sb0 changed the topic of #m-labs to: ARTIQ, Migen, MiSoC, Mixxeo & other M-Labs projects :: fka #milkymist :: Logs http://irclog.whitequark.org/m-labs
acathla has quit [Read error: Connection reset by peer]
acathla has joined #m-labs
acathla has quit [Changing host]
acathla has joined #m-labs
<sb0> whitequark, ping
<mithro> Built using MiSoC!
<cr1901_modern> Pretty sure he knows that already
<shuffle2> what is the point of this "dummy" module? ( rjo? ) https://github.com/jordens/redpid/blob/master/gateware/redpid.py#L186
<shuffle2> also here https://github.com/jordens/redpid/blob/659144e3ac33904e2c237f2f7c2b2e9e14cf9731/gateware/pitaya_ps.py#L268 , if len(slaves)==1, Signal will complain. hopefully max(2, len(slaves)) is fine?
<shuffle2> so er CSRs will show up as magical mmio? can i just start reading/writing some physical memory from arm?
<GitHub68> [sinara] marmeladapk pushed 1 new commit to master: https://github.com/m-labs/sinara/commit/26ceff3021e6eb63827ab69c678b9e8e7dfec908
<GitHub68> sinara/master 26ceff3 Paweł: Kasli PCB routing nearly finished
<shuffle2> hm, doesn't seem like that's the case. tried giving CSRBankArray 0, then using phys addr 0 on arm...
<shuffle2> https://ybin.me/p/89af96d425bb0e48#KfiOpwIZm4dplNHVfjvxGG/aSFdFbS/gp258k7oiK8g= just trying with simple test, does it look sane?
<shuffle2> that part itself simulates fine, i'm not sure how to simulate all the axi glue stuffs tho
<shuffle2> https://ybin.me/p/c16c4be97174d7b1#KfOSzlMQOuI58LhDshuJa0ZhKPrA4yFlwiZvcyZ6Vio= with the axi stuff (just stolen from redpid)
<cr1901_modern> "magical mmio?" Misoc has code to automatically create CSR registers attached to a CSR bus. The CSR bus is then connected to the wishbone bus
<shuffle2> redpid apparently just uses axi
<shuffle2> i believe i have the CSRs "hooked up", i'm just wondering how to actually access them from arm side of zynq
<shuffle2> axi has some mmio region, maybe?
<cr1901_modern> Oh I've no experience using a hard CPU w/ misoc. I have no idea how the arm CPU is exposed to the FPGA
<shuffle2> its just 0x40000000 i think, but kernel doesnt seem to want to map it via /dev/mem (need dtb entry or something i guess...gross)
mumptai_ has joined #m-labs
mumptai has quit [Ping timeout: 255 seconds]
balrog has quit [Ping timeout: 264 seconds]
balrog has joined #m-labs
<whitequark> sb0: pong
<whitequark> sb0: the ARP issue is a known issue with smoltcp
<whitequark> though I haven't expected it to manifest in passive mode
<whitequark> but yeah, doing massive ARP scans would do that alright
<sb0> is it quick to fix?
<whitequark> fairly
<whitequark> I actually started fixing it during the flight
<whitequark> so I'm halfway done
<GitHub36> [smoltcp] whitequark pushed 1 new commit to master: https://git.io/vFOCS
<GitHub36> smoltcp/master 0091191 whitequark: Rework TcpSocket::{send,recv} to remove need for precomputing size....
<whitequark> (the above is unrelated)
<whitequark> (but it lets you do streaming parsing in ionpak I guess)
<sb0> well the current solution in ionpak is good enough
<sb0> the artiq bugs, however, are obnoxious
<whitequark> I was stuck without internet and had nothing better to do, sorry
<GitHub23> [smoltcp] whitequark closed pull request #67: Do not send ICMPv4 responses to broadcasts (master...improve_dst_unreachable) https://git.io/vFqBj
<GitHub161> [smoltcp] whitequark pushed 1 new commit to master: https://git.io/vFOCp
<GitHub81> [smoltcp] whitequark commented on issue #67: Thanks! https://git.io/vFOCh
<GitHub161> smoltcp/master d1d80ca Dan Robertson: Do not send ICMPv4 responses to broadcasts...
<travis-ci> m-labs/smoltcp#361 (master - 0091191 : whitequark): The build passed.
<GitHub136> [artiq] sbourdeauducq pushed 1 new commit to master: https://github.com/m-labs/artiq/commit/6089d44fd0616505088e2e2e98e3a2f5195615a3
<GitHub136> artiq/master 6089d44 Sebastien Bourdeauducq: artiq_flash: update Sayma flash proxy bitstream name
<sb0> rjo, did you already figure out the openocd commands to select a particular sayma when multiple are connected?
<bb-m-labs> build #872 of artiq-board is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq-board/builds/872
<bb-m-labs> build #605 of artiq-win64-test is complete: Warnings [warnings python_unittest] Build details are at http://buildbot.m-labs.hk/builders/artiq-win64-test/builds/605 blamelist: Sebastien Bourdeauducq <sb@m-labs.hk>
<bb-m-labs> build #1758 of artiq is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq/builds/1758
<sb0> ftdi_location I guess
mumptai_ has quit [Quit: Verlassend]
bb-m-labs has quit [Quit: buildmaster reconfigured: bot disconnecting]
<GitHub47> [buildbot-config] sbourdeauducq pushed 1 new commit to master: https://git.io/vFO8x
<GitHub47> buildbot-config/master 3dc15ff Sebastien Bourdeauducq: do not use --target-file for flashing KC705...
bb-m-labs has joined #m-labs
<sb0> whitequark, how did you obtain names like pci-0000:00:14.0-usb-0:8.4:1.0 ?
<shuffle2> ah i forgot, /dev/mem is picky about O_SYNC, and hexdump doesn't use that...
<shuffle2> trying to read from the csrs hangs the board...fun
<sb0> ah udevadm info ...
<sb0> is there any correspondence between ID_PATH, which udev uses, and the output of lsusb -t, which openocd uses?
<sb0> except for the USB interface number they seem completely different. sigh!
<whitequark> try DEVPATH
<whitequark> I think that's what you're looking for
<rjo> shuffle2: to implement the api for the redpitaya host side software. yes, max(2, n) is fine. it will be optimized away. and yes. the CSRs appear on mmio. see the little script that sets a bunch of them.
<sb0> ah yes, seems to match
<shuffle2> https://ybin.me/p/2372505ffea1106a#toq0Tm2q/iZmNiBPfUnbjNCyUXRhz9KnYUM0+E5HUgM= this is my linux-side test (which currently hangs)
<rjo> sb0: yes. ftdi_location. last i checked we had saymas on 5:1, 5:2 and 3:10
<sb0> there are so many nomenclatures for device locations in linux...
<sb0> rjo, ack
<rjo> lsusb -t also has the numbers.
<rjo> iirc
<rjo> sb0: ~rj/src/openocd/sayma.cfg
<shuffle2> rjo | shuffle2: to implement the api for the redpitaya host side software. < what was this part in response to? the "dummy"?
<rjo> yes
<sb0> okay you can match them like this:
<sb0> udevadm info -q property -n /dev/ttyUSB_sayma2_0|grep DEVPATH
<sb0> DEVPATH=/devices/pci0000:00/0000:00:14.0/usb3/3-10/3-10:1.1/ttyUSB1/tty/ttyUSB1
<sb0> then 3-10 (-> 3:10) is the location that openocd wants
<sb0> the devnodes I made udev create correspond to the UART (non-JTAG) channels only
<sb0> rjo, is the openocd installed in /usr/local/bin up-to-date and able to flash saymas?
<shuffle2> rjo: i really dont understand why the dummy stuff is required :( it seems that also requires adding the CRG module...is that really another PS-required component?
<sb0> also I'm going to put all three boards on the relay, since there is the 1V8 bug and others
<sb0> whitequark, what did you use to program that small arduino-like board? do you still have the source and instructions somewhere?
<sb0> is it 5V compatible? we can just hook it up to the ATX control pin if so
<rjo> sb0: i didn't install that. no idea.
<rjo> shuffle2: the redpitaya software axpected things at certain mmio segments. the dummy things provide the offsets.
<rjo> shuffle2: take a look at the redpitaya software and gateware (from that time)
<whitequark> sb0: amazingly, yes
<whitequark> I do have the source, I thought about throwing it out no later than two days ago
<whitequark> it's Maple IDE
<sb0> and 5v?
<whitequark> let me just zip everything up
<whitequark> just configure it as open drain
<whitequark> I'm not sure if the pin is 5V tolerant, sec
<sb0> you can still blow an open drain pin if you exceed the maximum voltage
<whitequark> many are
<whitequark> I know
<whitequark> the IDE
<whitequark> the script
<whitequark> save it as whatever.pde and put into ~/sketchbook/whatever/
<whitequark> pins D5-D9 and D14 are 5V tolerant
<whitequark> the sketch uses D13 and D14 now
<whitequark> one of them is a LED I think
<whitequark> yeah LED is D13
<whitequark> so you can connect your thing to D14
<shuffle2> rjo: ok...so the hk(housekeeping) region should be at offset 0, ie 0x40000000. however the other region is at 0x40300000. im not following where this huge offset comes from. the redpid scripts to deal with csr maps seem to indicate the bank size is only 0x800
<shuffle2> in any case, since i only have 1 bank, i should be able to expect it at 0x40000000 afaict
key2_ has joined #m-labs
<shuffle2> i'm suspecting there is something else going on. any access to any part of this region just hangs
<GitHub151> [artiq] sbourdeauducq pushed 2 new commits to master: https://github.com/m-labs/artiq/compare/6089d44fd061...99b99d436d46
<GitHub151> artiq/master 99b99d4 Sebastien Bourdeauducq: DEVELOPER_NOTES: add instructions to select Sayma
<GitHub151> artiq/master 4a57b52 Sebastien Bourdeauducq: artiq_flash: add preinit-command option
sb0 has quit [Quit: Leaving]
sb0 has joined #m-labs
<shuffle2> accessing axi region also hangs even with PL unconfigured/not using axi :(
<rjo> shuffle2: csr banks are 0x800 iirc. the 0x300000 is the 3 "Sys" devices in front of redpid iirc.
<GitHub102> [artiq] sbourdeauducq commented on pull request #842 0bc168e: breakout https://github.com/m-labs/artiq/pull/842#discussion_r148219675
<GitHub194> [artiq] sbourdeauducq commented on pull request #842 0bc168e: Please look at the patches you are proposing. https://github.com/m-labs/artiq/pull/842#discussion_r148219837
<GitHub170> [artiq] sbourdeauducq commented on pull request #842 0bc168e: ditto https://github.com/m-labs/artiq/pull/842#discussion_r148220190
<bb-m-labs> build #873 of artiq-board is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq-board/builds/873
<sb0> bah i have 2n7002 around, i'll just use that
<sb0> arduino doesn't let you set pins as open drain easily. and there is less risk of frying the board with an external transistor
<bb-m-labs> build #606 of artiq-win64-test is complete: Warnings [warnings python_unittest] Build details are at http://buildbot.m-labs.hk/builders/artiq-win64-test/builds/606 blamelist: Sebastien Bourdeauducq <sb@m-labs.hk>
<bb-m-labs> build #1759 of artiq is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq/builds/1759
key2_ has quit [Ping timeout: 260 seconds]
sb0 has quit [Quit: Leaving]
mumptai has joined #m-labs
rohitksingh has joined #m-labs
sb0 has joined #m-labs
Gurty has quit [Remote host closed the connection]
<sb0> rjo, does openocd support both flash banks?
<sb0> your script has only one flash bank
<sb0> (/usr/local/share/openocd/scripts/board/sayma_amc.cfg)
<sb0> and the flash still doesn't work
<sb0> Error: Unknown flash device (ID 0x00000000)
<sb0> are the two flashes still supposed to be configured with:
<sb0> "flash bank xcu.spi0 jtagspi 0 0 0 0 xcu.proxy $XILINX_USER1",
<sb0> "flash bank xcu.spi1 jtagspi 0 0 0 0 xcu.proxy $XILINX_USER2",
<GitHub134> [smoltcp] dlrobertson commented on issue #66: After a rebase, tests should pass here. https://git.io/vFOye
<GitHub102> [artiq] sbourdeauducq commented on pull request #842 2cf530c: The FMC breakout board does not have a "J44A" marking on its silkscreen and even less on its future front panel, so referencing it does not make the reader's life easier. You should write this documentation for someone who is setting up their boards the first time, and you should have said the *bottom* connector. https://github.com/m-labs/artiq/pull/842#discussion_r148244915
<GitHub5> [artiq] sbourdeauducq commented on pull request #842 2cf530c: The VHDCI breakout board does not have a "J44A" marking on its silkscreen and even less on its future front panel, so referencing it does not make the reader's life easier. You should write this documentation for someone who is setting up their boards the first time, and you should have said the *bottom* connector. https://github.com/m-labs/artiq/pull/842#discussion_r148244915
<GitHub2> [artiq] sbourdeauducq pushed 2 new commits to master: https://github.com/m-labs/artiq/compare/99b99d436d46...33d339947c36
<GitHub2> artiq/master bfd36e7 Sebastien Bourdeauducq: artiq_flash: update for Sayma OpenOCD changes
<GitHub2> artiq/master 33d3399 Sebastien Bourdeauducq: doc: add information on how to connect Zotino
<rjo> sb0: openocd version, bitstream source, openocd script, etc please.
<bb-m-labs> build #874 of artiq-board is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq-board/builds/874
<rjo> sb0: works fine here.
Gurty has joined #m-labs
Gurty has quit [Changing host]
Gurty has joined #m-labs
<bb-m-labs> build #1760 of artiq is complete: Failure [failed python_unittest_2] Build details are at http://buildbot.m-labs.hk/builders/artiq/builds/1760 blamelist: Sebastien Bourdeauducq <sb@m-labs.hk>
<sb0> rjo, the openocd in the buildbot /usr/local/bin and the latest artiq_flash I just committed
<sb0> should be the latest openocd, i just compiled it
<sb0> bitstream is the latest from your repos (jordens/bscan_spi_bitstreams), /home/sbbscan_spi_xcku040-sayma.bit
<sb0> btw RTMs are on saymas 1 and 3
<sb0> please test on sayma3, I'm using sayma1 (and the lock)
<sb0> the media converter here strictly refuses to find any sort of carrier coming from sayma. sigh.
<sb0> that would fix the problem: https://ethernetfmc.com/
<GitHub142> [artiq] jordens pushed 1 new commit to master: https://github.com/m-labs/artiq/commit/e16ab69f01f176a1b2d522514643dfa883d93034
<GitHub142> artiq/master e16ab69 Robert Jordens: DEVELOPER_NOTES: fix syntax
<GitHub42> [artiq] sbourdeauducq closed pull request #842: add zotino example code (master...master) https://github.com/m-labs/artiq/pull/842
<GitHub39> [artiq] jordens pushed 1 new commit to master: https://github.com/m-labs/artiq/commit/62320432a559652bb436e522a9112fe29f653f47
<GitHub39> artiq/master 6232043 Robert Jordens: artiq_flash: fix sayma amc tap/pld index
<sb0> hm, openocd doesn't throw an error when you load the bitstream into the wrong fpga?
<rjo> how would it know? it's just a bitstream.
<sb0> it can check the status of DONE afterwards
<sb0> the xilinx impact tool does that
<rjo> could you try again?
<rjo> sure. those things would be nice.
<sb0> okay, it works, thanks
<rjo> that would be unsuccessful bitstream loads, not specifically "bitstream fpga mismatch"
<rjo> ack.
<sb0> okay, but then it already gives a clear hint of what the issue is
<rjo> true.
<sb0> _florent_, how did you set the sdram to 32-bit? just remove the additional pins in migen?
<sb0> or did you use the other sdram bank?
<rjo> sb0: did anyone reply to jeff? do we have an issue to track that?
<rjo> whitequark: ^^
<rjo> ... the icmp storm
<rjo> *arp storm
<sb0> I didn't reply, just forwarded it to whitequark
<sb0> I can open an issue and reply
<rjo> i was about to hit send.
<sb0> ok go ahead
<GitHub6> [smoltcp] jordens opened issue #68: ARP storm https://git.io/vFOxt
<sb0> whitequark, this leafmaple ide doesn't find the board
<bb-m-labs> build #875 of artiq-board is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq-board/builds/875
<sb0> it looks for usb id 1eaf:0003 but the board is 1eaf::0004
<sb0> sigh
<sb0> there is obviously something you didn't tell me because that ide's source code has zero mention of that 0004 usb id
<bb-m-labs> build #1761 of artiq is complete: Failure [failed python_unittest_2] Build details are at http://buildbot.m-labs.hk/builders/artiq/builds/1761 blamelist: Robert Jordens <rj@m-labs.hk>
<sb0> oh you have to press the buttons
<sb0> sigh
<sb0> bah even with the right usb ids it refuses to find it.
<sb0> mcu programming standard fare
rohitksingh1 has joined #m-labs
rohitksingh has quit [Ping timeout: 240 seconds]
<GitHub172> [smoltcp] dlrobertson commented on issue #68: Also related to #50 and #25 https://git.io/vF3et
<GitHub33> [smoltcp] dlrobertson commented on issue #68: Also related to #50 https://git.io/vF3et
<GitHub65> [smoltcp] dlrobertson commented on issue #68: Might also be related to #50 https://git.io/vF3et
<bb-m-labs> build #876 of artiq-board is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq-board/builds/876
<sb0> of course, the 32-bit sdram does not work at all
<sb0> (using ddram_32)
<bb-m-labs> build #1762 of artiq is complete: Failure [failed python_unittest_2] Build details are at http://buildbot.m-labs.hk/builders/artiq/builds/1762 blamelist: Robert Jordens <rj@m-labs.hk>
<GitHub95> [smoltcp] klickverbot commented on issue #68: We saw a similar issue in our network a couple of weeks ago (ARP requests for the same address more than once every millisecond, from a nominally idle KC705), but it was late at night trying to get some actual data, so we just restarted the core device and it was gone. Unfortunately, I can't seem to find the dump right now, but reproducing this might definitely be a bit non-straightforward, as the probl
<sb0> oh, and sayma1 just died. won't power up anymore and only the 12V LED is on.
<sb0> perfect!
<sb0> just a power cycle did that
key2 has quit [Quit: Page closed]
<sb0> relay and RTMs are on the survivors ...
<sb0> fixed the board. one of the solders was flaky.
<sb0> ok removing data/dqs/dm pins from ddram_64 in migen works around the bug
<sb0> now the rtm bridge doesn't work
<sb0> not reliably
<sb0> well not at all; reading on the other side of the bridge just locks up the bus ...
<sb0> _florent_, ^^^
<sb0> which is also what i remember from the tests in maryland
<sb0> _florent_, didn't we also agree that the bus shall never lock up but instead bus errors should be reported?
sb0 has quit [Quit: Leaving]
sb0 has joined #m-labs
<GitHub72> [smoltcp] sbourdeauducq commented on issue #3: Does anyone want to contribute funds to this? https://git.io/vF3RQ
rohitksingh1 has quit [Quit: Leaving.]
Gurty has quit [Ping timeout: 240 seconds]
Gurty has joined #m-labs
Gurty has quit [Changing host]
Gurty has joined #m-labs
<whitequark> sb0: regarding the leafmaple board, I remember now
<whitequark> there is an obnoxious USB DFU bug
<whitequark> sb0: aha right. a) you need dfu-util 0.6 or later, and I think their IDE doesn't ship that one
<whitequark> b) how I flashed it is, I put it into "perpetual bootloader mode" (hold BOOT and release RESET)
<whitequark> then tried to flash it via the IDE, which failed, but gave me the built firmware somewhere in /tmp
<whitequark> then manually flashed it via system dfu-util
<whitequark> standard embedded device fare...
<GitHub62> [smoltcp] whitequark commented on issue #3: I know @edef1c had some code they wanted to contribute to smoltcp. That would actually go a long way towards IPv6 support. https://git.io/vF3pT
<GitHub158> [smoltcp] whitequark commented on issue #68: This is the exact same issue as #25.... https://git.io/vF3pn
<GitHub53> [smoltcp] whitequark commented on pull request #65 700ddd0: No no no, this is very intentional! Sequence numbers are **not** totally ordered! They wrap! https://git.io/vF3hT
<GitHub42> [smoltcp] whitequark commented on issue #65: I'll need to take a closer look at these changes. I'm not yet completely convinced they're correct. But, they do seem to go in the right direction. https://git.io/vF3hY
<GitHub80> [smoltcp] whitequark commented on issue #57: Can you please rebase again? https://git.io/vF3h8
<GitHub186> [artiq] whitequark commented on issue #723: LLVM-C binding just [gained DIBuilder support](https://reviews.llvm.org/rL317135), which means we will retain the debug information when migrating to LLVM-C. https://github.com/m-labs/artiq/issues/723#issuecomment-341260558
<GitHub108> [misoc] enjoy-digital pushed 1 new commit to master: https://git.io/vFsTO
<GitHub108> misoc/master 1ac1d7d Florent Kermarrec: cores/liteeth_mini/mac/crc: fix crc_error generation
<bb-m-labs> build #267 of misoc is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/misoc/builds/267