Thanks Joseph! I now believe I'm seeing two different issues in 4.8 that are not in 4.4 and slow down my boot, which is very confusing. - #1: High number of kworkers (fixed in 4.8 using the two patches and bisected by Doug Smythies already) - #2: Unknown other issue. I've noticed that when compiling a simple cmake/ninja project using 6 compiler processes on my rather old 4-core CPU, some kernels show odd behaviour that might also slow down the boot process.
When compiling that cmake project on 4.4, 4.6-7afd16f or 4.7-rc1, issue #2 does not exist. While compilation takes place on 4.8 (no matter if kworker patches are applied or not), my mouse cursor becomes sluggish and it takes a long time to just switch to another window in Unity 7. Compilation speed is unaffected though, so maybe just some kind of task switching problem, not a CPU utilization issue. So that makes it quite difficult to tell if any given kernel is good or bad by just looking at the boot time. As far as I can tell, this is the current list of kernels and the issues they have, where #2 was detected using the cmake compilation behaviour and #1 by the number of kworker processes after boot (should be ~35): 4.8 + patches: #2 4.8: #1, #2 4.7-rc1: #1 4.6-7afd16f: none 4.4: none 4.8 has about the same wall clock boot time as 4.7-rc1 in my case, while 4.8+patches is ~5 seconds faster, but that's hard to tell, since boot times vary by 2-3 seconds anyway. So #1 apparently dominates the slow down. I've only been able to get a "perfect" <10sec boot time from GRUB to lightdm with 4.4 and 4.6-7afd16f. So the order of boot speed is like this right now for me: 4.4|4.6-7afd16f < 4.8+patches < 4.8|4.7-rc1 Sorry for the long-winded comment, I just feel uncomfortable calling this kernel "good" or "bad" without explaining why 4.7-rc1 might actually be "good" on my system concerning issue #2. 4.6-7afd16f appears to be as "good" as 4.4 in any case. -- You received this bug notification because you are a member of Kernel Packages, which is subscribed to linux in Ubuntu. https://bugs.launchpad.net/bugs/1626436 Title: [4.8 regression] boot has become very slow Status in linux package in Ubuntu: In Progress Status in linux source package in Yakkety: In Progress Bug description: With yakkety's recent update from linux 4.4 to 4.8 booting has become a lot slower. It's not one service in particular, but without "quiet" and "splash" you can now easily read every single line instead of that whole wall of text zipping by. It now takes over 20s instead of ~10 seconds to boot. This is even more dramatic when factoring out the recent boot hang of NetworkManager (bug 1622893) and disabling lightdm: sudo systemctl mask NetworkManager NetworkManager-wait-online lightdm then booting with 4.4 takes 1.5s and with 4.8 19.5s (!). Some excerps from systemd-analyze blame: 4.4: 474ms postfix@-.service 395ms lxd-containers.service 305ms networking.service 4.8: 4.578s postfix@-.service 7.300s lxd-containers.service 6.285s networking.service I attach the full outputs of critical-chain and analyze for 4.4 and 4.8 for reference. This is much less noticeable in the running system. There is no immediate feeling of sluggishness (although my system is by and large idle). I compared the time of sbuilding colord under similar circumstances (-j4, building on tmpfs, thus no hard disk delays; running with fully pre-loaded apt-cacher-ng thus no random network delays), and with 4.4 it takes 6.5 minutes and with 4.8 it takes 7.5. So that got a bit slower, but much less dramatically than during boot, so this is either happening when a lot of processes run in parallel, or is perhaps related to setting up cgroups. One thing I noticed that during sbuild in 4.8 "top" shows ridiculous loads (~ 250) under 4.8, while it's around 4 or 5 under 4.4. But that doesn't reflect in actual sluggishness, so this might be just an unrelated bug. ProblemType: Bug DistroRelease: Ubuntu 16.10 Package: linux-image-4.8.0-11-generic 4.8.0-11.12 ProcVersionSignature: Ubuntu 4.8.0-11.12-generic 4.8.0-rc6 Uname: Linux 4.8.0-11-generic x86_64 ApportVersion: 2.20.3-0ubuntu7 Architecture: amd64 AudioDevicesInUse: USER PID ACCESS COMMAND /dev/snd/pcmC0D0c: martin 3049 F...m pulseaudio /dev/snd/pcmC0D0p: martin 3049 F...m pulseaudio /dev/snd/controlC0: martin 3049 F.... pulseaudio Date: Thu Sep 22 09:42:56 2016 EcryptfsInUse: Yes MachineType: LENOVO 2324CTO ProcEnviron: TERM=linux PATH=(custom, no user) XDG_RUNTIME_DIR=<set> LANG=de_DE.UTF-8 SHELL=/bin/bash ProcFB: 0 inteldrmfb ProcKernelCmdLine: BOOT_IMAGE=/@/boot/vmlinuz-4.8.0-11-generic.efi.signed root=UUID=f86539b0-3a1b-4372-83b0-acdd029ade68 ro rootflags=subvol=@ systemd.debug-shell RelatedPackageVersions: linux-restricted-modules-4.8.0-11-generic N/A linux-backports-modules-4.8.0-11-generic N/A linux-firmware 1.161 SourcePackage: linux UpgradeStatus: No upgrade log present (probably fresh install) dmi.bios.date: 07/09/2013 dmi.bios.vendor: LENOVO dmi.bios.version: G2ET95WW (2.55 ) dmi.board.asset.tag: Not Available dmi.board.name: 2324CTO dmi.board.vendor: LENOVO dmi.board.version: 0B98401 Pro dmi.chassis.asset.tag: No Asset Information dmi.chassis.type: 10 dmi.chassis.vendor: LENOVO dmi.chassis.version: Not Available dmi.modalias: dmi:bvnLENOVO:bvrG2ET95WW(2.55):bd07/09/2013:svnLENOVO:pn2324CTO:pvrThinkPadX230:rvnLENOVO:rn2324CTO:rvr0B98401Pro:cvnLENOVO:ct10:cvrNotAvailable: dmi.product.name: 2324CTO dmi.product.version: ThinkPad X230 dmi.sys.vendor: LENOVO To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1626436/+subscriptions -- Mailing list: https://launchpad.net/~kernel-packages Post to : kernel-packages@lists.launchpad.net Unsubscribe : https://launchpad.net/~kernel-packages More help : https://help.launchpad.net/ListHelp
