/srv/irclogs.ubuntu.com/2012/03/23/#ubuntu-kernel.txt

ckingyawn08:00
* smb agrees to yawn08:12
ckingsmb, hrm, has mumble died again08:22
smbcking, Somewhere. Not sure here or there08:22
smbAt least you seemed to have gone silent all of a sudden08:23
ckingugh, let me kick it on my end08:23
smbBut taht could be me or you08:23
ckingor both ;-)08:23
smbtrue :)08:23
smbNow we have to wait for a third person to join08:23
ckingi'm getting connection refused now08:23
ckingah, now back in08:24
apwsmb, is it safe to update today?08:57
smbapw, No idea, I decided (oh well just not did) against it08:57
apwahh well, will give it a go, this is supposed to be B2 after all08:57
smbits friday, what can possibly go wrong...08:58
apwindeed08:58
apwanother shiney kernel yay08:58
* cking updated and it works fine 09:00
apwcking, then there is hope09:04
apwget any new unity features today ?  pretty sure we are expecting some09:05
smbapw, So sad we don't have new kernel features every Friday... :-P09:38
apwi claim landing them weds so that everything every09:38
apwbody lands on a thursday is out of date, is better09:38
apw.quit09:46
ppisatiapw: actually the stuff tim pushed covered only tools/10:58
apwppisati, yeah indeed.  those patches i do not recognise10:58
apwppisati, though they are as invasive as an invasive thing on invasive juice10:59
ppisati:)10:59
ppisatiyou mean the two i posted on people?11:00
apwppisati, yes11:13
apwppisati, it affects the kernel build rather than the packaging, so it is more invasive11:14
ppisatiapw: well, but it seems they do what they advertise11:17
ppisatiapw: the real problem is that i can't reproduce the breakage in chroot anymore11:18
apwppisati, those wern't the ones which tiggered the breakage, they were all changes to the packaging11:22
ppisatiapw: they were, the secondo one was reverted in O by Tim due to breakage11:23
ppisatiapw: 9b1520dfc90d3c7f6b21773600a1d48ed85cb33611:23
ppisatiapw: "Breaks cross compile in Oneiric x86 chroots."11:23
apwhmm and are they the same now11:24
ppisatiapw: had to do some modifications to the first one to make it apply11:26
ppisatiapw: but the funny thing is that, if i revert the revert in O11:26
ppisatiapw: kernel builds fine11:26
ppisatidoor bell, brb11:26
apwppisati, odd indeed.  perhaps the chroots are better now11:26
apwwe did use to use external compilers for this, outside the chroots from code saucery, and now we have proper packages from linaro ... so 11:27
apwppisati, and i note we left in the intrusive part ... so it can't have been tooo annoying else we'd have noticed11:28
apwso ... perhaps its time to consider wacking them back on, and let us find out again if its broken11:28
ppisatii'm able to cross compile in a P/amd64 chroot, armhf fails (but that's a qemu issue)11:34
=== smb` is now known as smb
* ppisati -> lunch13:07
* apw wanders to lunch13:25
tgardnercking, re: '[PATCH] ACPI, APEI, fix ERST table size checking' - have you noticed cases where those 2 tables are now different sizes, e.g., with UEFI ?13:27
ckingtgardner, not that I know of14:12
ckingtgardner, but I need to double check now that you've made me wonder14:13
ckingbut it is more to do with I need to check ACPI 514:13
tgardnercking, its definitely a righteous fix14:16
ckingbut it should make no difference since those structs are the same size14:16
ckingaka we got lucky it worked14:17
tgardnerindeed14:17
ckingtgardner, no difference in ACPI 5 either (which is kinda expected)14:18
* cking gets on the phone to whine about his dead laptop, wish me luck14:22
josephtcking: good luck indeed :)14:30
ckinglike being lost in a twisty maze w/o a map14:33
* apw cannot believe how warm it is here14:48
* smb sweats it14:52
* ppisati notes the weather indicator crashed14:59
smbProbably never tested to display more than 19°C...15:01
ppisaticould be15:01
ppisatibtw, how one is supposed to restart a crashed indicator?15:02
ppisatijust relaunch it?15:02
smbbe a good citizen a report a bug?15:02
ppisati21C btw15:03
smb25C! ok... I think the sun reached the sensor...15:03
ppisati:)15:03
smbactually not but it might be a bit sheltered out there on the balcony15:04
ogasawarasmb: has anyone approached you about some LVM and MD memory leak issue?15:09
smbogasawara, nope15:09
smbWell you, now sort of15:09
ogasawarasmb: hrm, sabdfl pinged me about and said Daviey had the details so was wondering if he may have pinged you15:10
ogasawarasmb: I'm just trying to figure out what the actual issue is15:10
tgardnerogasawara, apw mentioned that he was running a 32 bit kernel on a platform with gobs of RAM15:10
smbtgardner, Oh maybe thats all the same15:11
tgardnerand having issues because his 6 core CPU didn't support all of the CPUs15:11
apwogasawara, yes Daviey tallked to me15:11
=== bladernr_ is now known as bladernr_afk
apwogasawara, i am thinking about it, should be getting access to the box shortly15:11
ogasawaraapw: ack.  is there a bug filed?15:12
apwhttps://bugs.launchpad.net/ubuntu/+source/linux/+bug/96299215:12
ubot2`Launchpad bug 962992 in linux "OOM when using a large amount of RAM, on i386/smp, when high disk IO" [Undecided,Confirmed]15:12
apwbasically the machine is way over spec to work well with 32bit15:12
apwi am about to writ eit up and put it in the bug15:12
ogasawaraapw: ok thanks15:13
ogasawarawhoa, 32 gigs of ram?!15:13
* smb already had erased that from his head (that it initially talked about lvm/md)15:13
apwogasawara, 4215:14
ogasawarawhoa++15:14
apwyeah, no way i would expect that machine to work on 32bit, but the failure more is catasrophic15:15
apwso i am thinking we'll do something to mitigate it like drop the end of ram or something15:16
ckingthe e820 table is eye-popping on that bug15:20
smbcking, Yeah I think that makes the problem even worse on machines that are able to handle that much ram... or its something else15:21
smbbut on my server board the lowmem is not even enough to run memtest15:21
smb(a huge chunk taken away by bios tables)15:21
ckingBIOS-e820: 0000000100000000 - 0000000c40000000 (usable) - that's something you don't see everyday15:22
apwtgardner, hmm that isci.ko thing, if its isci.ko then that is a real scsi disk controller, and that is not in d-i17:29
apwapw@dm$ find /lib/modules/`uname -r` -name isci.ko17:29
apw/lib/modules/3.2.0-20-generic/kernel/drivers/scsi/isci/isci.ko17:29
apwapw@dm$ git grep isci.ko -- debian.master/d-i17:29
apwapw@dm$ 17:30
tgardnerapw, agreed, but its not a HW driver17:30
apwtgardner, i think isci is a h/w driver:17:31
apw        tristate "Intel(R) C600 Series Chipset SAS Controller"17:31
apw          This driver supports the 6Gb/s SAS capabilities of the storage17:31
tgardnerapw, hmm, maybe you're right. drivers/scsi/isci/init.c17:32
tgardnerit is doing a PCI register17:32
apwyeah i think its some new thing, that they have (stupidly) named very similarly to many other utterly different things17:33
tgardnerapw, ok< i had a patch to add it to d-i. I'll ppush it shortly.17:33
tgardnerI might as well add the firmware to the linux-firmware udeb at the same time17:34
apwyeah... bound to need the firmware ...17:34
bryceh-20 failed to boot for me just now; want a bug report or is this already known?17:38
tgardnerbryceh, ASPM ?17:41
tgardnerjsalisbury, ^^17:41
jsalisburybryceh, know issue: bug 96148217:44
ubot2`Launchpad bug 961482 in linux "3.2.0-19 kernel fails to boot (-18 OK)" [Critical,Confirmed] https://launchpad.net/bugs/96148217:44
jsalisburybryceh, There is a test kernel in the bug, or you may be able to boot with: pcie_aspm=force17:45
Sarvattbryceh: oh you skipped -19?17:46
brycehSarvatt, yeah this is my main desktop so I don't update it as often as the rest of my hw17:48
brycehjsalisbury, thanks that could  be it.  I have several other systems all running 19 happily (and being upgraded to 20 now), so guessing the desktop is unusual in some way.17:51
lamontsmb: any updates?17:57
smblamont, Err, no. Our chat ended in a way that led me to assume you wanted to check on the phone config ... So I did not do anything on it, yet17:59
smblamont, If you want another debug kernel, please type "yes" <enter> now. ;)18:00
lamontah.  I dug into it some and concluded that it's quite possible that the fine folks at cisco are compliant and that the registration may not come from 506018:01
brycehrebooting to test kernel, brb18:01
lamontthat is, registering is one thing, where to send the reply is found in the UDP headers18:01
lamontyes18:01
lamont:-p18:01
smblamont, :) Ok, well I agree that you can register from another port. But then you should not state otherwise in the headers18:02
=== tgardner is now known as tgardner-lunch
smblamont, Given that it is past b-o-c, I'll get back to it on Monday18:03
lamontsmb: I was meaning that one might actually register using a separate socket (and hence source address) than the socket that one has open to receive SIP traffic18:03
lamontbut dunno18:03
lamontthe biggest challenge is that nearly everyone uses 7640s with a cisco controller, rather than asterisk, as the SIP provider18:04
lamontmaking google not really my friend when it comes to finding answers18:04
smblamont, Its hard to say. Just gathered from the difference between working and non-working example and the rought hinting about some expectation handling on addresses goes wrong18:04
ckingrats, the ethernet cable popped out of the socket 95% the way through a 1 hour test and I lost my data capture from my multimeter. 18:27
=== tgardner-lunch is now known as tgardner
apwcking, that sucks18:51
* apw thinks its time to go and enjoy some british fare ... curry and beer18:51
apwlaters18:51
argesapw, have a good one18:52
smosersmb, you're on friday night?19:06
=== yofel_ is now known as yofel
smbsmoser, not really but the computer is while I am enyoing tv and a bit of wine. :)19:48
smosersmb, well, after you've had another glass of wine, you can have some fun looking at https://bugs.launchpad.net/ubuntu/+source/linux/+bug/96342020:21
ubot2`Launchpad bug 963420 in linux "https download performance significantly worse in precise than lucid" [Undecided,New]20:21
tgardnersmoser, cjwatson just uploaded a new ssl that (I think) takes advantage of HW encryption support. I wonder if that'll make a difference.20:42
tgardnerthe other difference between lucid and precise is that the HW encryption routines are built-in in precise.20:44
smosertgardner, but you wouldn't think that was there in lucid.20:44
tgardnerwell, these are virtual instances, right ?20:44
smoserie, precise shouldnt need that to catch up20:44
smoser(yes, lots could be going on i dont know about)20:44
smoserbut i can't seem to make precise win20:45
tgardneractually, I don't know for sure. I'm just thinking with my fingers.20:45
smoseryeah, i probalby missed the your intent above.20:47
smoserbut the key thing is... i just can't seem to get precise faster than lucid with https20:48
smoseracross multiple instances across multiple days... lucid always wins.20:48
smoserbut with http... they both hit ~ 80M/s even.20:48
tgardnersmoser, well, I'd be curious to know what crypto modules are loaded. that seems to be the root of the problem, right ?20:48
smoserand, i noted in the bug... on EC2, they seem similar.20:48
smoseri have no idea, tgardner 20:48
smoseryou're welcome to my instances if you'd like20:49
smoserto poke around.20:49
smoserive got to run her ein a minute though20:49
tgardnersmoser, as do I. when you get a chance just drop it in the bug report20:49
smoseryou have canonistack creds ?20:49
smoserjust launch some instances20:49
smoseri got to run, so monday probably for me.20:50
tgardnermaybe this weekend...20:50
* tgardner -> EOD20:50
=== rsalveti` is now known as rsalveti

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!