IRC log of #cubox of Fri 13 Oct 2017. All times are in CEST < Back to index

11:04 Ke> jnettlet[m]: also, what is the bricking warning on mcbin, can't you always use sdboot, if you mess up the spi-rom?
13:32 topi`> jnettlet[m]: do you remember is there a regulator that could be used to drop the +5V from the mPCIe slot on HB-Gate?
13:33 topi`> I have a strange mPCIe modem there which shows up on usb (as it should) but then 5 sec later USB disconnect
13:33 topi`> I'd like to see what happens if I manage to usbreset it
13:34 jnettlet[m]> Ke: yes you can always change the dipswitches to recover
13:34 jnettlet[m]> topi`: no there is no regulator for the 5v on the pcie bus.
13:34 topi`> damn
13:35 topi`> perhaps I try to boot a different kernel to see if it changes anything
13:35 Ke> any idea, what's SCP_BL2=`pwd`/RTOSDemo-cm3.bin and do I strictly need that binary?
13:35 Ke> from binaries-marvell
13:35 jnettlet[m]> I would make sure USB autopowersave is disabled
13:35 topi`> the older version of the same modem works 100%
13:35 topi`> this one has some new firmware I guess
13:38 jnettlet[m]> Ke: it is the RTOS demo binary. You shouldn't need it. I don't build atf with it. Just mv-ddr and uboot
13:38 topi`> I upgraded my mv-ddr
13:38 Ke> I can just leave the env undefined and build with the default instructions otherwise?
13:39 topi`> but I had one of the very first mcbins
13:39 Ke> http://wiki.macchiatobin.net/tiki-index.php?page=Build+from+source+-+Bootloader ie. these instructions
13:40 Ke> well I guess I can try
13:41 Ke> jnettlet[m]: thanks, you're awesome
13:42 topi`> jnettlet is the saviour of poor hw hackers' souls :)
13:56 jnettlet[m]> Ke: that binary is for power management. You need it for cpuidle and cpufreq support
13:57 Ke> how was it done in the legacy bootloaders that did not have it?
13:57 Ke> it was just missing?
13:58 jnettlet[m]> Yes. Ask topi he ran into that issue.
13:59 Ke> anyway, I remember this was also an issue with rk3399, apparently ARM forbids the SoC vendors to publish the power management code
14:00 Ke> but with rk3399 power management was just some KiB sized inline blob on ATF
14:01 Ke> rk3399 power management has since then been reverse engineered, as I understood
14:40 topi`> I had to sacrifice power mgmt blob for the newer mv-ddr blob
14:40 topi`> because I switched to a 16GB DIMM and it didn't work until I upgraded to the newest mv-ddr
14:41 topi`> power mgmt has always been a fairly closed-source affair
14:41 topi`> back when I worked for Nokia, everything from charging phones to the power management of SoCs was strictly closed source and a company secret
14:41 Ke> mv-ddr is now completely blobber or what?
14:42 Ke> blobbed?
14:42 topi`> I just grabbed the source code for mv-ddr and compiled it and it worked
14:42 Ke> sure
14:42 topi`> I can't remember what the license was
14:42 topi`> https://github.com/MarvellEmbeddedProcessors/mv-ddr-marvell
14:42 topi`> here it is
14:42 Ke> sure
14:45 topi`> oddly enough, that branch only seems to have a38x support. judging by the subdirs that exist ...
14:45 topi`> maybe I'm wrong. I can't remember on which machine I compiled it.
14:46 topi`> I don't have too many aarch64 machines around, anyways
14:46 topi`> ah, maybe on the Geekbox
14:46 Ke> yeah marvell doesn't do reasonable master branches
14:46 Ke> perhaps
14:47 topi`> I talked to a couple of Marvell guys at linuxconf 2007 Ottaa
14:47 topi`> ottawa
14:47 topi`> they seemed to be pretty much linux guys, but maybe they've left for better companies :)
14:47 topi`> of course, one can be a hardcore linux guy and still get git wrong
14:47 Ke> well I bet there are company policies
14:48 topi`> that's strange, given that they've trying to do this linux thing well since 2007
14:48 Ke> not many people get to dictate terms of work so that they never have to do stupid stuff
14:48 topi`> I think I still have those guys' businesscards somewhere...
14:49 Ke> topi`: are you saying I should get drunk and call them?
14:49 topi`> yeah avoiding stupid stuff in a corporate environment is ... avoiding the inevitable :D
14:49 topi`> firm after firm, I've become disillusioned
14:51 topi`> KE, please don't do it :D ages ago, was it in the first OScon ever, in 2001, I got drunk at some party organised by ActiveState and ended up getting thrown out of the party :D
14:51 topi`> I was too loud, and was pushing Eric S. Raymond too far with annoying questions
14:51 Ke> <3
14:51 topi`> poor Eric
14:51 topi`> I was also very young, so let's forget this
14:52 topi`> moral of the story: if there is a Finn at the party, hide all the free booze!
14:53 Ke> though kids these days are quitting alcohol
14:53 topi`> they have plenty of other (more powerful) stuff to play with
14:54 Ke> possibly, though in general there is a strong tendency for more orderly behaviour
14:54 topi`> damn I don't know what to do with this modem... it just does a harakiri and never comes back to the usb bus. SO I cannot just do a "usb reset" on the device, because it has disappeared
14:55 topi`> :/
14:56 topi`> I did a thing called "physical disconnect" and ho behold, there it is and I can even see the ttyACMx devices
14:56 topi`> this must be some buggy fw behavior
14:57 topi`> YES. as soon as I opened minicon on the modem's acm port (AT), I saw it was in the android bootloader prompt
14:57 topi`> so I had to type "boot" and watch UBOOT load something. The modem seems to run a MIPS core.
15:14 jnettlet[m]> topi`: is this a Telit modem by any chance?
15:15 topi`> nope. It's made by a singaporean company called Atel
15:16 topi`> my bad. Asiatelco Technologies, Shanghai
15:19 topi`> all I can say, if the device does a USB Disconnect after trying to boot its firmware, I'd say it's bu