**** BEGIN LOGGING AT Thu May 29 03:00:01 2014 May 29 05:43:11 panto : I'm having problems reading the sysfs file I created! May 29 05:43:35 not at the coding level, but rather conceptually May 29 06:21:22 panto : ping :) May 29 07:30:46 mranostay: care for a few minutes to give me a head start on how it is to be done in the kernel module? May 29 07:34:15 panto: can your remoteproc implementation load pasm-generated bin files? May 29 07:42:39 * panto yawns May 29 07:42:54 karki_, paste the link again May 29 07:42:59 Abhishek_, yes it can May 29 07:46:17 * vvu wishes good morning to everybody May 29 07:46:48 panto: for me the syscall PRU registers would have to change as they would be holding the sampled data May 29 07:47:00 morning vvu May 29 07:47:08 how is it going Abhishek_ ? May 29 07:47:19 vvu: To kernel mode now May 29 07:47:25 panto : Never mind! I got it to work :) Now my platform driver supports sysfs May 29 07:47:43 I mean the link to what you want me to enter - hardware request or something May 29 07:47:52 Abhishek_ : I don't think thats a good idea, won't you be breaking the ABI? May 29 07:47:54 Abhishek_, I don't get it May 29 07:48:06 why change the syscall? May 29 07:48:22 panto : alexanderhaim filled it yesterday :) May 29 07:48:39 the interface is low speed, you can't transfer the bulk of data with it May 29 07:49:08 the registers that would be holding the syscalls are actually holding the LA samples in my firmware. May 29 07:49:32 lemme see if I can shift them to a different bank of registers May 29 07:49:53 Abhishek_ : but that is how the compiler compiles stuff..... I don't think you can change that... May 29 07:50:40 or rather that is where the compiler expects stuff to be. May 29 07:51:33 panto: In my firmware R13-R21 hold the LA samples. I do not mean to transfer data with it, it's directly written to the memory May 29 07:52:14 karki_: The new compiler has already a different ABI IIRC, mranostay has added a note to his lighting example about it May 29 07:52:54 I don't think the registers used in function calls have changed. May 29 07:52:56 Abhishek_, you don't accept the downcall at any point in your code May 29 07:53:18 you test for the event from ARM and then issue an instruction May 29 07:53:32 you can easily save those registers and restore after you return May 29 07:54:55 panto : I want to start with firmware load and upcall/downcall support for my driver from tomorrow! May 29 07:55:22 I'll probably be bugging you a little for the firmware load part ;) May 29 07:59:20 * karki_ is now working on bin files and mmap, will push another set of commits in a few hours. May 29 08:45:56 karki_: for testing your module, are you compiling against the stock kernel on the BBB, or built your custom kernel and testing with it? May 29 08:59:19 stock kernel May 29 08:59:42 I have enough problems installing stuff ;) Don't want any more :p May 29 09:00:25 why would you want to recompile the kernel? (other than the fact that you'd feel great the first couple of times you did it ;) ) May 29 09:01:11 karki_: Because mranostay's patch for rproc driver isn't included in the stock May 29 09:01:24 Abhishek_ : It is May 29 09:01:37 the debian ones have it May 29 09:02:06 lemme check May 29 09:02:49 which patch? the pru-rproc one? May 29 09:03:02 panto 's driver ? May 29 09:03:31 nope, that's already there May 29 09:03:54 mranostay had added some code to make his lighting example work May 29 09:05:21 oh, yeah. I don't think thats there. May 29 09:05:28 but why'd you want it? May 29 09:06:07 are you shifting to rproc? May 29 09:08:28 the LA would need its own memory management in kernel space. May 29 09:08:49 and interrupt handling, userspace isn't the way forward May 29 09:17:36 essentially, I have a week to set up the LA core in kernel space. May 29 09:19:33 * Abhishek_ is considering that panto's driver needs to be refactored to provide a common infrastructure, and modules can link against it to add their functionality. But then it would set off deadlines. May 29 09:21:03 * karki_ considered that a month ago, but refrained from doing it as it puts him way off track :( May 29 09:23:21 Abhishek_ : How will you take care of the remote proc stuff on the PRU end if your app is written completely in assembly ? May 29 09:24:09 karki_: Yes, that is the issue on my side, and you would also have to take care of that while writing your C Code May 29 09:24:28 yeah :/ May 29 09:24:44 tell me if you are refactoring the code. I'd love that ;) May 29 09:25:07 I have a feeling my code is going to get messy and patchy. May 29 09:28:20 writing kernel space code was not my aim, although I knew that it had to go to the kernel space at some point as a stretch goal. May 29 09:29:29 Abhishek_ : neither was it mine, I didn't realize there was no userspace bridge for remote proc! May 29 09:29:42 So my schedule is as it is messed up. May 29 10:15:52 av500: do you have any hints on the gadget drivers not being loaded on startup when I build a new kernel? May 29 10:59:15 panto : what would be a decent size for the bin_attribute file? May 29 13:13:19 * karki_ hmm... May 29 13:23:32 * Abhishek_ fixed the issue with his custom kernel May 29 13:23:50 <_av500_> qhat was it? May 29 13:23:53 <_av500_> what May 29 13:24:28 av500: the issue was that the USB gadget drivers were not being loaded May 29 13:25:16 the modules had somehow got installed in /lib/modules/modules/3.8.13-xxxx instead of /lib/modules/3.8.13-xx May 29 13:25:27 <_av500_> ah May 29 13:26:04 I don't know why but ssh takes longer to connect to the system than it took with the stock kernel. May 29 13:26:54 could it be linked to ipv6 support compiled in? May 29 13:30:34 <_av500_> maybe May 29 13:30:56 <_av500_> I always disable that May 29 13:31:32 av500: how to disable? May 29 13:32:00 <_av500_> kernel config, no? May 29 13:32:17 no, disabling loading of module I thought. May 29 13:35:03 panto : I got bin_file to work as a part of sysfs May 29 13:36:27 :) May 29 13:48:27 av500: Does it suffice to disable CONFIG_IPV6 or I have to disable related options as well? May 29 14:21:29 Abhishek_: why disable ipv6 ? it is not used anyway as i remember May 29 14:22:11 dunno, but ssh connection establishing is slower on my compiled kernel as compared to stock May 29 14:23:23 v6 should not effect this May 29 14:26:59 vvu : +1 May 29 15:03:21 vvu : which linux processes run on physical memory (i.e.) no MMU? May 29 15:03:39 * karki_ wonders where rseethamraju is..... May 29 16:01:57 karki_: not many May 29 16:02:35 mranostay : So do certain kernel process just bypass the MMU? May 29 16:02:57 panto: ^ May 29 16:03:34 because I need to implement mmap for my sysfs entry, had to get some doubts cleared.... May 29 16:03:45 why would that matter? May 29 16:04:49 I want to know whether my driver will have (direct or offset-ed) access to the physical memory..... May 29 16:05:00 or will it by KVM May 29 16:05:10 kernel virtual memory May 29 16:08:21 panto mranostay ds2 : Also the UIO reserves a magic number in the DDR, so the program on the PRU side has no problem communicating to the client program via DDR. How do you think I should implement that in remote proc? May 29 16:51:17 I'm not here yet :| May 29 16:51:34 I'll answer in a couple hours May 29 17:53:26 karki_: i have no clue regarding these kind of things May 29 17:53:29 did not go into that level yet :)) May 29 17:53:42 :'( May 29 19:41:50 Do I need to be root to execute ioctl on a device? May 29 19:43:33 thinks so May 29 19:43:56 * vvu neded su to do ioctl in OS Lab May 30 02:52:48 panto : pint May 30 02:52:52 *ping **** ENDING LOGGING AT Fri May 30 03:00:00 2014