fact (which was previously supported)

14:47 <_florent_> https://github.com/enjoy-digital/litex/commit/1d04b1dd83cf864d7f48cb4fe34e09a37412414c

14:49 <_florent_> somlo: with this, consecutive sdcard_inits are working fine (with the BIOS)

14:53 <_florent_> somlo: but there is probably still an issue (this time i think in the gateware) since i also reproduce the miss-alignment of data on the first reads

15:06 <keesj> what is the bench in litesdcard https://github.com/enjoy-digital/litesdcard/tree/master/bench ?

15:06 <keesj> I was serching for a python script to send sd commands (there used to be one)

15:07 <keesj> it looks like .. pretty cool setup with sirgok/pulseview

15:13 <_florent_> keesj: i'm indeed creating hardware benchs to ease development of the different cores

15:13 <_florent_> https://usercontent.irccloud-cdn.com/file/NkmVAdYo/architecture.png

15:15 <_florent_> that's the idea with LiteSDCard bench, but the Pulseview/sigrok capture part is not yet fully validated

15:17 <_florent_> for now i've been mostly testing with an external analyzer on the Pmod connector, but the idea is to try to get rid of it and re-sample Pmod signals directly with the FPGA and stream it to the Host

16:57 <keesj> but.. is litesdcard the dut or the SDCARD? (I need to dump an eMMC and would fancy doing this with the fpga)

17:02 <somlo> _florent_: the data returned by `sdcard_read 0` after `sdcard_init` is different from the very first time, once you start ejecting, reinserting, and re-running the `sdcard_init` + `sdcard_read 0` sequence. No more outright failure upon `sdcard_init`, that's true

17:03 <somlo> but there's still something in the gateware that returns different results starting with the second time after eject+reinsert

17:03 <somlo> (maybe that's what you meant by "mis-alignment of data on first read")?

17:05 <somlo> either way, with the linux driver it still only works the first time the card is inserted (so whatever happens after *re*insertion is problematic still)

17:23 <_florent_> keesj: LiteSDCard is the DUT yes, there is a logic analyzer probing it internally and the PMOD is also probed (to verify the PHY/Protocol)

17:24 <_florent_> somlo: yes that's what i meant with mis-alignment, i'm looking at this

17:26 <somlo> _florent_: got it, thanks (apologies for coming off dumber than usual, experiencing unusually high levels of multitasking over here :)

18:59 <_florent_> somlo: the misalignment seems to be causes by some garbage data we get from the SDCard at the end of the initialization. I added some logic to ensure we flush this garbage data before doing the DMA:

18:59 <_florent_> https://github.com/enjoy-digital/litesdcard/commit/9e267a5a17e955326ec84059f8d94aa738b5c422

19:00 <_florent_> this seems to be working on my setup and with the BIOS (insert, sdcard_init, sdcard_read 0, eject, etc... sequence)

19:00 <_florent_> can you do a test?

19:42 <somlo> _florent_: thanks, working on it!

20:35 <somlo> _florent_: superficially, with the existing linux driver on nexys4ddr, it's still failing to re-initialize the sdcard upon reinsertion -- still only works the first time the card is inserted

20:40 <somlo> but I can confirm that from the bios perspective, initializing the card and reading any given block is repeatable and predictable, regardless of how many times we eject and reinsert/reinitialize the card

20:42 <somlo> since my minimal gateware reproducer is now gone, I'll dig into the driver over the weekend and either figure out what it's doing wrong, or come up with another reproducer :)