#qi-hardware IRC log for Thursday, 2012-11-08

arossDOTmewpwrak: The problem seams to be the GPT partition table. I reformatted the 2gb sdcard to ext4 with its msdos partition table. My NN could mount it and mount another 32gb sdcard formatted ext4, GPT00:10
arossDOTmestill do post any replies and I will look at them one day  in the log. thanks bb02:20
wpwrakarossDOTme: sounds a bit strange that the partition table would have such an effect. but then, i've never experimented with GPT so far, so maybe there's a world of fun waiting out there of which i don't know yet :)03:04
viricWhy the nanonote does not implement a usb-storage device? only usbnet14:24
viricor can I switch that easily?14:24
pabs3on the openmoko its just loading a different gadget driver, probably the same on other devices14:25
viricso, either one or the other14:26
viriccan't the kernel do as if it was a hub with two devices?14:26
pcercueiviric, it can14:31
pcercueithe problem of mass-storage mode is that you have to dedicate a full partition to it14:32
pcercueiand the OS has to be apart14:32
pcercueinow the problem is that the jz4740 USB driver seems to have problems with that gadget driver14:33
viricyes, clear. I remember seeing that setup in some devices14:33
viricah really? meh14:34
pabs3viric: theres a composite usb device stuff in Linux but it is quite limited, even on Android you can't have usbnet plus adb at the same time14:37
arossDOTmewpwrak: Opps "My NN could mount it and mount another 32gb sdcard formatted ext4, GPT" I mean 32gb ext4 msdos sd card mounted but the one with GPT did not mount.14:52
wpwrakah, so GPT is what causes the trouble14:53
wpwrakand not ext414:53
arossDOTmeyes14:53
arossDOTme2:** am for you14:53
arossDOTmeIt would be awesome to by able to connect a usb flash drive(s) :D.So is it the case of some one having the time and will to fix this driver?14:55
wpwrak"fixing" the driver may be a tad difficult. the hardware doesn't support usb host. you could perhaps write a sw-based usb host driver, but so far, nobody has done that.14:57
arossDOTmeok14:57
wpwrak(hw) that is, the chip would have a usb host interface. but it's not connected to the outside world. and connecting to it would be very difficult.14:58
paul_boddieI seem to recall that the combination gadget drivers can't reserve enough endpoints from the USB system, or something like that.15:20
paul_boddieThey probably have to advertise multiple endpoints with their own descriptors, and the jz4740 won't let them. At least that's what my interpretation of the problem was.15:22
pcercueithe composite gadget works on jz474015:23
pcercueiI tried it with networking + serial15:23
paul_boddieIt didn't work for me when I tried it on the NanoNote. Maybe something has been fixed recently.15:24
wpwrakthe USB device controller has ... hmm ... manual says 5 or 16 EPs15:27
wpwrakerr, 6 or 16. let's see ...15:27
wpwrakapparently it's 1615:28
paul_boddieWell, I don't know why it didn't work, then. I was sure that the driver mentioned some really small number like ep0, ep1 (in), ep1 (out) and ep2 (bulk), or something.15:31
wpwrakthe manual claims that "The core has up to 3 IN endpoints and/or up to 2 OUT endpoints in addition to Endpoint 0."15:32
wpwrakbut then happily goes on explaining how to use all those 16 (!) endpoints15:32
wpwrakmaybe there's an additional restriction i didn't spot when skimming the pages15:33
paul_boddieOn the topic of USB Host, I was doing some stuff tunnelling USB/IP to a PC and also to a USB Host-capable Arduino device, but I became quite tired of (1) the Microsoft-level USB/IP protocol (it just bundles up kernel structures and sends them over the wire), (2) trying to issue ioctls from user-space on a Linux host, (3) debugging the driver (although I think I cracked that problem, mostly), (4) trying to get the USB controller on the other devi15:34
paul_boddie...(5) dealing with the fact that USB hubs require the Frankenstein treatment to have "normal" out-of-the-box hub behaviour like what you'd have for any sane interconnectivity technology.15:35
wpwrak;-)15:37
paul_boddieI do know a lot more about USB now, however. ;-)15:37
wpwrakusb is fun. gives you that warm fuzzy feeling that those doomsayers may have a point with hell and eternal damnation15:38
wpwrakbtw, a more straightforward approach could be to just do your own protocol, like i did with dirtpan15:42
viric16 EPs looks like a lot15:45
wpwrakyeah. dunno why, at the same time, they say that we only have 1+2+3.15:46
viricI didn't know that could have a hardware limit.15:47
viricmaybe more usb stuff can be delegated to the cpu?15:47
viricand then have as much EPs as wanted15:47
paul_boddieI guess it all comes down to buffers and latency. You could probably have a number limited only by the addressing, but the hardware would probably need to be capable of shoving the data in the right buffers in a timely fashion.15:48
wpwrakviric: you don't have access to the lower levels. so you'd have to implement the whole USB device in software. kinda like v-usb.15:49
viricv-usb?15:50
wpwrakif imgtfy worked better today, i'd give you a link ...15:56
viricif what?16:01
wpwraklmgtfy.com16:01
viric:)16:02
paul_boddiev-usb is that Atmel-based software-only implementation of USB 1.1, isn't it?16:04
viricah I think I remember it16:07
whitequarkwpwrak: http://lmgtfy.com/?q=v-usb17:46
whitequarkand I did not look up anything prior to posting this link.17:47
wpwrakyeah, i'm too lazy to type it off the screen. for some reason, copy & paste doesn't want to work.17:47
wpwrakhmm. now to get a cross-compilation environment for the ben with SDL and SDL_gfx in less than two hours ?17:49
wpwraklet's see if building a single package works ...17:50
wpwrakhmm. but no way to make a static binary :-(18:23
kristianpaulagain? ( cross-compilatio env)18:33
wpwrakit's an eternal struggle ...18:33
wpwraki have a working cross-compiler. now i tried to add SDL. that actually works, but ... i also need my binaries to be static (don't want to reinstall the whole ben)18:34
viricwpwrak: nix18:35
wpwrakwithout SDL, static binaries are fine. SDL, on the other hand, depends on stuff that doesn't support static binaries.18:35
viric:)18:35
wpwrakviric: the master of all niches ;-)18:35
viricI've been told that gcc from 4.5 to 4.7 has serious mips bugs18:36
virichttps://www.linux-mips.org/cgi-bin/mesg.cgi?a=linux-mips&i=20120906140843.GA17068%40linux-mips.org18:36
wpwrakhmm, fun :)18:37
viricI'm having lots of troubles with a mips64 kernel18:38
viricI'm building a 4.7.2 toolchain now18:38
arossDOTmeI refomated my 32gb card so its also msdos part tablet ext4 and it now mounts on my NN. BTW before it was GPT with protective MBR. in other words GParted defaults.21:06
arossDOTmewpwrak: ^21:07
wpwrakkewl. a reasonably happy ending then :)21:27
viricformatting sd cards?21:35
viricsd cards come with properly aligned partitions21:35
viricfactory determined. For optimal performance.21:35
arossDOTmeI only change the partition table to gpt becuase I like superor stuff but 1mb could not be used. With the msdos partion table I have no odd left over unused 1mb.21:38
arossDOTmeI don't know: "sd cards come with properly aligned partitions"21:39
arossDOTmeI assume that its nothing that can't be undone.21:39
viricyou could have noted where the partitions started :)21:45
viricand repeat that in msdos partitions :)21:45
viricThe manufacturers place the fastest flash memory where the FAT tables reside, that require most modification. That, based on the from-factory partitions.21:45
viric(fastest, or that can be written more times, or with more extra blocks, ...)21:46
viricAnd the rest of the flash is expected to be written linearly. If you break that (and does not use FAT for example :) you may have a bad performance.21:46
viricMoreover, if you don't align the partitions to the flash page size or the erase block size, you may have a much higher performance penalty21:47
arossDOTmeoh didn't know that. darn. I hate fat. oh dear21:47
viricthat also depends on the quality of the card, of course.21:48
viriclwn had a good page about it21:48
virichttps://lwn.net/Articles/428584/21:49
arossDOTmeumm when switching back to msdos part table I tried trying out gdisk to convert the table for gpt to msdos. that didn't seam to work. So I went for gparted change part table to msdos format ext4 and again disable the journal21:50
arossDOTmeby not work I mean gparted moan about it.21:50
arossDOTmethanks for the article. I will read that :)21:50
arossDOTmeSo what command,app do I use to make a note  partitions started? thanks for putting up with my newbieness.22:16
arossDOTmeI guess formatting back to fat32 ant going to tell me where the original partition started...22:18
arossDOTmeoh it took around 1hour to copy 15,16ish GB on to it. If I remember correctly :/.22:20
arossDOTmeThis would not have anything to do with my prev sd card getting files corrupted/errors? msdos ext4 disabled journal.22:24
arossDOTmere copying: I forget the numbers of what speeds to expect. should look them up, now I think of it...22:25
--- Fri Nov 9 201200:00

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