Tuesday, 15 December 2015

BFS 467, linux-4.3-ck3

Announcing an updated BFS for linux-4.3 based kernels.

BFS by itself:
4.3-sched-bfs-467.patch

-ck branded linux-4.3-ck3 patches:
4.3-ck3

After my initial enthusiasm regarding the improved throughput on the previous BFS release, I unfortunately had some reports of regressions in interactive behaviour for the first time in a while, both on this forum and through other channels. So for the first time in a while, I've released a -ck3 with yet another updated BFS to address this since I can't stand having a dodgy release out for any extended period.

With this release what I've done is reinstate an old tunable that used to be on my scheduler patches many years ago

/proc/sys/kernel/interactive

This has two settings, 1 for on and 0 for off. By default - of course since this is BFS - it is set to 1. What it does is prioritise latency over throughput in mode 1 and vice versa in mode 0. In addition to addressing the latency issues in the previous kernel, mode 1 actually completely turns off all soft affinity scheduling in the kernel, for the lowest possible latencies all round, so this may be the first kernel with an improvement in latency in a while too.

Bear in mind none of these changes make any difference on uniprocessor kernels so there is no need for UP users to update unless they need the build fixes that came with BFS466+.

Amusingly enough, linux-4.3.2 wouldn't boot for me for unrelated reasons so I'm using 4.3.0-ck3 myself. So if you can't get 4.3.2 to boot, roll back.

Enjoy!
お楽しみください
Posted by at
Labels: , , ,

39 comments:

  1. Oleksandr Natalenko16 December 2015 at 01:46

    4.3.3+0.467 booted OK for me.

    ReplyDelete
    Replies
    1. Anonymous19 December 2015 at 03:11

      Has you closed pf-kernel forum due to the success of Xanmod kernel?

      Delete
    2. Oleksandr Natalenko19 December 2015 at 04:06

      I've pointed the exact reason in announcement.

      Delete
    3. Anonymous19 December 2015 at 04:40

      I can really understand this step. Of course it is not your job to take over all discussions for particular patch issues, that should be directed to their maintainers directly, normally... ^^
      On the other hand I liked your forum for it's dialog function and as a source of inspiration and exchange. Thank you for your previous work on that area, it was not useless at all! I wish there was a feedback function within your new blog for the posts, so that others' comments are visible, but that would not lead to your goal, I assume.

      Anyway, best regards, keep on your appreciated "core" work,
      Manuel Krause

      Delete
    4. Anonymous19 December 2015 at 05:09

      After nine months of silence, and after letting people to post the compiled kernels day after day. At least you would give a couple of days to let people to recover the work done, or deleting some posts, or simply making some advice. It has been the major disrecpectful behaviour to the users that I have ever seen. If you dislike something, don't blame all people. Good bye, and txh x 1^10000000000000000.

      Delete
    5. Anonymous19 December 2015 at 05:36

      This is getting too off-topic and I hope we will stay fair && friendly. @Anonymous: It is post-factum's decision, what he wants to maintain or not.
      The kernels won't get forgotten and/or can be rebuilt each time you want. For safety reasons it's useful to keep up with most recent kernels and you should not need all the old things (even if they were good).

      BR, Manuel Krause

      Delete
    6. Anonymous19 December 2015 at 06:29

      Thank you, Manuel Krause, to be so polite.

      Delete
    7. Oleksandr Natalenko19 December 2015 at 07:46

      Thanks for feedback.

      Delete
    8. Anonymous19 December 2015 at 09:48

      Xanmod, if you read this, please come back!

      Delete
    9. Anonymous19 December 2015 at 22:45

      "Sometimes it is necessary to be lonely in order to prove that you are right." Vladimir Putin.

      Delete
    10. Anonymous20 December 2015 at 21:54

      Pf-kernel is not longer active?

      Delete
    11. Anonymous21 December 2015 at 11:49

      Xan has opened his forum. Con Kolivas BFS is the best patch ever built!

      Delete
    12. Oleksandr Natalenko21 December 2015 at 17:14

      Anonymous should read my announcement more carefully.

      Delete
    13. Anonymous21 December 2015 at 20:45

      Thanks for feedback.

      Delete
  2. Anonymous16 December 2015 at 04:27

    4.3.3 and bfs467 also booted OK for me. :)

    ReplyDelete
  3. Anonymous16 December 2015 at 05:21

    @post-factum & @Anonymous:
    Any issues with s2ram or router as with -VRQ?
    How do you judge the current latency/throughput related approach?

    Thanks,
    Manuel Krause

    ReplyDelete
    Replies
    1. Oleksandr Natalenko16 December 2015 at 08:19

      No issues so far, need to test more.

      Delete
    2. Anonymous16 December 2015 at 11:01

      Perfect for me after four hours using it. :)

      Delete
  4. ooo16 December 2015 at 06:15

    boots fine against 4.3.3 for me too, and I haven't noticed issues so far. It actually feels snappier than 4.3.0-ck1.

    Thanks, and great to hear you're still continuing to improve bfs :)

    ReplyDelete
    Replies
    1. ck16 December 2015 at 06:49

      Yes as I said in the announce post, with interactive enabled, this kernel should have a slight interactive improvement over any previous BFS to date.

      Delete
    2. Anonymous16 December 2015 at 06:58

      Hi, Con,
      if I read the code changes correctly, setting interactive to 0 would lead to the BFS 466 behavour?

      BR, Manuel Krause

      Delete
    3. ck16 December 2015 at 07:01

      Yes that is correct.

      Delete
  5. jwh722 December 2015 at 02:17

    Fwiw, 4.3.3-ck bfs467 UP is running a-ok on my old netbook; still getting the (upstream I presume) warnings at boot:
    WARNING: CPU: 0 PID: 65 at drivers/gpu/drm/drm_atomic.c:491 drm_atomic_check_only+0xed/0x3ef()

    ReplyDelete
    Replies
    1. Anonymous22 December 2015 at 07:38

      @jwh7: Are you the one, who reported this bootup WARNING earlier on here? I get this too, and I blame the relevant driver for it, not BFS. In my case it's the i915 for my intel GM45 integrated gfx compiled as module. I don't remember exactly whether this goes away if you compiled it into the kernel. But what I remember is, that in this case you would get different error messages if you resumed from suspend to disk (gfx reset). And when using TuxOnIce for that, reliability can decrease. (Some unpredictable timing issue.) Have you filed a bugreport on bugs.freedesktop.org already, or attached to one existing?
      I can only hope 4.4 would correct these things. At least it shows only once per boot and doesn't harm normal operation.

      Best regards,
      Manuel Krause

      Delete
    2. jwh726 December 2015 at 09:10

      Sorry Manuel I was out of town... yes that was me. I agree; by upstream I meant of the kernel, not BFS. :) But ya I have seen lots of Intel DRM changes for the i915 driver going into 4.4. /fingers crossed
      I had not file a bug report as I originally thought it was the same warning filed at: https://bugs.freedesktop.org/show_bug.cgi?id=93104
      ...but later realized it was not. Given the 4.4 merges and that its just a warning, I hadn't worried about it yet (and the holidays haven't helped either). :)

      Delete
    3. Anonymous27 December 2015 at 07:07

      No problem. ;-) Just only want to thank you for coming back and clarifying this off-topic.
      The DRM BUGs are so confusing -- regarding the warning I also attached to one -- and it most probably isn't the right one and isn't answered. And I've several other tabs open with possibly related issues' candidates. Almost useless to dig through all of that stuff.
      I agree, fingers crossed for kernel 4.4.

      BR Manuel Krause

      Delete
  6. Anonymous23 December 2015 at 01:54

    running on 4.3.3 flawlessly and faster than ever. thanks so much. much appreciated. never stop xD

    ReplyDelete
    Replies
    1. Anonymous27 December 2015 at 05:38

      4.3.3 with 467 ck3 was flawless for me for 10 days uptime, but have now had two CPU soft lockups [BFS]. Systemd decided to wipe out my journels both times, so I don't have the exact error unfortunately.

      Delete
    2. Anonymous29 December 2015 at 04:02

      ^^ had one lockup yesterday on 4.3.3.
      my logs are on ramdisk so i can't provide further information unfortunately

      Delete
    3. Anonymous30 December 2015 at 04:03

      Well, it's good to know I'm not alone. Wish I had more debug info. I went back to vanilla 4.2.8 for now just in case this is somehow related to the 4.3.3 upstream and not ck/bfs.

      Delete
  7. Anonymous25 January 2016 at 16:40

    Will that patch work on linux-4.4 ?

    ReplyDelete
  8. Alfred Chen3 February 2016 at 20:16

    For those who want to try bfs on 4.4, my bfs port and enhancement patch for 4.4 have updated at http://cchalpha.blogspot.com/2016/01/v440-vrq2-released.html

    Have fun with 4.4 release.

    BR Alfred

    ReplyDelete
    Replies
    1. Anonymous9 February 2016 at 14:05

      Thank You Very Much.

      Delete
    2. Anonymous12 February 2016 at 02:38

      from your words:
      "Holding back some feature commits like removal of SMT_NICE code."

      Why not more smt_nice feature, what issues with it?
      I have had it as enabled in all older kernels with bfs.

      BR NicCo

      Delete
  9. Holger Hoffstätte3 March 2016 at 07:11

    After trying to use 4.4.x with CFS for a while I couldn't take it anymore and decided to forward-port the last BFS v467 with cherry-picked changes by Alfred (thanks!), but without his VRQ changes. This gives us 4.4.x with "plain" BFS, with excellent interactive results and no problems with stock suspend/resume (YMMV depending on weird extensions). I've sent this to Con and he mentioned I should post here in the meantime until gets around to do an official release.

    You can find the all-in-one patch here:
    https://raw.githubusercontent.com/hhoffstaette/kernel-patches/master/4.4/bfs-467.patch

    Simply patch that into a plain 4.4.x and you're good to go.

    Please see the additional changelog in the header for the (very few) changes I had to make to merge everything. All credits for the original 4.4 patch go to Alfred Chen.

    While this was made against 4.4.3, it also applies cleanly and works fine with 4.4.4-rc1 as of last night.

    Enjoy. :)

    ReplyDelete
    Replies
    1. Alfred Chen6 March 2016 at 10:55

      Hi, Holger,
      Do you still have suspend/resume issue with the latest vrq3 patch from mine?

      BR Alfred

      Delete
    2. Holger Hoffstätte7 March 2016 at 10:42

      I only ever tried the first v466-based vrq version and reverted when it had the same interactivity problems under load as plain bfs-466. Never tried suspend/resume myself, just read the comments by others.

      Delete
  10. Anonymous3 March 2016 at 10:11

    Thanks so much!

    ReplyDelete
  11. ulenrich4 March 2016 at 11:45

    Very thanx Holger,
    because I am a git ignorant I could not find the commit at Chens repository. Bfs runs without issue with linux-4.4rc1 on an old Mac-mini of 2009

    ReplyDelete

Subscribe to: Post Comments (Atom)