On 2013/9/5 12:58, Wanpeng Li wrote: > Hi Jianguo, > On Thu, Sep 05, 2013 at 11:54:00AM +0800, Jianguo Wu wrote: >> On 2013/9/5 11:37, Wanpeng Li wrote: >> >>> On Thu, Sep 05, 2013 at 11:04:22AM +0800, Jianguo Wu wrote: >>>> Hi Wanpeng, >>>> >>>> On 2013/9/5 10:11, Wanpeng Li wrote: >>>> >>>>> Hi Jianguo, >>>>> On Wed, Sep 04, 2013 at 09:30:22PM +0800, Jianguo Wu wrote: >>>>>> Since commit d39d33c332(thp: enable direct defrag), defrag is enable >>>>>> for all transparent hugepage page faults by default, not only in >>>>>> MADV_HUGEPAGE regions. >>>>>> >>>>>> Signed-off-by: Jianguo Wu <wujianguo@xxxxxxxxxx> >>>>>> --- >>>>>> mm/huge_memory.c | 6 ++---- >>>>>> 1 file changed, 2 insertions(+), 4 deletions(-) >>>>>> >>>>>> diff --git a/mm/huge_memory.c b/mm/huge_memory.c >>>>>> index a92012a..abf047e 100644 >>>>>> --- a/mm/huge_memory.c >>>>>> +++ b/mm/huge_memory.c >>>>>> @@ -28,10 +28,8 @@ >>>>>> >>>>>> /* >>>>>> * By default transparent hugepage support is enabled for all mappings >>>>> >>>>> This is also stale. TRANSPARENT_HUGEPAGE_ALWAYS is not configured by default in >>>>> order that avoid to risk increase the memory footprint of applications w/o a >>>>> guaranteed benefit. >>>>> >>>> >>>> Right, how about this: >>>> >>>> By default transparent hugepage support is disabled in order that avoid to risk >>> >>> I don't think it's disabled. TRANSPARENT_HUGEPAGE_MADVISE is configured >>> by default. >>> >> >> Hi Wanpeng, >> >> We have TRANSPARENT_HUGEPAGE and TRANSPARENT_HUGEPAGE_ALWAYS/TRANSPARENT_HUGEPAGE_MADVISE, >> TRANSPARENT_HUGEPAGE_ALWAYS or TRANSPARENT_HUGEPAGE_MADVISE is configured only if TRANSPARENT_HUGEPAGE >> is configured. >> >> By default, TRANSPARENT_HUGEPAGE=n, and TRANSPARENT_HUGEPAGE_ALWAYS is configured when TRANSPARENT_HUGEPAGE=y. >> >> commit 13ece886d9(thp: transparent hugepage config choice): >> >> config TRANSPARENT_HUGEPAGE >> - bool "Transparent Hugepage Support" if EMBEDDED >> + bool "Transparent Hugepage Support" >> depends on X86 && MMU >> - default y >> >> +choice >> + prompt "Transparent Hugepage Support sysfs defaults" >> + depends on TRANSPARENT_HUGEPAGE >> + default TRANSPARENT_HUGEPAGE_ALWAYS >> > > mmotm tree: > > grep 'TRANSPARENT_HUGEPAGE' .config > CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE=y > CONFIG_TRANSPARENT_HUGEPAGE=y > # CONFIG_TRANSPARENT_HUGEPAGE_ALWAYS is not set > CONFIG_TRANSPARENT_HUGEPAGE_MADVISE=y > > distro: > > grep 'TRANSPARENT_HUGEPAGE' config-3.8.0-26-generic > CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE=y > CONFIG_TRANSPARENT_HUGEPAGE=y > # CONFIG_TRANSPARENT_HUGEPAGE_ALWAYS is not set > CONFIG_TRANSPARENT_HUGEPAGE_MADVISE=y > Hi Wanpeng, I'm a little confused, at mm/Kconfig, TRANSPARENT_HUGEPAGE is not configured by default. and in x86_64, linus tree: $make defconfig $grep 'TRANSPARENT_HUGEPAGE' .config CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE=y # CONFIG_TRANSPARENT_HUGEPAGE is not set Do i misunderstand something here？ Thanks > >> Thanks, >> Jianguo Wu >> >>> Regards, >>> Wanpeng Li >>> >>>> increase the memory footprint of applications w/o a guaranteed benefit, and >>>> khugepaged scans all mappings when transparent hugepage enabled. >>>> Defrag is invoked by khugepaged hugepage allocations and by page faults for all >>>> hugepage allocations. >>>> >>>> Thanks, >>>> Jianguo Wu >>>> >>>>> Regards, >>>>> Wanpeng Li >>>>> >>>>>> - * and khugepaged scans all mappings. Defrag is only invoked by >>>>>> - * khugepaged hugepage allocations and by page faults inside >>>>>> - * MADV_HUGEPAGE regions to avoid the risk of slowing down short lived >>>>>> - * allocations. >>>>>> + * and khugepaged scans all mappings. Defrag is invoked by khugepaged >>>>>> + * hugepage allocations and by page faults for all hugepage allocations. >>>>>> */ >>>>>> unsigned long transparent_hugepage_flags __read_mostly = >>>>>> #ifdef CONFIG_TRANSPARENT_HUGEPAGE_ALWAYS >>>>>> -- >>>>>> 1.8.1.2 >>>>>> >>>>>> -- >>>>>> To unsubscribe, send a message with 'unsubscribe linux-mm' in >>>>>> the body to majordomo@xxxxxxxxx. For more info on Linux MM, >>>>>> see: http://www.linux-mm.org/ . >>>>>> Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a> >>>>> >>>>> -- >>>>> To unsubscribe, send a message with 'unsubscribe linux-mm' in >>>>> the body to majordomo@xxxxxxxxx. For more info on Linux MM, >>>>> see: http://www.linux-mm.org/ . >>>>> Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a> >>>>> >>>>> >>>> >>>> >>> >>> >>> . >>> >> >> > > -- > To unsubscribe, send a message with 'unsubscribe linux-mm' in > the body to majordomo@xxxxxxxxx. For more info on Linux MM, > see: http://www.linux-mm.org/ . > Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a> > > -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>