02:33 <Bleepshop> _florent_: Did you happen to extract a copy of the gateware from your RV901T before you reflashed it ?

08:54 <Melkhior> _florent_ Found the issue with Ethernet I think, device tree issue, PR litex #877 to patch json2dts

12:38 <_florent_> Melkhior: Thanks, I remember spending a few hours on this also recently: https://github.com/enjoy-digital/litex/commit/7abfbd9825188d1f6d97453838e18ed7af5526a7#diff-6bac543d87df6681b46e337de366d43284e9d06de14e43108fe407fb29c4c69f

12:40 <_florent_> I'll have a close look at #877 tomorrow to see what has changed since the last time I tested Ethernet with IRQs

12:51 <Melkhior> _florent_ From my understanding, LInux somehow look-up the 'interrupt-parent' property where it causes trouble ; I just moved it down into each device that needs it... and since then I've been able to compiler bash and some of binutils (ongoing) natively in Litex from/to a NFS filesystem.

12:51 <Melkhior> Network is not very stable for SSH (which loses the connection quite often, 'screen' is mandatory there), but NFS is quite tolerant to bad network and the SoC keeps going on through the '[15481.843109] nfs: server 192.168.X.Y not responding, still trying / [15482.914571] nfs: server 192.168.X.Y OK ' messages

12:51 <Melkhior> My switch tells me the interface is set to Gigabit, that probably is too fast for the four 100 MHz Vexriscv (and just one get the interrupt it seems:

12:51 <Melkhior> root@buildroot:~# cat /proc/interrupts |grep eth0

12:51 <Melkhior>   2: 437050 0 0 0 SiFive PLIC 2 eth0

12:51 <Melkhior> )

12:51 <Melkhior> but I don't know how to tells Linux to fallback to Fast Ethernet, and I've yet to try forcing the switch to see what happens.

12:51 <Melkhior> Anyway I now have an Ethernet interface doing what I want it to do :-)

12:51 <Melkhior> Thanks again for the great work!

12:53 <Melkhior> Last note: I've created the interface with 'nrxslots=16' don't know if/how much it helps.

13:31 <Melkhior> It seems forcing the switch to 100 MBits a) causes no issue whatsoever b) is faster/more reliable for my use case, I'm guessing I'm dropping a lot fewer packets

16:43 <Melkhior> 3 hours of non-stop compiling on NFS with the 100 Mbit setting, not a single SSH disconnection or 'nfs server not responding' :-)

16:43 <Melkhior> binutils built & installed

16:46 <Melkhior> kudos for the stability, and the Qmtech Wukong seems like a good cheap option for Litex on Artix-7 100T at this time!

16:46 <Melkhior> Only missing the keyboard, I now see the interrupt in /proc/interrupts (fixed by the Ethernet fix), but it stubbornly stays at 0, probably need to set the type somehow ('IRQ_TYPE_EDGE_FALLING')

16:46 <Melkhior> Kudos again

20:11 <sajattack[m]> <dkozel "Is there a way to load a new bit"> dkozel, I've been trying to figure this out too but haven't been able to get it working let me know if you do first.

20:12 <sajattack[m]> I set up the spiflash the way florent said but it just reads 0xff and fails to erase while programming

20:12 <sajattack[m]> maybe we can figure it out together