#qi-hardware IRC log for Sunday, 2013-10-13

kristianpaulfreertos? :o00:47
DocScrutinizer05o.O02:40
DocScrutinizer05rogue site?02:41
apeleteahah, just built jz-3.11 vanilla (without my musb hacks), and the pagealloc memory corruption appears in it too: http://paste.debian.net/57073/17:37
larscnow that is unexpected17:38
apeletetried to debug the thing this whole afternoon, and after reading backtraces in gdb I suspected that the problem was not in musb and I was right17:38
apeleteI don't know where it's coming from exactly though, cause gdb won't let me break on the exact function that produce the memory corruption error :-(17:40
larscyou should build your kernel with KALLSYMS17:40
larscthen we'll see where it is triggered17:40
apeleteyeah, forgot to build jz-3.11 with KALLSYMS, but I think I recorded a boot session with more information a few days ago17:42
apeletelarsc: there you have it, recorded on october 10 -> http://paste.debian.net/57081/17:43
larschm, maybe your memory is bad17:45
larscso it should all be 0xaa17:45
larscbut you got some 0xae17:45
larscin there17:45
apeletelarsc: what do you mean by "memory is bad" ?17:46
apeleteyou mean my nanonote may be faulty ?17:47
apeletethe memory corruption only shows up if DEBUG_PAGEALLOC is enabled by the way17:47
larscyes17:47
larscit will only be detected if DEBUG_PAGEALLOC is enabled17:47
larsccause there is a function checking if it is 0xaa17:47
apeletelarsc: hmm, since I'm booting from sdcard, maybe it's the card that's faulty, what do you think ?17:50
larscno17:50
apeleteah no, it must be the nanonote memory17:50
apeletepagealloc is done on memory, didn't think before asking :)17:51
larscyep, there is one function that sets the memory to all 0xaa17:51
larscwhen the memory is freed17:52
larsconce it is allocated again it checks whether it is still all 0xaa17:52
apeleteyes, I think it's check_poison_mem() from debug-pagealloc.c you're talking about17:53
apeletestill, that's bad. I hope the device won't be damaged though17:54
apeleteI need it to continue hacking on musb :-(17:54
larscit could also be a user after free somewhere17:55
apeletewill ignore that problem for now, and focus on getting the glue layer init function right17:58
apeletelarsc: this is what I'm trying to do currently:18:00
apelete/* The musb_platform_init() call:18:00
apelete *   - adjusts musb->mregs18:00
apelete *   - sets the musb->isr18:00
larscok18:00
apeletelarsc: any advice on how to proceed here ?18:00
larscSo mregs should probably come out of the mmio platform_resource18:01
larscsame for the interrupt number (platform_get_irq())18:01
larscnot sure about what the isr is supposed to do exactly18:02
larscor if there is some kind of musb default isr18:02
apelete<larsc> So mregs should probably come out of the mmio platform_resource18:03
apeleteI don't get it, could you be more specific please ?18:03
larscI think you are supposed to call musb_init_controller18:07
larschm, no actually18:07
larscI think you don't need to do anything for mregs18:07
larscas it should already be correct18:07
apeleteso that leaves musb->isr to take care of18:08
larscAs far as I can see none of the existing drivers implement the init callback18:09
larscah, no they do18:09
larscso some init functions add a offset to mregs18:10
larscbut I don't think we need to do this18:10
larsc about what the isr is supposed to do exactly18:11
larsc20:02 < larsc> or if there is some kind of musb default isr18:11
larscups18:11
larscux500_musb_interrupt seems to be a good example18:12
apeletewill try that18:14
apeletelarsc: ok, will try to get the interrupt settings right, then we'll see. thanks for the advices18:17
wpwrakDocScrutinizer05: btw, some more EE design verification: http://lists.en.qi-hardware.com/pipermail/discussion/2013-October/010329.html20:41
apeletelarsc mth: just added interruot regs initialisation in musb init function -> http://seketeli.fr/git/~apelete/qi-kernel.git/commit/?h=jz4740-musb&id=6de03fc47ef850a34b4bcdf1d63f7b9fb015b94a22:48
apeletebut something's amiss -> http://paste.debian.net/57163/:22:49
apelete[    1.320000] musb-hdrc musb-hdrc.0.auto: can't request region for resource [mem 0x13040000-0x1304ffff]22:49
apelete[    1.330000] musb-hdrc: probe of musb-hdrc.0.auto failed with error -1622:49
apeleteenabled a few printk to get a clearer picture -> http://paste.debian.net/57164/:22:50
apelete[    1.560000] bus: 'platform': really_probe: probing driver musb-hdrc with device musb-hdrc.0.auto22:50
apelete[    1.570000] musb-hdrc musb-hdrc.0.auto: can't request region for resource [mem 0x13040000-0x1304ffff]22:50
apelete[    1.580000] musb-hdrc: probe of musb-hdrc.0.auto failed with error -1622:50
apeleteseems like there is still something wrong with platform data...22:52
apeleteanyway, need some sleep right now, maybe we can discuss it in a few hours22:56
pcercueithis memory area has probably already been requested elsewhere22:58
Yogi_Bearinvitation http://www.tatuuu.com.br ONLY QI-HARDWARE23:14
DocScrutinizer05wpwrak: sorry, just wake up23:37
DocScrutinizer05woke23:38
DocScrutinizer05ugh a wall of text23:38
DocScrutinizer05wpwrak: sorry, takes a while to what that NCIS in TV and then wait til the coffeee kicks in23:53
DocScrutinizer05to watch*23:53
--- Mon Oct 14 201300:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!