<rqou>
lain: do pcie switches exist that are v3 on the uplink and v2 on the downlink? is this allowed?
<lain>
no idea
<lain>
haven't done much with pcie yet
<rqou>
because a pcie card that is x1 v3 on the edge and hosts two x1 v2 on mpcie connectors would be quite useful
<rqou>
dual wifi cards without unnecessarily occupying dual slots
<lain>
you could probably just use a v3 to v3
<lain>
and then it'll auto downgrade to v2
<lain>
I can't imagine that would have an issue
<lain>
(but I haven't read the spec)
<rqou>
afaik this type of card doesn't exist in general
<rqou>
even though it should be easy to make one
<lain>
why do you want dual wifi though :o
<rqou>
one for 2.4 ghz, one for 5 ghz
<rqou>
(as an AP, not a client)
<lain>
hm
<lain>
can modern cards not be AP at both simultaneously?
<lain>
I've never tried
<rqou>
most can't be on both bands at once
<lain>
ah
<rqou>
if you scroll up i mentioned i was using an ath9k n card on 2.4ghz and an ath10k ac card on 5ghz
<lain>
ah ok
<rqou>
because trying to do vht80 on 2.4ghz probably isn't allowed :P
<rqou>
and would be a dick move if it was
<lain>
lul
<rqou>
i was actually running ht40 on 2.4ghz just to be a bad wireless neighbor :P
<rqou>
yeah apparently vht80 wouldn't even be possible on 2.4ghz in the united states
<rqou>
you could theoretically do it in other countries
<rqou>
occupying literally the entire 2.4ghz band at once :P
<lain>
rip
<rqou>
hmm i haven't checked whether vht80+80 equipment exists yet
<rqou>
ludicrous 160mhz of rf bandwidth
<fpgacraft2>
<nmesisgeek> b/c 802.11ac is too slow?
<rqou>
much increased probability of causing laserbeams to appear on weather radar! :P
<lain>
lol
<rqou>
wikipedia claims that 160mhz could give you up to 3gbps for one client
<rqou>
vht80 is already supposed to give you just over a gigabit
<rqou>
also wikipedia claims the particular 802.11ac chipset i have can't do TxBF :(
<rqou>
offtopic: i really want to know who came up with the brilliant (/s) hack of inputting 2fa codes by appending them to the password
<fpgacraft2>
<nmesisgeek> rqou: at $WORK we do 2fa by re-prompting for the password
<fpgacraft2>
<nmesisgeek> basically pretending the first one failed
<rqou>
that's terrible too
<fpgacraft2>
<nmesisgeek> integrating with legacy vpn? not sure of other options
<fpgacraft2>
<nmesisgeek> our custom apps designed for 2fa prompt directly for the token
<rqou>
not having legacy? :P
<rqou>
hmm, TLS client certificate auth with the client certificate on a smartcard and requiring a password is technically 2fa
<rqou>
probably supported even more poorly
<rqou>
in general the UX for 2fa still sucks massively
<rqou>
the biggest thing I don't get is SMS-based 2fa
<rqou>
vs totp: more work, more expensive, less secure, less inclusive
<rqou>
less reliable
<whitequark>
it's defending against "ex-boyfriend stalking his ex's facebook" while not requiring a complicated token setup
<whitequark>
it works for that, barely
<whitequark>
trying to make it work for anything else is futile
<rqou>
so does totp?
<whitequark>
totp needs some "app" and then you lose your phone and can't post pictures of your food anymore
<whitequark>
sms-based 2fa works even on dumbphones
<whitequark>
and a mobile phone # is often required anyway due to antispam
<whitequark>
you *can* realistically mandate sms 2fa. you cannot mandate totp
<rqou>
but not all mobile phone plans include SMS
<whitequark>
I seriously doubt there's a mobile phone plan that doesn't permit you to receive SMS anywhere in the world, possibly excluding the US
<rqou>
I used to have a plan like that many years ago
<whitequark>
"many years ago" doesn't count
<rqou>
it was after having mobile data
<rqou>
I had a customized plan that was voice and mobile data only, no SMS
<fpgacraft2>
<laincat> anybody else remember the Ogo devicE?
<rqou>
idk if you can still have one like that
<fpgacraft2>
<laincat> on cingular in the US, dunno if it was elsewhere
<rqou>
I remember being on Cingular :P
<rqou>
having only voice and no data nor SMS
<rqou>
idk if this has changed, but _receiving_ a SMS on a crappy plan could cost up to 10 cents each
<fpgacraft2>
<laincat> that is still true for some plans yes :(
<rqou>
apparently that style of billing exists in China too, except it's RMB0.10 instead of USD$0.10
<lain>
rqou: am I allowed to punch the seed for fpgacraft2 into a slime chunk finder website? if so, would you mind sharing the seed? :>
<rqou>
how about a challenge: back-calculate the seed given the terrain you have access to :P
<rqou>
i believe this is actually a well-posed problem
<rqou>
(i'm too lazy to look it up right now :P )
<rqou>
anyways, another very irritating thing is services that are obviously using TOTP under the hood but have some kind of "value add" that makes it not work with normal TOTP tools
<rqou>
(looking at you battle.net and symantec VIP)
<lain>
haha
<lain>
and, yes, that is annoying :/
<lain>
I like totp with yubikey, it works very well
<rqou>
yeah, that's what i'm using too
<lain>
actually, the other day I had a thought about my yubikey - I use the openpgp smartcard in it, protected by a strong, randomly-generated PIN
<lain>
I have to enter the PIN, which is then verified, and then in order to actually sign something I have to tap the device itself
<lain>
it would be nice if you had to tap it /to initiate pin verification/
<rqou>
mine doesn't even require a tap
<rqou>
i assume you're using the yubikey 4?
<lain>
not that it really matters, it locks out after like 3 failed pin attempts
<lain>
yes
<lain>
(at which point you have I think 3 more attempts with the admin PIN, which in my case is even longer than the user PIN, and I think if you fail all 3 admin attempts it self-wipes)
<lain>
you have to configure the tap requirement when you setup the openpgp keys, guessing you can enable it later though
<rqou>
ime the yubikey/openpgp cards work pretty well, but the software stack works like shit
<rqou>
right now my desktop is refusing to ever prompt me for a pin
<lain>
yes, I think all openpgp-related software suffers from usability concerns :P
<rqou>
it'll probably fix itself after i reboot
<lain>
ah, I haven't had much issue aside from figuring out how to get it all working
<rqou>
i also have gnupg marked as held in apt right now because upgrading makes hotplugging the yubikey not work at all
<lain>
one time the stack did hose itself though, but restarting the gpg-agent process fixed it
<lain>
lool
<rqou>
i really should file a bug about that, but _effort_ :P
<lain>
I'm on windows, the situation is pretty good here at least
<rqou>
i've only ever gotten things worse by restarting gpg-agent
<lain>
it even works over RDP
<rqou>
i don't understand why this is so damn complicated
<lain>
since you can share smartcards over RDP, if you desire
<lain>
because linux developers don't like usability :V
<lain>
just look at git
<rqou>
git at least has a simple enough object model that fits in your brain
<lain>
I dunno, when you hit those edge cases stuff falls apart pretty quickly
<rqou>
whereas anything smartcard-related seems to have a huge pile of layers that do nothing
<lain>
google is full of magical incantations for fixing various git problems, most of which have an astounding number of hidden side-effects
<rqou>
"I think so, although this does mean any XSS on webex.com would allow remote code execution. If they think that is an acceptable risk, it's okay with me."
<rqou>
"This means that if a site is not *.webex.com or *.webex.com.cn, then the user must click OK for code execution to happen."
<lain>
oh.. /wow/
<lain>
like seriously wow
<rqou>
but it's fine because the user consented by clicking OK :P
<lain>
"pwn me? sure whatever I just want the googles to work.
<lain>
I was very surprised to read a while back that japan has horrible computer literacy rates
digshadow has quit [Ping timeout: 255 seconds]
<davidc__>
rqou: yeah, I bet you could totally pwn tons of people with bigcorp-webex.com
digshadow has joined ##openfpga
digshadow has quit [Client Quit]
digshadow has joined ##openfpga
promach has quit [Ping timeout: 276 seconds]
digshadow has quit [Ping timeout: 255 seconds]
promach has joined ##openfpga
<cyrozap>
rqou: I RE'd Symantec VIP Access a few years ago and wrote a (now fairly-popular) Python program in order to avoid using their app :P
<cyrozap>
lain: "like seriously wow" shows up with rainbow colors...
ZipCPU|Laptop has joined ##openfpga
<cyrozap>
rqou: Also, Valve's Steam Guard *also* uses TOTP, but does encodes the resulting OTP into a 5 character string instead of a 6 digit code because it's Valve and *of course* they would do that.
<lain>
cyrozap: that was intentional :>
<cyrozap>
lain: how... how do?
<lain>
`prism' extension for weechat
<lain>
/prism <words>
<lain>
it also lets you start the rainbows with a trigger char, like /prism -c : test :test gives
<lain>
test test
<lain>
it's magical
<cyrozap>
wow
<lain>
it has a variety of other settings
<lain>
like "eye-destroying colours"
<lain>
this
<lain>
which is a great way to get colours banned on a channel :P
<mtp>
oh dang
<mtp>
i dig it
<mtp>
i wrote a rainbow for my client
<mtp>
but i'm also an emacs wank
<lain>
:D
digshadow has joined ##openfpga
digshadow has quit [Ping timeout: 255 seconds]
<rqou>
cyrozap: the symantec vip thing was you? nice
<rqou>
i've definitely used it
<rqou>
never encountered the steam guard one because i don't have 2fa on steam
* lain
hijacks rqou's steam account
<rqou>
i personally don't see why i would bother much
<rqou>
i have 2fa on my runescape account though even though i no longer really play it
<rqou>
because my account is "OG" and has the 10-year cape, so it's somewhat of a target
<lain>
some cyber cafes / game places have basically every game installed and you just sign into your steam account. I wouldn't do that without 2fa. also some people leave payment creds saved in steam store (I wouldn't do that in general though, because why)
<rqou>
hmm never used a cyber cafe
<rqou>
those places always felt really shady
<lain>
I used to frequent some local ones with friends, was fun times
<rqou>
hmm i always associated cyber cafes with east asia
<lain>
usually they were violating the hell out of the ToS/EULAs though :P
<rqou>
so HK/CN/JP
<lain>
daemontools to mount games from a central nas on like 10Gbit connections to all the PCs
<rqou>
i never really thought about ones in the US
<lain>
good excuse to build out an overpowered storage network
<rqou>
like i'm working on right now? :P
<lain>
hehe
<rqou>
although i have a semi-serious use case of slowly cleaning up and deduping my files
<rqou>
i have lots of machines with lots of files that are screwed up in all sorts of ways
<rqou>
e.g. i have a hdd that has at least one and possibly two "tumor" ubuntu partitions
<rqou>
when i needed to upgrade to the next major version, instead of doing a dist-upgrade
<rqou>
i resized the existing partition until it had 0 bytes free and created a new partition
<lain>
rofl
<rqou>
and did a clean install
<lain>
you can do some awesome tricks in that regard with zfs on freebsd
<rqou>
how about zfs on linux? :P
<lain>
I dunno, does it support root on zfs?
<rqou>
i don't know, maybe?
<lain>
with snapshots and such
<rqou>
i currently have a btrfs root though
<lain>
you can do things like have a single dataset for your OS (separate dataset for your user files, etc)
<rqou>
i haven't lost any data yet (knocks on wood)
<lain>
and then if you want to upgrade but be able to quickly roll back, or try a new build or etc
<rqou>
i do that using btrfs
<lain>
you can just snapshot the OS dataset, do the install, and then you can roll back to snapshot if you need
<lain>
or you can clone the snapshot, escalating it into its own copy, and then you can dual/triple/N-boot
<rqou>
i actually did a ubuntu to debian "crossgrade" using btrfs snapshots
<rqou>
it might be the cause of some mysterious bugs still
<lain>
ah nice
<rqou>
i basically created a whole new subvolume and deboostraped debian into it
<rqou>
keeping the existing home subvolume
<rqou>
but apparently ubuntu does/did some weird things to some files that somehow configure the gui stack
<rqou>
so there were mysterious bugs occasionally
<rqou>
like text appearing anywhere from 1x to 4x normal size
<rqou>
depending on i don't know what
<rqou>
most of the bugs got perturbed away over time
<rqou>
:P
<lain>
rofl
<rqou>
it seems that linux desktop never "fixes" bugs
<lain>
the win10 / server 2016 upgrade process is sort of like this also
<rqou>
they just perturb them away
<lain>
it effectively images the current installation, some other magic, so you can roll back
<lain>
not as powerful, but it's neat stuff
<rqou>
hmm i just remembered i need to file a networkmanager bug
<lain>
too bad the windows update team is, imo, messing up lots of stuff :x
<rqou>
it'll probably be a wontfix though :(
<lain>
what's the bug, out of curiosity?
<rqou>
mac address randomization doesn't work correctly with wl.ko
<rqou>
yeah
<rqou>
"get your proprietary shit outta here"
<lain>
lol
<rqou>
i wonder when windows will fix the "need to update the updater to install updates" meme/problem? :P
<lain>
indeed
<rqou>
apt has literally never given me this problem
<lain>
well that seems ok in win10
<rqou>
win10 failed to upgrade properly in my vm without a bit of babysitting
<lain>
I haven't had issues with stable win10, but the insider preview fast track is a mess for me lately
<rqou>
you needed to poke various parts of the shutdown/reboot process or else it would instantly fail to load the kernel
<lain>
it's been totally unusable
<lain>
so many feedbacks to file
<rqou>
i guess microsoft can claim that "Red Hat VirtIO SCSI" is not a supported hdd controller :P
GenTooMan has quit [Read error: Connection reset by peer]
<lain>
the latest amusing insider preview fast track bug is: VM attempts to install latest build, bugchecks on every reboot, gives up, restores itself from image, then immediately tries to install the latest build again before I can blacklist that update
<lain>
so I've just stopped using that vm for now
<lain>
probably about time to wipe it anyway, vs2017 RC kinda horked a few things too
<rqou>
so can anyone explain to me wtf texlive is? i have it installed because various shit needs it to build docs or whatever, but its only purpose seems to make apt-get upgrade take forever
<lain>
lol
<rqou>
like it downloads an extra gig of stuff
<lain>
good lord
<rqou>
sagemath was also a pretty bad offender back when i had that installed
<rqou>
their cdn was super slow too
<rqou>
and yes, it was definitely on their end
<rqou>
because back when i was on campus i did a dist-upgrade once and everything else downloaded from "mirrors.ocf.berkeley.edu" in seconds
<rqou>
and then i had to wait like 5 minutes for sagemath to download
<lain>
yeah I love when youtube is like "Problems loading? Find out why!" and it says my ISP is too slow
<lain>
... on a gigabit link to a 10gbit tier 1 Level3 leased line
<lain>
YEAH OK GOOGLE
<lain>
the problem is always a bum cache server, too
<rqou>
on campus i had a 1 gigabit link to the "mirrors.ocf.berkeley.edu" machine
<lain>
usually if you flush dns cache and get lucky and dns points you to a different cache server, it'll fix itself
<lain>
but google is happy to blame ISPs for their own cache server issues
<rqou>
one time i had a friend who had already graduated show up on campus
<rqou>
he said that he needed to clean-reinstall his mac
<rqou>
and the campus connection was faster than his work connection
<lain>
haha
digshadow has joined ##openfpga
<rqou>
this friend also had a brilliant trick for making the campus wifi work better
<rqou>
the trick was: if you hear some very rapid clicking and swearing in korean coming from the lounge area, walk up to the (PoE) wifi AP and unplug it
<rqou>
wait until the swearing intensifies and a group of korean grad students leaves the lounge
<rqou>
then plug the AP back in
<lain>
rofl
<rqou>
apt is now installing texlive-lang-polish
<rqou>
i don't know why polish needs special logic
<cyrozap>
rqou: Thank you for reminding me that Runescape exists--I just logged into my account that I probably haven't accessed since middle school. 6-digit password, no email associated to the account, with ~37 hours of total play time :P
<rqou>
lool
<lain>
lol
<rqou>
glad it didn't get hacked in the intervening years
<whitequark>
... hm
<cyrozap>
I'm actually amazed it wasn't
<whitequark>
I had a Runescape account since middle school too
<whitequark>
*stare*
<rqou>
wait, russia had runescape?
<whitequark>
no?
<rqou>
or were you not in russia when you were in middle school?
<whitequark>
why would this have any bearing on whether i can play runescape
<rqou>
oh right, runescape doesn't geoblock
<whitequark>
I don't think geoblocking was *invented* back then
<rqou>
they do geoblock now for buying membership
<rqou>
because people were using SEK to buy membership and evade VAT
<whitequark>
also, do you think geoblocking would have stopped me?
<rqou>
they've now since moved to RUR to buy membership to both evade VAT and get lower prices :P
<rqou>
yeah, i suppose not
<lain>
whitequark doesn't afraid of anything
<whitequark>
no you just kinda... have to figure out this shit
<whitequark>
I think the literal first thing I had to learn when I got access to the internet is how to pirate software (safely)
* whitequark
still has adoration for keygen music
<rqou>
hmm that reminds me that i still need to check out the ida 6.8 i picked up at 33c3
<rqou>
i still haven't dared to run it
<lain>
lol
<whitequark>
lol rqou, isn't it a bit more dangerous for you to admit that than for me? :p
<mtp>
true story, i got into doing sysadmin shit so i could get around the school's internet filter
<rqou>
aaand upgrading debian broke it
* lain
has a legit ida license
<lain>
but no hex-rays :<
<rqou>
"attempt to index field '?' (a nil value)
* whitequark
would just write a goddamn decompiler, it's not that hard
<lain>
eh
<whitequark>
seriously, the amount of whining about hexrays in infosec is ridiculous
<lain>
the decompilation isn't the hard part
<whitequark>
what is?
<lain>
it's all the built-in knowledge
<lain>
it's not so much hard as cumbersome
<whitequark>
mmh
<lain>
falls under "work I don't care to do myself" :P
<lain>
but then, reading the disassembly is easy enough anyway
<lain>
something ida is sorely lacking imo is a collaboration system
<lain>
there is collabREate, but that doesn't work on the latest versions
<lain>
ida seems like a mess under the hood, also :/
<lain>
understandable given its history, but it's time for some better alternatives to pop up, I think
<whitequark>
radare seems like a horrible mess from what i looked
<lain>
that's unfortunate
<whitequark>
not sure how it compares to ida but it was quite bad, enough to dissuade me from touching it
<lain>
hopper seems ok but not as featureful as ida
<lain>
the r2 webui is an interesting idea, I wonder if that enables collaboration?
<rqou>
i am this close to flagging awesome wm as "held" forever
<rqou>
it always seems to frivolously break things
<lain>
I still don't know how I survived running a *nix desktop years ago
<lain>
it seems so fragile
<whitequark>
rqou: I gave up on awesome and use i3 now
<whitequark>
it has given me no headache whatsoever for years now
<rqou>
aaand i accidentally caused firefox to forget my previous session
<lain>
hrm so honestly, those of you who use *nix for desktop/workstation, what distro, what wm, and how's the ratio of "fixing *nix" to "actually using it to do work", in your experience?
<rqou>
rip 300+ tabs :P
<lain>
(honestly wondering, not trying to incite a riot)
<lain>
:P
<rqou>
also, awesome gained a new bug where reloading it somehow causes the titlebar to double in size
<whitequark>
lain: debian, i3, maybe ten minutes of fixing *nix per week at most
<lain>
looks like i3 also specifically targets bsd, that's good news for me
<lain>
it's not just a linux wm that has been ported over
<lain>
I wonder if there's anything good on the wayland side of things yet
<whitequark>
I also use i3 with kde junk for the rest of desktop functions
<whitequark>
wifi, etc
<whitequark>
most of those 10 minutes is restarting plasma when it hangs :p
<whitequark>
(wayland) i think there's a port of i3
<lain>
wifi is another pain point for *nix imo, but with gui tools it's not so bad
<lain>
wayland seems "better," but I need to look into it more, I don't know the real pros/cons
<lain>
other than "wayland lacks double-digit years of cruft"
<whitequark>
wayland is fairly sane as far as display servers go
<whitequark>
my main complaint is that stability of intel GPUs actually seems to have went down with more recent kernels and/or more recent chromium
<whitequark>
until I told chromium to not use GL for most stuff I was getting daily GPU hangs
<lain>
ouch
<whitequark>
having them bring down the entire desktop would be... unhelpful
<whitequark>
nevertheless this would not hold me on x11
<rqou>
hmm i can get ff to use intel gpu stuff and it works fine
<whitequark>
rqou: i think it's something about chromium
<whitequark>
it also leaks gpu buffers like a bucket
<whitequark>
fun fact: OOM killer doesn't really consider GPU buffers as "used memory"
<whitequark>
fun fact #2: overallocating GPU buffers doesn't result in other memory going into swap
<whitequark>
it took me *months* to *diagnose* the problem because the system just ground to a halt every time that happened
<lain>
hahaha
<whitequark>
fucking millenials and their programmable graphics pipeline
<rqou>
the only problem i seem to have with intel gpus is awful vsync tearing
<whitequark>
wayland is literally a solution to that
<rqou>
hmm i should clean my ff profile occasionally
<rqou>
i have tons of experimental flags enabled
<cyrozap>
lain: I'm running Arch Linux, i3 window manager, and Intel graphics, and I spend *maybe* a few hours a month fixing package upgrade issues, but that's to be expected since Arch is rolling-release.
DocScrutinizer05 has quit [Disconnected by services]
DocScrutinizer05 has joined ##openfpga
maaku has quit [Quit: No Ping reply in 180 seconds.]
maaku has joined ##openfpga
<whitequark>
oh god
<whitequark>
I just realized that Arch Linux's logo is an, well, arch
<whitequark>
I've learned about Arch Linux at *least* six years ago
pie_ has quit [Ping timeout: 240 seconds]
<lain>
oh shit I never noticed that either
<whitequark>
I tried to make a joke about an arch and then it hit me
<whitequark>
I blame that one friend with Arch who insisted on calling it "Arse Linux"
<rqou>
i always got the impression that the only* difference between arch and debian sid was that arch was used by weebs and debian was used by germans (*not actually only :P )
<whitequark>
I don't even know the deal with debian and germans
<cyrozap>
rqou: lol
<cyrozap>
One of the primary reasons I've stuck with Arch is that it lets me write my own package-building scripts, so everything not in the official repos can still be installed and removed cleanly.
<rqou>
quote from my roommate while playing XCOM 2: "step one, find the turret on Shodan" :P
amclain has quit [Quit: Leaving]
scrts has quit [Ping timeout: 252 seconds]
scrts has joined ##openfpga
kuldeep_ has quit [Remote host closed the connection]
kuldeep_ has joined ##openfpga
azonenberg_work has quit [Ping timeout: 240 seconds]
kuldeep_ has quit [Remote host closed the connection]
kuldeep_ has joined ##openfpga
Bike has quit [Quit: leaving]
massi has joined ##openfpga
pie_ has joined ##openfpga
promach has quit [Ping timeout: 276 seconds]
<whitequark>
lol
<cr1901_modern>
whitequark: Good idea to stay away from r2. It's UI completely violates principle of least surprise for me
<lain>
cr1901_modern: the web ui?
<cr1901_modern>
lain: No, the command line. Haven't tried the web ui, and tbh don't really care to. How to use r2 is underdocumented anyway. And many of the older archs (which is what I'm most interested in) are horribly broken.
* cr1901_modern
is reminded he owes them a test case tho
pie_ has quit [Ping timeout: 240 seconds]
promach has joined ##openfpga
pie_ has joined ##openfpga
pie_ has quit [Changing host]
pie_ has joined ##openfpga
<cyrozap>
cr1901_modern: Ugh, the 8051 disassembler uses tons of vsprintf's, which *segfaults* when compiled with _FORTIFY_SOURCE=2 (default on Arch, meant to harden code against certain attacks).
<cyrozap>
I've been using Hopper for ARM RE and it's been really useful. The only thing I don't like about it is that it isn't FOSS :P
<cr1901_modern>
For SNES work I use someone else's disassembler, which, far as I can tell, is superior to r2's (and can handle the SNES's really stupid addressing scheme)
pie_ has quit [Ping timeout: 240 seconds]
pie_ has joined ##openfpga
promach has quit [Ping timeout: 276 seconds]
promach has joined ##openfpga
pie_ has quit [Changing host]
pie_ has joined ##openfpga
pie_ has quit [Ping timeout: 245 seconds]
pie_ has joined ##openfpga
eduardo__ has joined ##openfpga
eduardo_ has quit [Ping timeout: 240 seconds]
m_t has joined ##openfpga
massi has quit [Quit: Leaving]
scrts has quit [Ping timeout: 255 seconds]
scrts has joined ##openfpga
amclain has joined ##openfpga
digshadow has quit [Quit: Leaving.]
kuldeep_ is now known as kuldeep
Bike has joined ##openfpga
carl0s has joined ##openfpga
azonenberg_work has joined ##openfpga
carl0s has quit [Quit: Leaving]
digshadow has joined ##openfpga
kristianpaul has quit [Ping timeout: 258 seconds]
kristianpaul has joined ##openfpga
kristianpaul has joined ##openfpga
kristianpaul has quit [Changing host]
pie_ has quit [Changing host]
pie_ has joined ##openfpga
m_t has quit [Quit: Leaving]
<rqou>
lain: I don't know if you're allowed to say anything about this, but do you know why for Skylake systems ECC is only available with the C-series PCH and not the consumer Z/B/H ones?
<rqou>
i though the memory controller was supposed to be in the CPU so this shouldn't matter?
espes_ has quit [Ping timeout: 240 seconds]
digshadow has quit [Quit: Leaving.]
digshadow has joined ##openfpga
<lain>
rqou: I haven't looked but most likely it's just some silly product differentiation thing
<lain>
as best I can tell all of them have ecc, but it's fused off in consumer chipsets
<lain>
or at least, the ones that also have embedded variants almost always have ecc
<rqou>
TIL my minifridge has terrible temperature evenness
<rqou>
the eggs i had stored near the top of the fridge are frozen
<rqou>
but the milk in the bottom is still perfectly liquid
<rqou>
for the europeans who are confused: in the USA eggs need to be refrigerated because they're washed and have their protective membrane removed
<lain>
lol
<rqou>
lain: i'm confused why the chipset affects ecc support at all when the memory controller is in the cpu and not the chipset
<lain>
rqou: no idea, guessing it's just differentiation? is one an "embedded" chipset and the other consumer or something?
<lain>
intel does weird stuff
<rqou>
C series are the "server" chipsets that support xeon (product differentiation in the ME and/or microcode)
<rqou>
afaik Z/B/H are the normal consumer stuff
<rqou>
this is special for skylake too
<rqou>
on older gens you can run xeon cpus on a X-series "HEDT" (high end desktop) consumer chipset
<rqou>
also, some X99 (broadwell) chipset motherboards support ecc and some don't
<rqou>
idk how that even happens
<lain>
in general intel doesn't share the "why" of anything, they barely share the "what" :P
<rqou>
unless you get access to a "super-sku" pch? :P
<rqou>
or do those no longer exist?
<lain>
wuzzat
<rqou>
or are you not allowed to say? :P :P
<lain>
like, I think you're vastly ovwrestimating how much info is in these nda docs :P
<lain>
I barely have enough info to design a functional board here
<rqou>
on some older gens there were "super-sku" PCH chips that had no fuses set but could impersonate any of the different pch variants using some bits set in the bios flash
<lain>
ahh neat
<lain>
yeah I would have access to that info
<lain>
but I've only been working with SoCs so far
<lain>
so, chipset on same die
<rqou>
actually same die or just same package?
<rqou>
as a multi-chip module?
<lain>
same die
<lain>
it's all in one die
<rqou>
wow i didn't know that existed now
<lain>
yeah for a couple generations now
<rqou>
even the Boot Guard efuses? :P
<rqou>
or not allowed to talk about that either? :P :P
<lain>
boot guard? which is that
<rqou>
the feature where "the hardware" (the ME?) does a RSA signature check on the bios before the x86 even boots
<lain>
oh
<lain>
it changee names often I guess
<rqou>
naming things is one of the hardest problems in CS after all :P
<lain>
mmm I forget the details but the socs I've been working with do have some pretty advanced boot protection stuff
<rqou>
two hard problems in CS: cache invalidation, naming things, and off-by-one errors :P
<lain>
in particular, ME boots before the x86 cores, it can e.g. load microcode updates before ever executing code on the x86
<lain>
lol yeah
<rqou>
yeah i know about that part
<rqou>
waiting for the reveal of an intel sigcheck bug in the me bootrom :P
<lain>
same
<rqou>
unfortunately intel isn't nintendo so this won't be so easy :P :P
<lain>
or sony, yeah
<rqou>
re naming things: at the last mtvre i discovered that "session" is one of the most overloaded terms ever in Linux
<lain>
lol
<rqou>
i thought that "session" meant "the thing you get when you call setsid()"
<rqou>
but apparently that's not what anybody means most of the time when they say session
<rqou>
apparently the "Linux desktop stuff" has multiple other ideas of what a "session" is
<rqou>
and some of them are the same some of the time
<lain>
I wonder how much of the behavior of apollo lake socs is fused vs. bios configuration
<lain>
I haven't looked at the bios writers guide much, but I suspect a lot of the behavior is fused
<lain>
(like, the difference between the apollo lake pentium/celeron consumer skus vs. the atom embedded skus)
<rqou>
i mean, there's the super sketchy concept of "soft straps"
<lain>
they're the same die afaik, and same package, except that the atom skus have integrated heat spreader (IHS)
<lain>
yeah
<rqou>
btw it's surprising how much of the random flash config documentation leaks all the time
<lain>
atom skus support ecc, but the cpu and gpu frequencies are different, as are the frequency scaling
<lain>
although that reminds me, frequency scaling stuff is handled by some mcu iirc
<rqou>
since when did atom skus support ecc?
<lain>
yet another processor in the processor
<lain>
rqou: looong time
<rqou>
my existing one doesn't
<lain>
there's embedded atom skus
<lain>
and server atom
<rqou>
oh herp derp
<rqou>
i had a j1900 as my old nas cpu
<rqou>
that's a "celeron"
<lain>
ah eyah
<rqou>
but it's a bay trail
<lain>
meaning it's an atom
<lain>
under a different name
<lain>
from what I've seen, celeron is the low-end consumer atom, pentium is the high-end consumer atom, and atom is the embedded/server atom, depending on sku, but there are *also* consumer atoms called atom, which is somewhat confusing :P
<lain>
the difference between "consumer atom" and "consumer atom branded as celeron/pentium" seems to be price, not feature
<rqou>
so before the j1900 i had a d510 atom as the nas box which also does not have ecc
<lain>
yeah
<lain>
only the embedded and server atom skus have ecc as an option
<rqou>
so d510 is neither?
<lain>
guess not
<rqou>
but the "pentium" g4400 has ecc
<lain>
ark says embedded options are available for the d510