#tyndur · 2010-04-06
191 Nachrichten · Seite 1 von 1 · Original-CSV herunterladen
- eddybXanClic: started to revive my ehci driver
- eddybno irq on virtualbox :((
- XanClicVBox has a bug there
- eddybit has?
- XanClicTry the IRQ below
- eddyblike 10 instead of 11?
- XanClicExactly
- eddybin vmware it goes pretty well
- eddybi get an irq when a device connects...
- eddybbut the schedule won't start
- eddybalso, it reports frame list rollover, but i got periodic schedule disabled(maybe i've switched the bits by mistake???)
- eddybthe interrupt thing makes no difference :|
- XanClicWell, for me it did ;-)
- XanClicAnd for some others I know, too
- eddybwtf?
- eddyb3 pci devices report interrupt number 11 :O
- XanClicNever heard of IRQ sharing?
- eddybhow much memory space OHCI uses?
- XanClicHm, 4 kB, I think
- eddybs/memory/mmio\/io/
- eddyb0x1000
- eddybexactly the number of mmio space one of the other devices with irq 11 got
- eddybthe device with irq 10 got 0x20 of io, 0x40000 of mmio and another 0x4000 of mmio
- XanClicIt's a network device
- XanClicI already checked tha
- XanClict
- eddybbut i don't get any irq, even if i add "-1" :(
- eddyboh, and also "<eddyb> okey, this is really stupid. but the beggining of my memory region i use in ehci driver got "SQLite format 3" :|" << this is quite strange and i think it's related to pedigree
- eddybdon't tell me...
- eddybthere's only a handler/irq allowed in pedigree
- XanClicThat's what I'd call a "fail" *g*
- eddybwhich can be == fail
- eddybheh, in the same time :P
- eddybthe problem is...
- eddybfrom the handler registration and first request, there's just a couple of instructions :|
- eddybBINGO
- eddyb+1 means 12, which is mouse irq. the good thing is that i can move the mouse and make it complete transactions :P
- eddybdamn non-working irq...
- eddybit does set the connect status changed bit, and i can see new devices
- eddybXanClic: afaik, you haven't touched ehci yet... you had problems with ohci, then?
- eddybbtw, no unhandled irq showed up
- XanClicNo, I've also touched EHCI
- eddybaah
- eddybno irqs at all
- eddyb== nothing reached the ISR
- eddybso it's more internal, if there is a problem
- eddybargh, linux can ahdnle it
- eddyb*handle
- XanClicOf course
- XanClicLinux remaps the IRQs
- eddybxD
- eddybvia some PCI nifty config stuff, i suppose?
- XanClicVia APIC
- eddybah, ok
- eddybapic is kinda unfunctional in pedigree
- eddybhmm...
- eddybso it can be that without that remapping you say linux is doing, no interrupt from ehci will reach even the ISR ?
- XanClicYes, because VBox gives you the wrong IRQ number
- XanClicAnd it is 10 for sure, it's not just me who detected that *g*
- eddybbut as i said, no interrupt reach PIC driver
- eddyb*reaches
- XanClicPedigree's got a driver for VBox' network card?
- kwolfe1000
- eddybsorry, afaik, i've enabled no network card
- kwolfXanClic: Paloxena has the same driver once you apply Matt's latest patch
- eddyband i made PIC driver print every irq(except 0 and 8)
- XanClickwolf: And once I've implemented CDI, yes ;-)
- eddybXanClic: so, what now? i told you it doesn't even reach ISRs
- XanClicWell, I don't know.
- XanClicYou have several problems to which I just answer "works for me" ;-)
- kwolfXanClic: I was talking about one of the Paloxena versions that have CDI ;-)
- XanClickwolf: But those don't have USB ;-)
- kwolfUntil you finish USB for CDI. ;-)
- XanClicI guess, that's no repeat-until but a repeat-forever *g*
- XanClici.e., I doubt it'll be ever finished
- eddybgood
- eddybotherwise i'd have to throw my work into trash://
- kwolfWell, yes, wrong word. An OS is never finished. ;-)
- XanClicNo, I meant I think I continue with my CDI.storage attempts ;-)
- XanClic(Triple I, yeah)
- kwolf*sigh* Why that?
- XanClicI don't know how to implement USB efficient when using different address spaces
- XanClicI don't even know how to implement it inefficiently *g*
- kwolfWe probably need to have you around at the next hackathon to sort things out
- kwolfShouldn't be all that hard, though
- XanClicYou won't accomplish that *g*
- eddybXanClic: OOI, on which version of virtualbox irq 10 works for you?
- XanCliceddyb: 3.1.4, afair
- eddybthat's quite recent :(((
- kwolfXanClic: I meant the implementation. I'm afraid getting you there is too hard indeed. ;-)
- eddybdamn poor radio internet :((
- eddyb3.1.2 r56127 here
- XanClicI had the same problem in 3.1.2, too ;-)
- eddybXanClic: do you have any source code for reference?
- eddybi'm getting sick of vm bugs
- XanClichttp://gitorious.org/paloxena/kernel3/blobs/master/src/kernel/drivers/network-controller/ehci.c maybe
- XanClicAnd http://gitorious.org/paloxena/kernel3/blobs/master/src/kernel/drivers/network-controller/ehci.c#line140 in particular for VBox *g*
- eddyb"(NN) e1000: Fuehre Reset der Karte durch" << stuck at there :|
- eddybwhy is it in network controllers?
- XanClicBecause all my USB HCs are network controllers for the OS
- XanClicThey send packets, just as a network card does *g*
- Homixhm funny^^ ;)
- eddybpci_outw(pci->bus, pci->dev, pci->func, 0x04, 0x06); << what does this do?
- XanClicIt writes 0x06 into the PCI configuration space at offset 0x04
- XanClicA word
- eddybok, but what's the effect?
- XanClic=> Enabling bus master capability and MMIO
- eddybah, k
- eddybdelay() does delay in ms?
- XanClicNo, in µs, afair
- XanClicAnd heart_beat_count is the system timer which is incremented every ns
- eddybare those waits really necessary?
- XanClicWhy not?
- XanClicThose are done just once
- eddybno, i mean not having them can cause problems?
- XanClicWell, I don't know
- eddybif you'd see my code, you'd say C++ makes things go water shrink :P
- eddybexcept for the waits and the big C stuff, i see nothing different
- eddybhmm...
- eddybdo waits do wonders?
- eddybbut still no irq :((
- Jarvixhi
- XanClicmoin, Jarvix
- quokmoin :)
- XanClicmoin, quok
- kwolfMoin quok
- eddybgreat, manually waiting, i get XactErr set :(
- XanClicNice, I get that one, too ;-)
- eddyblol?
- eddybVirtualBox?
- XanClicYe
- XanClics
- XanClicThough I receive an interrupt
- eddybwhich i don't :/
- eddybhmm, maybe i can make pedigree boot from this usb stick...
- eddybit would really help :|
- eddyb(considered now we have on screen log - half of it, actually - :P)
- eddybbtw, what ping really does in ehci? it's like instead sending a real packet, it sends just something small to see if the device responds?
- XanClichttp://wiki.osdev.org/USB#PING_Transaction_Protocol
- eddybi have specfobia :P
- XanClicThat's no spec
- eddybokey, i've read it before
- eddybi just wanted to see if my understanding of it is accurate :P
- XanClicWell, it's not whether the device responds but whether it's able to accept (or send) data
- eddybbut can it be used as a ping facillity? as in test if the device can be reached
- XanClicNo
- XanClicWell, you could, if you were able to send raw USB packets, but you aren't
- eddybah, so it's just a protection mechanism :(
- XanClicProtection?
- XanClicIt's been introduced to remove useless traffic
- Jarvixhi
- eddybbrb, testing a bootable pedigree
- eddyb*usb stick with pedigree
- eddybwell
- eddybI can boot pedigree from usb stick
- eddybbut
- eddyball UHCIs and the EHCIs got IRQ 255 :((
- eddybXanClic: does that mean I need APIC driver to make it work?
- XanClicDon't know
- eddybpedigree got some unused LAPIC routines, but it looks like they're just for IPI :|
- XanClicYou'd need I/O APIC for PCI
- eddybgreat, IoApic.cc got only the license header and the #include "IoApic.h" :(
- eddybdid "<eddyb> great, IoApic.cc got only the license header and the #include "IoApic.h" :(" reached the channel?
- XanClicYes
- eddybor did anyone say anything while i left?
- XanClicNo
- eddybokey
- eddybi've just read http://markmail.org/message/hjmwydqx6bk2jm6e
- eddybmaybe it'd help playing with the BIOS?
- eddybheh
- eddybBIOS update on this crap really did wonders
- eddybit now assigns irqs properly
- eddybbut pedigree still hangs after some EHCI message :|
- eddybi mean freeze
- eddybi got a while(portStatus&PORT_RESET); in my code, can it freeze pedigree so not even a F12(used to get into debugger) will get thru?
- eddybdammit, i sound like a total n00b
- eddybsure it can. at that point interrupts are disabled by something in ata and i enable them only after XD
- eddybwhich means there's a problem at port reset, btw
- XanClicSorry to ask those questions, but you marked portStatus as volatile and the MMIO is uncachable?
- eddybask matt/james
- eddyb-EWORKSINEMULATORS
- eddybso i guess yes
- XanClicSo it's at least volatile
- eddybah, and FYI, empty ports always return 0xffffffff, so it sounds like some other error
- eddyblooking over my usb-hcd/main.cc file, i saw probeXhci and it made me laught :))
- taljethXanClic: How are you sure that it's volatile? Emulators usually reset that bit immediately, so it might just have the right value from the beginning. (Well, qemu usually does, no idea for the rest of them)
- XanClicHm, well
- XanClicRight
- eddybi added some more debugging, brb
- eddybmy second guess
- eddybthe second port (with the usb stick) got 0x1803