=== jldugger is now known as pwnguin | ||
cooloney_ | NCommander: sorry, i missed the Mobile team IRC meeting last night, | 09:48 |
---|---|---|
cooloney_ | NCommander: is there any important thing for me? | 09:48 |
asac | cooloney_: hi | 10:37 |
lool | ogra: Do you manage to run the current versatile lucid linux-image kernel in qemu-system-arm? | 11:45 |
ogra | lool, havent tried yet but i plan to add a download function for the package to the lucid rootstock so i'll test soon | 11:46 |
lool | ogra: Would you mind pinging me when you look into this? | 11:46 |
ogra | will do | 11:46 |
lool | I couldn't get any video output or serial console output from it after Uncompressing linjux | 11:46 |
ogra | do you have probs with it ? | 11:46 |
ogra | oh | 11:46 |
lool | ogra: You don't have the modules for the kernels on your people.ubuntu.com pages? | 11:51 |
ogra | i dont think so | 11:51 |
lool | Sadly your cortex-a8 versatile kernel is missing some modules (not sure which) and mountall doesn't like that | 11:52 |
lool | These are also missing on x86, so they might be purposedly configured as modules in Ubuntu kernels in general | 11:53 |
ogra | right, i'll switch to the packaged lucid kernel for all cortex-a8 variants soon | 11:53 |
ogra | but indeed that requires that it works | 11:53 |
asac | ericm_: apw: dove kernel failed ( i guess you noticed) | 12:55 |
apw | asac, FTBFS ? | 12:56 |
ericm_ | asac, 'm lookin into that | 12:56 |
asac | yes | 12:56 |
asac | ftbfs | 12:56 |
asac | thanks ericm_ | 12:56 |
apw | ericm_, asac i've already uploaded a fix for it | 12:56 |
apw | https://edge.launchpad.net/ubuntu/+source/linux-mvl-dove/2.6.32-200.6 | 12:56 |
ericm_ | missing module ecb | 12:57 |
ericm_ | command exited with status 1 | 12:57 |
ericm_ | make[2]: *** [do-binary-udebs] Error 2 | 12:57 |
ericm_ | make[1]: *** [binary-udebs] Error 2 | 12:57 |
ericm_ | make: *** [binary-arch] Error 2 | 12:57 |
ericm_ | dpkg-buildpackage: error: /usr/bin/fakeroot debian/rules binary-arch gave error exit status 2 | 12:57 |
ericm_ | apw, thanks | 12:57 |
asac | apw: great. sorry for the noise ;) | 12:58 |
apw | ericm_, np ... just a stupid d-i interaction that we don't test well before upload | 12:58 |
apw | i've tested udeb generation on this one | 12:58 |
asac | sounds good | 12:58 |
ericm_ | cool | 12:58 |
apw | asac, np better you tell me and i'm already doing it than not | 12:58 |
cooloney | asac: i am back | 12:58 |
apw | i have a lot of kernels in flight right now | 12:59 |
asac | cooloney: hi. | 13:00 |
asac | cooloney: wanted to touch base with you on NEON status | 13:00 |
asac | cooloney: so dmart pointed out that just exporting diffferent hwcaps based on board revision might not be good enough | 13:00 |
asac | e.g. user space apps could trash the board | 13:00 |
asac | any ideas? | 13:01 |
ogra | asac, hmm ? the patch was already uploaded :) | 13:01 |
ogra | its looking at the board revision number | 13:01 |
ogra | and only enables it on 3.0's | 13:01 |
asac | right. but what does it do | 13:01 |
asac | does it fully disable NEON if board is too old in kernel | 13:02 |
ogra | export the neon flag | 13:02 |
asac | or just tweaks hwcaps | 13:02 |
ogra | it disables the cpu flag | 13:02 |
ogra | - if ((fmrx(MVFR1) & 0x000fff00) == 0x00011100) | 13:02 |
ogra | + if (((fmrx(MVFR1) & 0x000fff00) == 0x00011100) | 13:02 |
ogra | + && cpu_is_mx51_rev(CHIP_REV_3_0) > 0) | 13:02 |
ogra | elf_hwcap |= HWCAP_NEON; | 13:02 |
ogra | hwcaps | 13:02 |
cooloney | ogra: right | 13:03 |
cooloney | asac: ogra is right, | 13:03 |
asac | right. so thats the point | 13:04 |
asac | someone might argue its a security issue (DoS) that user space apps can run NEON instructions without getting a SIGILL on boards that dont support it | 13:05 |
asac | that was the argument made yesterday | 13:05 |
cooloney | asac: generally, there is no 'neon' when you run 'cat /proc/cpuinfo' on TO2 or TO1 | 13:05 |
asac | cooloney: yes, but apps can still do it ... e.g. it doesnt cause a SIGILL | 13:06 |
lool | Worse, it can actually hang the board | 13:06 |
asac | lool: thats the consequence e.g. the DoS i talked about above | 13:07 |
asac | cooloney: how hard would it be to make it so the kernel behaves as if CONFIG_NEON=n if its an old board? | 13:08 |
cooloney | asac and lool, right, probablly | 13:08 |
ericm_ | asac, is the gvfs crash issue filed on LP? | 13:09 |
cooloney | asac: if CONFIG_NEON=y, it will enable some assemble code which is hard to do dynamical disable. | 13:09 |
cooloney | asac: but I will check them later | 13:10 |
NCommander | cooloney, what does this code do specifically? | 13:10 |
ogra | cooloney, would you see a prob if we enabled CONFIG_TIMER_STATS and CONFIG_HPET_TIMER on imx51 and dove ? powertop would like to have these | 13:11 |
ericm_ | CONFIG_TIMER_STATS has been enabled on Dove | 13:12 |
ericm_ | lemme check HPET | 13:12 |
ogra | ah, imx51 seems to not have it | 13:12 |
cooloney | ogra: OK, i will check them on imx51 | 13:12 |
ogra | want a bug for it ? | 13:12 |
ericm_ | ogra, sure - file both for imx51 and dove so we can check | 13:13 |
ogra | well, probably better since i can link it to the spec | 13:13 |
ogra | right | 13:13 |
ericm_ | ogra, not sure HPET_TIMER or HIGH_RES_TIMERS | 13:13 |
ericm_ | but HPET_TIMER sounds to me like x86? | 13:13 |
ogra | powertop moans about HPET_TIMER | 13:13 |
ogra | well ... | 13:13 |
ogra | its an inel app :) | 13:14 |
ogra | *intel | 13:14 |
ericm_ | right, checked again: arch/x86/Kconfig:config HPET_TIMER | 13:14 |
ericm_ | mmm... maybe we need to figure out the dependency of powertop over HPET_TIMER now | 13:14 |
ogra | ok, then we only wasnt TIMER_STATS | 13:14 |
ogra | no | 13:14 |
ogra | its HPET_TIMER is just a suggestion it makes | 13:15 |
ericm_ | ok | 13:15 |
ogra | TIMER_STATS is actually needed for it to work | 13:15 |
ericm_ | ogra, ok - file one for dove as well so we can check if that's working all right in 200.6 | 13:16 |
ericm_ | thanks | 13:16 |
ogra | will we have new uploads before next alpha ? | 13:17 |
cooloney | NCommander: those .S code of NEON is in arch/arm/kernel/entry-armv.S | 13:17 |
ogra | else i wont target it to a milestone | 13:17 |
cooloney | NCommander: generally, it will handle some NEON code and call neon version handler for VFP | 13:18 |
NCommander | cooloney, my ARM ASM is a bit rusty, but we should be able to stuff a check in there to dynamically determine if its a TO1/TO2 board, and jump past the NEON access code | 13:19 |
cooloney | NCommander: right, I know that, but still need to find a way to determine that. heh | 13:21 |
NCommander | ogra, is there a way in userland to determine if its a TO1/2 board, or if its a TO3 board | 13:21 |
ogra | NCommander, did you test suspend/resume on the dove = | 13:21 |
ogra | NCommander, /proc/cpuinfo | 13:22 |
ogra | the revision string | 13:22 |
NCommander | ogra, not yet | 13:22 |
ogra | NCommander, would be nice if you did and checked if it works from deskop menu (that would prove the pm-utils scripts work correctly) | 13:22 |
NCommander | ogra, assuming I can even get to the desktop | 13:23 |
ogra | on imx51 there are no issues so i'm inclined to say they work fine but i need confirmation from dove too | 13:23 |
ogra | i thought the desktop works now for a while | 13:23 |
ogra | at least thats what i heard about the recent liveimages | 13:23 |
ogra | if you manage to click before the panel crashes it should work as a test :) | 13:24 |
ogra | NCommander, i see "dove kernel to be uploaded with CONFIG_HIBERNATION set: TODO" on the pm spec, didnt that happen already ? | 13:25 |
ericm_ | suspend/resume on dove still have a regression, yet I believe it's because I'm still using a DDR2 DIMM here | 13:26 |
ericm_ | NCommander, you have a DDR3 DIMM with you, which you may help test a bit | 13:26 |
NCommander | ericm_, I have DDR2 systems. | 13:26 |
NCommander | I think GrueMaster has a DDR3 Dove | 13:26 |
ogra | ericm_, but was CONFIG_HIBERNATION set in the config ? | 13:27 |
ogra | i see no changelog entry saying so | 13:27 |
ogra | ericm_, Bug #502983 | 13:28 |
ubot4 | Launchpad bug 502983 in linux-mvl-dove "CONFIG_HIBERNATION needs to be set for dove kernels" [Undecided,Confirmed] https://launchpad.net/bugs/502983 | 13:28 |
ericm_ | ogra, never succeeded in resuming from hibernation, yet Marvell engineer thought it works | 13:31 |
ogra | well, either say its unsupported and close the bug or lets fix it :) | 13:32 |
ericm_ | ogra, same reason - I'd suspect a regression in DDR re-initialization which only valid for DDR3 | 13:32 |
ogra | i just want the spec item gone :) | 13:32 |
ericm_ | ogra, think I'll fix it | 13:32 |
ogra | ok | 13:32 |
ericm_ | anyone knows how to upload the existing crash report to LP? | 13:34 |
NCommander | ericm_, use ubuntu-bug *crash-file* on the machine it happened | 13:34 |
ogra | apport has some commandline switch to attach to bugs iirc | 13:34 |
ericm_ | NCommander, 'k | 13:35 |
ogra | NCommander, for your "[mcasadevall] check cpufreq functionallity on all boards (1 day): TODO" | 13:35 |
ogra | root@babbage2:~# echo powersave >/sys/devices/system/cpu/cpu0/cpufreq/scaling_governor | 13:36 |
ogra | root@babbage2:~# cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor | 13:36 |
ogra | powersave | 13:36 |
ogra | root@babbage2:~# cat /proc/cpuinfo |grep Bogo | 13:36 |
ogra | BogoMIPS: 159.90 | 13:36 |
NCommander | cooloney, you can get the machine_id for an ARM chip with the mrc instruction, or by using the read_cpuid_id function | 13:36 |
NCommander | (in the kernel of course) | 13:36 |
ogra | root@babbage2:~# echo ondemand >/sys/devices/system/cpu/cpu0/cpufreq/scaling_governor | 13:36 |
ogra | root@babbage2:~# cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor | 13:36 |
ogra | ondemand | 13:36 |
ogra | root@babbage2:~# cat /proc/cpuinfo |grep Bogo | 13:36 |
ogra | BogoMIPS: 799.53 | 13:36 |
NCommander | ogra, I assume I need a lucid userland for that? | 13:36 |
ogra | NCommander, can you do the same on a dove | 13:36 |
ogra | yes | 13:36 |
NCommander | ugh | 13:36 |
* NCommander has been trying not to use a lucid userland but a lucid chroot | 13:37 | |
ogra | come on, boot a liveimage | 13:37 |
NCommander | ogra, let me grab one and I'll test | 13:37 |
ogra | takes a minute to test that and we're done with the spec item | 13:37 |
ogra | and make asac happy :) | 13:37 |
ogra | lets get below trendline before sprint ! :) | 13:37 |
NCommander | \ooo/ | 13:38 |
NCommander | ogra, download in progress | 13:38 |
ogra | download ? | 13:38 |
ogra | you dont have one locally ? | 13:38 |
NCommander | ogra, I need to get a recent live image | 13:38 |
NCommander | I'm zsyncing it | 13:38 |
* NCommander 3> zsync | 13:39 | |
ogra | GrueMaster, ! | 13:39 |
* NCommander has his ia64 dualbooting Debian and Ubuntu | 13:41 | |
* NCommander feels he's playing with powers that should better be left alone ... | 13:41 | |
* ogra points NCommander to #ubuntu-ia64 :P | 13:41 | |
NCommander | ogra, its lonely there :-P | 13:41 |
ogra | i can imagine | 13:42 |
ericm_ | Mmmm...., cpufreq on dove actually turned on by command line "cpufreq_enable" | 13:42 |
ogra | can we turn that on by default ? | 13:42 |
ericm_ | ogra, sure | 13:43 |
ogra | in kernel i mean | 13:43 |
ogra | afaik initramfs tries to switch them | 13:43 |
ericm_ | yes, it's just controlled by a simple variable | 13:43 |
ogra | to speed up the boot | 13:43 |
ericm_ | I'll try to file a bug report and patch for that | 13:43 |
ogra | i.e. we start with performance until the rootfs is up and then switch to ondemand iirc | 13:43 |
ogra | thanks | 13:44 |
ericm_ | np | 13:44 |
ogra | ericm_, can you give me the bugnumber if you are done so i can add it to the spec | 13:44 |
ogra | (done filing i mean) | 13:44 |
ericm_ | sure | 13:44 |
ericm_ | moment | 13:44 |
ogra | thanks :) | 13:44 |
asac | chroot: cannot run command `debootstrap/debootstrap': Exec format error | 13:47 |
asac | still getting that even with binfmt_misc | 13:47 |
ogra | try a distro kernel | 13:47 |
ogra | i'm pretty sure there are bits missing in the vanilla build | 13:48 |
ogra | there is a procps value that needs to be set, vanilla might not allow that | 13:48 |
asac | ogra: you mean the kernel i am running? | 13:48 |
ogra | yes | 13:48 |
asac | ogra: i am running an ubuntu kernel | 13:49 |
asac | Linux tinya 2.6.31-17-generic #54-Ubuntu SMP Thu Dec 10 16:20:31 UTC 2009 i686 GNU/Linux | 13:49 |
ericm_ | ogra, bug 513254 | 13:49 |
ubot4 | Launchpad bug 513254 in linux-mvl-dove "[dove] CPUFREQ isn't turned on by default" [Undecided,New] https://launchpad.net/bugs/513254 | 13:49 |
ogra | thanks so much ! | 13:50 |
ericm_ | np | 13:50 |
ogra | asac, so i'd say the pm spec looks a lot better now | 13:54 |
asac | ogra: you updated the items? | 13:55 |
asac | already talked to jk? | 13:55 |
ogra | yes | 13:55 |
ogra | nope | 13:55 |
ogra | just DONE'd the low hanging fruits | 13:55 |
asac | ok i have that on my list still | 13:55 |
asac | great!! | 13:55 |
* asac will upgrade laptop to lucid today he thinks | 13:56 | |
* ogra too i want lucid at the sprint | 13:56 | |
ogra | but i'm scared a bit | 13:57 |
asac | ogra: i will let you know ;) | 13:57 |
asac | how it works out | 13:57 |
ogra | ok | 13:57 |
asac | have to finish my calls today first | 13:57 |
asac | but then i can risk the upgrade ;) | 13:58 |
ogra | i'm not even on latest karmic though | 13:58 |
ogra | i should probably first enable -updates and go to the most recent packages :) | 13:58 |
* ogra does that | 13:58 | |
asac | for me its really scary because i didnt upgrade that late in the past | 13:58 |
asac | always feels safer to upgrade in small chunks than going for the big jump ;) | 13:59 |
asac | even though the ground might be more solid | 13:59 |
ogra | yeah, same here | 13:59 |
ogra | i usually upgrade during A1 | 13:59 |
asac | right. thats what i feel comfortable with too | 13:59 |
ogra | oooouff | 14:01 |
ogra | 287 packages | 14:01 |
=== bizkut-miau is now known as bizkut-redhat | ||
armin76 | asac: get me a pass for mwc :) | 14:27 |
suihkulokki | armin76: unless you regularry wear a tie mwc is very boring | 14:30 |
armin76 | suihkulokki: but i can get boards! :P | 14:31 |
armin76 | zumbi: you going to mwc? | 14:32 |
zumbi | armin76: mwc? what is it? i am going to fosdem, btw | 14:37 |
zumbi | suihkulokki: fosdem this year? | 14:37 |
armin76 | zumbi: mobile world congress | 14:38 |
suihkulokki | neither.. | 14:38 |
suihkulokki | see you both in debconf10 ;) | 14:39 |
zumbi | suihkulokki: we'll miss you then | 14:39 |
NCommander | suihkulokki, ooh, lucky. MWC will be fun | 14:43 |
* NCommander is looking forward to commuting to his second debconf | 14:44 | |
=== _bjf is now known as bjf | ||
=== ogra_ is now known as ogra | ||
zumbi | armin76: i do not think i am going to mwc, but if you get tickets I can partner you, just to kill the curiosity of the newer fancy screen devices | 15:37 |
zumbi | armin76: btw, attending debconf this year? | 15:37 |
* ogra dist-upgrades to see if he can reproduce plars' nautilus crash | 15:51 | |
plars | ogra: yes, it's not good. Trying to see if I can manually get a backtrace right now, but not having much luck finding the right -dbg packages to pull in | 15:53 |
ogra | plars, i definately dont get nautilus in my upgrade | 15:54 |
ogra | http://paste.ubuntu.com/363941/ | 15:54 |
ogra | i suspect its rather one of the libs | 15:55 |
ogra | libglib2.0-0 smells like a good candidate here | 15:55 |
ogra | plars, oh | 15:55 |
ogra | plars, can you check the version of libgtk2.0-common vs the libgtk2.0 package ? | 15:56 |
ogra | i think common is arch: all and gtk itself ftbfs | 15:56 |
ogra | so that could be our issue | 15:56 |
plars | yes, could be | 15:58 |
plars | libgtk2.0-0 2.19.3-1ubuntu4 | 15:58 |
plars | libgtk2.0-common 2.19.4-1ubuntu1 | 15:58 |
GrueMaster | ogra: what? | 15:58 |
ogra | aha | 15:58 |
ogra | GrueMaster, nothing anymore, i wanted to ask you for a quick cpufreq test on dove but it turned out its not yet enabled in kernel | 15:59 |
GrueMaster | ok | 15:59 |
* GrueMaster needs to remember to have a discussion on time zones during sprint. | 15:59 | |
ogra | GrueMaster, yeah, i only checked my clock after i pinged, sorry :) | 16:01 |
plars | GrueMaster: I thought we agreed to abolish them? | 16:09 |
GrueMaster | I must have missed that memo (like so many others). | 16:10 |
armin76 | zumbi: nope | 16:35 |
armin76 | why do i want to go to a debian conf? :P | 16:36 |
zumbi | armin76: to meet NCommander et al., so you can ask for boards more effective | 16:41 |
NCommander | zumbi, you going to be at debconf 10? | 16:42 |
armin76 | zumbi: you can ask on my behalf :D | 16:42 |
zumbi | NCommander: probably, i am thinking to get tickets rather soonish than later that it'll be omre expensive. I am unsure if dates are already fixed | 16:43 |
NCommander | zumbi, the dates are fixed to my knowledge | 16:43 |
NCommander | armin76, just wait for the new Sheevaplug ;-) | 16:43 |
zumbi | armin76: they know me, so they do not send me anything.. too slugish.. :-P | 16:43 |
armin76 | NCommander: i don't want to! | 16:44 |
zumbi | NCommander: how are new sheevas? do they come with eSATA? or add wifi? | 16:44 |
* zumbi kicks armin76 to NYC | 16:45 | |
NCommander | zumbi, I dunno, haven't seem then yet | 16:45 |
* NCommander steals armin76's SPARC and ia64 systems | 16:45 | |
NCommander | bahaha | 16:45 |
zumbi | http://www.linuxfordevices.com/c/a/News/Marvell-Plug-Computer-30-and-Armada-300-and-610/ | 16:47 |
* zumbi wants http://www.linuxfordevices.com/c/a/News/Mindspeed-Transcede-T4000-and-T4020/ | 16:48 | |
zumbi | armin76: last processors are shown at mwc. If you persuade somebody for tickets | 17:04 |
plars | ogra: downgrading libgtk2.0-common didn't help anything either | 17:09 |
plars | ogra: in case you missed my last... | 17:11 |
plars | ogra: downgrading libgtk2.0-common didn't help anything either | 17:11 |
ogra | yes, i know | 17:12 |
ogra | see -devel | 17:12 |
ogra | i did that | 17:12 |
* ogra somehow had a complete network breakdown over here | 17:12 | |
plars | ogra: ah, I saw that you had tried glib and gnome-session, wasn't sure about others | 17:14 |
ogra | i i actually upgraded gtk :) | 17:15 |
ogra | manually pulling debs from LP | 17:15 |
ogra | but its surely not get | 17:15 |
ogra | gtk | 17:16 |
ogra | sigh, that focus stealing of update-manager is annoying | 17:16 |
plars | ogra: got it | 17:31 |
plars | ogra: gvfs | 17:31 |
plars | ogra: looks to be probably libgvfscommon0 | 17:31 |
NCommander | Apple finally announced a tablet | 18:35 |
NCommander | O_o; | 18:35 |
Stskeeps | it's kinda disappointing. | 18:35 |
ojn | haha | 18:35 |
Stskeeps | and embarassing from a ui invention and technical pov (check their pixel doubling to fit native iphone apps..) | 18:36 |
ojn | stskeeps: what else could they have done? | 18:36 |
ojn | I bet most apps will be updated to handle the new screen, it's just a stopgap thing. | 18:37 |
Stskeeps | ojn: it looks like a cheap copy of moblin. | 18:37 |
persia | Pity they went for such a low DPI. Historically they've been pretty good at making sure stuff renders sharply. | 18:37 |
ojn | persia: apple has always been low-dpi. they're behind most other PC desktops/laptops on dpi as well. | 18:37 |
ojn | stskeeps: ah, you're a maemo guy. that explains the bitterness. :) | 18:38 |
persia | ojn: Hrm? That's not been my experience. Mind you, I went and got the special 300DPI handheld from Sharp back when it came out, but in terms of mainstream stuff they seem to go for higher DPI than other companies. | 18:38 |
ojn | persia: not for laptops. Took them forever to come out with a 1440x900 15" laptop, and they're still at that resolution for those screens. 17" has a highres option though. | 18:39 |
persia | ojn: Hrm. Strange. I haven't looked at their laptops in a *long* time, but the marketing materials still talk about high-DPI. | 18:39 |
ojn | everything is relative. It's quite appropriate DPI for most things, I'm not complaining. But thinkpads and some other brands have quite a bit higher resolution screens. Too high for some use, I would say. But nice for programming | 18:40 |
persia | Oh well, I guess the low DPI is in line with other things then :) | 18:40 |
persia | There is no such thing as too much DPI : you just get better rendering :) | 18:40 |
ojn | right, and bigger fonts. and apps that handle that nicely (and that's the catch, not all do) | 18:41 |
persia | Those are bugs :) | 18:41 |
persia | But the fonts should be the same size. Those are supposed to be defined in ems not pixels. | 18:41 |
persia | Or points, but that's related to ems. | 18:42 |
=== beckyb_ is now known as beckyb | ||
armin76 | rabeeh: get me a pass for mwc2010! | 19:13 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!