**** BEGIN LOGGING AT Wed Dec 23 02:59:58 2015
Dec 23 09:43:04 <freemangordon> Pali: BTW, what was the final decision re ATAGS?
Dec 23 09:43:23 <freemangordon> I've lost the end of the conversation
Dec 23 09:43:55 <Pali> freemangordon: looks like via reserve memory hook from arch/arm/mach-omap2 we should be able to prevent linux kernel to overwrite memory where are original ATAGs stored
Dec 23 09:44:40 <freemangordon> so it can be dumped to a file on a later stage?
Dec 23 09:44:48 <Pali> and then from rx51 specific board hook (already have one for emmc and one for thumb2 errata) we should be able to read them and store back to /proc/atags
Dec 23 09:45:29 <Pali> everything just from rx51 code without touching global arm/boot code
Dec 23 09:45:37 <freemangordon> great
Dec 23 09:45:37 <Pali> but I did not tried to play with it yet
Dec 23 09:45:54 <Pali> this solution will be accepted (once somebody write and test it)
Dec 23 09:47:00 <freemangordon> h,, I'll see if I feel relaxed in the incomming holidays, may look at it
Dec 23 09:47:54 <zGrr> moin :)
Dec 23 09:48:04 <freemangordon> :)
Dec 23 10:04:43 <freemangordon> Pali: does latest n900 kernel on github compiles?
Dec 23 10:04:55 <freemangordon> and does it have DT enabled?
Dec 23 10:05:11 <Pali> freemangordon: do not know, I have not updated it for a long time
Dec 23 10:05:16 <Pali> but I think it should
Dec 23 10:05:22 <freemangordon> ok
Dec 23 10:05:44 <freemangordon> Pali: which one to use, 4.2?
Dec 23 10:05:54 <Pali> probably last version
Dec 23 10:06:01 <freemangordon> ok
Dec 23 10:06:19 <freemangordon> hmm, seems my CPU heatsink needs to get cleaned :)
Dec 23 10:07:33 <freemangordon> CPU hit 78 deg while compiling the zImage, before being stopped with ctrl-c :)
Dec 23 10:16:02 <zGrr> are there any news related to neo900?
Dec 23 10:16:19 <zGrr> i mean, some progress?
Dec 23 10:54:05 <freemangordon> Pali: any ide where to put rx51_reserve()? in board-generic.c? in common.c? in pdata-quirks.c?
Dec 23 10:54:14 <freemangordon> *idea
Dec 23 10:54:54 <Pali> freemangordon, search for nokia_n900_legacy_init
Dec 23 10:55:07 <freemangordon> it is in pdata
Dec 23 10:55:15 <freemangordon> pdata-quirks.c
Dec 23 10:55:16 <Pali> and for n900_boards_compat
Dec 23 10:55:32 <freemangordon> it is boards-generic :)
Dec 23 10:55:43 <freemangordon> *board-generic
Dec 23 10:55:50 <Pali> I think reserve memory needs to be instead omap_reserve in "DT_MACHINE_START(OMAP3_N900_DT, "Nokia RX-51 board")"
Dec 23 10:55:56 <freemangordon> sure
Dec 23 10:56:13 <freemangordon> the point is that - where to put that replacement function
Dec 23 10:56:20 <freemangordon> in which file
Dec 23 10:57:25 <freemangordon> I think the best place is board-generic.c, so it can be defined as static, but I see no function definitions in that fil
Dec 23 10:57:30 <Pali> in my opinion put new function into pdata-quirks.c
Dec 23 10:57:53 <freemangordon> and declare it where? in common.h?
Dec 23 10:57:59 <Pali> and assign it to .reserve =
Dec 23 10:58:39 <Pali> header file common.h
Dec 23 10:58:52 <freemangordon> I know where to assign it, but it needs to be extern, not static if defined anywhere outside if board-generic
Dec 23 10:59:00 <freemangordon> *of
Dec 23 10:59:13 <Pali> yes, it must be non-static
Dec 23 10:59:30 <Pali> "extern" keyword for C function is not needed and is not used
Dec 23 10:59:53 <freemangordon> but if it is declared in common.h, whine not define it in common.c then?
Dec 23 10:59:58 <freemangordon> *while
Dec 23 11:00:05 <freemangordon> shit
Dec 23 11:00:06 <Pali> because it is quirkk function
Dec 23 11:00:07 <freemangordon> WHY
Dec 23 11:00:20 <freemangordon> so is omap_reserve()
Dec 23 11:00:44 <freemangordon> will ask on linux-omap
Dec 23 11:00:53 <Pali> basically this is n900 specific function and correct place is in pdata-quirks.c
Dec 23 11:01:34 <Pali> but you must assign pointer to that function in file board-generic.c and that file includes only common.h
Dec 23 11:04:14 <freemangordon> Pali: but there *is* device specific data in board-generic.c -   struct n900_boards_compat, why not define that function there as well. And make it static.
Dec 23 11:04:52 <freemangordon> or put it in common.c - if it is declared in common.h
Dec 23 11:06:22 <Pali> that board-generic.c does not contains board specific functions
Dec 23 11:06:32 <Pali> but put it here...
Dec 23 11:06:40 <freemangordon> oh, wait, the *is* at least one function defined in board-generic -   omap_generic_init()
Dec 23 11:06:59 <freemangordon> but it contains board specific data, so I'll put it there
Dec 23 11:07:03 <Pali> omap maintainers will accept it or not
Dec 23 11:07:46 <freemangordon> :nod:
Dec 23 11:08:45 <freemangordon> Pali: do we have atags address and size?
Dec 23 11:09:53 <Pali> we have offset
Dec 23 11:10:01 <freemangordon> 0ь100?
Dec 23 11:10:07 <freemangordon> 0x100?
Dec 23 11:10:08 <Pali> and size shold be smaller then one page
Dec 23 11:10:13 <freemangordon> ok
Dec 23 11:10:37 <Pali> memblock_reserve(PHYS_OFFSET, PAGE_SIZE);
Dec 23 11:10:46 <freemangordon> yeah
Dec 23 11:13:09 <Pali> read this thread: http://www.spinics.net/lists/arm-kernel/msg464340.html
Dec 23 11:13:22 <Pali> (all messages in "Next by thread" links)
Dec 23 11:27:03 <freemangordon> Pali: already did it, I am in CC on linux-omap ML :)
Dec 23 11:27:11 <Pali> ok :-)
Dec 23 11:28:19 <freemangordon> but this memblock API is not very understandable, WTF is the difference between memblock_free and memblock_remove?!?
Dec 23 14:03:04 <freemangordon> Pali: what about copying atags data to some __init structure instead of doing memblock_reserve()?
Dec 23 14:03:57 <freemangordon> as it turns out, memblock_reserved() memory cannot be given back to the system once  paging_init() is called
Dec 23 14:04:43 <Pali> freemangordon, you can call directly function save_atags()
Dec 23 14:06:26 <freemangordon> Pali: hmm, tell me then why this is not done yet? I men - I've lost the roots of the problem.
Dec 23 14:06:30 <freemangordon> *mean
Dec 23 14:07:00 <Pali> problem is that I'm not sure if you really can read that memory where are atags stored
Dec 23 14:07:06 <Pali> if they are not rewritten yet...
Dec 23 14:07:34 <freemangordon> static char __initdata atags_copy[BOOT_PARAMS_SIZE];
Dec 23 14:07:41 <Pali> you need to read them before some memory/page init functions are called
Dec 23 14:08:37 <freemangordon> Pali: save_atags is called from setup_machine_tags
Dec 23 14:08:58 <Pali> yes, but that pointer points to DTS blob!
Dec 23 14:09:02 <Pali> not to atags memory
Dec 23 14:10:24 <freemangordon> oh, I see
Dec 23 14:14:53 <freemangordon> Pali: but then, what needs to be done? I mean - DT anyways creates /pros/atags, correct? We can;t overwrite it.
Dec 23 14:15:16 <Pali> DT does not create /proc/atags
Dec 23 14:15:25 <Pali> *DT booted kernel
Dec 23 14:15:58 <Pali> and we want to read ATAGs passed by nolo to kernel also for rx51 device booted in DT mode
Dec 23 14:16:23 <freemangordon> but what does  init_atags_procfs then?
Dec 23 14:16:53 <Pali> on DT booted kernel nothing
Dec 23 14:17:13 <Pali> but if you are using my tree, it contains special patch which store ATAGs into DT structure
Dec 23 14:17:27 <Pali> and then kernel unpack ATAGs from DT structure and store them into /proc/atags
Dec 23 14:17:34 <Pali> but this patch was rejected
Dec 23 14:17:37 <freemangordon> because  atags_copy contains nothing?
Dec 23 14:17:46 <Pali> something like that
Dec 23 14:17:57 <Pali> do not rememeber exacly
Dec 23 14:18:04 <Pali> save_atags is not called in DT booted kernel
Dec 23 14:20:10 <freemangordon> it should be, at least from what I see in the code http://lxr.free-electrons.com/source/arch/arm/kernel/atags_parse.c#L227 . unless that if fails
Dec 23 14:21:21 <freemangordon> I know it is not called with the atags from the bootloader
Dec 23 14:21:27 <freemangordon> but still it is called
Dec 23 14:46:11 <freemangordon> Pali: what do you think about http://pastebin.com/MHzUvusK ?
Dec 23 14:56:41 <Pali> freemangordon, no need to #ifdef CONFIG_ATAGS_PROC
Dec 23 14:56:56 <Pali> just include atags.h or kernel/atags.h
Dec 23 14:56:57 <freemangordon> why?
Dec 23 14:57:02 <freemangordon> you can;t
Dec 23 14:57:16 <Pali> save_atags() is noop when CONFIG_ATAGS_PROC is not defined
Dec 23 14:57:18 <freemangordon> what needs to be included is ../kernel/atags.h
Dec 23 14:57:28 <freemangordon> which is ugly
Dec 23 14:57:54 <freemangordon> I know it is noop, but still I define one more static function, which is not needed
Dec 23 14:58:02 <Pali> if you use function from kernel you should include header file
Dec 23 14:58:33 <freemangordon> I don;t believe #include "../kernel/atags.h" will be upstreamed :)
Dec 23 14:59:30 <Pali> without correct header too...
Dec 23 14:59:34 <Pali> better ask on ML
Dec 23 14:59:46 <freemangordon> lets see what will say Russell, I'll fix those nits later
Dec 23 14:59:48 <Pali> but first check if it is working
Dec 23 15:00:18 <Pali> ok
Dec 23 15:00:25 <freemangordon> and this is the problem, I don;t have device with booting upstream kernel :)
Dec 23 15:00:42 <freemangordon> can I boot qemu to userspace?
Dec 23 15:01:06 <freemangordon> Pali: ^^^
Dec 23 15:01:19 <Pali> yes
Dec 23 15:01:29 <Pali> small userspace with busybox is enough
Dec 23 15:01:40 <freemangordon> do you have something prepared?
Dec 23 15:01:58 <Pali> I have script which build userspace :-)
Dec 23 15:02:07 <Pali> and start qemu
Dec 23 15:02:10 <freemangordon> good for you :)
Dec 23 15:02:17 <freemangordon> care to share it?
Dec 23 15:02:45 <Pali> ok, I'm going to clean it and send it to you
Dec 23 15:02:55 <Pali> now it depends on my PC :-)
Dec 23 15:02:56 <freemangordon> thanks
Dec 23 15:03:06 <freemangordon> send it as well :)
Dec 23 15:04:11 <freemangordon> Pali: don;t hurry, my GF has birthday today, so I don;t think I will have much more time to spend on that today
Dec 23 15:14:54 <DocScrutinizer05> hi folks
Dec 23 15:17:31 <Pali> freemangordon: done, see email
Dec 24 01:38:42 <Maxdamantus[r]> Hm. jonwil might be interested to know that the guy from the Google security team responding to my report has reproduced the issue and filed a bug to "the team".
Dec 24 01:39:15 <Maxdamantus[r]> Dunno if I can log in to TMO over this internet connection.
**** ENDING LOGGING AT Thu Dec 24 02:59:59 2015