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 <[email protected]>
---
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
- * 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
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 <[email protected]>
>> ---
>> 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
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 [email protected]. For more info on Linux MM,
>> see: http://www.linux-mm.org/ .
>> Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
>
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to [email protected]. For more info on Linux MM,
> see: http://www.linux-mm.org/ .
> Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
>
>
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 <[email protected]>
>>>> ---
>>>> 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
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 [email protected]. For more info on Linux MM,
>>>> see: http://www.linux-mm.org/ .
>>>> Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
>>>
>>> --
>>> To unsubscribe, send a message with 'unsubscribe linux-mm' in
>>> the body to [email protected]. For more info on Linux MM,
>>> see: http://www.linux-mm.org/ .
>>> Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
>>>
>>>
>>
>>
>
>
> .
>
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 <[email protected]>
>>>>>> ---
>>>>>> 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 [email protected]. For more info on Linux MM,
>>>>>> see: http://www.linux-mm.org/ .
>>>>>> Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
>>>>>
>>>>> --
>>>>> To unsubscribe, send a message with 'unsubscribe linux-mm' in
>>>>> the body to [email protected]. For more info on Linux MM,
>>>>> see: http://www.linux-mm.org/ .
>>>>> Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
>>>>>
>>>>>
>>>>
>>>>
>>>
>>>
>>> .
>>>
>>
>>
>
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to [email protected]. For more info on Linux MM,
> see: http://www.linux-mm.org/ .
> Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
>
>