#qi-hardware IRC log for Sunday, 2013-01-06

jayaris this the right place to ask a harddrive question?07:33
Action: jayar taps the mic07:34
jayaris this thing on?07:35
jayartesting testing... civilance... civilance...07:35
jayarwell dangsauce07:36
jayarcarry on, idlers07:36
wpwraknowadays they have really earth-shattering insults. "idlers". boah.11:54
qi-bot[commit] kyak: ben-cyrillic: use phonetic keymap by default (master) http://qi-hw.com/p/openwrt-packages/fa463c313:08
whitequarkwpwrak: you know what makes me sad?13:21
whitequarkall these new high-speed interfaces. they're completely impenetrable for DIY13:22
whitequarkDVI, HDMI, USB>2 (or >= maybe), DisplayPort, SATA13:22
DocScrutinizer05and WTF?13:24
DocScrutinizer05friggin bot's statistics are useless13:24
larscwpwrak talks as much as everybody else combined13:24
larsckind of13:25
whitequarkwell not everyone, sum(2..5) = 113:25
DocScrutinizer05nfc what those statistics mean at all13:25
wpwrakdamn. my absolute majority is being threatened !13:37
wpwrakneed to talk more then :)13:38
wpwrakwhitequark: (fancy new interfaces) yeah, they're tricky. not completely impenetrable but considerably more difficult to play with13:39
whitequarkfortunately USB is as backwards-compatible as it is realistically possible13:40
wpwrakyeah, a pain at any speed, but at least manageable while slow13:41
whitequarkwhy? USB is pretty nice13:41
whitequarkwhen you are programming a device at least :)13:41
whitequark(well, theoretically it is fully backwards compatible, but I've seen more than once noncompliant device working on 2.0 and failing on 3.0.)13:42
whitequarkin fact let me check if v-usb-based USBASP still works...13:44
whitequarkif only I had a clue where the board is.13:46
whitequarkyeah works perfectly13:48
whitequark[1183600.307007] usb 3-3: new low-speed USB device number 32 using xhci_hcd13:48
whitequarkwell, kudos to USB Consortium then.13:48
whitequarkSTM32VLDISCOVERY: fail13:50
whitequarkthey abused USB Mass Storage on their programmer to display a 64k "block device" with a pseudo-FAT where four .url file is stored with links to the ST website13:50
larscwell they have to make sure it works13:50
larscthe USB consortium13:50
whitequarklarsc: there is more than one example when something ought to work, a consortium existed to ensure that, and it actually doesn't13:51
whitequarkfor example, I dunno, Bluetooth13:51
whitequarkthe fucking thing never works for me. every single time I need to use BT I spend hours trying to make it work and then usually give up.13:52
whitequarkbluez in Debian is completely unable to talk with bluez in Android. perfect.13:52
larscyea, it is always sad when that happens13:53
wpwrakwhitequark: USB mass storage for a virtual FAT is a very good idea. i'm actually surprised it seems to have taken so long to be implemented (NXP have it too)13:53
whitequarkwpwrak: I'm not if this was intentional or not on their side, but this device violates sheer amount of requirements in the spec13:54
wpwrakwhitequark: what sucks is that you can't have USB mass storage on a USB low-speed device13:54
whitequarkit's a high-speed device btw13:54
wpwrakah, like what ?13:54
whitequarkit hardly works even on 2.0, requiring a kernel patch and a quirk enable13:54
whitequarkjust to enumerate13:54
whitequark(the patch is in mainline for a long time, but still)13:55
whitequarkthen, the mass-storage port doesn't work in linux anyway13:55
wpwrakperhaps they had deadlines to meet, leaving no time to read those boring specs ;-)13:55
whitequarkbut you can abuse SCSI vendor-specific commands to actually program the target13:55
whitequarkand in 3.0, it is just stuck in an endless reset loop13:56
whitequarkwpwrak: oh, it's a high-speed device btw13:56
whitequark[1183860.020531] xhci_hcd 0000:00:14.0: xHCI xhci_drop_endpoint called with disabled ep ffff88010cd8078013:56
whitequark[1183860.020549] xhci_hcd 0000:00:14.0: xHCI xhci_drop_endpoint called with disabled ep ffff88010cd807c013:56
whitequark[1183860.179003] usb 3-3: reset full-speed USB device number 34 using xhci_hcd13:56
whitequark[1183860.190426] xhci_hcd 0000:00:14.0: xHCI xhci_drop_endpoint called with disabled ep ffff88010cd8078013:56
whitequark[1183860.190436] xhci_hcd 0000:00:14.0: xHCI xhci_drop_endpoint called with disabled ep ffff88010cd807c013:56
whitequark[1183860.349773] usb 3-3: reset full-speed USB device number 34 using xhci_hcd13:56
whitequarkI've seen similar behavior with Samsung's bootloader in their phones13:56
larscbut is that a bug in the device or in the kernel?13:58
wpwrakheh. perhaps they should have been less ambitious ;-)13:58
wpwrak(vendor-specific commands) the beauty of using USB storage is that this is about the only way to get data to a USB device that's OS-agnostic.13:59
wpwrakfor anything else you need some sort of driver13:59
whitequarkwpwrak: but they screwed them up too13:59
whitequarksomething with error codes, I don't remember, it was two years ago14:00
wpwrakwell, one you decide to foul it up, you may as well go all the way ;-)14:00
whitequarkso you need even more quirks, or libusb14:00
larscwpwrak: usb keyboard works as well ;)14:00
whitequarkoh yes, HID requests.14:00
whitequarkyou can even send them from an unprivileged account in Windows14:00
wpwraklarsc: how do you send a file over HID, in an OS-neutral way ? :)14:01
larsctell people to open up a texteditor, then send the characters ;)14:01
wpwrakhuh ?14:01
viricwith a proper keyboard layout chosen.14:02
wpwrakdoes HID have a snoop option ? e.g., for law enforcement :)14:02
wpwrakviric: the problem is that you're the device. so how do you see what other HID devices are doing ?14:02
larscwell you can of course only send one file14:02
viricno idea.14:03
whitequarklarsc: what?14:03
wpwraklarsc: and then it self-destructs ?14:03
viricin ps2 times, barcode readers were put between the keyboard and the computer :)14:03
whitequarkviric: you can do this with HID too14:03
whitequarkthere are examples for atmel chips on the net14:03
larscwpwrak: well you can send the same file multiple times14:04
wpwrakbar code is easy. pretent you're a keyboard and just send the numbers.14:04
larscyou need a push button on the device to start the transaction14:04
wpwraklarsc: wait ... how does the data get sent to the device ? i mean, what does it pretend to be and by what mechanism does it receive the data ?14:05
larscwpwrak: the device only sends data out14:05
wpwrakah, okay. that's trivial14:05
larscno backchannel14:05
wpwrakno firmware updates then14:06
whitequarklarsc: wpwrak implemented a firmware update via mass storage14:06
wpwrakback to USB storage and its full-speed requirement :-(14:06
larscexcept maybe using morse code on the capslock key14:06
wpwrakwhitequark: no, i haven't yet. but i'd want to. without paying the premium for full-speed.14:06
wpwraklarsc: yes, the LEDs are a backchannel :)14:07
whitequarkwpwrak: linux has a quirk which modifies bulk endpoints to be interrupt for low-speed devices14:07
wpwrakwhitequark: that one no longer exists (tried it)14:07
whitequarkbut you probably have no luck with windows and even more os x14:07
whitequarkoh, that sucks14:07
whitequarkit was a nice quirk.14:07
larscI think we even looked up the commit that removed it some time ago14:12
larscso it is still there14:17
wpwrakthe internet never forgets :)14:17
larscI mean the quirk14:17
larscit's still in the kernel14:17
wpwrakah .. no, i don't think the quirk is still there14:21
wpwrakor at least it doesn't work14:22
wpwrakanyway, given that the objective is compatibility, relying on linux-specific quirks wouldn't be such a great idea14:23
wpwrakafter all, if we just care about linux, making a little libusb-based utility is trivial14:24
larsclibusb works also on windows and macos14:27
whitequarklarsc: but not when code signing is enabled14:28
whitequarki.e. >= Vista14:28
whitequarkand/or 64-bit kernel, which forces code signing to be enabled14:28
larscah, didn't knew that14:29
whitequarkoh wait, they fixed it14:29
larscis this because of the libusb kernel module?14:29
whitequarkFor 64bit Windows Vista/7/2008/2008R2, the version should be or later.14:29
wpwrakwindows - designed to suck and delivering on its premises since day one :)14:29
whitequarkthat's very neat14:29
whitequark As of version V1.2.0.0, a valid digital signature is embedded inside libusb0.sys for AMD/Intel x86_64 version of Windows so that the users can install the driver as well under 64bit x86_64 version of Windows Vista/7/2008/2008R2.14:30
wpwrakah, nice14:30
whitequarkwpwrak: code signing in windows is awesome14:30
whitequarkthat should've been there from day one14:30
whitequarkyou can override it with a boot.ini switch even for 64-bit kernels btw14:31
whitequarkit does make sense in fact that it is intentionally hard for users, because malware.14:31
larscI must say I like the new Linux kernel module signing14:32
Action: whitequark doesn't like mindless bashing of windows14:32
whitequarklarsc: the EFI Secure Boot one?14:32
larsccreate a new public-private key pair, build your kernel, sign your modules and delete the private key14:32
larscmakes it much harder for rootkits14:33
whitequarklinux doesn't have a shortage of kernel holes14:33
whitequarkjust look at all the android phones, which are trivially routed, often without physical access14:34
whitequarkhalf of them is due to vendor stupidity. the other half is in linux.14:34
larscit's for the case where the attacker is already root14:34
larscand tries to hide himself14:35
whitequarkbut if you're root, isn't it trivial to patch kernel memory?14:35
whitequarkunless you have selinux or something like that14:35
whitequarkwhere root isn't an actual root14:35
larscstill makes it harder14:36
whitequarkmarginally IMO14:37
larscyea, I didn't really though about that you can still patch the kernel easily14:40
qi-bot[commit] Werner Almesberger: README: add ubbctl and ubb-jtag (master) http://qi-hw.com/p/ben-blinkenlights/85ed49314:40
qi-bot[commit] Werner Almesberger: ubbctl/ubbctl.c: add copyright header (master) http://qi-hw.com/p/ben-blinkenlights/fe96b3b14:40
qi-bot[commit] Werner Almesberger: ubbctl/ubbctl.c (main): move pin status display to separate function (master) http://qi-hw.com/p/ben-blinkenlights/e6a0e4214:40
qi-bot[commit] Werner Almesberger: ubbctl/ubbctl.c (main): add command line processing and usage display (master) http://qi-hw.com/p/ben-blinkenlights/33085ce14:40
qi-bot[commit] Werner Almesberger: ubbctl/ubbctl.c (show_pins): also show level seen at pin (master) http://qi-hw.com/p/ben-blinkenlights/77fd69614:40
qi-bot[commit] Werner Almesberger: ubbctl/ubbctl.c (show_pins): indicate function pins with "F" instead of "FN" (master) http://qi-hw.com/p/ben-blinkenlights/5297de414:40
qi-bot[commit] Werner Almesberger: ubbctl/ubbctl.c: add setting of UBB signals (DAT0=1, etc.) (master) http://qi-hw.com/p/ben-blinkenlights/246a8a014:40
qi-bot[commit] Werner Almesberger: ubbctl/ubbctl.c: add actions "on" and "off" to control nPWR (master) http://qi-hw.com/p/ben-blinkenlights/fd5707e14:40
wpwrakwhitequark: (mindless bashing)  bash, beat your windows every day, it will know why :)14:57
wpwrakof course, in my case "my" windows would be a diffuse ambient condition14:58
qi-bot[commit] Werner Almesberger: ubbctl/ubbctl.c: new option -c for continuous display (master) http://qi-hw.com/p/ben-blinkenlights/5fadbce16:49
qi-bot[commit] Werner Almesberger: ubbctl/README: short documentation (master) http://qi-hw.com/p/ben-blinkenlights/ed1923916:49
DocScrutinizer05fyi, CA warning: http://googleonlinesecurity.blogspot.com/2013/01/enhancing-digital-certificate-security.html17:55
wpwrakDocScrutinizer05: that's bee on heise news a few days ago. certificate-based "trust" must die :)19:53
DocScrutinizer05wpwrak: I didn't say I invented it today19:59
DocScrutinizer05better late than never, eh?19:59
DocScrutinizer05well, TURKTRUST20:00
DocScrutinizer05"ooops we handed out a master cert by accident" - LOL20:01
DocScrutinizer05"ooops again, even two"20:02
DocScrutinizer05"but thanks for letting us know, we wouldn't have noticed otherwise"20:02
wpwrakah, you hadn't read it before ;-)20:02
DocScrutinizer05I've read it 2h ago20:03
DocScrutinizer05when somebody informed me about the issue, since it is relevant for maemo-CSSU20:04
wpwrakit potentially affects everyone using SSL20:05
kyakwpwrak: can i use one instance of ubbctl to set ping status, and another instance of ubbctl to get pin status?20:08
wpwrakkyak: yes, you can run as many ubbctls in parallel as you like. they don't conflict with each other. that it, unless they try to perform contradictory changes on the same pins20:11
kyaknice :)20:11
wpwrakbut even then it's not certain you would get an invalid result20:11
wpwrak(at least the static result should be identical, no matter how many concurrent setters you have. there may be transient states that differ between concurrent runs of ubbctl and unordered sequential runs, though)20:16
DocScrutinizer05wpwrak: sure it affects everyone, but not everyone is responsible to fix the issue for ~60k other users20:35
DocScrutinizer05that's the kind of relevance it has for maemo-CSSU20:36
wpwrakheh :)20:36
DocScrutinizer05and actually I had no time to read any Heise news lately, since my plate is already more than filled with migrating a complete infra (*.maemo.org) to a yet-to-be-defined new hoster and financial basis20:41
DocScrutinizer05which gives me a headache every now and then, when I have to explain difference between a vhost and a RX300 rack20:42
kyakwpwrak: i'm packaging libubb and ubbctl. However, i noticed that you only provide libubb.a and you link ubbctl statically20:58
kyakhow about providing an libubb.so and linking ubbctl against it? I will then package both and ubbctl with depend on libubb20:59
wpwrakyeah, i should probably do that. though i hate the version number management :)21:09
kyakbtw, it's working great21:09
wpwrakbut that'll have to wait until tomorrow. got a barbecue to go to now.21:09
wpwrakkewl :-) and thanks for packaging !21:10
kyaknp, have fun!21:10
wpwrakthanks ! :)21:52
--- Mon Jan 7 201300:00

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