<!-- Some styling for better description lists --><style type='text/css'>dt { font-weight: bold;float: left;display:inline;margin-right: 1em} dd { display:block; margin-left: 2em}</style>
ParkerR: *umm
<br> *device
<br> *disconnect
<br> Shit I cant type today lol
<br> I remember an old xc5000 bug I reported where it was freeing a non existant instance and had a similar crash (but froze up the entire system). Wonder if it's possibly related
b-rad: with what device?
<br> 4.17.2 contains broken 01595 dualhd support
<br> the patch i pushed to fix that didn't make it until 4.18-rc1
ParkerR: It is dual HD
<br> Ahh good to know
<br> b-rad, Thanks!
b-rad: you'll notice it announces ts2, but never actually creates it
<br> so the tear down poops itself
<br> s/create/register/
ParkerR: Heh here's the xc5000 I was thinking of https://github.com/torvalds/linux/commit/856260a57cdfa5c121c7b7a6e816409bab07885c#diff-c2b7d8fe59e79b9f245c3b677e97d2a1
<br> Also I think this was related https://github.com/torvalds/linux/commit/4961a5323f5d873e2170c5ef4f48538930e6df3e#diff-c2b7d8fe59e79b9f245c3b677e97d2a1
b-rad: i don't think thats, related, but i had to sort out something similar in lgdt3306a and si2157 with my analog stuff
ParkerR: No I meant to the other xc5000 lol
<br> Just reminiscing on older bugs
b-rad: good ole bugs
<br> if you build 4.17 yourself my fix is a single line
ParkerR: b-rad, So did 4.17.2 partially pull in the second tuner patches causing that or was the break unrelated?
b-rad: they were merged correctly, but then em28xx had stack reduction retooling and an in the mix an important variable somehow got swapped with hardcode value
ParkerR: Ouch
<br> Oh so the dual tuner is merged in 4.17.2 or 4.18 rc?
b-rad: well technically it's merged in 4.17.2, just totally broken :/
ParkerR: That single line fix or or is there something else in 4.17 killing it?
b-rad: 4.18 rc it's fixed, i just confirmed that this morning for someone else
ParkerR: *it or
<br> Cool cool
<br> I might doa comile from git
b-rad: nope, the single line fix should be it
ParkerR: b-rad, Congrats on the merge! (Albeit a rocky one it seems lol)
b-rad: c'est la vie :)
<br> analog next and that one is still being difficult on ryzen
<br> i'm starting to very much dislike ryzen pcie
ParkerR: The DUal HD in particular?
b-rad: quadhd's now
ParkerR: Ahh
b-rad: dualhd does not have analog
<br> all our cx231xx stuff does now though
ParkerR: Ahh yeah. I meant in trerms of issues with Ryzen
<br> All PCIe based?
b-rad: usb is fine on ryzen, only pcie is a hassle
ParkerR: Ahh ok
b-rad: cx231xx is usb
<br> SoloHD is basically a 955Q without analog
ParkerR: That's weird. Never even thought of newer PCIe implementations breaking stuff
b-rad: ryzen is too fast
ParkerR: TOO AST
<br> &FAST
b-rad: xeon exhibits similar issues
ParkerR: Ahh so the timings arent quite right?
b-rad: those two platforms act a little racey, so it definitely feels timing related
ParkerR: b-rad, And we're off to the races :) https://i.imgur.com/GCEMEOU.png
<br> b-rad, :/ https://gist.github.com/parkerlreed/3bdf0ed6b0bf907d6c7c42e13840f204 So the first issue was when I tried starting a second stream the first died completely. Couldnt restart. Unplugged and I got that. No error message related to the first card ying upon trying to use second
<br> Linux inspiron15 4.18.0-rc1-g81e97f01371f #1 SMP PREEMPT Wed Jun 20 12:06:01 EDT 2018 x86_64 GNU/Linux
<br> *dying
b-rad: hmmmm interesting, i'll add this to my queue
ParkerR: Thanks. Yeah seems odd. Fresh pull from git master
b-rad: i've been heavily testing 01595 lately using the tip drivers from last week, albeit on backported systems, and have not seen that
<br> and i specifically test for disconnect oopses because i hate them
ParkerR: So yeah I have 95. Weird
<br> Heh
<br> b-rad, Actually whats the single line fix from 4.17.2 so I can at least check my sources and see if its there
b-rad: https://git.linuxtv.org/media_tree.git/commit/drivers/media/usb/em28xx?id=01affb000e00cfa0a9e9954476ef50962eb8b168
ParkerR: Yep so thats right
<br> I see another dvb->i2c_client_tuner = dvb_module_probe("si2157", NULL, where the i2c is still hardcoded to 0x60. Is that intended?
<br> Line 1245
<br> Err 1247
<br> https://git.linuxtv.org/media_tree.git/tree/drivers/media/usb/em28xx/em28xx-dvb.c?id=01affb000e00cfa0a9e9954476ef50962eb8b168#n1247
b-rad: it's correct, since the dualhd has a second tuner it has another pid besides the default value
<br> for the second tuner
ParkerR: Ahh ok
<br> b-rad, Following the loginc https://gist.github.com/parkerlreed/430f95b76dd4e1a3513d36bcc5257868 The first disconenct says #1 for the first but nothing for the second. And the second freeing device seems to be where it crashes.
<br> *logic
b-rad: well that sure sounds to me like that patch isn't applied
ParkerR: Is there a separate verbose mode I can do for em28xx or do I have to do a whole debug kernel?
b-rad: do you see registration of the second adapter?
ParkerR: Yeah thats the weird part
<br> Umm lemme see
<br> Didnt realize my first paste was cut off, woops https://gist.github.com/parkerlreed/732af6a5827b34df6101b56113798732
<br> [ 136.974704] em28xx 1-3:1.0: DVB: registering adapter 1 frontend 0 (LG Electronics LGDT3306A VSB/QAM Frontend)...
b-rad: yeh, looks fine
<br> i'll try and repro and fix this up if i can
ParkerR: And my line 1393 is deifnitely addr so I know that patch was there
<br> Cheers
tjader: What do UCB, postBER and preBER mean on dvb-fe-tool -m?
b-rad: @ParkerR, confirming your issue
<br> i see oops on unplug
<br> i also see crazy amount of i2c errors the moment i do a second w_scan
<br> i have another couple em28xx devices i think, i'll try them, but this is abnormal and new to me
ParkerR: b-rad, Ahh damn. Might be what I ran into trying to tune the second tuner
<br> Thanks for looking into it
b-rad: no i2c failures on tuner0 during scan by self, 3 i2c failures on tuner1 scan by self, total shtf when accessing both
ParkerR: Yep. I tried using both in mpv when it shat out
b-rad: same result using media_tree
<br> 4.15 with media tree dated april 18 is fine, as is 3.10 and 4.1 using backported drivers from tip media_tree
ParkerR: b-rad, Any luck?
b-rad: my 4.15 build did have the disconnect oops, but both tuners worked
<br> looking into now