azonenberg_work has quit [Ping timeout: 260 seconds]
digshadow has quit [Quit: Leaving.]
digshadow has joined ##openfpga
pie_ has joined ##openfpga
pie__ has quit [Ping timeout: 245 seconds]
promach has joined ##openfpga
lain__ is now known as lain
pie__ has joined ##openfpga
pie_ has quit [Ping timeout: 240 seconds]
pie__ has quit [Ping timeout: 240 seconds]
pie_ has joined ##openfpga
m_w has quit [Quit: leaving]
<cr1901_modern> azonenberg: I guess I'll make a XIlinx forum acct and ask. Clock skew is a good guess, but I wish they'd explain where the extra 9ns came from
digshadow has quit [Quit: Leaving.]
<cyrozap> azonenberg: I think most people just write a utility to generate pcap data and either write that to a Linux FIFO or pipe it into Wireshark directly.
<azonenberg> cyrozap: well the thing is, i want it real time
<azonenberg> s.t. i can push a button and watch traffic streaming into wireshark as i do that
<cyrozap> Yeah, that's what I meant
<azonenberg> Wireshark can open pcap-format data on a pipe?
<cyrozap> Yup
<azonenberg> Interesting, that might be an easy option
<azonenberg> Make a simple utility that lets you configure capture filters
<azonenberg> then writes to the fifo
<azonenberg> down the road i want to do a proper integration so i can use wireshark capture filters and push them over jtag to limit what i'm sniffing
<azonenberg> rather than doing my own UI for that
pie_ has quit [Quit: Leaving]
<azonenberg> But any wireshark integration is better than none b/c it will save me the trouble of writing my own GUI
<azonenberg> And i hate doing guis :p
<cyrozap> I do it all the time to tcpdump from my OpenWrt router and pipe it over an SSH tunnel to Wireshark on my local machine.
<azonenberg> I will have to patch wireshark to do the NoC link layer etc for sure
<azonenberg> probably will have to register a custom protocol type with the libpcap folks etc
<azonenberg> But still, that should def be on my radar
<cyrozap> I think you can just write a plugin for that
<azonenberg> You can write a plugin for the decoding
<azonenberg> But you need to define a new encapsulation
<azonenberg> Since it's not in 802.3 framing
<azonenberg> they have predefined values for HDLC, PPP, 802.3, 802.11, USB, etc
<azonenberg> So i just have to ask them to allocate a value for Antikernel RPC / Antikernel DMA
<cyrozap> Ah
<cyrozap> I see
<azonenberg> Or i can use one of the "reserved experimental" values but then i cant share the pcaps with anyone else or they wont decode right
<azonenberg> i do expect to take that route for initial prototyping
_whitelogger has joined ##openfpga
<lain> just translate it to 802.3 framing :3
* lain laze
<cr1901_modern> "pipe it into Wireshark directly."?
<rqou> fake 802.3 and add a new ethertype?
<lain> ^
<rqou> you can even use branding like Antikernel over Ethernet :P
<jn__> AoE will always be Age of Empires :P
digshadow has joined ##openfpga
<lain> AoE is ata over ethernet to me
<lain> :P
<rqou> wait that exists?
<lain> yes
<rqou> why?
<lain> much simpler than iscsi
<lain> the entire spec is like 2 pages
<rqou> i guess
<rqou> wait, iscsi runs over ip right? it's not a layer 3 protocol by itself?
<rqou> can i do iscsi over ip over infiniband just to be a troll?
<cr1901_modern> I can't decide whether SCSI is difficult or not. I don't understand it, but I know someone who write a SCSI driver in 6502 asm for his custom SBC
<cr1901_modern> So obviously it must be grokkable by a single person
<trap15> it's not too bad
<trap15> yeah I thought I had, had to double-check. I've written a (fairly basic) SCSI driver so I could do CD access on PC-FX. admittedly most of that work is against the SCSI controller itself rather than the protocol, but the protocol isn't particularly complicated
<lain> scsi is great
digshadow has quit [Quit: Leaving.]
<lain> I've implemented spc, sbc, mmc
<lain> scsi specs should be used to educate people on how to write specs
<rqou> spc/sbc?
<lain> spc is primary commands
<lain> sbc block commands
<lain> mmc multimedia
<rqou> ah
<lain> (cd/dvd/etc)
<rqou> so maybe you can explain to me how some software "magically" can read stuff like the burst cutting area from an optical disk?
<rqou> every time i try to look at how this metadata is obtained, it seems to be shrouded in secrecy due to relevance to copy protections
<rqou> are these just extra commands somewhere?
<lain> it's likely defined in mmc yeah, the draft specs are free fwiw
<lain> you can do raw sector reads and wuch
<lain> such*
<rqou> and other weirdness like "read table of contents" or "set book type?"
<lain> but a lot is optional, some drives won't implement that
<lain> yeah
<lain> that's all in mmc
<trap15> yup
<rqou> alright, how about "css authenticate" or "aacs authenticate?" :P
<lain> yes
<lain> those are commands iirc
digshadow has joined ##openfpga
<rqou> undocumented? :P
<lain> no
<lain> fully dpc'd
<lain> doc'd*
<lain> grab thespec
<trap15> I'm looking at my old driver again now, I must've been cracked out or something. all in V810 assembly with no comments at all and no symbolic defs. ugh
<rqou> wait v810?
<rqou> are you in japan?
<rqou> or just an otaku? :P
<trap15> lol no, I was just playing with a jp console that was severely underloved
<trap15> and was also CD based with no protection, so easy unofficial dev
<rqou> "easy"
<trap15> but yeah the options were either use a really hacky GCC 2.95 fork with V810 support (it miscompiled a lot of my code), or write assembly
<rqou> except for the stack of cds you will end up going through :P
<trap15> so... I wrote assembly
<trap15> yeah, fair
<rqou> if you love broken gcc, you should have a crack at re-bootstrapping gcc for mips r5900
<trap15> D:
<cr1901_modern> trap15: v810 support was added to gcc4 by PlanetVB community
<lain> rqou: ^
<cr1901_modern> not that this helps you now :P
<lain> an older mmc draft
<trap15> cr1901_modern: yeah I heard about that, I intended to try it at some point but never did
<lain> I forget where but there's a table of all commands
<lain> well, the commands are also available in vsrious formsts from t10.org separately, but without the docs :P
<cr1901_modern> Someone will prob do an LLVM port at some point as well- v810 is just "classic MIPS with interlocking and bit-string isns" IIRC
<lain> I have some scripts that parse those listings for analyzing scsi traffic, to put real names to commsnds
<trap15> nah, v810 is all custom stuff
<lain> also typing while treadmilling is difficult
<rqou> wait, a nerd exercising? :P
<rqou> i thought only azonenberg did that :P :P
<lain> lol
<lain> gotta exorcise all the fat daemons
<cr1901_modern> trap15: Yes you're correct. I meant that figuratively, sorry :P.
<cr1901_modern> It reminds* me of MIPS
<cr1901_modern> but better
<trap15> yeah it's not too dissimilar, but it's got some encoding eccentricities iirc
<rqou> cr1901_modern: what about sh?
<rqou> :P
<trap15> and the ABI is jank
<rqou> apparently all the japanese cpus ended up with f*cked/bitrotten ABIs
<trap15> superh is quite awful in several but I like it
<trap15> several ways*
<trap15> yeah that's not surprising
<trap15> V810's jank ABI is due to a number of instructions having hard-wired register numbers
<trap15> like the string ops have fixed registers
<rqou> i mean, avr has that and their abi is just fine
<rqou> x86 too
<cr1901_modern> I wonder how the string ops are actually implemented in silicon
<trap15> hmm, maybe I'm being picky when I say jank ABI then. as I would say x86 ABI is jank probably
<rqou> lain: dumb question that isn't obvious from the spec: do cd rom "time" or sector addresses map to a fixed physical location on the disk?
<azonenberg_work> Sooo something just occurred to me
<cr1901_modern> I guess v810 is microcoded or has dedicated FSM states to handle the RISC ops a bit string insn translates to
<rqou> e.g. "sector <nnn> is <xyz> mm out from the center?"
<azonenberg_work> a LOT of 7 series complex IP blocks have a bunch of config registers documented since they're used for the DRP
<azonenberg_work> I wonder if this would provide a foot in the door for things like the PLL
<azonenberg_work> or MMCM
<azonenberg_work> or even transceivers
<rqou> i mean, you can always poke the ip block hdl, right?
<rqou> also, i have no 7-series devices right now; donations welcome :P :P
<azonenberg_work> Lol
<trap15> cr1901_modern: I believe it's like z80 where it just doesn't advance the PC unless the conditions are met. since it
<trap15> 's just doing a compare at the end, it just checks the conditions
<trap15> that'd just be a guess though
<rqou> lain: i'm basically curious what the disk drive firmware must be thinking when it sees:
<rqou> a) someone tricking it into reading a gd-rom
<rqou> b) those promotional rectangular credit-card shaped cd-roms that existed at one point in time
<azonenberg_work> in case B, i think they only wrote to a circular region within the square
<azonenberg_work> and the outer bit was just fluff
<azonenberg_work> writing starts from th ecenter
<azonenberg_work> you can see on a partially written CD-R a slight change in surface where the data stops
<rqou> hmm i've never actually owned one of those, so i don't know how full they are
<cr1901_modern> trap15: z80 has bit string instructions?
<rqou> i suppose if it's just marketing bullshit it can probably fit in the center region :P
<trap15> cr1901_modern: errr, no I'm thinking of the wrong CPU
<trap15> I'm thinking of 8086's REPZ/REPNZ stuff
<cr1901_modern> ahhh. Probably. In addition, I seem to recall that the the bit string insn registers are automatically updated each bit that is traversed (which is how interrupt information is saved)
<azonenberg_work> the duller area is where the bits are, smooth is unprogrammed
<rqou> wtf
<rqou> what product was that promoting?
<trap15> lmao
<rqou> i'm surprised there are no mechanical stress problems
<trap15> yeah I'm kinda curious what the front looks like
<azonenberg_work> curious too, they only have a back pic
<cr1901_modern> I could've gone my whole life without seeing that.
<rqou> "Irregularly shaped, non-rotationally-symmetric discs with an offset center of mass may also cause damaging vibration if played in computer CD drives, which can operate at a much higher rotational velocity than stand-alone audio CD players. "
<rqou> traip15, you sure you're not a closet otaku? :P
<trap15> I like games, does that count?
<trap15> I swear I barely watch anime ;p
<cr1901_modern> You lost half your manga at the Canadian border...
<rqou> um...
<rqou> is it _that_ specific manga material that is illegal in canada but legal in the united states?
<davidc___> yeah, you gotta be fairly twisted for them to take your comics...
<cr1901_modern> rqou: No, I don't think so
<trap15> no, but they sure weren't happy about me crossing the border with all of my belongings in a car
<trap15> so they roughed me up a bit and sent me back. good experience, would visit again
<trap15> they took a few of my arcade boards too I think? I'm not actually sure what all they took
<cyrozap> < cr1901_modern > "pipe it into Wireshark directly."?
<cyrozap> cr1901_modern: You can use a POSIX pipe to send pcap data to wireshark
<cyrozap> e.g., `foo | wireshark -i -` (I think, haven't done it in a while)
<cr1901_modern> cyrozap: That's what I thought, but I don't think I'd normally invoke wireshark (or a GUI program, for that matter) as the destionation of a pipe.
<cyrozap> cr1901_modern: Yeah, that's why I usually like to do mkfifo and have wireshark read from that
<cyrozap> so killing wireshark won't kill your capture, and vice-versa
<cr1901_modern> I am not long for this night...
* cr1901_modern tries to sleep to make up for staying up all night yesterday
<lain> rqou: yes, time values map to specific locations based on the TOC and such
<rqou> and when tricked into reading a gd-rom?
<lain> rqou: sectors can be addressed by LBA (data discs) or by track/time/frame
DocScrutinizer05 has quit [Disconnected by services]
DocScrutinizer05 has joined ##openfpga
<lain> I dunno, I assume raw sector reads
<rqou> but if you tell the drive "i want sector 12345" and the drive goes to where it expects that to be
<rqou> and then the drive firmware sees "wtf, this isn't the sector i wanted"
<rqou> it just seeks again and tries to correct?
<lain> if you do a raw read it won't even look at EDC/ECC
<lain> or at least, you can tell it not to
<rqou> so gd-roms have approximately 2x more data than a normal cd-rom
<lain> and I would guess the physical markers that allow it to locate a track are the same in gdrom
<rqou> how does addressing the extra data work?
<rqou> the physical markers are afaik the same
<rqou> but the "sector <nnn> to <xxx> mm from the center" mapping is all off
<lain> there are free ecma specs for the physical cdrom and dvdrom and etc design, including the EDC, ECC, track layout, data layout, amplifier design, etc
<lain> oooh this reminds me of a fascinating thing
<lain> lemme dig it up
<rqou> right, but it doesn't specify what happens when sega hacks their cd firmware to spin the motor half as fast
<lain> you can tell the drive what speed to run at
<lain> the velocities are well-defined in the spec
<rqou> so from what i understand sega took those numbers and divided them by 2
<rqou> so that their gd-roms had twice as much data as normal because the data was denser
<lain> cute
<rqou> but apparently you can trick a normal cd drive into reading a gd-rom
<rqou> and most of them will be able to stay in sync still
<azonenberg_work> Welllp
<rqou> i assume the drive firmware just gets really confused seeking?
<azonenberg_work> THAT would explain my weird transmit buffer behavior
<rqou> afaik usually people don't care about seeking; they just read the entire disk from the beginning to the end
<rqou> good job azonenberg_work :P
<lain> looool
* azonenberg_work makes note to add signoff checklist entry
<azonenberg_work> "verify polarity on silkscreen matches PCB"
<rqou> eh, it's just silkscreen; ship it :P
<azonenberg_work> lol
<azonenberg_work> good news is this is a characterization board
<azonenberg_work> only one will ever be made
<lain> ah right
<lain> I was looking for an amusing thing for rqou
<lain> hrm... it's definitely in my bookmarks somewhere...
<lain> but there's like 10,000 things in there... and I'm not seeing it in the expected categories, so it may be from The Before Times
<lain> aw man... well, this wasn't it, but... found a bookmark lost to time. archive.org doesn't have a copy
<lain> I really need to find/write a thing to scrape sites I bookmark for archival purposes
<azonenberg_work> Hmmm so my HV driver works
<azonenberg_work> but my LV driver is derping out hard
<azonenberg_work> the greenpak isnt enabling the relay to drive it
<azonenberg_work> Typoing the net name would help
<rqou> wtf ubuntu 16.04.2 delayed again
<rqou> for the third time
<lain> 2017 year of the linux desktop
<lain> </troll>
<rqou> i'm not even running a desktop
<rqou> this is a server
<lain> is it ubuntu server?
<rqou> yeah
<lain> is that... is that good?
<lain> I've honestly never used it, it seemed awful
<azonenberg_work> You arent missing much
<rqou> more supported than running debian sid on a server :P
<azonenberg_work> :p
<lain> lol
<lain> I used debian until I moved to freebsd
<azonenberg_work> Run debian jessie on your server then
<rqou> i need kernel 4.8
<azonenberg_work> May not be bleeding edge but it'll work
<lain> debian sta(b)le
<lain> :3
<rqou> has a sunrise point pcie isolation bugfix
<azonenberg_work> Wait till latest debian comes out then? :p
* azonenberg_work is on 3.16 lol
<rqou> good luck with a 3.16 on a kaby lake :P
<lain> man... this is such a great story, too
<lain> where did it goooo
<azonenberg_work> woop
<azonenberg_work> rqou: just finished first run testing of the starshipraider output buffer
<azonenberg_work> Looks good from 1.0 up to 5V at 100 Mbps
<azonenberg_work> through just the OBUF
<lain> google you cheeky fucks
<rqou> sustained short circuit duration? :P
<azonenberg_work> About to push to 200 then will reset things and try through the protection
<lain> you allow sorting results by date but you don't let me specify which direction
<azonenberg_work> rqou: so far i'm only testing the raw buffer sans protection
<azonenberg_work> working my way up
<rqou> idk if slew rate will matter for weird/buggy targets?
<azonenberg_work> That said, through a ~50 ohm resistor the max fault current is relatively low
<azonenberg_work> And i can probably add short detection at some point
<azonenberg_work> Eye at 200 Mbps 5V looks good...
<azonenberg_work> a bit more jitter but i blame my scope
<azonenberg_work> (still on the unprotected raw)
<rqou> i wonder what happens now when you try driving a marginal-timing ancient buggy nmos piece of crap :P
<rqou> where the slew rate must be in the right range
<azonenberg_work> Still a nice wide eye at 1.2V 200 Mbps
<azonenberg_work> Now to try through the protection
<azonenberg_work> After i go inside and let my hands warm up
<azonenberg_work> i dont think i can unscrew a SMA like this
<azonenberg> http://i.imgur.com/7niLn2d.jpg (wow this is blurry, need to reshoot the pic)
<azonenberg> but this is 1.2V at 200 Mbps through the LV driver, it looks a lot cleaner at 3.25 (with the LV driver) and 3.5 (with the HV driver)
<azonenberg> and just as nice at 5V with the HV driver
<azonenberg> Going to try through the protection once my hands thaw
<rqou> i "accidentally" decided to look in drivers/pci/quirks.c :P
<azonenberg> Or possibly tomorrow depending on how much time i have
<azonenberg> In any case it's looking promising so far
<rqou> must suppress the urge to strangle hardware/firmware people :P :P
<azonenberg> Then if i have time over the weekend i wanna bang up a full PRBS loopback tester on the zybo so i can go >> 200 Mbps
<azonenberg> (without seeing an eye)
<rqou> alternately, mail it to me and force me to do something more useful? :P :P
<rqou> huh, i just realized that the .hk whois server doesn't return mailing addresses
amclain has quit [Quit: Leaving]
laintoo has quit [Ping timeout: 240 seconds]
<azonenberg_work> rqou: ok did further testing
<azonenberg_work> OK results (some visible eye closing though) down to 1.2V at 100 Mbps through protection
<azonenberg_work> Trying 200 now
<lain> huh. my home router just rebooted. aaaand of course whatever caused it didn't emit a syslog packet to my syslog server...
<lain> I should setup something to log the serial output of it
<azonenberg_work> I really need to add a switch or something to my BERT design
<rqou> lol lain
<azonenberg_work> I have a PLL core in there that has runtime reconfig enabled
<rqou> and you mock my various broken server/workstation linux boxes :P
<azonenberg_work> But the reconfig settings i feed it are hard coded :p
<lain> rqou: first time I've ever had this happen
<lain> it's been rock solid since day 1 of installation
<azonenberg_work> even if i have to soft reset i need to fix this...
<lain> not sure what's up :P
<rqou> is this an x86 box or some embedded crap?
<lain> it's an edgerouter-x
<lain> very good router, runs a custom variant of vyatta
<lain> which is a variant of debian
<lain> but made for routers
<lain> it's /very/ good :P
<rqou> but it's an arm thing?
<lain> hardware packet processing acceleration, etc
<lain> mips
<lain> specifically it's uhhh
<lain> cavium octeon I think?
<rqou> ok, less crappy than i expected :P
<rqou> not some bullshit 200-400mhz arm or mips thing
<lain> no
<lain> it's not terribly fast, I think it's around 550 MIPS or something
<lain> but
<lain> hardware offload everything so it doesn't matter
<lain> looks like it's a quad core
<rqou> ugh, hardware offload tends to be buggy and/or doesn't plug into the linux networking stack very well
<lain> works fine here
<azonenberg_work> Wow there is actually some eye left
<lain> ubnt engineers are good
<azonenberg_work> at 1.2V 200 Mbps
<azonenberg_work> not bad at all
* rqou is saying this as i'm using a meraki box as a edge router :P
<azonenberg_work> visible jitter and eye closing
<lain> hmmm fascinating, maybe this /isn't/ an octeon like the others
<azonenberg_work> But it might be usable (and that's not even accounting for my scope trigger jitter)
<rqou> i used to be using a bay trail x86 as an edge router
<lain> this is actually a mediatek chipset, wow
<rqou> but then one of the housemates got employed at meraki :P
<lain> lol
<lain> so apparently this is a mediatek mt7621at @ 880 MHz, dual core
<rqou> not even working on the actual product
<lain> TIL
<azonenberg_work> welp, i think at this point
<rqou> this housemate was an SRE working on the meraki cloud crap
<azonenberg_work> i've done about as much testing as i can with the sillyscope on this board
<rqou> it was a lot less "legit" than i expected
<lain> the er-lite I have at the datacenter is cavium octeon
<azonenberg_work> Lol
<lain> running at 1 GHz
<rqou> he said that he was glad he wasn't on the product team so that he didn't have to have weekly meetings with broadcom
<rqou> :P
<lain> lol
laintoo has joined ##openfpga
<lain> the ubnt devs are pretty transparent
<lain> it's nice that, via the forums, you can directly talk with the people hacking on the firmware
<lain> and if all else fails, some people ported freebsd to this thing :P
<rqou> yeah i've heard of ubiquity
<lain> but for $50 it's a hell of a little router
<rqou> i didn't look too much into it because i had just gotten off of openwrt on consumer trash
<lain> haha
<lain> looks like there's a firmware update
<lain> might as well.
<rqou> you didn't hear this from me, but a broadcom coworker had his homelab powered by old broadcom SVK boxes
<lain> lol
<rqou> i wonder what kind of cpu power you need to handle nat at 2gbps?
<rqou> comcast has such an internet plan here if you have $$$$$$$$$$
<lain> the er-lite can do full line rate at 1gbps for max packet size
<lain> and 1Mpps for 64byte packets
<lain> in practice I'm using it to push full line rate through a gigabit uplink with nat and like 100 firewall rules with no noticeable slowdown
<lain> multiple vlans
<rqou> wait, the firewall rules can be offloaded?
<azonenberg> rqou: 2 Gbps? um
<lain> either that or it's just fast enough to do it in software, I dunno :D
<azonenberg> do you need a 10gbit nic on your PC?
<rqou> yeah i think so
<lain> it does offload a lot though, those cavium chips are quite powerful
<azonenberg> or is it two bonded 1g links?
<azonenberg> honestly i dont see a point in going faster than 1g for home internet for the time being
<azonenberg> even 100M
<azonenberg> because you usually will be rate-limited upstream and unable to actually saturate it
<azonenberg> heck, a lot of smaller websites may have one gigabit or 100M pipe shared by all concurrent clients
<rqou> gigabit pro
<rqou> #300/mo
<lain> honestly 100Mbps is plenty for what I do. late-night msdn will serve me full gigabit though which is fun
<lain> most stuff won't do >100
<azonenberg> Yeah i can pull isos from a few mirrors at >1G if i am on a fast enough pipes
<azonenberg> but its not worth the money b/c the servers cant keep up
<azonenberg> i'd rather have a full symmetric 50M pipe
<lain> so unless you're doing a lot of different things at once, there's not much real benefit beyond latency, which you could get with, say, gigabit link rate but 100M throughput for example
<rqou> downloading from "mirrors.ocf.berkeley.edu" while on campus is lots of fun
<azonenberg> With a static ipv4 and /56 ipv6
<rqou> when i lived in the dorms berkeley dorms were gigabit line rate capped to 100M throughput both ways
<rqou> dhcp-assigned static ipv4 but no ipv6
<rqou> on-campus lan ports (if live) are uncapped
<rqou> hmm, comcast makes it quite difficult to measure residential internet speeds
<rqou> because they let you burst a lot of traffic
<lain> yeah
<lain> also all the major ISPs boost throughput to speed test sites
<rqou> unfortunately afaik comcast is the only option in this area
<rqou> every other alternative i've looked at is much slower
<rqou> e.g. at&t u-verse thinks that they can sell us 6mbps
<rqou> as their fastest option
<rqou> honestly at this point that's barely even enough for "your grandma" :P
<lain> wow
<rqou> the excuse i've heard is that NIMBY-types are making it impossible to put in more telco pedestals in berkeley
<rqou> and the wiring here is ww2-levels of ancient
<rqou> so your only option is docsis
<rqou> and the only docsis option is comcast
<rqou> also at&t's u-verse offerings in general suck
<rqou> a) they're far too asymmetrical
<lain> yes
<rqou> b) their stupid 2-wire dsl box limits your performance
<lain> we're on uverse because comcast sucks, though they seem to be improving of late
<rqou> you should find a zero-day to pull the 802.1x cert out of the box
<lain> but we have 32M down (it *says* 24 but we get 32, it links at like 52 for the hdtv/phone overhead)
<rqou> and then ditch it :P
<lain> but only like 4M up
<lain> which is pitiful
<lain> yeah I've been meaning to experiment with high-speed SDR stuff and see what all goes on on the DSL side of the modem :P
<rqou> although i've heard that "ditch it" is also hard because vdsl2 is a dead-end technology that only at&t and 2wire even use
<rqou> wait lain why do you even get hdtv?
<lain> rqou: I live with my parents
<lain> I don't watch tv, but they do :P
<rqou> my parents don't have hdtv :P
<rqou> huh, apparently the berkeley (city, not university) library has gigabit internet
<rqou> via california's CENIC educational network
<rqou> aka "why can't we hook this public library to the university a block away that actually has good internet connectivity?" :P :P
<rqou> it's almost like that's actually a smart idea!
<rqou> in general the state of consumer/residential internet connectivity in the US is amazingly terrible
<lain> I really need to get those dishes setup
<lain> so I can just beam 100M to my house, from the dc :>
<rqou> microwave or free-space optics?
<lain> 5.8 GHz
<rqou> hmm don't you need a license to do that?
<lain> no
<lain> that's the same ISM band wifi uses
<rqou> normally you need "type approval" for equipment
<rqou> you can't just change out the antenna with a dish
<lain> ...
<lain> it's an ubnt dish
<lain> it's designed for purpose and certified as such
<rqou> ah ok
<lain> no license required :P
<lain> the datacenter is also a WISP
<rqou> i didn't realize that was allowed
<lain> so they have a tower
<lain> just need to get mike to climb it and install my dish, and hopefully I can get a link
<rqou> free-space optics don't need a license because they're above the fcc-regulated max frequency :P
<lain> hahaha
<rqou> afaik
<rqou> there may be additional laser safety requirements though
<lain> but lasers of high enough power to require a license, probably?
<lain> yeah
<lain> I don't know if I have optical LoS
<fpgacraft2> <nmesisgeek> i've seen free space optics w/ eye safe LEDs
<lain> that would be neat to mess with though
<fpgacraft2> <nmesisgeek> not even coherent
<fpgacraft2> <nmesisgeek> but at 10 Mbps or so
<rqou> i mean, hobbyists have been building that since the 80s
<lain> I've also seen like... atmospheric bounce optics
<rqou> anyways, right now basically all of my wifi equipment is technically illegal
<rqou> due to no type approval
<lain> lol
<lain> ubnt also sells some 3.something GHz dishes which is neat, you /do/ need an operator license for that though, and you have to file a site declaration when you install it
<lain> but that's not a big deal
<lain> main purpose of that is to avoid ISM congestion
<lain> or go whole-hog, get the airfibre 24GHz
<lain> (yes, twenty-four jiggahertz)
<rqou> i mean, as long as you can live without encryption you can operate under ham radio rules
<lain> they have a couple of those at the DC for backhaul to secondary sites
<rqou> according to my father apparently a number of years ago china realized they needed to do a crackdown of people illegally operating wireless links as a WISP
<rqou> because (given LoS) microwave links work _really well_ in dense urban areas like that
<lain> haha
<rqou> in general though china doesn't really care about spectrum enforcement very much
<lain> china doesn't care about much of anything :P
<rqou> people even operate US FRS/GMRS radios illegally there
<rqou> because they're cheap and common and afaik china doesn't actually have a similarly convenient radio service
<rqou> iirc canada had such a problem with us skiing tourists bringing over FRS radios that canada was forced to harmonize the band plan :P
<lain> woo
<lain> router updated
<rqou> it's kinda funny how international travel forced people to rethink band plans
<lain> FUCK
<lain> I FOUND IT
<lain> rqou: ^
<nats`> plop
<lain> azonenberg: you might also enjoy that ^
<rqou> hmm, i read the beginning and already assumed it was a scrambler issue
mifune has joined ##openfpga
mifune has joined ##openfpga
<rqou> why do people keep designing crappy scramblers?
<rqou> neat story overall though
mifune has quit [Read error: Connection reset by peer]
mifune has joined ##openfpga
mifune has joined ##openfpga
<rqou> i've never heard of ibm 8b/10b 64b/66b having these types of problems
<lain> they're designed not to
<lain> but yeah
<lain> the cd scrambler is interesting
<lain> the odds of that corruption happening are so slim that it was ignored
<rqou> the genius guys who designed SDI also screwed up in a similar way
<nats`> but 8b/10b was licensed
<lain> and the reasons for the scrambler doing what it does is related to amplifier technology of the day, prettymuch
<nats`> + there is one drawback of 8b/10b it's the need of LUT
<rqou> SDI can generate up to 19 consecutive zeros
<nats`> so if you don't need a clock recovery there are cheaper alternatives
<rqou> hence necessitating the creation of special "video SFPs"
<rqou> nats`: oh that's pretty interesting
<rqou> i didn't know that
<nats`> for example TMDS
<rqou> so that's why all of the "video" scramblers are different
<rqou> because they have embedded clocks
<nats`> TMDS was made for two reason: 1) no royalties 2) no clock recovery needed
<nats`> royalties is no more aproblem now
<rqou> but EFM/EFM+ do require clock recovery
<nats`> but TMDS can be calculated where 8b10b need to be "LUTed"
<fpgacraft2> <nmesisgeek> TMDS was made to reduce EMI
<fpgacraft2> <nmesisgeek> It was not intended for clock recovery
<fpgacraft2> <nmesisgeek> it's meant to have a source synchronous clock alongside then PLL up
<nats`> TMDS is not really good at EMI reduction in fact :)
<rqou> lol
<nats`> but yes like I said no clock recovery needed
<fpgacraft2> <nmesisgeek> i didnt say it worked, i said that was the goal :p
<rqou> in general i've somehow found that "video" people are not really good at anything
<nats`> in HDMI2 they even added a scrambler over TMDS because they wanted to spread the noise :D
<fpgacraft2> <nmesisgeek> matthiasm in ##fpga will rant at you all day long about such :p
<rqou> except maybe obfuscation (of both ideas and files (DRM))
<rqou> e.g. i have no idea why protocols like SDI need to exist
<nats`> yep I discussed a lot with matthiasm about that :)
<nats`> rqou because of existing system
<nats`> it's like ADSL
<fpgacraft2> <nmesisgeek> i dont either, why cant you just use raw frames over 10gbase-r?
<nats`> there was BNC everywhere
<rqou> i agree with azonenberg
<nats`> you couldn't ask studio to change everything when going to digital
<fpgacraft2> <nmesisgeek> you had to replace all of the expensive electronics
<fpgacraft2> <nmesisgeek> is it so muc hwork to redo the cable plant too?
<lain> never underestimate the cost of replacing building wiring
<nats`> yep but's it's easier than breaking wall
<lain> ethernet is built around this concept for example
<lain> there's a lot of things ethernet can't/won't do because of this
<nats`> do you imagine the cost of cutting through a street to put new cable
<rqou> except how nobody seems to want to deploy 10gbase-t?
<lain> if you look at the design decisions relating to this it's pretty interesting
<fpgacraft2> <nmesisgeek> who runs hd-sdi through a street?
<fpgacraft2> <nmesisgeek> also yes, nobody uses 10gbase-T b/c everything is multimode now
<nats`> it's often the case because TV studio are often dispatched through building
<lain> optics are the future
<fpgacraft2> <nmesisgeek> in the future i expect all protocols to be 8b10b or 64b66b high speed diff serial
<rqou> 128b/132b? :P
<nats`> 8b10b is basically dead
<fpgacraft2> <nmesisgeek> over 100 ohm differential pairs or fiber
<nats`> 64b66b is already not optimised enough
<lain> hm
<nats`> I was impressed by last radio technologie
<nats`> with fucking huge QAM
<rqou> yeah those are pretty neat
<lain> isn't docsis up to like 4096-qam now?
<rqou> 8192-QAM?
<fpgacraft2> <nmesisgeek> docsis is nuts
<fpgacraft2> <nmesisgeek> mostly because they have so little upstream b/w and the modems are braindead :p
<rqou> at least docsis is better than vdsl
<fpgacraft2> <nmesisgeek> dsl needs to die
<rqou> vdsl (or at least the u-verse way of deploying it) is even more braindead
<fpgacraft2> <nmesisgeek> i'm waiting for the day when i can just get a CWDM SFP for my cisco
<fpgacraft2> <nmesisgeek> and plug into a demarc point from my ISP
<fpgacraft2> <nmesisgeek> with a short singlemode patch cord
<rqou> i thought isps were deploying PON though
<fpgacraft2> <nmesisgeek> and be done with all of this "modem" nonsense
<nats`> I already have FTTH here :D
<rqou> because it saves cables
<fpgacraft2> <nmesisgeek> yes, they are
<fpgacraft2> <nmesisgeek> i said what i wanted :p
<nats`> no PON at home for me
<nats`> in france they deploy the two
<fpgacraft2> <nmesisgeek> i want 1000base-LX to the house throttled to whatever i pay for
<nats`> P2P and PON
<rqou> yeah, but you live in europe
<lain> PON?
<fpgacraft2> <nmesisgeek> or better yet, uncapped and billed at 95% usage
<rqou> we're in the land of making america great again :P
<nats`> GPON/EPON
<lain> ah
<nats`> basically each consummer has his share of same fiber
<rqou> i would be reasonably happy with PON
<nats`> I don't like pon because it is a HUB
<fpgacraft2> <nmesisgeek> i wouldnt, because it means i have to use an isp-provided modem
<nats`> you can sniff everything
<fpgacraft2> <nmesisgeek> which has been a major failure point for me
<fpgacraft2> <nmesisgeek> i want to just run 802.3 to the ISP
<rqou> hmm, my comcast docsis modem seems to work just fine
<fpgacraft2> <nmesisgeek> gaaah i wish i could just phone cogent or level3 and be like "hey, hook me up"
<rqou> at&t/2wire devices are dreadful though
<fpgacraft2> <nmesisgeek> how much does an AS number and PI space cost now? :p
<rqou> according to my father 2wire devices are crap for a "brilliant" reason
<fpgacraft2> <nmesisgeek> to help cisco sell more hardware?
<rqou> supposedly at&t basically squeezed 2wire to death
<rqou> but they can't kill 2wire
<nats`> sadly I'll have to pay for my internet in few weeks :p
<rqou> because nobody else really makes vdsl2 CPEs
<rqou> so now at&t has basically turned 2wire into a parasite that they _must_ keep on life support
<rqou> or otherwise at&t won't have a CPE
<nats`> I'm not sure you get it correctly
<nats`> ISP can't just remove copper line like that
<nats`> it's a fucking hard and expensive work
<rqou> yeah they can, just google "verizon fios complains" :P
<nats`> I'm working for an ISP....
<nats`> I have a pretty good idea how difficult it is
<lain> rqou: att deploys motorola modems nowadays
<rqou> oh?
<lain> I don't think any new installations are 2wire
<rqou> that's new
<fpgacraft2> <nmesisgeek> my comcast modem is a cisco
<lain> that's been the case for a few years
<lain> and 2wire isn't 2wire anymore it's PACE
<fpgacraft2> <nmesisgeek> dont know if linksys or real cisco business unit
<lain> but that's just buyouts
<fpgacraft2> <nmesisgeek> i mean IMO it isnt a cisco if it's not a teal-colored sheet metal 19" wide case :p
<nats`> nmesisgeek certainly huawei :p
<rqou> my comcast modem is an arris
<lain> azonenberg: ASNs aren't too expensive, IP space isn't bad either if you're willing to go v6-only :P
<fpgacraft2> <nmesisgeek> yeah my big problem is getting an isp who will do that
<rqou> hmm i should tell my father to poke at&t to upgrade the modem for his separate business line
<rqou> maybe to one that doesn't suck as much
<fpgacraft2> <nmesisgeek> i have no idea what it costs to get a direct pipe to a tier 1/2 provider but it cant be cheap
<lain> azonenberg: also technically level3 *will* just hook you up, but you're going to pay for them to trench it to you, which is $$$$$$$$
<fpgacraft2> <nmesisgeek> i'm not even worried about the cost of trenching
<rqou> lain: does the motorola modem have bridge mode?
<fpgacraft2> <nmesisgeek> that's a one time cost, maybe as much as a nice ferrari or something
<fpgacraft2> <nmesisgeek> i'm a lot more concerned about the cheap car worth of monthly bill :p
<lain> rqou: no idea, we're still on an old 2wire. I'm waiting for it to kick the bucket so I can request an upgrade, because the 2wires don't have ipv6 at all - only the motos
<lain> azonenberg: I mean, for a 150m run you may be looking at $10k+ trenching fees alone
<rqou> oh? i recall it did
<fpgacraft2> <nmesisgeek> my cisco modem is in bridge mode, my border router has a native routable v4 address
<lain> azonenberg: but the monthly fee will be $500 for like a DS1 lol
<rqou> whatever, this is for my father's work that is now Not My Job :P
<lain> varies a lot by region
<fpgacraft2> <nmesisgeek> lain: Yes, i'm aware (and i'm >> 150m from a fiber)
<fpgacraft2> <nmesisgeek> i was talking more like the $10K/mo it'd cost for an oc48
<fpgacraft2> <nmesisgeek> at that point i really dont care how much the fiber cost to trench :p
<rqou> ime most docsis deployments are bridge mode
<lain> depending on how well-connected your region is you might be able to get 100Mbps for like $1k/mo
<nats`> that's fun because when I listen to you I remember how lucky we are in france :D
<rqou> somehow dsl people just like to f*ck it up
<fpgacraft2> <nmesisgeek> (and it'd probably be on a pole, not a trench, until you reach the hosue)
<nats`> unlimited 4G on smartphone for like 20€ per month
<nats`> FTTH for 30€ per month
<lain> nats`: is it actually unlimited?
<nats`> :D
<nats`> yep it is
<lain> because here in the US there's "unlimited"
<nats`> when you reacha high threshold they shape the rate
<lain> but it doesn't mean unlimited at all
<nats`> but they don't increase the price it stays the same
<lain> it's either: you drop to 2G or 3G beyond a certain amount, or they charge you extra
<rqou> i don't get why i live like 5 blocks from one of the best research universities in the world and we still have crappy as shit internet
<fpgacraft2> <nmesisgeek> rqou: because you are willing to pay what you're paying for that crap
<fpgacraft2> <nmesisgeek> and it's cheaper for them to offer it
<lain> rqou: they want all the internet for themselves :3
<nats`> the ISP I work for put a high limit at 50GB....
<fpgacraft2> <nmesisgeek> capitalist logic
<nats`> before shaping your access
<rqou> i mean, $FANCY_SCHOOL basically owns their own fiber and their own uplink connections
<nats`> and the FTTH/DSL are really unlimited
<nats`> no shapping nothing like that
<rqou> i have no idea if they even need to pay for it :P
<lain> yeah, ftth/dsl is unlimited here
<lain> but 4g isn't
<rqou> ime ffth and dsl might be unlimited
<rqou> 4g definitely isn't
<rqou> docsis varies
<lain> att has tried to institute a cap on dsl
<lain> I don't know if they enforce it
<fpgacraft2> <nmesisgeek> my comcast business is uncapped afaik but its also $$$/mo
<fpgacraft2> <nmesisgeek> somewhere in the $150ish range for voice + data
<fpgacraft2> <nmesisgeek> iirc
<rqou> my father now has both a comcast resi and an at&t business line
<rqou> there's various reasons (e.g. redundancy) for that
<rqou> i should just move to europe :P :P
<rqou> at least i get good internet there
<lain> EU is even more surveillance-state-bullshit than the US
<lain> except for maybe NY area
<lain> :P
<nats`> ?
<rqou> obviously they're not doing enough, because they still have terrorists :P :P
<nats`> sadly it's what politics think -_-
<rqou> politicians think? :P
<rqou> definitely not POTUS
<nats`> yep
<rqou> POTUS just tweets :P
digshadow has quit [Quit: Leaving.]
<rqou> hmm some devs are finally working to make linux containers slightly less of a disaster security-wise
digshadow has joined ##openfpga
m_t has joined ##openfpga
<rqou> note to self: computer chassis parts are made off metal and block rf :P
<nats`> today I'll put my pcie spartan 6 board in my computer :)
Bike has quit [Quit: leaving]
_whitelogger has joined ##openfpga
<eduardo_> rqou: I dont get it. You do nothing with FPGAs, but you hang out on this channel. Why?
<rqou> hey, I do do stuff with fpgas
<rqou> nothing visible yet though
<rqou> also, this is a great "random hardware hacking" channel
mifune has quit [Ping timeout: 260 seconds]
<rqou> well, i just semi-bricked the nas box i'm building :P
<rqou> i accidentally disabled running legacy video option roms
<rqou> but didn't disable the csm altogether
<rqou> the anti-footgun feature doesn't kick in in this case and you can't see anything in the bios
<rqou> welp, going to wait for my uefi gop-supporting card to show up
<rqou> sleep time
<lain> lol
<lain> no serial console? ;)
<rqou> not a server board
<rqou> it's not actually bricked
<rqou> as linux boots the drm layer initializes and you can see the console again
<rqou> but now you can't change any settings
<lain> whoops :3
azonenberg_work has quit [Ping timeout: 256 seconds]
openfpga-bb has quit [Ping timeout: 264 seconds]
openfpga-bb has joined ##openfpga
azonenberg has quit [Ping timeout: 256 seconds]
azonenberg_work has joined ##openfpga
azonenberg has joined ##openfpga
azonenberg_work has quit [Ping timeout: 256 seconds]
azonenberg_work has joined ##openfpga
mifune has joined ##openfpga
mifune has joined ##openfpga
mifune has quit [Quit: Leaving]
<eduardo_> rqou: you asked for a series 7 FPGA board for "random hacking" . Still interested?
scrts has quit [Ping timeout: 240 seconds]
scrts has joined ##openfpga
promach has quit [Quit: Leaving]
<felix_> not exactly sure how similar v810 and v850 are, but i kinda liked v850e. definitely one of the not too messy isas
eduardo__ has joined ##openfpga
eduardo_ has quit [Ping timeout: 252 seconds]
Bike has joined ##openfpga
azonenberg_work has quit [Ping timeout: 258 seconds]
azonenberg_work has joined ##openfpga
openfpga-bb has quit [Ping timeout: 240 seconds]
openfpga-bb has joined ##openfpga
digshadow has quit [Quit: Leaving.]
Hootch has joined ##openfpga
digshadow has joined ##openfpga
Hootch has quit [Quit: Leaving]
pie_ has joined ##openfpga
<openfpga-github> [yosys] azonenberg pushed 26 new commits to master: https://git.io/vDwO5
<openfpga-github> yosys/master 7e08e37 Steffen Vogel: Fix compilation on OS X in order to support both MacPorts and Homebrew
<openfpga-github> yosys/master 19f3627 Steffen Vogel: Allow standard tools to be overwritten in make invocation
<openfpga-github> yosys/master 44b47b5 Steffen Vogel: use Homebrew only if installed
pie_ has quit [Changing host]
pie_ has joined ##openfpga
m_t has quit [Quit: Leaving]
<pointfree> I've been looking for a better way to way to describe RE'd things and found this: http://kaitai.io/ "Kaitai Struct is a declarative language used for describe various binary data structures, laid out in files or in memory: i.e. binary file formats, network stream packet formats, etc."
<pointfree> This was at FOSDEM 2017, https://fosdem.org/2017/schedule/event/om_kaitai/
<pointfree> The library of kaitai structs repo contains hardware not just image formats etc... https://github.com/kaitai-io/kaitai_struct_formats
<pointfree> Could this be the way to document bitstream formats? e.g, the Greenpak or PSoC?
<pie_> i tried to use it but didnt get very far due to lack of docs and ended up using Construct IIRC
<pie_> something like that
<pie_> for a firmware RE project
<pointfree> It would be cool to have support for mapping data into memory with kaitai structs in radare2.
<pie_> we didnt finish but one of the guys ended up writing a script for IDA to do some really basic memory mappingstuff the firmwre loader did (so not radare, cant cmment)
<pie_> but that was severalmonths ago
* felix_ wonders when ida will migrate away from legacy python to python3
<pie_> kek
<pie_> we should just write a better IDA
<felix_> well, there are some projects in that direction, but sadly all of them are ways less convenient to use :/
<cyrozap> pointfree: I don't think that project would be very useful for describing bitstreams, imo. Generally, by the time a DSL like that becomes specific enough to describe every format, it succumbs to the inner-platform effect: https://en.wikipedia.org/wiki/Inner-platform_effect
<pie_> to be fair, i havent gotten around to playing with/using ida much yet, but i get the feeling UI is probably one of the easiest plaes to improve :I
<pie_> *places
<felix_> sure
<pie_> cyrozap, whats that? (cant open browser rn, need RAM lol)
<pointfree> cyrozap: Good point.
<cyrozap> pie_: Basically, it just becomes a crappy programming language that no one uses.
<pie_> ah
<cyrozap> In most cases, you're better off just writing docs and a parser reference implementation in a normal programming language.
<pie_> yeah i always wonder about those "yet another application specific" languages that sometimes people/companies put in their software...
<pie_> i mean sure sometimes its a good thing but not most of the time when a proper "scripting" language would do
<pointfree> Before your comment, cyrozap, I was also going to say I'm concerned that it's not going to be very expressive for a human. If the scope is limited to mapping structures into memory and does only that I guess it would be okay. To be a good DSL the problem scope should be limited.
<pie_> one of the dudes on our little project wrote an ld script for i dont remember what reason
<pie_> not sure if thats relevant
* pie_ never actually looked into what those do
<cyrozap> That said, these sorts of things *are* useful in the case of ProtoBufs/Cap'n Proto/MessagePack/etc., where the focus is very narrow and it's designed for a specific problem domain (in the case of ProtoBufs: ensuring backwards-compatible, efficient, and reliable communication between systems written in different languages).
<openfpga-github> [openfpga] azonenberg pushed 2 new commits to master: https://git.io/vDwWQ
<openfpga-github> openfpga/master da9be6a Andrew Zonenberg: tests: Now PARing quietly by default
<openfpga-github> openfpga/master 717a1fb Andrew Zonenberg: gp4par: Fixed signed/unsigned mismatch
<cyrozap> pointfree: My thoughts exactly.
<pointfree> Also, what about describing analog formats! This only supports digital!
<pie_> :D
<cyrozap> This is pretty cool and also on-topic: https://github.com/forflo/yodl
<openfpga-bb> build #66 of openfpga is complete: Failure [failed test_openfpga_normal] Build details are at https://openfpga-dashboard.antikernel.net/builders/openfpga/builds/66 blamelist: Andrew Zonenberg <azonenberg@drawersteak.com>
<openfpga-bb> build #67 of openfpga is complete: Failure [failed test_openfpga_normal] Build details are at https://openfpga-dashboard.antikernel.net/builders/openfpga/builds/67 blamelist: Andrew Zonenberg <azonenberg@drawersteak.com>
<azonenberg> well duuuh its failing CI tests
<azonenberg> i need more devkits
<azonenberg> i can't use one in a project withotu breaking CI :p
<pointfree> cyrozap: That is cool. It could go in the list of project on the front page of the openfpga wiki.
<balrog> be careful with cap'n proto; as nice as it is, it uses C++14 features that aren't in the microsoft compilers
<pointfree> protip: https://pdfgrep.org/ can give you nice clean regmaps in a snap:
<pointfree> pdfgrep ': 0x' 001-82120_PSoC5LP_Registers_TRM_001-82120_0C_V.pdf | sed 's/\s\+/\n/g' | grep -v '^$' | paste - - | column -t | sort -k2 -u
<cr1901_modern> cap'n proto cuts the roof of my mouth. As does C++.
<cyrozap> pointfree: Nice!
<davidc___> I sorta discourage capnproto ATM
<davidc___> (I used it for a project of mine)
<davidc___> windows support is always $X months away
<cr1901_modern> This is why MinGW is good :P?
<davidc___> cr1901_modern: hard to do when the app is also a gui app
<cr1901_modern> What do you mean? MinGW has qt/gtk ports
<qu1j0t3> asking 1 more time are there some analog computing aficionados in here? i thought i remembered some discussion of same
<rqou> secure boot is broken beyond belief
<rqou> uefi is also broken beyond belief
<rqou> TPMs are also also broken beyond belief
<rqou> how did we get from "just perform some group/field operations on algebraic structures" to "giant stack many layers deep of code that does nothing except add more potential to break"?
<pointfree> qu1j0t3: In all seriousness I am interested in analog computers. I first learned of the PSoC while looking into FPAA's. I've been reading: http://www.analogmuseum.org/library/handbook_of_analog_computation.pdf but I don't claim to be an expert on the topic.
<qu1j0t3> pointfree: pm?
<pointfree> qu1j0t3: pm.