<hno>
wingrime, u-boot do not seem to like all SD cards. With some cards if fails to detect the card. Quite likely a timing issue somewhere, but haven't been able to look into it. All cards I have access to works just fine.
<Turl>
wingrime: your patch seems to remove half of the PM ops hm
* hramrach
is away ;-)
rellla has joined #linux-sunxi
<wingrime>
Turl: all that ops are stud
<wingrime>
Turl: just cleanup )
<wingrime>
Turl: pm not nesessory must have that ops
<wingrime>
only pm_valid / pm_enter have needed stuff
<hramrach>
wingrime: it's better to do the cleanup in a separate patch then so that it's obvious what your patch changes
<wingrime>
huo: most strange thing that card work on other a10
<wingrime>
without issues
<wingrime>
on a13 but 1/3 chance to boot
<Turl>
wingrime: I'd leave pm_wake one, it's nice to debug why device wakes by itself sometimes
<wingrime>
Trul: nope you don't need that I make debug print with reason
<wingrime>
collect2: error: ld returned 1 exit status
<wingrime>
VLC still not builds
<hramrach>
install dbus development files
<wingrime>
try)
<wingrime>
mnemoc: are you here?
<mnemoc>
wingrime: not really
<mnemoc>
hramrach: I can't take github pull requests. I would be disrespecting the devs in the ML. please submit it there
<hramrach>
mnemoc: the nand is copy of what was sent to the ml
<mnemoc>
hramrach: cool
<hramrach>
the leds is just backport with 2 lines of new code
<hramrach>
or really code pulled from non-applicable patch
<wingrime>
hramrach: don't steal patch
<hramrach>
wingrime: I am not stealing any patches
<wingrime>
hramrach: anyway wait I remake patch
<hramrach>
cool
<mnemoc>
wingrime: if it's short and doesn't require disccusing. tell me
<wingrime>
mnemoc: are you saw our conversation?
<wingrime>
mnemoc: what do you think about Suspend ....
<mnemoc>
wingrime: no, i haven't read the backlog. busy packaging the stuff of the house, moving to a differnt appartment next week
<hramrach>
ugh
<hramrach>
well, good luck with that ;-)
<wingrime>
good luck
* mnemoc
hates moving :|
<wingrime>
mnemoc: any way suspend works
theborger has quit [Ping timeout: 260 seconds]
<hramrach>
mnemoc: if you think the cpu led trgger backport needs discussion I will send it together wit hthe other led pieces when I am done with them
rellla has quit [Remote host closed the connection]
<Turl>
mnemoc: packing is not too problematic, the problem is unpacking
<techn__>
wingrime:I think these needs to be added console_lock();fb_set_suspend(info, 1); console_unlock();
<techn__>
and same to resume
<techn__>
but 1 to 0
<wingrime>
techn: if you send patch it will be better
<techn__>
Havent yet had time to try that patch
<techn__>
so cant yet know if I can reproduce that problem
<wingrime>
I too
<wingrime>
hramrach: you can do?
<mnemoc>
wingrime: great news!
<mnemoc>
hramrach: can you test the led thing in 3.0 too?
<wingrime>
mnemoc: you can see patches
<wingrime>
techn:you have time test suspend?
<mnemoc>
hramrach: sorry. didn't see it. thanks!
<mnemoc>
wingrime: :)
<mnemoc>
this cpu leds thing is already mainlined, right?
<mnemoc>
hramrach: wingrime: ---^
techn_ has joined #linux-sunxi
theborger has joined #linux-sunxi
techn__ has quit [Ping timeout: 252 seconds]
<hramrach>
mnemoc: only tested the cpu trigger on 3.4
<hramrach>
yes, it's already mainlined. the third patch just does not apply because it removes previous cpu trigger which we never had
<hramrach>
techn_: where does the console_lock() go?
<hramrach>
anyway, the board is half locked up atm and git ahd the great ide ato recompress the repo
<Turl>
cpu trigger?
<Turl>
what does it do?
<hramrach>
lihgt the led when cpu is not idle
<Turl>
where is it?
* Turl
doesn't see patch on ML
<hramrach>
I just put it on github. It's trivial backport from upstream
<mnemoc>
Turl: it's on the stage branches
<wingrime>
who sends patch to upstream ?
<mnemoc>
mripard and Turl mostly
<Turl>
hramrach: ah, I see it now :)
<wingrime>
Turl: after my cleanup it possible drop standby.bin and use stupid suspend
<Turl>
did you move standby.bin code into the kernel?
<wingrime>
Turl: I saw that kernel can allocate data and function in SRAM
<wingrime>
Turl: nope
<wingrime>
Turl: I make simple STUD that commented
<wingrime>
simply go idle
<wingrime>
if send to upstrem
<Turl>
wingrime: but then how does it suspend?
<Turl>
the code to switch clocks etc for suspend is on that binary
<wingrime>
cpu_do_idle();
<Turl>
yes but that doesn't do proper suspend
<wingrime>
perfect first implemention for mainlaing
<wingrime>
but it can be wakeup only by UART with console_no_suspend
wingrime has quit [Read error: Connection reset by peer]
wingrime has joined #linux-sunxi
<wingrime>
only UART with no_console_suspend in cmdline can wakeup it
<wingrime>
Turl: what do you think, we can push mainline that simple support ?
<Turl>
so you support both too
<wingrime>
Turl: I done this for make it differnet late
<wingrime>
Turl: if remove comments it will be different
<wingrime>
Turl: next step for it : move code from standby.bin for interupt controller and lradc
<torqu3e>
guys am trying to get the UVC drivers going on an a13, and am stuck here File not found: /lib/modules/3.0.42+/build/.config at ./scripts/make_kconfig.pl line 33, <IN> line 4.
<torqu3e>
lsusb, shows the cam connected, I can check the UID to have everything for uvc to work with this
<torqu3e>
dmesg does not show the cam being picked up since the drivers aren't there
<wingrime>
Turl: I mean save intcontroller registers, enable only wakeup interupts , than wake
<hramrach>
torqu3e: you need to configure teh kernel
<hramrach>
that is, copy the config of the kernel you have in .config, add the option you want
<torqu3e>
there is no .config in that location
<Turl>
wingrime: I see, both do the same hm
<hramrach>
torqu3e: you are doing something weird
<wingrime>
Turl: also kernel support SRAM functions by attributes , data , so possible make same thing with without "bin"
<hramrach>
uvc is in the kernel so jsut get kernel source an enable it if not enabled in yout build
<wingrime>
Turl : what do you think, about mainline my simple implemention ?
<hramrach>
decide if sunxi does mem or standby and don't pretend the other
<torqu3e>
can you point me to something which gives a few more steps, am just about venturing into kernel compiles :)
<Turl>
wingrime: many things changed on mainline to support multiplatform
<torqu3e>
the uvc page says it should be in source, that does not seem to be the case here, probably was not compiled with it :-/
<wingrime>
Turl: explan
<Turl>
wingrime: I think the proper way to do suspend these days is a cpuidle driver
<Turl>
mripard: can you confirm/deny? ^
<wingrime>
Turl: wakeup possible only by interupt
<wingrime>
Turl: and we anyway need save/restore int controller regs and wakeup source regs
<Turl>
currently there is little hardware support in mainline
<wingrime>
It better than nothng ))
<Turl>
so maybe adding suspend for 3.10 ("mem") would be nice, so we don't have to spend a lot of time debugging stuff later and we can add specific device suspend support when submitting the drivers themselves
<Turl>
mnemoc, mripard: thoughts?
<wingrime>
Turl: I think same
<wingrime>
Turl: If we will have suspend we will be able debug it on early development stage
<wingrime>
45t[3~
<Turl>
wingrime: talk with mripard when he is around and find out if pm on arch-sunxi is still the right way to implement it
<Turl>
so you don't waste time working on something that can't be mainlined
<Turl>
s/arch-sunxi/mach-sunxi/
<mnemoc>
mainline is all about drivers so mach-sunxi/pm is probably not welcomed
<Turl>
mnemoc: yet tegra has it, with some dt code
* Turl
scratches head
<wingrime>
mnemoc: posible drop binary image
<mnemoc>
Turl: maybe pre-multiplatform?
<Turl>
mnemoc: but it has dt glue :| idk
<mnemoc>
hramrach: any idea how was "partition-generic.c" named in 3.0?
<mnemoc>
hramrach: fs/partitions/check.c ?
<hramrach>
didn't look into backpoting that part further to 3.0
<hramrach>
but check.c is also used in the patches
<mnemoc>
fs/partitions/check.c:struct hd_struct *add_partition(struct gendisk *disk, int partno,
<hramrach>
looks like the place
<hramrach>
it even applies verbatim to the file
<mnemoc>
Applying: Remove nand partition handling cruft. <--- I'm fighting this one now
<hramrach>
that's completely sunxi specific
<wingrime>
mnemoc: do you have Aw git log ?
<hramrach>
and not really needed. just removes leftover junk
<mnemoc>
wingrime: check my personal github
<wingrime>
mnemoc: why sun5i don't have "crane" andorid config ?
<mnemoc>
wingrime: nuclear
<wingrime>
mnemoc: you should add cpufreq and suspend to sun5i defconfig....
<hramrach>
send patch ;-)
<wingrime>
but interactive and smartassv2 not posible to build as module
<wingrime>
need make 2 functions exportable in sched
<mnemoc>
wingrime: send patches :)
<mnemoc>
and please, pick your favorite subsystem/driver and review what's new in allwinner's branches
<wingrime>
if i have new HW....
<mnemoc>
there are a10-dev, a13-dev and a10s-dev branches in my github
<mnemoc>
it's not only about A20/A10
<mnemoc>
err
<mnemoc>
it's not only about A20/A31
<wingrime>
I don't think that aw make many patches for a10/a13
<mnemoc>
there is a lot of history newer than our start point
<mnemoc>
qware's release was based on allwinner's android4.0*-v0.8
<wingrime>
will see tomorrow
<mnemoc>
awesome
<wingrime>
is nand[abcd] set default only for android for now ?
<hramrach>
it's still the default
<hramrach>
but hopefully can be removed in the future
<hramrach>
well, probably on 3.8+ there will be no such cruft anyway so even android will have to update the paths
<wingrime>
hramrach: why not use it only in "crane"
torqu3e_ has joined #linux-sunxi
<hramrach>
it defaults to yes in Kconfig
torqu3e has quit [Read error: Connection reset by peer]
torqu3e_ is now known as torqu3e
rz2k has quit [Read error: Connection reset by peer]
jelly-home has quit [Ping timeout: 252 seconds]
hurtigbuffer has joined #linux-sunxi
christopher has joined #linux-sunxi
<mnemoc>
wingrime: it's default yes because the change would break linux installations as well
hurtigbuffer is now known as jelly-home
christopher is now known as Guest2298
Guest2298 is now known as anunnaki
<libv>
first open source compiled vertex shader runs on the lima driver :)
<wingrime>
lima can simple 2d for windowsmanager ?
<libv>
the dual texture vertex shader for q3a just ran on some walls grabbed from a q3a frame
gzamboni has quit [Ping timeout: 260 seconds]
<mnemoc>
wingrime: lima (as mali) is 3d-only, to become part of mesa3d
<wingrime>
mnemoc: windows can be shown as texures
<wingrime>
mnemoc: with ortogonal projection
<mnemoc>
the A10, A20 and A31 have a dedicated 2d unit called g2d
<hramrach>
which may even work, at least on a20
<hramrach>
techn_: I don't see how you get the fb_info in disp_normal_suspend
<mripard>
Turl: I'd really like to focus on the core stuff, before moving to non-essential things like pm
<hramrach>
and dev_fb has no _suspend
<techn_>
hramrach: I think you should notify change for both fb_info's.. they are in global array
torqu3e has quit [Read error: Connection reset by peer]
<mripard>
proper suspend is hard to achieve, we will have a lot of effort to put in this, I think effort on the USB/MMC/etc would be better
<mripard>
at least for a short term view
<hramrach>
well, they are actually in global struct that has arrays of size 2 it seems
<mripard>
mripard: really, for 3.10, all I want is a proper clocks support
<mripard>
of course, I'll take any patches that comes
<mripard>
(given that they are good enough)
torqu3e has joined #linux-sunxi
<wingrime>
mripard: simple suspend easy
<wingrime>
just stop CPU and wait IRQ
<wingrime>
mripard: this patch: clk: arm: sunxi: Add a new clock driver for sunxi SOCs ?
<mripard>
wingrime: then I think like Turl suggested that cpuidle is the way to go, so write a cpuidle that does that if you want :)
<mripard>
what's the question in your last sentence?
hramrach has quit [Remote host closed the connection]
<wingrime>
is that you want to see in mainline
<wingrime>
?
<mripard>
yes, it's the most urgent thing to merge
<mripard>
we begin to have quite a few drivers that are almost complete and require Turl's driver to be merged
<techn_>
sorry for huge patchset :(
<mripard>
Turl: (no pressure :))
<wingrime>
mripard: mainline rules so hard to archive ?
gzamboni has joined #linux-sunxi
<mripard>
it depends on what you write in the first place
<mripard>
once you get use to it, it's not hard at all
<mripard>
but in that case, Turl's not the one to blame
<mripard>
the maintainer doesn't review his patch
<mripard>
he did a first round of patches, got some reviews from other people, sent a second version, and... nothing
hramrach has joined #linux-sunxi
<wingrime>
I can't run vlc
<wingrime>
even without X stoped
<wingrime>
and in fb_con
<wingrime>
so what is wrong?
wingrime has quit [Read error: Operation timed out]