Announcement

Collapse
No announcement yet.

Linux 2.6.37-rc1 Kernel Is Here; Can Build Without BKL

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Linux 2.6.37-rc1 Kernel Is Here; Can Build Without BKL

    Phoronix: Linux 2.6.37-rc1 Kernel Is Here; Can Build Without BKL

    As anticipated, the 2.6.37 merge window closed yesterday and the first release candidate for the Linux 2.6.37 kernel is now available. Major changes that were pushed into the Linux 2.6.37 kernel include support for building the kernel without the Big Kernel Lock (BKL), many graphics DRM improvements, and more of the responsiveness patches...

    http://www.phoronix.com/vr.php?view=ODc0Mw

  • Tommy666
    replied
    Originally posted by hax0r View Post
    Did anybody tried compiling without Big Kernel Lock, this is what I get:
    Code:
      CHK     include/linux/version.h
      UPD     include/linux/version.h
      CHK     include/generated/utsrelease.h
      UPD     include/generated/utsrelease.h
      HOSTCC  scripts/kallsyms
      CC      scripts/mod/empty.o
      HOSTCC  scripts/mod/mk_elfconfig
      MKELF   scripts/mod/elfconfig.h
      HOSTCC  scripts/mod/file2alias.o
      HOSTCC  scripts/mod/modpost.o
      HOSTCC  scripts/mod/sumversion.o
      HOSTCC  scripts/pnmtologo
      HOSTCC  scripts/conmakehash
      HOSTCC  scripts/bin2c
      HOSTLD  scripts/mod/modpost
      CC      kernel/bounds.s
      GEN     include/generated/bounds.h
      CC      arch/x86/kernel/asm-offsets.s
    In file included from /usr/src/linux-2.6.37-rc1/arch/x86/include/asm/suspend_64.h:10:0,
                     from /usr/src/linux-2.6.37-rc1/arch/x86/include/asm/suspend.h:4,
                     from arch/x86/kernel/asm-offsets_64.c:20,
                     from arch/x86/kernel/asm-offsets.c:4:
    /usr/src/linux-2.6.37-rc1/arch/x86/include/asm/i387.h: In function ?irq_ts_save?:
    /usr/src/linux-2.6.37-rc1/arch/x86/include/asm/i387.h:324:2: error: implicit declaration of function ?kernel_locked?
    make[1]: *** [arch/x86/kernel/asm-offsets.s] Error 1
    make: *** [prepare0] Error 2
    [[email protected] linux-2.6.37-rc1]# make
      CHK     include/linux/version.h
      CHK     include/generated/utsrelease.h
      CC      arch/x86/kernel/asm-offsets.s
    In file included from /usr/src/linux-2.6.37-rc1/arch/x86/include/asm/suspend_64.h:10:0,
                     from /usr/src/linux-2.6.37-rc1/arch/x86/include/asm/suspend.h:4,
                     from arch/x86/kernel/asm-offsets_64.c:20,
                     from arch/x86/kernel/asm-offsets.c:4:
    /usr/src/linux-2.6.37-rc1/arch/x86/include/asm/i387.h: In function ?irq_ts_save?:
    /usr/src/linux-2.6.37-rc1/arch/x86/include/asm/i387.h:324:2: error: implicit declaration of function ?kernel_locked?
    make[1]: *** [arch/x86/kernel/asm-offsets.s] Error 1
    make: *** [prepare0] Error 2
    [[email protected] linux-2.6.37-rc1]#
    I get the same error....

    Leave a comment:


  • smitty3268
    replied
    Originally posted by NomadDemon View Post
    iam not native english, i already reed giant lock, but dont understand anything [nothing seems to be more clear]

    both, benchmarks and explanations
    A lock is a synchronization primitive. Basically you place it around blocks of code, and then it becomes impossible for 1 thread to enter the block until the first has exited. The idea is to keep 1 thread from deleting or modifying some shared data before another thread has finished using it. In a userspace program that is only an issue for multi-threaded applications. In the kernel, there are all sorts of threads, events, and hardware interrupts that can cause an issue.

    A long time ago, devs originally create 1 large lock and used it everywhere in the kernel. That was simple to do, because essentially grabbing the lock switched all the code inside into a single-threaded mode. But it has issues with scalability - you want as many different threads as possible able to work. There's no reason the memory subsystem and the Ext4 filesystem should both be blocking each other. So the solution is to create 2 locks, one for each section. The complication is that so much complicated code had come to rely on the single lock that it was almost impossible to predict the behavior of the kernel when you messed with that global lock. So people have been slowly going through section by section of the kernel to try and factor out what codes needs to share the same lock and what code can be separated out. A lot of the code under the global lock didn't even need to be, and was just there because the original coder thought it might be an issue.

    So I'm not really sure how benchmarks will change. It will definitely increase the scalability of anything that likes to hit a bunch of kernel code. But on the other hand, this has been going on piece by piece for many years. So a lot of the benefits were already done and in previous kernels.

    Leave a comment:


  • hax0r
    replied
    Did anybody tried compiling without Big Kernel Lock, this is what I get:
    Code:
      CHK     include/linux/version.h
      UPD     include/linux/version.h
      CHK     include/generated/utsrelease.h
      UPD     include/generated/utsrelease.h
      HOSTCC  scripts/kallsyms
      CC      scripts/mod/empty.o
      HOSTCC  scripts/mod/mk_elfconfig
      MKELF   scripts/mod/elfconfig.h
      HOSTCC  scripts/mod/file2alias.o
      HOSTCC  scripts/mod/modpost.o
      HOSTCC  scripts/mod/sumversion.o
      HOSTCC  scripts/pnmtologo
      HOSTCC  scripts/conmakehash
      HOSTCC  scripts/bin2c
      HOSTLD  scripts/mod/modpost
      CC      kernel/bounds.s
      GEN     include/generated/bounds.h
      CC      arch/x86/kernel/asm-offsets.s
    In file included from /usr/src/linux-2.6.37-rc1/arch/x86/include/asm/suspend_64.h:10:0,
                     from /usr/src/linux-2.6.37-rc1/arch/x86/include/asm/suspend.h:4,
                     from arch/x86/kernel/asm-offsets_64.c:20,
                     from arch/x86/kernel/asm-offsets.c:4:
    /usr/src/linux-2.6.37-rc1/arch/x86/include/asm/i387.h: In function ?irq_ts_save?:
    /usr/src/linux-2.6.37-rc1/arch/x86/include/asm/i387.h:324:2: error: implicit declaration of function ?kernel_locked?
    make[1]: *** [arch/x86/kernel/asm-offsets.s] Error 1
    make: *** [prepare0] Error 2
    [[email protected] linux-2.6.37-rc1]# make
      CHK     include/linux/version.h
      CHK     include/generated/utsrelease.h
      CC      arch/x86/kernel/asm-offsets.s
    In file included from /usr/src/linux-2.6.37-rc1/arch/x86/include/asm/suspend_64.h:10:0,
                     from /usr/src/linux-2.6.37-rc1/arch/x86/include/asm/suspend.h:4,
                     from arch/x86/kernel/asm-offsets_64.c:20,
                     from arch/x86/kernel/asm-offsets.c:4:
    /usr/src/linux-2.6.37-rc1/arch/x86/include/asm/i387.h: In function ?irq_ts_save?:
    /usr/src/linux-2.6.37-rc1/arch/x86/include/asm/i387.h:324:2: error: implicit declaration of function ?kernel_locked?
    make[1]: *** [arch/x86/kernel/asm-offsets.s] Error 1
    make: *** [prepare0] Error 2
    [[email protected] linux-2.6.37-rc1]#

    Leave a comment:


  • NomadDemon
    replied
    iam not native english, i already reed giant lock, but dont understand anything [nothing seems to be more clear]

    both, benchmarks and explanations

    Leave a comment:


  • droidhacker
    replied
    Originally posted by NomadDemon View Post
    whats the difference between BKL and nonBKL ?
    any charts? tests? something?
    Are you asking for an explanation or benchmarks?

    If you want an explanation, wikipedia under "giant lock" has an explanation that might help you understand it better.

    Leave a comment:


  • NomadDemon
    replied
    whats the difference between BKL and nonBKL ?
    any charts? tests? something?

    Leave a comment:


  • smitty3268
    replied
    Originally posted by pingufunkybeat View Post
    I don't know if it will result in performance improvement, but I believe that this was not the major motivation. The major motivation is that it was one big scary voodoo which nobody really understood anymore.
    There were certain places where i think it limited performance, but whether that will be noticeable by a desktop user I'm not sure. I think it was important for things like RT linux as well. The real gain is that now driver and subsystem maintainers can work in their code without having to worry about messing up the BKL in weird ways for everyone else that is using it across the whole kernel.

    Leave a comment:


  • nkef
    replied
    Is there any progress with Linux responsiveness problems
    as reported in:

    https://bugzilla.kernel.org/show_bug.cgi?id=12309
    https://bugs.launchpad.net/ubuntu/+s...ux/+bug/131094

    Leave a comment:


  • pingufunkybeat
    replied
    I think the problem is that V4L manually locks and unlocks the BKL, which will not work if BKL is not used. It will have to be ported, like most core subsystems have been already.

    I don't know if it will result in performance improvement, but I believe that this was not the major motivation. The major motivation is that it was one big scary voodoo which nobody really understood anymore.

    Leave a comment:

Working...
X