03:29 <somlo> _florent_: I noticed something weird: with LiteETH commit 2827e99 ("mac/sram: Fix length_lsb slicing (+1)", currently the latest in LiteETH master) litex/rocket can't get a dhcp lease under linux and busybox

03:29 <somlo> it's weird because I can netboot, and under linux and busybox I can manually set an IP address and ping (haven't tried anything fancier)

03:30 <somlo> but if I try and get a dhcp lease, it just keeps printing out "sending discover", forever

03:31 <somlo> that is unless I revert 2827e99, which is when udhcpc starts working again

03:31 <somlo> So I'm wondering if the linux driver needs updated to match 2827e99, or if that commit is in itself a regression of some sort

03:32 <somlo> sorry I haven't dug any deeper yet, just wanted to start by giving you the vague high-level behavior description first before I go to sleep over here :)

03:34 <somlo> oh and sniffing for udp packets on the LAN shared with the litex/rocket board, there are no dhcp requests being broadcast out

03:35 <somlo> anyway, that's all I have for now, I can dig deeper tomorrow morning

07:32 <_florent_> somlo: Hi, thanks for the feedback. I was trying to fix Linux-On-LiteX-VexRiscv CI for the SDS1104XE (using Hybrid Mac) but without access to my hardware. It should be good now with https://github.com/enjoy-digital/liteeth/commit/9b03fc14de8aa2454d967c16bcf7c97cf894b9ee

17:03 <tpw_rules> what are some other mPCIe FPGA cards that aren't the acorn cle-215 and friends?

17:46 <DerekKozel[m]> What're you looking for tpw_rules ?

17:47 <DerekKozel[m]> The Numato Aller probably counts as "friends", but is a commercially available component.

17:48 <DerekKozel[m]> The PicoEVB was/is interesting and is a direct relative of the CLE-215.

17:48 <tpw_rules> i guess i should specify that the reason i am interested in something else is that has nontrivial i/o and is shorter

17:48 <DerekKozel[m]> BitWare has the 250-M2D which has a Kintex FPGA on it in an M.2 formfactor

17:50 <DerekKozel[m]> I don't know of anything that has more GPIO, and most of the height is heatsink that you'd need to deal with one way or another. I have the Aller inside a Thunderbolt enclosure so it's thinness is required, but the enclosure is a big heatsink.

17:50 <DerekKozel[m]> There are non M.2 form-factors that are small and have loads of IO if that's a direction you could go

17:51 <tpw_rules> sorry, shorter as in not 2280 size

17:51 <DerekKozel[m]> Ah, less long

17:51 <tpw_rules> yes ;P

17:51 <DerekKozel[m]> the Pico EVB is less long, not more IO though

17:51 <DerekKozel[m]> https://concurrenteda.com/index.php/fpga-modules/artix-modules/te0711-datasheet

17:51 <tpb> Title: TE0711 - Artix-7 High I/O & USB (at concurrenteda.com)

17:52 <DerekKozel[m]> This is less long and has more IO pins, but more $$ and not M.2

17:53 <tpw_rules> i wanted m.2 so i could hook it up to a small pc with a fast interface. maybe a zynq type device would be an alternative, but i think their CPUs are relatively slower compared to even low end x86, right?

17:53 <tpw_rules> also for Reasons™ i can't buy anything made in china. made in taiwan is okay though

17:54 <DerekKozel[m]> There's a load of FPGA boards out there, given Reasons I'd guess you can afford more of the options than me

17:55 <tpw_rules> like i said i don't know too much about the ARM CPU power

17:55 <DerekKozel[m]> Can you spin a board based off the PicoEVB? It'd be "easy" to route more pins to a high density connector

17:59 <tpw_rules> that is an idea

18:02 <DerekKozel[m]> RHS Research published the design files under GPL and they're nice to talk to. I've sent a few emails over the years and they like intelligent questions and comments.