questions about memory hotplug

questions about memory hotplug

[Larry Bassel]

We want to handle the following 2 use cases we have on
some of our (ARM) platforms:

1. A platform where part of the memory may be powered
off. The location and size of this memory is not known
until the kernel parses the memory tags (there aren't
any non-standard tags used, but the memory layout and
a memory bank size which is obtained from HW are used
to figure out where and how large this memory is). All
of this memory must either be on or off. For a given
configuration the location of this memory can't be moved.

2. A (different) platform where part of the memory
is occasionally needed as a large physically contiguous 
block, but usually is not (and then should be usable
by the kernel as normal memory). When the memory is
needed for the contiguous block, the pages in this
range currently being used by the kernel will
need to be migrated out.

The size of this is known at compile time and can be
placed at any reasonable place in memory (but should
be SPARSEMEM section aligned -- its size may however
not be a power of 2 and thus this memory could
span more than one section). This memory will not be
powered off, but must either be completely used for
one purpose or the other.

The size of memory in question for #1 is generally
much larger than that of #2.

Memory hotplug/hotremove (logical and physical for #1,
only logical for #2) approximately solves these problems,
but there is some functionality we need that AFAIK is
not present:

* The memory in #1 and #2 above must be in a movable
zone so that the chance of migration is maximized.
I'm familiar with the kernelcore= and movablecore=
commandline options, but they don't do what is necessary
here, because we need control on where the movable zone
is formed as well as the size. Also the location and size
of these special memory areas is not known until the
kernel comes up (in #1 it is conceivable that the bootloader
could locate the memory that can be powered on and off and
pass it in via some commandline option, but AFAICT this won't
work at all for #2).

One could hack up find_zone_movable_pfns_for_nodes() presumably,
but I wonder if there is an already existing way of doing
this (or at least a clean extension to the current
functionality that someone might suggest).

Would CONFIG_ARCH_POPULATES_NODE_MAP help here? Does anyone
use this? It doesn't seem to be in any defconfig or Kconfig
on 3.0 (or earlier versions I've looked at).

Perhaps CMA (which is not merged yet and AFAIK still has some
issues on ARM) might handle #2 better than memory
hotplug/hotremove. Or is there a better way to handle #2
than either CMA or memory hotplug?

Thanks.

Larry

-- 
Sent by an employee of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.

questions about memory hotplug

[Larry Bassel]

We want to handle the following 2 use cases we have on
some of our (ARM) platforms:

1. A platform where part of the memory may be powered
off. The location and size of this memory is not known
until the kernel parses the memory tags (there aren't
any non-standard tags used, but the memory layout and
a memory bank size which is obtained from HW are used
to figure out where and how large this memory is). All
of this memory must either be on or off. For a given
configuration the location of this memory can't be moved.

2. A (different) platform where part of the memory
is occasionally needed as a large physically contiguous 
block, but usually is not (and then should be usable
by the kernel as normal memory). When the memory is
needed for the contiguous block, the pages in this
range currently being used by the kernel will
need to be migrated out.

The size of this is known at compile time and can be
placed at any reasonable place in memory (but should
be SPARSEMEM section aligned -- its size may however
not be a power of 2 and thus this memory could
span more than one section). This memory will not be
powered off, but must either be completely used for
one purpose or the other.

The size of memory in question for #1 is generally
much larger than that of #2.

Memory hotplug/hotremove (logical and physical for #1,
only logical for #2) approximately solves these problems,
but there is some functionality we need that AFAIK is
not present:

* The memory in #1 and #2 above must be in a movable
zone so that the chance of migration is maximized.
I'm familiar with the kernelcore= and movablecore=
commandline options, but they don't do what is necessary
here, because we need control on where the movable zone
is formed as well as the size. Also the location and size
of these special memory areas is not known until the
kernel comes up (in #1 it is conceivable that the bootloader
could locate the memory that can be powered on and off and
pass it in via some commandline option, but AFAICT this won't
work at all for #2).

One could hack up find_zone_movable_pfns_for_nodes() presumably,
but I wonder if there is an already existing way of doing
this (or at least a clean extension to the current
functionality that someone might suggest).

Would CONFIG_ARCH_POPULATES_NODE_MAP help here? Does anyone
use this? It doesn't seem to be in any defconfig or Kconfig
on 3.0 (or earlier versions I've looked at).

Perhaps CMA (which is not merged yet and AFAIK still has some
issues on ARM) might handle #2 better than memory
hotplug/hotremove. Or is there a better way to handle #2
than either CMA or memory hotplug?

Thanks.

Larry

-- 
Sent by an employee of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

questions about memory hotplug

[Larry Bassel]

We want to handle the following 2 use cases we have on
some of our (ARM) platforms:

1. A platform where part of the memory may be powered
off. The location and size of this memory is not known
until the kernel parses the memory tags (there aren't
any non-standard tags used, but the memory layout and
a memory bank size which is obtained from HW are used
to figure out where and how large this memory is). All
of this memory must either be on or off. For a given
configuration the location of this memory can't be moved.

2. A (different) platform where part of the memory
is occasionally needed as a large physically contiguous 
block, but usually is not (and then should be usable
by the kernel as normal memory). When the memory is
needed for the contiguous block, the pages in this
range currently being used by the kernel will
need to be migrated out.

The size of this is known at compile time and can be
placed at any reasonable place in memory (but should
be SPARSEMEM section aligned -- its size may however
not be a power of 2 and thus this memory could
span more than one section). This memory will not be
powered off, but must either be completely used for
one purpose or the other.

The size of memory in question for #1 is generally
much larger than that of #2.

Memory hotplug/hotremove (logical and physical for #1,
only logical for #2) approximately solves these problems,
but there is some functionality we need that AFAIK is
not present:

* The memory in #1 and #2 above must be in a movable
zone so that the chance of migration is maximized.
I'm familiar with the kernelcore= and movablecore=
commandline options, but they don't do what is necessary
here, because we need control on where the movable zone
is formed as well as the size. Also the location and size
of these special memory areas is not known until the
kernel comes up (in #1 it is conceivable that the bootloader
could locate the memory that can be powered on and off and
pass it in via some commandline option, but AFAICT this won't
work at all for #2).

One could hack up find_zone_movable_pfns_for_nodes() presumably,
but I wonder if there is an already existing way of doing
this (or at least a clean extension to the current
functionality that someone might suggest).

Would CONFIG_ARCH_POPULATES_NODE_MAP help here? Does anyone
use this? It doesn't seem to be in any defconfig or Kconfig
on 3.0 (or earlier versions I've looked at).

Perhaps CMA (which is not merged yet and AFAIK still has some
issues on ARM) might handle #2 better than memory
hotplug/hotremove. Or is there a better way to handle #2
than either CMA or memory hotplug?

Thanks.

Larry

-- 
Sent by an employee of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.

Re: questions about memory hotplug

[Shaohua Li]

On Sat, Jul 30, 2011 at 06:12:31AM +0800, Larry Bassel wrote:
> We want to handle the following 2 use cases we have on
> some of our (ARM) platforms:
> 
> 1. A platform where part of the memory may be powered
> off. The location and size of this memory is not known
> until the kernel parses the memory tags (there aren't
> any non-standard tags used, but the memory layout and
> a memory bank size which is obtained from HW are used
> to figure out where and how large this memory is). All
> of this memory must either be on or off. For a given
> configuration the location of this memory can't be moved.
> 
> 2. A (different) platform where part of the memory
> is occasionally needed as a large physically contiguous 
> block, but usually is not (and then should be usable
> by the kernel as normal memory). When the memory is
> needed for the contiguous block, the pages in this
> range currently being used by the kernel will
> need to be migrated out.
> 
> The size of this is known at compile time and can be
> placed at any reasonable place in memory (but should
> be SPARSEMEM section aligned -- its size may however
> not be a power of 2 and thus this memory could
> span more than one section). This memory will not be
> powered off, but must either be completely used for
> one purpose or the other.
> 
> The size of memory in question for #1 is generally
> much larger than that of #2.
> 
> Memory hotplug/hotremove (logical and physical for #1,
> only logical for #2) approximately solves these problems,
> but there is some functionality we need that AFAIK is
> not present:
> 
> * The memory in #1 and #2 above must be in a movable
> zone so that the chance of migration is maximized.
> I'm familiar with the kernelcore= and movablecore=
> commandline options, but they don't do what is necessary
> here, because we need control on where the movable zone
> is formed as well as the size. Also the location and size
> of these special memory areas is not known until the
> kernel comes up (in #1 it is conceivable that the bootloader
> could locate the memory that can be powered on and off and
> pass it in via some commandline option, but AFAICT this won't
> work at all for #2).
> 
> One could hack up find_zone_movable_pfns_for_nodes() presumably,
> but I wonder if there is an already existing way of doing
> this (or at least a clean extension to the current
> functionality that someone might suggest).
> 
> Would CONFIG_ARCH_POPULATES_NODE_MAP help here? Does anyone
> use this? It doesn't seem to be in any defconfig or Kconfig
> on 3.0 (or earlier versions I've looked at).
> 
> Perhaps CMA (which is not merged yet and AFAIK still has some
> issues on ARM) might handle #2 better than memory
> hotplug/hotremove. Or is there a better way to handle #2
> than either CMA or memory hotplug?
is such memory always at the end of memory? if yes, such memory
can be added to MOVABLE_ZONE.
or if such memory can be at the middle of memory, maybe we can add
a new mirgate type, the migrate type is movable but other type allocation
can't fallback into.

Thanks,
Shaohua

Re: questions about memory hotplug

[Shaohua Li]

On Sat, Jul 30, 2011 at 06:12:31AM +0800, Larry Bassel wrote:
> We want to handle the following 2 use cases we have on
> some of our (ARM) platforms:
> 
> 1. A platform where part of the memory may be powered
> off. The location and size of this memory is not known
> until the kernel parses the memory tags (there aren't
> any non-standard tags used, but the memory layout and
> a memory bank size which is obtained from HW are used
> to figure out where and how large this memory is). All
> of this memory must either be on or off. For a given
> configuration the location of this memory can't be moved.
> 
> 2. A (different) platform where part of the memory
> is occasionally needed as a large physically contiguous 
> block, but usually is not (and then should be usable
> by the kernel as normal memory). When the memory is
> needed for the contiguous block, the pages in this
> range currently being used by the kernel will
> need to be migrated out.
> 
> The size of this is known at compile time and can be
> placed at any reasonable place in memory (but should
> be SPARSEMEM section aligned -- its size may however
> not be a power of 2 and thus this memory could
> span more than one section). This memory will not be
> powered off, but must either be completely used for
> one purpose or the other.
> 
> The size of memory in question for #1 is generally
> much larger than that of #2.
> 
> Memory hotplug/hotremove (logical and physical for #1,
> only logical for #2) approximately solves these problems,
> but there is some functionality we need that AFAIK is
> not present:
> 
> * The memory in #1 and #2 above must be in a movable
> zone so that the chance of migration is maximized.
> I'm familiar with the kernelcore= and movablecore=
> commandline options, but they don't do what is necessary
> here, because we need control on where the movable zone
> is formed as well as the size. Also the location and size
> of these special memory areas is not known until the
> kernel comes up (in #1 it is conceivable that the bootloader
> could locate the memory that can be powered on and off and
> pass it in via some commandline option, but AFAICT this won't
> work at all for #2).
> 
> One could hack up find_zone_movable_pfns_for_nodes() presumably,
> but I wonder if there is an already existing way of doing
> this (or at least a clean extension to the current
> functionality that someone might suggest).
> 
> Would CONFIG_ARCH_POPULATES_NODE_MAP help here? Does anyone
> use this? It doesn't seem to be in any defconfig or Kconfig
> on 3.0 (or earlier versions I've looked at).
> 
> Perhaps CMA (which is not merged yet and AFAIK still has some
> issues on ARM) might handle #2 better than memory
> hotplug/hotremove. Or is there a better way to handle #2
> than either CMA or memory hotplug?
is such memory always at the end of memory? if yes, such memory
can be added to MOVABLE_ZONE.
or if such memory can be at the middle of memory, maybe we can add
a new mirgate type, the migrate type is movable but other type allocation
can't fallback into.

Thanks,
Shaohua

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

questions about memory hotplug

[Shaohua Li]

On Sat, Jul 30, 2011 at 06:12:31AM +0800, Larry Bassel wrote:
> We want to handle the following 2 use cases we have on
> some of our (ARM) platforms:
> 
> 1. A platform where part of the memory may be powered
> off. The location and size of this memory is not known
> until the kernel parses the memory tags (there aren't
> any non-standard tags used, but the memory layout and
> a memory bank size which is obtained from HW are used
> to figure out where and how large this memory is). All
> of this memory must either be on or off. For a given
> configuration the location of this memory can't be moved.
> 
> 2. A (different) platform where part of the memory
> is occasionally needed as a large physically contiguous 
> block, but usually is not (and then should be usable
> by the kernel as normal memory). When the memory is
> needed for the contiguous block, the pages in this
> range currently being used by the kernel will
> need to be migrated out.
> 
> The size of this is known at compile time and can be
> placed at any reasonable place in memory (but should
> be SPARSEMEM section aligned -- its size may however
> not be a power of 2 and thus this memory could
> span more than one section). This memory will not be
> powered off, but must either be completely used for
> one purpose or the other.
> 
> The size of memory in question for #1 is generally
> much larger than that of #2.
> 
> Memory hotplug/hotremove (logical and physical for #1,
> only logical for #2) approximately solves these problems,
> but there is some functionality we need that AFAIK is
> not present:
> 
> * The memory in #1 and #2 above must be in a movable
> zone so that the chance of migration is maximized.
> I'm familiar with the kernelcore= and movablecore=
> commandline options, but they don't do what is necessary
> here, because we need control on where the movable zone
> is formed as well as the size. Also the location and size
> of these special memory areas is not known until the
> kernel comes up (in #1 it is conceivable that the bootloader
> could locate the memory that can be powered on and off and
> pass it in via some commandline option, but AFAICT this won't
> work at all for #2).
> 
> One could hack up find_zone_movable_pfns_for_nodes() presumably,
> but I wonder if there is an already existing way of doing
> this (or at least a clean extension to the current
> functionality that someone might suggest).
> 
> Would CONFIG_ARCH_POPULATES_NODE_MAP help here? Does anyone
> use this? It doesn't seem to be in any defconfig or Kconfig
> on 3.0 (or earlier versions I've looked at).
> 
> Perhaps CMA (which is not merged yet and AFAIK still has some
> issues on ARM) might handle #2 better than memory
> hotplug/hotremove. Or is there a better way to handle #2
> than either CMA or memory hotplug?
is such memory always at the end of memory? if yes, such memory
can be added to MOVABLE_ZONE.
or if such memory can be at the middle of memory, maybe we can add
a new mirgate type, the migrate type is movable but other type allocation
can't fallback into.

Thanks,
Shaohua

Re: questions about memory hotplug

[Larry Bassel]

On 30 Jul 11 17:30, Shaohua Li wrote:
> On Sat, Jul 30, 2011 at 06:12:31AM +0800, Larry Bassel wrote:
> > We want to handle the following 2 use cases we have on
> > some of our (ARM) platforms:
> > 
> > 1. A platform where part of the memory may be powered
> > off. The location and size of this memory is not known
> > until the kernel parses the memory tags (there aren't
> > any non-standard tags used, but the memory layout and
> > a memory bank size which is obtained from HW are used
> > to figure out where and how large this memory is). All
> > of this memory must either be on or off. For a given
> > configuration the location of this memory can't be moved.
> > 
> > 2. A (different) platform where part of the memory
> > is occasionally needed as a large physically contiguous 
> > block, but usually is not (and then should be usable
> > by the kernel as normal memory). When the memory is
> > needed for the contiguous block, the pages in this
> > range currently being used by the kernel will
> > need to be migrated out.
> > 
> > The size of this is known at compile time and can be
> > placed at any reasonable place in memory (but should
> > be SPARSEMEM section aligned -- its size may however
> > not be a power of 2 and thus this memory could
> > span more than one section). This memory will not be
> > powered off, but must either be completely used for
> > one purpose or the other.
> > 
> > The size of memory in question for #1 is generally
> > much larger than that of #2.
> > 
> > Memory hotplug/hotremove (logical and physical for #1,
> > only logical for #2) approximately solves these problems,
> > but there is some functionality we need that AFAIK is
> > not present:
> > 
> > * The memory in #1 and #2 above must be in a movable
> > zone so that the chance of migration is maximized.
> > I'm familiar with the kernelcore= and movablecore=
> > commandline options, but they don't do what is necessary
> > here, because we need control on where the movable zone
> > is formed as well as the size. Also the location and size
> > of these special memory areas is not known until the
> > kernel comes up (in #1 it is conceivable that the bootloader
> > could locate the memory that can be powered on and off and
> > pass it in via some commandline option, but AFAICT this won't
> > work at all for #2).
> > 
> > One could hack up find_zone_movable_pfns_for_nodes() presumably,
> > but I wonder if there is an already existing way of doing
> > this (or at least a clean extension to the current
> > functionality that someone might suggest).
> > 
> > Would CONFIG_ARCH_POPULATES_NODE_MAP help here? Does anyone
> > use this? It doesn't seem to be in any defconfig or Kconfig
> > on 3.0 (or earlier versions I've looked at).
> > 
> > Perhaps CMA (which is not merged yet and AFAIK still has some
> > issues on ARM) might handle #2 better than memory
> > hotplug/hotremove. Or is there a better way to handle #2
> > than either CMA or memory hotplug?
> is such memory always at the end of memory? if yes, such memory
> can be added to MOVABLE_ZONE.

In use case #1 yes, maybe not in #2 (we can arrange it to be
at the end of memory, but then might waste memory as it may
not be aligned on a SPARSEMEM section boundary and so would
need to be padded).

> or if such memory can be at the middle of memory, maybe we can add
> a new mirgate type, the migrate type is movable but other type allocation
> can't fallback into.
> 
> Thanks,
> Shaohua

Larry

> 
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to majordomo@kvack.org.  For more info on Linux MM,
> see: http://www.linux-mm.org/ .
> Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
> Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

-- 
Sent by an employee of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.

Re: questions about memory hotplug

[Larry Bassel]

On 30 Jul 11 17:30, Shaohua Li wrote:
> On Sat, Jul 30, 2011 at 06:12:31AM +0800, Larry Bassel wrote:
> > We want to handle the following 2 use cases we have on
> > some of our (ARM) platforms:
> > 
> > 1. A platform where part of the memory may be powered
> > off. The location and size of this memory is not known
> > until the kernel parses the memory tags (there aren't
> > any non-standard tags used, but the memory layout and
> > a memory bank size which is obtained from HW are used
> > to figure out where and how large this memory is). All
> > of this memory must either be on or off. For a given
> > configuration the location of this memory can't be moved.
> > 
> > 2. A (different) platform where part of the memory
> > is occasionally needed as a large physically contiguous 
> > block, but usually is not (and then should be usable
> > by the kernel as normal memory). When the memory is
> > needed for the contiguous block, the pages in this
> > range currently being used by the kernel will
> > need to be migrated out.
> > 
> > The size of this is known at compile time and can be
> > placed at any reasonable place in memory (but should
> > be SPARSEMEM section aligned -- its size may however
> > not be a power of 2 and thus this memory could
> > span more than one section). This memory will not be
> > powered off, but must either be completely used for
> > one purpose or the other.
> > 
> > The size of memory in question for #1 is generally
> > much larger than that of #2.
> > 
> > Memory hotplug/hotremove (logical and physical for #1,
> > only logical for #2) approximately solves these problems,
> > but there is some functionality we need that AFAIK is
> > not present:
> > 
> > * The memory in #1 and #2 above must be in a movable
> > zone so that the chance of migration is maximized.
> > I'm familiar with the kernelcore= and movablecore=
> > commandline options, but they don't do what is necessary
> > here, because we need control on where the movable zone
> > is formed as well as the size. Also the location and size
> > of these special memory areas is not known until the
> > kernel comes up (in #1 it is conceivable that the bootloader
> > could locate the memory that can be powered on and off and
> > pass it in via some commandline option, but AFAICT this won't
> > work at all for #2).
> > 
> > One could hack up find_zone_movable_pfns_for_nodes() presumably,
> > but I wonder if there is an already existing way of doing
> > this (or at least a clean extension to the current
> > functionality that someone might suggest).
> > 
> > Would CONFIG_ARCH_POPULATES_NODE_MAP help here? Does anyone
> > use this? It doesn't seem to be in any defconfig or Kconfig
> > on 3.0 (or earlier versions I've looked at).
> > 
> > Perhaps CMA (which is not merged yet and AFAIK still has some
> > issues on ARM) might handle #2 better than memory
> > hotplug/hotremove. Or is there a better way to handle #2
> > than either CMA or memory hotplug?
> is such memory always at the end of memory? if yes, such memory
> can be added to MOVABLE_ZONE.

In use case #1 yes, maybe not in #2 (we can arrange it to be
at the end of memory, but then might waste memory as it may
not be aligned on a SPARSEMEM section boundary and so would
need to be padded).

> or if such memory can be at the middle of memory, maybe we can add
> a new mirgate type, the migrate type is movable but other type allocation
> can't fallback into.
> 
> Thanks,
> Shaohua

Larry

> 
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to majordomo@kvack.org.  For more info on Linux MM,
> see: http://www.linux-mm.org/ .
> Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
> Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

-- 
Sent by an employee of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

questions about memory hotplug

[Larry Bassel]

On 30 Jul 11 17:30, Shaohua Li wrote:
> On Sat, Jul 30, 2011 at 06:12:31AM +0800, Larry Bassel wrote:
> > We want to handle the following 2 use cases we have on
> > some of our (ARM) platforms:
> > 
> > 1. A platform where part of the memory may be powered
> > off. The location and size of this memory is not known
> > until the kernel parses the memory tags (there aren't
> > any non-standard tags used, but the memory layout and
> > a memory bank size which is obtained from HW are used
> > to figure out where and how large this memory is). All
> > of this memory must either be on or off. For a given
> > configuration the location of this memory can't be moved.
> > 
> > 2. A (different) platform where part of the memory
> > is occasionally needed as a large physically contiguous 
> > block, but usually is not (and then should be usable
> > by the kernel as normal memory). When the memory is
> > needed for the contiguous block, the pages in this
> > range currently being used by the kernel will
> > need to be migrated out.
> > 
> > The size of this is known at compile time and can be
> > placed at any reasonable place in memory (but should
> > be SPARSEMEM section aligned -- its size may however
> > not be a power of 2 and thus this memory could
> > span more than one section). This memory will not be
> > powered off, but must either be completely used for
> > one purpose or the other.
> > 
> > The size of memory in question for #1 is generally
> > much larger than that of #2.
> > 
> > Memory hotplug/hotremove (logical and physical for #1,
> > only logical for #2) approximately solves these problems,
> > but there is some functionality we need that AFAIK is
> > not present:
> > 
> > * The memory in #1 and #2 above must be in a movable
> > zone so that the chance of migration is maximized.
> > I'm familiar with the kernelcore= and movablecore=
> > commandline options, but they don't do what is necessary
> > here, because we need control on where the movable zone
> > is formed as well as the size. Also the location and size
> > of these special memory areas is not known until the
> > kernel comes up (in #1 it is conceivable that the bootloader
> > could locate the memory that can be powered on and off and
> > pass it in via some commandline option, but AFAICT this won't
> > work at all for #2).
> > 
> > One could hack up find_zone_movable_pfns_for_nodes() presumably,
> > but I wonder if there is an already existing way of doing
> > this (or at least a clean extension to the current
> > functionality that someone might suggest).
> > 
> > Would CONFIG_ARCH_POPULATES_NODE_MAP help here? Does anyone
> > use this? It doesn't seem to be in any defconfig or Kconfig
> > on 3.0 (or earlier versions I've looked at).
> > 
> > Perhaps CMA (which is not merged yet and AFAIK still has some
> > issues on ARM) might handle #2 better than memory
> > hotplug/hotremove. Or is there a better way to handle #2
> > than either CMA or memory hotplug?
> is such memory always at the end of memory? if yes, such memory
> can be added to MOVABLE_ZONE.

In use case #1 yes, maybe not in #2 (we can arrange it to be
at the end of memory, but then might waste memory as it may
not be aligned on a SPARSEMEM section boundary and so would
need to be padded).

> or if such memory can be at the middle of memory, maybe we can add
> a new mirgate type, the migrate type is movable but other type allocation
> can't fallback into.
> 
> Thanks,
> Shaohua

Larry

> 
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to majordomo at kvack.org.  For more info on Linux MM,
> see: http://www.linux-mm.org/ .
> Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
> Don't email: <a href=mailto:"dont@kvack.org"> email at kvack.org </a>

-- 
Sent by an employee of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.

Re: questions about memory hotplug

[Larry Bassel]

On 29 Jul 11 15:12, Larry Bassel wrote:
> 
> Would CONFIG_ARCH_POPULATES_NODE_MAP help here? Does anyone
> use this? It doesn't seem to be in any defconfig or Kconfig

I want to clarify this, I meant used on ARM -- I see it is being used
on other architectures (and that movablecore= and kernelcore=
require this config option).

> on 3.0 (or earlier versions I've looked at).
> 

Larry

-- 
Sent by an employee of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.

Re: questions about memory hotplug

[Larry Bassel]

On 29 Jul 11 15:12, Larry Bassel wrote:
> 
> Would CONFIG_ARCH_POPULATES_NODE_MAP help here? Does anyone
> use this? It doesn't seem to be in any defconfig or Kconfig

I want to clarify this, I meant used on ARM -- I see it is being used
on other architectures (and that movablecore= and kernelcore=
require this config option).

> on 3.0 (or earlier versions I've looked at).
> 

Larry

-- 
Sent by an employee of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

questions about memory hotplug

[Larry Bassel]

On 29 Jul 11 15:12, Larry Bassel wrote:
> 
> Would CONFIG_ARCH_POPULATES_NODE_MAP help here? Does anyone
> use this? It doesn't seem to be in any defconfig or Kconfig

I want to clarify this, I meant used on ARM -- I see it is being used
on other architectures (and that movablecore= and kernelcore=
require this config option).

> on 3.0 (or earlier versions I've looked at).
> 

Larry

-- 
Sent by an employee of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.

Re: questions about memory hotplug

[Shaohua Li]

On Tue, 2011-08-02 at 01:08 +0800, Larry Bassel wrote:
> 
> In use case #1 yes, maybe not in #2 (we can arrange it to be
> at the end of memory, but then might waste memory as it may
> not be aligned on a SPARSEMEM section boundary and so would
> need to be padded).
then maybe the new migrate type I suggested can help here for the
non-aligned memory. Anyway, let me do an experiment.

Thanks,
Shaohua

Re: questions about memory hotplug

[Shaohua Li]

On Tue, 2011-08-02 at 01:08 +0800, Larry Bassel wrote:
> 
> In use case #1 yes, maybe not in #2 (we can arrange it to be
> at the end of memory, but then might waste memory as it may
> not be aligned on a SPARSEMEM section boundary and so would
> need to be padded).
then maybe the new migrate type I suggested can help here for the
non-aligned memory. Anyway, let me do an experiment.

Thanks,
Shaohua

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

questions about memory hotplug

[Shaohua Li]

On Tue, 2011-08-02 at 01:08 +0800, Larry Bassel wrote:
> 
> In use case #1 yes, maybe not in #2 (we can arrange it to be
> at the end of memory, but then might waste memory as it may
> not be aligned on a SPARSEMEM section boundary and so would
> need to be padded).
then maybe the new migrate type I suggested can help here for the
non-aligned memory. Anyway, let me do an experiment.

Thanks,
Shaohua

Re: questions about memory hotplug

[Shaohua Li]

On Tue, 2011-08-02 at 09:09 +0800, Shaohua Li wrote:
> On Tue, 2011-08-02 at 01:08 +0800, Larry Bassel wrote:
> > 
> > In use case #1 yes, maybe not in #2 (we can arrange it to be
> > at the end of memory, but then might waste memory as it may
> > not be aligned on a SPARSEMEM section boundary and so would
> > need to be padded).
> then maybe the new migrate type I suggested can help here for the
> non-aligned memory. Anyway, let me do an experiment.
so your problem is to offline memory in arbitrary address and size (eg,
might not be at the end, and maybe smaller than a section)

I had a hack. In my machine, I have DMA, DMA32, and NORMAL zone.
At boot time, I mark 500M~600M ranges as MOVABLE_NOFB. the range is in
DMA32 and not section size aligned. MOVABLE_NOFB is a new migrate type I
added. That range memory is movable, but other type of allocation can't
fallback into such ranges. so such range memory can only be used by
userspace.
I then run a memory stress test and do memory online/offline for the
range at runtime, the offline always success.
Does this meet your usage? If yes, I'll cook it up a little bit.


---
 include/linux/gfp.h    |    5 +-
 include/linux/mmzone.h |    7 +-
 mm/memory_hotplug.c    |  117 ++++++++++++++++++++++++++++++++++---------------
 mm/page_alloc.c        |   94 +++++++++++++++++++++++++++++++++------
 mm/vmstat.c            |    1 
 5 files changed, 171 insertions(+), 53 deletions(-)

Index: linux/include/linux/gfp.h
===================================================================
--- linux.orig/include/linux/gfp.h	2011-08-03 14:59:47.000000000 +0800
+++ linux/include/linux/gfp.h	2011-08-03 15:24:52.000000000 +0800
@@ -155,8 +155,9 @@ static inline int allocflags_to_migratet
 		return MIGRATE_UNMOVABLE;
 
 	/* Group based on mobility */
-	return (((gfp_flags & __GFP_MOVABLE) != 0) << 1) |
-		((gfp_flags & __GFP_RECLAIMABLE) != 0);
+	if ((gfp_flags & __GFP_MOVABLE) != 0)
+		return MIGRATE_MOVABLE_NOFB;
+	return (gfp_flags & __GFP_RECLAIMABLE) != 0;
 }
 
 #ifdef CONFIG_HIGHMEM
Index: linux/include/linux/mmzone.h
===================================================================
--- linux.orig/include/linux/mmzone.h	2011-08-03 14:59:47.000000000 +0800
+++ linux/include/linux/mmzone.h	2011-08-03 15:24:52.000000000 +0800
@@ -39,9 +39,10 @@
 #define MIGRATE_RECLAIMABLE   1
 #define MIGRATE_MOVABLE       2
 #define MIGRATE_PCPTYPES      3 /* the number of types on the pcp lists */
-#define MIGRATE_RESERVE       3
-#define MIGRATE_ISOLATE       4 /* can't allocate from here */
-#define MIGRATE_TYPES         5
+#define MIGRATE_MOVABLE_NOFB  3
+#define MIGRATE_RESERVE       4
+#define MIGRATE_ISOLATE       5 /* can't allocate from here */
+#define MIGRATE_TYPES         6
 
 #define for_each_migratetype_order(order, type) \
 	for (order = 0; order < MAX_ORDER; order++) \
Index: linux/mm/page_alloc.c
===================================================================
--- linux.orig/mm/page_alloc.c	2011-08-03 14:59:47.000000000 +0800
+++ linux/mm/page_alloc.c	2011-08-03 15:25:29.000000000 +0800
@@ -627,7 +627,7 @@ static void free_pcppages_bulk(struct zo
 			page = list_entry(list->prev, struct page, lru);
 			/* must delete as __free_one_page list manipulates */
 			list_del(&page->lru);
-			/* MIGRATE_MOVABLE list may include MIGRATE_RESERVEs */
+			/* MIGRATE_MOVABLE list may include MIGRATE_RESERVE and MIGRATE_MOVABLE_NOFBs */
 			__free_one_page(page, zone, 0, page_private(page));
 			trace_mm_page_pcpu_drain(page, 0, page_private(page));
 		} while (--to_free && --batch_free && !list_empty(list));
@@ -828,10 +828,11 @@ struct page *__rmqueue_smallest(struct z
  * the free lists for the desirable migrate type are depleted
  */
 static int fallbacks[MIGRATE_TYPES][MIGRATE_TYPES-1] = {
-	[MIGRATE_UNMOVABLE]   = { MIGRATE_RECLAIMABLE, MIGRATE_MOVABLE,   MIGRATE_RESERVE },
-	[MIGRATE_RECLAIMABLE] = { MIGRATE_UNMOVABLE,   MIGRATE_MOVABLE,   MIGRATE_RESERVE },
-	[MIGRATE_MOVABLE]     = { MIGRATE_RECLAIMABLE, MIGRATE_UNMOVABLE, MIGRATE_RESERVE },
-	[MIGRATE_RESERVE]     = { MIGRATE_RESERVE,     MIGRATE_RESERVE,   MIGRATE_RESERVE }, /* Never used */
+	[MIGRATE_UNMOVABLE]   = { MIGRATE_RECLAIMABLE, MIGRATE_MOVABLE,   MIGRATE_RESERVE, MIGRATE_RESERVE },
+	[MIGRATE_RECLAIMABLE] = { MIGRATE_UNMOVABLE,   MIGRATE_MOVABLE,   MIGRATE_RESERVE, MIGRATE_RESERVE},
+	[MIGRATE_MOVABLE]     = { MIGRATE_RECLAIMABLE, MIGRATE_UNMOVABLE, MIGRATE_RESERVE, MIGRATE_RESERVE},
+	[MIGRATE_MOVABLE_NOFB] = { MIGRATE_MOVABLE, MIGRATE_RECLAIMABLE, MIGRATE_UNMOVABLE, MIGRATE_RESERVE}, /* Never used */
+	[MIGRATE_RESERVE]     = { MIGRATE_RESERVE,     MIGRATE_RESERVE,   MIGRATE_RESERVE, MIGRATE_RESERVE}, /* Never used */
 };
 
 /*
@@ -923,6 +924,7 @@ __rmqueue_fallback(struct zone *zone, in
 	struct page *page;
 	int migratetype, i;
 
+	BUG_ON(start_migratetype == MIGRATE_MOVABLE_NOFB);
 	/* Find the largest possible block of pages in the other list */
 	for (current_order = MAX_ORDER-1; current_order >= order;
 						--current_order) {
@@ -994,8 +996,16 @@ static struct page *__rmqueue(struct zon
 	struct page *page;
 
 retry_reserve:
+	/* FIXME: Speed up a little bit */
 	page = __rmqueue_smallest(zone, order, migratetype);
 
+	if (!page && migratetype == MIGRATE_MOVABLE_NOFB) {
+		/* Don't change other migrate type to MIGRATE_MOVABLE_NOFB, so
+		 * we shortcut here */
+		migratetype = MIGRATE_MOVABLE;
+		goto retry_reserve;
+	}
+
 	if (unlikely(!page) && migratetype != MIGRATE_RESERVE) {
 		page = __rmqueue_fallback(zone, order, migratetype);
 
@@ -1044,7 +1054,8 @@ static int rmqueue_bulk(struct zone *zon
 			list_add(&page->lru, list);
 		else
 			list_add_tail(&page->lru, list);
-		set_page_private(page, migratetype);
+		/* __rmqueue might do fallback, so reread the migrate type */
+		set_page_private(page, get_pageblock_migratetype(page));
 		list = &page->lru;
 	}
 	__mod_zone_page_state(zone, NR_FREE_PAGES, -(i << order));
@@ -1300,10 +1311,11 @@ again:
 	if (likely(order == 0)) {
 		struct per_cpu_pages *pcp;
 		struct list_head *list;
+		int migratetype_pcp = migratetype == MIGRATE_MOVABLE_NOFB? MIGRATE_MOVABLE : migratetype;
 
 		local_irq_save(flags);
 		pcp = &this_cpu_ptr(zone->pageset)->pcp;
-		list = &pcp->lists[migratetype];
+		list = &pcp->lists[migratetype_pcp];
 		if (list_empty(list)) {
 			pcp->count += rmqueue_bulk(zone, 0,
 					pcp->batch, list,
@@ -5558,7 +5570,8 @@ __count_immobile_pages(struct zone *zone
 	if (zone_idx(zone) == ZONE_MOVABLE)
 		return true;
 
-	if (get_pageblock_migratetype(page) == MIGRATE_MOVABLE)
+	if (get_pageblock_migratetype(page) == MIGRATE_MOVABLE ||
+	    get_pageblock_migratetype(page) == MIGRATE_MOVABLE_NOFB )
 		return true;
 
 	pfn = page_to_pfn(page);
@@ -5601,7 +5614,7 @@ bool is_pageblock_removable_nolock(struc
 	return __count_immobile_pages(zone, page, 0);
 }
 
-int set_migratetype_isolate(struct page *page)
+static int __set_migratetype(struct page *page, int migratetype)
 {
 	struct zone *zone;
 	unsigned long flags, pfn;
@@ -5647,8 +5660,8 @@ int set_migratetype_isolate(struct page
 
 out:
 	if (!ret) {
-		set_pageblock_migratetype(page, MIGRATE_ISOLATE);
-		move_freepages_block(zone, page, MIGRATE_ISOLATE);
+		set_pageblock_migratetype(page, migratetype);
+		move_freepages_block(zone, page, migratetype);
 	}
 
 	spin_unlock_irqrestore(&zone->lock, flags);
@@ -5657,6 +5670,11 @@ out:
 	return ret;
 }
 
+int set_migratetype_isolate(struct page *page)
+{
+	return __set_migratetype(page, MIGRATE_ISOLATE);
+}
+
 void unset_migratetype_isolate(struct page *page)
 {
 	struct zone *zone;
@@ -5665,12 +5683,18 @@ void unset_migratetype_isolate(struct pa
 	spin_lock_irqsave(&zone->lock, flags);
 	if (get_pageblock_migratetype(page) != MIGRATE_ISOLATE)
 		goto out;
-	set_pageblock_migratetype(page, MIGRATE_MOVABLE);
-	move_freepages_block(zone, page, MIGRATE_MOVABLE);
+	/* Fixme: should restore previous type */
+	set_pageblock_migratetype(page, MIGRATE_MOVABLE_NOFB);
+	move_freepages_block(zone, page, MIGRATE_MOVABLE_NOFB);
 out:
 	spin_unlock_irqrestore(&zone->lock, flags);
 }
 
+int set_migratetype_movable_nofb(struct page *page)
+{
+	return __set_migratetype(page, MIGRATE_MOVABLE_NOFB);
+}
+
 #ifdef CONFIG_MEMORY_HOTREMOVE
 /*
  * All pages in the range must be isolated before calling this.
@@ -5815,3 +5839,47 @@ void dump_page(struct page *page)
 	dump_page_flags(page->flags);
 	mem_cgroup_print_bad_page(page);
 }
+
+/* -------------- test ----------------------------------- */
+static int test_init(void)
+{
+	unsigned long start = 500*1024*1024L;
+	unsigned long end = 600*1024*1024L;
+	struct page *page;
+
+	/* both start and size must be pageblock and max buddy aligned */
+	while (start < end) {
+		int pfn = start >> PAGE_SHIFT;
+
+		page = pfn_to_page(pfn);
+		set_migratetype_movable_nofb(page);
+		start += pageblock_nr_pages * PAGE_SIZE;
+	}
+	return 0;
+}
+late_initcall(test_init);
+
+extern int isolate_memory(u64 start, u64 size);
+#define MODULE_PARAM_PREFIX "test."
+static int isolate_mm_set(const char *val, struct kernel_param *kp)
+{
+	if (!strncmp(val, "isolate", strlen("isolate") - 1))
+		isolate_memory(500*1024*1024L, 100*1024*1024L);
+	if (!strncmp(val, "use", strlen("use") - 1)) {
+		unsigned long start = 500*1024*1024L;
+		unsigned long end = 600*1024*1024L;
+		struct page *page;
+
+		/* both start and size must be pageblock and max buddy aligned */
+		while (start < end) {
+			int pfn = start >> PAGE_SHIFT;
+
+			page = pfn_to_page(pfn);
+			unset_migratetype_isolate(page);
+			start += pageblock_nr_pages * PAGE_SIZE;
+		}
+	}
+	return 0;
+}
+module_param_call(isolate_mm, isolate_mm_set, NULL, NULL, 0200);
+#undef MODULE_PARAM_PREFIX
Index: linux/mm/vmstat.c
===================================================================
--- linux.orig/mm/vmstat.c	2011-08-03 14:59:47.000000000 +0800
+++ linux/mm/vmstat.c	2011-08-03 15:24:52.000000000 +0800
@@ -612,6 +612,7 @@ static char * const migratetype_names[MI
 	"Unmovable",
 	"Reclaimable",
 	"Movable",
+	"Movable-nofb",
 	"Reserve",
 	"Isolate",
 };
Index: linux/mm/memory_hotplug.c
===================================================================
--- linux.orig/mm/memory_hotplug.c	2011-08-03 14:59:47.000000000 +0800
+++ linux/mm/memory_hotplug.c	2011-08-03 15:24:52.000000000 +0800
@@ -864,15 +864,9 @@ check_pages_isolated(unsigned long start
 	return offlined;
 }
 
-static int __ref offline_pages(unsigned long start_pfn,
-		  unsigned long end_pfn, unsigned long timeout)
+static int __ref offline_pages_check(unsigned long start_pfn,
+	unsigned long end_pfn)
 {
-	unsigned long pfn, nr_pages, expire;
-	long offlined_pages;
-	int ret, drain, retry_max, node;
-	struct zone *zone;
-	struct memory_notify arg;
-
 	BUG_ON(start_pfn >= end_pfn);
 	/* at least, alignment against pageblock is necessary */
 	if (!IS_ALIGNED(start_pfn, pageblock_nr_pages))
@@ -883,28 +877,14 @@ static int __ref offline_pages(unsigned
 	   we assume this for now. .*/
 	if (!test_pages_in_a_zone(start_pfn, end_pfn))
 		return -EINVAL;
+	return 0;
+}
 
-	lock_memory_hotplug();
-
-	zone = page_zone(pfn_to_page(start_pfn));
-	node = zone_to_nid(zone);
-	nr_pages = end_pfn - start_pfn;
-
-	/* set above range as isolated */
-	ret = start_isolate_page_range(start_pfn, end_pfn);
-	if (ret)
-		goto out;
-
-	arg.start_pfn = start_pfn;
-	arg.nr_pages = nr_pages;
-	arg.status_change_nid = -1;
-	if (nr_pages >= node_present_pages(node))
-		arg.status_change_nid = node;
-
-	ret = memory_notify(MEM_GOING_OFFLINE, &arg);
-	ret = notifier_to_errno(ret);
-	if (ret)
-		goto failed_removal;
+static int __ref offline_pages_migrate(unsigned long start_pfn,
+	unsigned long end_pfn, unsigned long timeout, long *offlined_pages)
+{
+	unsigned long pfn, expire;
+	int ret, drain, retry_max;
 
 	pfn = start_pfn;
 	expire = jiffies + timeout;
@@ -914,10 +894,10 @@ repeat:
 	/* start memory hot removal */
 	ret = -EAGAIN;
 	if (time_after(jiffies, expire))
-		goto failed_removal;
+		goto out;
 	ret = -EINTR;
 	if (signal_pending(current))
-		goto failed_removal;
+		goto out;
 	ret = 0;
 	if (drain) {
 		lru_add_drain_all();
@@ -934,7 +914,7 @@ repeat:
 		} else {
 			if (ret < 0)
 				if (--retry_max == 0)
-					goto failed_removal;
+					goto out;
 			yield();
 			drain = 1;
 			goto repeat;
@@ -946,11 +926,52 @@ repeat:
 	/* drain pcp pages , this is synchrouns. */
 	drain_all_pages();
 	/* check again */
-	offlined_pages = check_pages_isolated(start_pfn, end_pfn);
-	if (offlined_pages < 0) {
+	*offlined_pages = check_pages_isolated(start_pfn, end_pfn);
+	if (*offlined_pages < 0) {
 		ret = -EBUSY;
-		goto failed_removal;
+		goto out;
 	}
+out:
+	return ret;
+}
+
+static int __ref offline_pages(unsigned long start_pfn,
+		  unsigned long end_pfn, unsigned long timeout)
+{
+	unsigned long nr_pages;
+	long offlined_pages;
+	int ret, node;
+	struct zone *zone;
+	struct memory_notify arg;
+
+	if (offline_pages_check(start_pfn, end_pfn))
+		return -EINVAL;
+
+	lock_memory_hotplug();
+
+	zone = page_zone(pfn_to_page(start_pfn));
+	node = zone_to_nid(zone);
+	nr_pages = end_pfn - start_pfn;
+
+	/* set above range as isolated */
+	ret = start_isolate_page_range(start_pfn, end_pfn);
+	if (ret)
+		goto out;
+
+	arg.start_pfn = start_pfn;
+	arg.nr_pages = nr_pages;
+	arg.status_change_nid = -1;
+	if (nr_pages >= node_present_pages(node))
+		arg.status_change_nid = node;
+
+	ret = memory_notify(MEM_GOING_OFFLINE, &arg);
+	ret = notifier_to_errno(ret);
+	if (ret)
+		goto failed_removal;
+
+	ret = offline_pages_migrate(start_pfn, end_pfn, timeout, &offlined_pages);
+	if (ret)
+		goto failed_removal;
 	printk(KERN_INFO "Offlined Pages %ld\n", offlined_pages);
 	/* Ok, all of our target is islaoted.
 	   We cannot do rollback at this point. */
@@ -996,6 +1017,32 @@ int remove_memory(u64 start, u64 size)
 	end_pfn = start_pfn + PFN_DOWN(size);
 	return offline_pages(start_pfn, end_pfn, 120 * HZ);
 }
+
+int isolate_memory(u64 start, u64 size)
+{
+	unsigned long start_pfn, end_pfn;
+	long offlined_pages;
+	int ret;
+
+	start_pfn = PFN_DOWN(start);
+	end_pfn = start_pfn + PFN_DOWN(size);
+
+	if (offline_pages_check(start_pfn, end_pfn))
+		return -EINVAL;
+	lock_memory_hotplug();
+
+	/* set above range as isolated */
+	ret = start_isolate_page_range(start_pfn, end_pfn);
+	if (ret)
+		goto out;
+	ret = offline_pages_migrate(start_pfn, end_pfn,
+		120 * HZ, &offlined_pages);
+	if (ret)
+		undo_isolate_page_range(start_pfn, end_pfn);
+out:
+	unlock_memory_hotplug();
+	return ret;
+}
 #else
 int remove_memory(u64 start, u64 size)
 {

Re: questions about memory hotplug

[Shaohua Li]

On Tue, 2011-08-02 at 09:09 +0800, Shaohua Li wrote:
> On Tue, 2011-08-02 at 01:08 +0800, Larry Bassel wrote:
> > 
> > In use case #1 yes, maybe not in #2 (we can arrange it to be
> > at the end of memory, but then might waste memory as it may
> > not be aligned on a SPARSEMEM section boundary and so would
> > need to be padded).
> then maybe the new migrate type I suggested can help here for the
> non-aligned memory. Anyway, let me do an experiment.
so your problem is to offline memory in arbitrary address and size (eg,
might not be at the end, and maybe smaller than a section)

I had a hack. In my machine, I have DMA, DMA32, and NORMAL zone.
At boot time, I mark 500M~600M ranges as MOVABLE_NOFB. the range is in
DMA32 and not section size aligned. MOVABLE_NOFB is a new migrate type I
added. That range memory is movable, but other type of allocation can't
fallback into such ranges. so such range memory can only be used by
userspace.
I then run a memory stress test and do memory online/offline for the
range at runtime, the offline always success.
Does this meet your usage? If yes, I'll cook it up a little bit.


---
 include/linux/gfp.h    |    5 +-
 include/linux/mmzone.h |    7 +-
 mm/memory_hotplug.c    |  117 ++++++++++++++++++++++++++++++++++---------------
 mm/page_alloc.c        |   94 +++++++++++++++++++++++++++++++++------
 mm/vmstat.c            |    1 
 5 files changed, 171 insertions(+), 53 deletions(-)

Index: linux/include/linux/gfp.h
===================================================================
--- linux.orig/include/linux/gfp.h	2011-08-03 14:59:47.000000000 +0800
+++ linux/include/linux/gfp.h	2011-08-03 15:24:52.000000000 +0800
@@ -155,8 +155,9 @@ static inline int allocflags_to_migratet
 		return MIGRATE_UNMOVABLE;
 
 	/* Group based on mobility */
-	return (((gfp_flags & __GFP_MOVABLE) != 0) << 1) |
-		((gfp_flags & __GFP_RECLAIMABLE) != 0);
+	if ((gfp_flags & __GFP_MOVABLE) != 0)
+		return MIGRATE_MOVABLE_NOFB;
+	return (gfp_flags & __GFP_RECLAIMABLE) != 0;
 }
 
 #ifdef CONFIG_HIGHMEM
Index: linux/include/linux/mmzone.h
===================================================================
--- linux.orig/include/linux/mmzone.h	2011-08-03 14:59:47.000000000 +0800
+++ linux/include/linux/mmzone.h	2011-08-03 15:24:52.000000000 +0800
@@ -39,9 +39,10 @@
 #define MIGRATE_RECLAIMABLE   1
 #define MIGRATE_MOVABLE       2
 #define MIGRATE_PCPTYPES      3 /* the number of types on the pcp lists */
-#define MIGRATE_RESERVE       3
-#define MIGRATE_ISOLATE       4 /* can't allocate from here */
-#define MIGRATE_TYPES         5
+#define MIGRATE_MOVABLE_NOFB  3
+#define MIGRATE_RESERVE       4
+#define MIGRATE_ISOLATE       5 /* can't allocate from here */
+#define MIGRATE_TYPES         6
 
 #define for_each_migratetype_order(order, type) \
 	for (order = 0; order < MAX_ORDER; order++) \
Index: linux/mm/page_alloc.c
===================================================================
--- linux.orig/mm/page_alloc.c	2011-08-03 14:59:47.000000000 +0800
+++ linux/mm/page_alloc.c	2011-08-03 15:25:29.000000000 +0800
@@ -627,7 +627,7 @@ static void free_pcppages_bulk(struct zo
 			page = list_entry(list->prev, struct page, lru);
 			/* must delete as __free_one_page list manipulates */
 			list_del(&page->lru);
-			/* MIGRATE_MOVABLE list may include MIGRATE_RESERVEs */
+			/* MIGRATE_MOVABLE list may include MIGRATE_RESERVE and MIGRATE_MOVABLE_NOFBs */
 			__free_one_page(page, zone, 0, page_private(page));
 			trace_mm_page_pcpu_drain(page, 0, page_private(page));
 		} while (--to_free && --batch_free && !list_empty(list));
@@ -828,10 +828,11 @@ struct page *__rmqueue_smallest(struct z
  * the free lists for the desirable migrate type are depleted
  */
 static int fallbacks[MIGRATE_TYPES][MIGRATE_TYPES-1] = {
-	[MIGRATE_UNMOVABLE]   = { MIGRATE_RECLAIMABLE, MIGRATE_MOVABLE,   MIGRATE_RESERVE },
-	[MIGRATE_RECLAIMABLE] = { MIGRATE_UNMOVABLE,   MIGRATE_MOVABLE,   MIGRATE_RESERVE },
-	[MIGRATE_MOVABLE]     = { MIGRATE_RECLAIMABLE, MIGRATE_UNMOVABLE, MIGRATE_RESERVE },
-	[MIGRATE_RESERVE]     = { MIGRATE_RESERVE,     MIGRATE_RESERVE,   MIGRATE_RESERVE }, /* Never used */
+	[MIGRATE_UNMOVABLE]   = { MIGRATE_RECLAIMABLE, MIGRATE_MOVABLE,   MIGRATE_RESERVE, MIGRATE_RESERVE },
+	[MIGRATE_RECLAIMABLE] = { MIGRATE_UNMOVABLE,   MIGRATE_MOVABLE,   MIGRATE_RESERVE, MIGRATE_RESERVE},
+	[MIGRATE_MOVABLE]     = { MIGRATE_RECLAIMABLE, MIGRATE_UNMOVABLE, MIGRATE_RESERVE, MIGRATE_RESERVE},
+	[MIGRATE_MOVABLE_NOFB] = { MIGRATE_MOVABLE, MIGRATE_RECLAIMABLE, MIGRATE_UNMOVABLE, MIGRATE_RESERVE}, /* Never used */
+	[MIGRATE_RESERVE]     = { MIGRATE_RESERVE,     MIGRATE_RESERVE,   MIGRATE_RESERVE, MIGRATE_RESERVE}, /* Never used */
 };
 
 /*
@@ -923,6 +924,7 @@ __rmqueue_fallback(struct zone *zone, in
 	struct page *page;
 	int migratetype, i;
 
+	BUG_ON(start_migratetype == MIGRATE_MOVABLE_NOFB);
 	/* Find the largest possible block of pages in the other list */
 	for (current_order = MAX_ORDER-1; current_order >= order;
 						--current_order) {
@@ -994,8 +996,16 @@ static struct page *__rmqueue(struct zon
 	struct page *page;
 
 retry_reserve:
+	/* FIXME: Speed up a little bit */
 	page = __rmqueue_smallest(zone, order, migratetype);
 
+	if (!page && migratetype == MIGRATE_MOVABLE_NOFB) {
+		/* Don't change other migrate type to MIGRATE_MOVABLE_NOFB, so
+		 * we shortcut here */
+		migratetype = MIGRATE_MOVABLE;
+		goto retry_reserve;
+	}
+
 	if (unlikely(!page) && migratetype != MIGRATE_RESERVE) {
 		page = __rmqueue_fallback(zone, order, migratetype);
 
@@ -1044,7 +1054,8 @@ static int rmqueue_bulk(struct zone *zon
 			list_add(&page->lru, list);
 		else
 			list_add_tail(&page->lru, list);
-		set_page_private(page, migratetype);
+		/* __rmqueue might do fallback, so reread the migrate type */
+		set_page_private(page, get_pageblock_migratetype(page));
 		list = &page->lru;
 	}
 	__mod_zone_page_state(zone, NR_FREE_PAGES, -(i << order));
@@ -1300,10 +1311,11 @@ again:
 	if (likely(order == 0)) {
 		struct per_cpu_pages *pcp;
 		struct list_head *list;
+		int migratetype_pcp = migratetype == MIGRATE_MOVABLE_NOFB? MIGRATE_MOVABLE : migratetype;
 
 		local_irq_save(flags);
 		pcp = &this_cpu_ptr(zone->pageset)->pcp;
-		list = &pcp->lists[migratetype];
+		list = &pcp->lists[migratetype_pcp];
 		if (list_empty(list)) {
 			pcp->count += rmqueue_bulk(zone, 0,
 					pcp->batch, list,
@@ -5558,7 +5570,8 @@ __count_immobile_pages(struct zone *zone
 	if (zone_idx(zone) == ZONE_MOVABLE)
 		return true;
 
-	if (get_pageblock_migratetype(page) == MIGRATE_MOVABLE)
+	if (get_pageblock_migratetype(page) == MIGRATE_MOVABLE ||
+	    get_pageblock_migratetype(page) == MIGRATE_MOVABLE_NOFB )
 		return true;
 
 	pfn = page_to_pfn(page);
@@ -5601,7 +5614,7 @@ bool is_pageblock_removable_nolock(struc
 	return __count_immobile_pages(zone, page, 0);
 }
 
-int set_migratetype_isolate(struct page *page)
+static int __set_migratetype(struct page *page, int migratetype)
 {
 	struct zone *zone;
 	unsigned long flags, pfn;
@@ -5647,8 +5660,8 @@ int set_migratetype_isolate(struct page
 
 out:
 	if (!ret) {
-		set_pageblock_migratetype(page, MIGRATE_ISOLATE);
-		move_freepages_block(zone, page, MIGRATE_ISOLATE);
+		set_pageblock_migratetype(page, migratetype);
+		move_freepages_block(zone, page, migratetype);
 	}
 
 	spin_unlock_irqrestore(&zone->lock, flags);
@@ -5657,6 +5670,11 @@ out:
 	return ret;
 }
 
+int set_migratetype_isolate(struct page *page)
+{
+	return __set_migratetype(page, MIGRATE_ISOLATE);
+}
+
 void unset_migratetype_isolate(struct page *page)
 {
 	struct zone *zone;
@@ -5665,12 +5683,18 @@ void unset_migratetype_isolate(struct pa
 	spin_lock_irqsave(&zone->lock, flags);
 	if (get_pageblock_migratetype(page) != MIGRATE_ISOLATE)
 		goto out;
-	set_pageblock_migratetype(page, MIGRATE_MOVABLE);
-	move_freepages_block(zone, page, MIGRATE_MOVABLE);
+	/* Fixme: should restore previous type */
+	set_pageblock_migratetype(page, MIGRATE_MOVABLE_NOFB);
+	move_freepages_block(zone, page, MIGRATE_MOVABLE_NOFB);
 out:
 	spin_unlock_irqrestore(&zone->lock, flags);
 }
 
+int set_migratetype_movable_nofb(struct page *page)
+{
+	return __set_migratetype(page, MIGRATE_MOVABLE_NOFB);
+}
+
 #ifdef CONFIG_MEMORY_HOTREMOVE
 /*
  * All pages in the range must be isolated before calling this.
@@ -5815,3 +5839,47 @@ void dump_page(struct page *page)
 	dump_page_flags(page->flags);
 	mem_cgroup_print_bad_page(page);
 }
+
+/* -------------- test ----------------------------------- */
+static int test_init(void)
+{
+	unsigned long start = 500*1024*1024L;
+	unsigned long end = 600*1024*1024L;
+	struct page *page;
+
+	/* both start and size must be pageblock and max buddy aligned */
+	while (start < end) {
+		int pfn = start >> PAGE_SHIFT;
+
+		page = pfn_to_page(pfn);
+		set_migratetype_movable_nofb(page);
+		start += pageblock_nr_pages * PAGE_SIZE;
+	}
+	return 0;
+}
+late_initcall(test_init);
+
+extern int isolate_memory(u64 start, u64 size);
+#define MODULE_PARAM_PREFIX "test."
+static int isolate_mm_set(const char *val, struct kernel_param *kp)
+{
+	if (!strncmp(val, "isolate", strlen("isolate") - 1))
+		isolate_memory(500*1024*1024L, 100*1024*1024L);
+	if (!strncmp(val, "use", strlen("use") - 1)) {
+		unsigned long start = 500*1024*1024L;
+		unsigned long end = 600*1024*1024L;
+		struct page *page;
+
+		/* both start and size must be pageblock and max buddy aligned */
+		while (start < end) {
+			int pfn = start >> PAGE_SHIFT;
+
+			page = pfn_to_page(pfn);
+			unset_migratetype_isolate(page);
+			start += pageblock_nr_pages * PAGE_SIZE;
+		}
+	}
+	return 0;
+}
+module_param_call(isolate_mm, isolate_mm_set, NULL, NULL, 0200);
+#undef MODULE_PARAM_PREFIX
Index: linux/mm/vmstat.c
===================================================================
--- linux.orig/mm/vmstat.c	2011-08-03 14:59:47.000000000 +0800
+++ linux/mm/vmstat.c	2011-08-03 15:24:52.000000000 +0800
@@ -612,6 +612,7 @@ static char * const migratetype_names[MI
 	"Unmovable",
 	"Reclaimable",
 	"Movable",
+	"Movable-nofb",
 	"Reserve",
 	"Isolate",
 };
Index: linux/mm/memory_hotplug.c
===================================================================
--- linux.orig/mm/memory_hotplug.c	2011-08-03 14:59:47.000000000 +0800
+++ linux/mm/memory_hotplug.c	2011-08-03 15:24:52.000000000 +0800
@@ -864,15 +864,9 @@ check_pages_isolated(unsigned long start
 	return offlined;
 }
 
-static int __ref offline_pages(unsigned long start_pfn,
-		  unsigned long end_pfn, unsigned long timeout)
+static int __ref offline_pages_check(unsigned long start_pfn,
+	unsigned long end_pfn)
 {
-	unsigned long pfn, nr_pages, expire;
-	long offlined_pages;
-	int ret, drain, retry_max, node;
-	struct zone *zone;
-	struct memory_notify arg;
-
 	BUG_ON(start_pfn >= end_pfn);
 	/* at least, alignment against pageblock is necessary */
 	if (!IS_ALIGNED(start_pfn, pageblock_nr_pages))
@@ -883,28 +877,14 @@ static int __ref offline_pages(unsigned
 	   we assume this for now. .*/
 	if (!test_pages_in_a_zone(start_pfn, end_pfn))
 		return -EINVAL;
+	return 0;
+}
 
-	lock_memory_hotplug();
-
-	zone = page_zone(pfn_to_page(start_pfn));
-	node = zone_to_nid(zone);
-	nr_pages = end_pfn - start_pfn;
-
-	/* set above range as isolated */
-	ret = start_isolate_page_range(start_pfn, end_pfn);
-	if (ret)
-		goto out;
-
-	arg.start_pfn = start_pfn;
-	arg.nr_pages = nr_pages;
-	arg.status_change_nid = -1;
-	if (nr_pages >= node_present_pages(node))
-		arg.status_change_nid = node;
-
-	ret = memory_notify(MEM_GOING_OFFLINE, &arg);
-	ret = notifier_to_errno(ret);
-	if (ret)
-		goto failed_removal;
+static int __ref offline_pages_migrate(unsigned long start_pfn,
+	unsigned long end_pfn, unsigned long timeout, long *offlined_pages)
+{
+	unsigned long pfn, expire;
+	int ret, drain, retry_max;
 
 	pfn = start_pfn;
 	expire = jiffies + timeout;
@@ -914,10 +894,10 @@ repeat:
 	/* start memory hot removal */
 	ret = -EAGAIN;
 	if (time_after(jiffies, expire))
-		goto failed_removal;
+		goto out;
 	ret = -EINTR;
 	if (signal_pending(current))
-		goto failed_removal;
+		goto out;
 	ret = 0;
 	if (drain) {
 		lru_add_drain_all();
@@ -934,7 +914,7 @@ repeat:
 		} else {
 			if (ret < 0)
 				if (--retry_max == 0)
-					goto failed_removal;
+					goto out;
 			yield();
 			drain = 1;
 			goto repeat;
@@ -946,11 +926,52 @@ repeat:
 	/* drain pcp pages , this is synchrouns. */
 	drain_all_pages();
 	/* check again */
-	offlined_pages = check_pages_isolated(start_pfn, end_pfn);
-	if (offlined_pages < 0) {
+	*offlined_pages = check_pages_isolated(start_pfn, end_pfn);
+	if (*offlined_pages < 0) {
 		ret = -EBUSY;
-		goto failed_removal;
+		goto out;
 	}
+out:
+	return ret;
+}
+
+static int __ref offline_pages(unsigned long start_pfn,
+		  unsigned long end_pfn, unsigned long timeout)
+{
+	unsigned long nr_pages;
+	long offlined_pages;
+	int ret, node;
+	struct zone *zone;
+	struct memory_notify arg;
+
+	if (offline_pages_check(start_pfn, end_pfn))
+		return -EINVAL;
+
+	lock_memory_hotplug();
+
+	zone = page_zone(pfn_to_page(start_pfn));
+	node = zone_to_nid(zone);
+	nr_pages = end_pfn - start_pfn;
+
+	/* set above range as isolated */
+	ret = start_isolate_page_range(start_pfn, end_pfn);
+	if (ret)
+		goto out;
+
+	arg.start_pfn = start_pfn;
+	arg.nr_pages = nr_pages;
+	arg.status_change_nid = -1;
+	if (nr_pages >= node_present_pages(node))
+		arg.status_change_nid = node;
+
+	ret = memory_notify(MEM_GOING_OFFLINE, &arg);
+	ret = notifier_to_errno(ret);
+	if (ret)
+		goto failed_removal;
+
+	ret = offline_pages_migrate(start_pfn, end_pfn, timeout, &offlined_pages);
+	if (ret)
+		goto failed_removal;
 	printk(KERN_INFO "Offlined Pages %ld\n", offlined_pages);
 	/* Ok, all of our target is islaoted.
 	   We cannot do rollback at this point. */
@@ -996,6 +1017,32 @@ int remove_memory(u64 start, u64 size)
 	end_pfn = start_pfn + PFN_DOWN(size);
 	return offline_pages(start_pfn, end_pfn, 120 * HZ);
 }
+
+int isolate_memory(u64 start, u64 size)
+{
+	unsigned long start_pfn, end_pfn;
+	long offlined_pages;
+	int ret;
+
+	start_pfn = PFN_DOWN(start);
+	end_pfn = start_pfn + PFN_DOWN(size);
+
+	if (offline_pages_check(start_pfn, end_pfn))
+		return -EINVAL;
+	lock_memory_hotplug();
+
+	/* set above range as isolated */
+	ret = start_isolate_page_range(start_pfn, end_pfn);
+	if (ret)
+		goto out;
+	ret = offline_pages_migrate(start_pfn, end_pfn,
+		120 * HZ, &offlined_pages);
+	if (ret)
+		undo_isolate_page_range(start_pfn, end_pfn);
+out:
+	unlock_memory_hotplug();
+	return ret;
+}
 #else
 int remove_memory(u64 start, u64 size)
 {


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

questions about memory hotplug

[Shaohua Li]

On Tue, 2011-08-02 at 09:09 +0800, Shaohua Li wrote:
> On Tue, 2011-08-02 at 01:08 +0800, Larry Bassel wrote:
> > 
> > In use case #1 yes, maybe not in #2 (we can arrange it to be
> > at the end of memory, but then might waste memory as it may
> > not be aligned on a SPARSEMEM section boundary and so would
> > need to be padded).
> then maybe the new migrate type I suggested can help here for the
> non-aligned memory. Anyway, let me do an experiment.
so your problem is to offline memory in arbitrary address and size (eg,
might not be at the end, and maybe smaller than a section)

I had a hack. In my machine, I have DMA, DMA32, and NORMAL zone.
At boot time, I mark 500M~600M ranges as MOVABLE_NOFB. the range is in
DMA32 and not section size aligned. MOVABLE_NOFB is a new migrate type I
added. That range memory is movable, but other type of allocation can't
fallback into such ranges. so such range memory can only be used by
userspace.
I then run a memory stress test and do memory online/offline for the
range at runtime, the offline always success.
Does this meet your usage? If yes, I'll cook it up a little bit.


---
 include/linux/gfp.h    |    5 +-
 include/linux/mmzone.h |    7 +-
 mm/memory_hotplug.c    |  117 ++++++++++++++++++++++++++++++++++---------------
 mm/page_alloc.c        |   94 +++++++++++++++++++++++++++++++++------
 mm/vmstat.c            |    1 
 5 files changed, 171 insertions(+), 53 deletions(-)

Index: linux/include/linux/gfp.h
===================================================================
--- linux.orig/include/linux/gfp.h	2011-08-03 14:59:47.000000000 +0800
+++ linux/include/linux/gfp.h	2011-08-03 15:24:52.000000000 +0800
@@ -155,8 +155,9 @@ static inline int allocflags_to_migratet
 		return MIGRATE_UNMOVABLE;
 
 	/* Group based on mobility */
-	return (((gfp_flags & __GFP_MOVABLE) != 0) << 1) |
-		((gfp_flags & __GFP_RECLAIMABLE) != 0);
+	if ((gfp_flags & __GFP_MOVABLE) != 0)
+		return MIGRATE_MOVABLE_NOFB;
+	return (gfp_flags & __GFP_RECLAIMABLE) != 0;
 }
 
 #ifdef CONFIG_HIGHMEM
Index: linux/include/linux/mmzone.h
===================================================================
--- linux.orig/include/linux/mmzone.h	2011-08-03 14:59:47.000000000 +0800
+++ linux/include/linux/mmzone.h	2011-08-03 15:24:52.000000000 +0800
@@ -39,9 +39,10 @@
 #define MIGRATE_RECLAIMABLE   1
 #define MIGRATE_MOVABLE       2
 #define MIGRATE_PCPTYPES      3 /* the number of types on the pcp lists */
-#define MIGRATE_RESERVE       3
-#define MIGRATE_ISOLATE       4 /* can't allocate from here */
-#define MIGRATE_TYPES         5
+#define MIGRATE_MOVABLE_NOFB  3
+#define MIGRATE_RESERVE       4
+#define MIGRATE_ISOLATE       5 /* can't allocate from here */
+#define MIGRATE_TYPES         6
 
 #define for_each_migratetype_order(order, type) \
 	for (order = 0; order < MAX_ORDER; order++) \
Index: linux/mm/page_alloc.c
===================================================================
--- linux.orig/mm/page_alloc.c	2011-08-03 14:59:47.000000000 +0800
+++ linux/mm/page_alloc.c	2011-08-03 15:25:29.000000000 +0800
@@ -627,7 +627,7 @@ static void free_pcppages_bulk(struct zo
 			page = list_entry(list->prev, struct page, lru);
 			/* must delete as __free_one_page list manipulates */
 			list_del(&page->lru);
-			/* MIGRATE_MOVABLE list may include MIGRATE_RESERVEs */
+			/* MIGRATE_MOVABLE list may include MIGRATE_RESERVE and MIGRATE_MOVABLE_NOFBs */
 			__free_one_page(page, zone, 0, page_private(page));
 			trace_mm_page_pcpu_drain(page, 0, page_private(page));
 		} while (--to_free && --batch_free && !list_empty(list));
@@ -828,10 +828,11 @@ struct page *__rmqueue_smallest(struct z
  * the free lists for the desirable migrate type are depleted
  */
 static int fallbacks[MIGRATE_TYPES][MIGRATE_TYPES-1] = {
-	[MIGRATE_UNMOVABLE]   = { MIGRATE_RECLAIMABLE, MIGRATE_MOVABLE,   MIGRATE_RESERVE },
-	[MIGRATE_RECLAIMABLE] = { MIGRATE_UNMOVABLE,   MIGRATE_MOVABLE,   MIGRATE_RESERVE },
-	[MIGRATE_MOVABLE]     = { MIGRATE_RECLAIMABLE, MIGRATE_UNMOVABLE, MIGRATE_RESERVE },
-	[MIGRATE_RESERVE]     = { MIGRATE_RESERVE,     MIGRATE_RESERVE,   MIGRATE_RESERVE }, /* Never used */
+	[MIGRATE_UNMOVABLE]   = { MIGRATE_RECLAIMABLE, MIGRATE_MOVABLE,   MIGRATE_RESERVE, MIGRATE_RESERVE },
+	[MIGRATE_RECLAIMABLE] = { MIGRATE_UNMOVABLE,   MIGRATE_MOVABLE,   MIGRATE_RESERVE, MIGRATE_RESERVE},
+	[MIGRATE_MOVABLE]     = { MIGRATE_RECLAIMABLE, MIGRATE_UNMOVABLE, MIGRATE_RESERVE, MIGRATE_RESERVE},
+	[MIGRATE_MOVABLE_NOFB] = { MIGRATE_MOVABLE, MIGRATE_RECLAIMABLE, MIGRATE_UNMOVABLE, MIGRATE_RESERVE}, /* Never used */
+	[MIGRATE_RESERVE]     = { MIGRATE_RESERVE,     MIGRATE_RESERVE,   MIGRATE_RESERVE, MIGRATE_RESERVE}, /* Never used */
 };
 
 /*
@@ -923,6 +924,7 @@ __rmqueue_fallback(struct zone *zone, in
 	struct page *page;
 	int migratetype, i;
 
+	BUG_ON(start_migratetype == MIGRATE_MOVABLE_NOFB);
 	/* Find the largest possible block of pages in the other list */
 	for (current_order = MAX_ORDER-1; current_order >= order;
 						--current_order) {
@@ -994,8 +996,16 @@ static struct page *__rmqueue(struct zon
 	struct page *page;
 
 retry_reserve:
+	/* FIXME: Speed up a little bit */
 	page = __rmqueue_smallest(zone, order, migratetype);
 
+	if (!page && migratetype == MIGRATE_MOVABLE_NOFB) {
+		/* Don't change other migrate type to MIGRATE_MOVABLE_NOFB, so
+		 * we shortcut here */
+		migratetype = MIGRATE_MOVABLE;
+		goto retry_reserve;
+	}
+
 	if (unlikely(!page) && migratetype != MIGRATE_RESERVE) {
 		page = __rmqueue_fallback(zone, order, migratetype);
 
@@ -1044,7 +1054,8 @@ static int rmqueue_bulk(struct zone *zon
 			list_add(&page->lru, list);
 		else
 			list_add_tail(&page->lru, list);
-		set_page_private(page, migratetype);
+		/* __rmqueue might do fallback, so reread the migrate type */
+		set_page_private(page, get_pageblock_migratetype(page));
 		list = &page->lru;
 	}
 	__mod_zone_page_state(zone, NR_FREE_PAGES, -(i << order));
@@ -1300,10 +1311,11 @@ again:
 	if (likely(order == 0)) {
 		struct per_cpu_pages *pcp;
 		struct list_head *list;
+		int migratetype_pcp = migratetype == MIGRATE_MOVABLE_NOFB? MIGRATE_MOVABLE : migratetype;
 
 		local_irq_save(flags);
 		pcp = &this_cpu_ptr(zone->pageset)->pcp;
-		list = &pcp->lists[migratetype];
+		list = &pcp->lists[migratetype_pcp];
 		if (list_empty(list)) {
 			pcp->count += rmqueue_bulk(zone, 0,
 					pcp->batch, list,
@@ -5558,7 +5570,8 @@ __count_immobile_pages(struct zone *zone
 	if (zone_idx(zone) == ZONE_MOVABLE)
 		return true;
 
-	if (get_pageblock_migratetype(page) == MIGRATE_MOVABLE)
+	if (get_pageblock_migratetype(page) == MIGRATE_MOVABLE ||
+	    get_pageblock_migratetype(page) == MIGRATE_MOVABLE_NOFB )
 		return true;
 
 	pfn = page_to_pfn(page);
@@ -5601,7 +5614,7 @@ bool is_pageblock_removable_nolock(struc
 	return __count_immobile_pages(zone, page, 0);
 }
 
-int set_migratetype_isolate(struct page *page)
+static int __set_migratetype(struct page *page, int migratetype)
 {
 	struct zone *zone;
 	unsigned long flags, pfn;
@@ -5647,8 +5660,8 @@ int set_migratetype_isolate(struct page
 
 out:
 	if (!ret) {
-		set_pageblock_migratetype(page, MIGRATE_ISOLATE);
-		move_freepages_block(zone, page, MIGRATE_ISOLATE);
+		set_pageblock_migratetype(page, migratetype);
+		move_freepages_block(zone, page, migratetype);
 	}
 
 	spin_unlock_irqrestore(&zone->lock, flags);
@@ -5657,6 +5670,11 @@ out:
 	return ret;
 }
 
+int set_migratetype_isolate(struct page *page)
+{
+	return __set_migratetype(page, MIGRATE_ISOLATE);
+}
+
 void unset_migratetype_isolate(struct page *page)
 {
 	struct zone *zone;
@@ -5665,12 +5683,18 @@ void unset_migratetype_isolate(struct pa
 	spin_lock_irqsave(&zone->lock, flags);
 	if (get_pageblock_migratetype(page) != MIGRATE_ISOLATE)
 		goto out;
-	set_pageblock_migratetype(page, MIGRATE_MOVABLE);
-	move_freepages_block(zone, page, MIGRATE_MOVABLE);
+	/* Fixme: should restore previous type */
+	set_pageblock_migratetype(page, MIGRATE_MOVABLE_NOFB);
+	move_freepages_block(zone, page, MIGRATE_MOVABLE_NOFB);
 out:
 	spin_unlock_irqrestore(&zone->lock, flags);
 }
 
+int set_migratetype_movable_nofb(struct page *page)
+{
+	return __set_migratetype(page, MIGRATE_MOVABLE_NOFB);
+}
+
 #ifdef CONFIG_MEMORY_HOTREMOVE
 /*
  * All pages in the range must be isolated before calling this.
@@ -5815,3 +5839,47 @@ void dump_page(struct page *page)
 	dump_page_flags(page->flags);
 	mem_cgroup_print_bad_page(page);
 }
+
+/* -------------- test ----------------------------------- */
+static int test_init(void)
+{
+	unsigned long start = 500*1024*1024L;
+	unsigned long end = 600*1024*1024L;
+	struct page *page;
+
+	/* both start and size must be pageblock and max buddy aligned */
+	while (start < end) {
+		int pfn = start >> PAGE_SHIFT;
+
+		page = pfn_to_page(pfn);
+		set_migratetype_movable_nofb(page);
+		start += pageblock_nr_pages * PAGE_SIZE;
+	}
+	return 0;
+}
+late_initcall(test_init);
+
+extern int isolate_memory(u64 start, u64 size);
+#define MODULE_PARAM_PREFIX "test."
+static int isolate_mm_set(const char *val, struct kernel_param *kp)
+{
+	if (!strncmp(val, "isolate", strlen("isolate") - 1))
+		isolate_memory(500*1024*1024L, 100*1024*1024L);
+	if (!strncmp(val, "use", strlen("use") - 1)) {
+		unsigned long start = 500*1024*1024L;
+		unsigned long end = 600*1024*1024L;
+		struct page *page;
+
+		/* both start and size must be pageblock and max buddy aligned */
+		while (start < end) {
+			int pfn = start >> PAGE_SHIFT;
+
+			page = pfn_to_page(pfn);
+			unset_migratetype_isolate(page);
+			start += pageblock_nr_pages * PAGE_SIZE;
+		}
+	}
+	return 0;
+}
+module_param_call(isolate_mm, isolate_mm_set, NULL, NULL, 0200);
+#undef MODULE_PARAM_PREFIX
Index: linux/mm/vmstat.c
===================================================================
--- linux.orig/mm/vmstat.c	2011-08-03 14:59:47.000000000 +0800
+++ linux/mm/vmstat.c	2011-08-03 15:24:52.000000000 +0800
@@ -612,6 +612,7 @@ static char * const migratetype_names[MI
 	"Unmovable",
 	"Reclaimable",
 	"Movable",
+	"Movable-nofb",
 	"Reserve",
 	"Isolate",
 };
Index: linux/mm/memory_hotplug.c
===================================================================
--- linux.orig/mm/memory_hotplug.c	2011-08-03 14:59:47.000000000 +0800
+++ linux/mm/memory_hotplug.c	2011-08-03 15:24:52.000000000 +0800
@@ -864,15 +864,9 @@ check_pages_isolated(unsigned long start
 	return offlined;
 }
 
-static int __ref offline_pages(unsigned long start_pfn,
-		  unsigned long end_pfn, unsigned long timeout)
+static int __ref offline_pages_check(unsigned long start_pfn,
+	unsigned long end_pfn)
 {
-	unsigned long pfn, nr_pages, expire;
-	long offlined_pages;
-	int ret, drain, retry_max, node;
-	struct zone *zone;
-	struct memory_notify arg;
-
 	BUG_ON(start_pfn >= end_pfn);
 	/* at least, alignment against pageblock is necessary */
 	if (!IS_ALIGNED(start_pfn, pageblock_nr_pages))
@@ -883,28 +877,14 @@ static int __ref offline_pages(unsigned
 	   we assume this for now. .*/
 	if (!test_pages_in_a_zone(start_pfn, end_pfn))
 		return -EINVAL;
+	return 0;
+}
 
-	lock_memory_hotplug();
-
-	zone = page_zone(pfn_to_page(start_pfn));
-	node = zone_to_nid(zone);
-	nr_pages = end_pfn - start_pfn;
-
-	/* set above range as isolated */
-	ret = start_isolate_page_range(start_pfn, end_pfn);
-	if (ret)
-		goto out;
-
-	arg.start_pfn = start_pfn;
-	arg.nr_pages = nr_pages;
-	arg.status_change_nid = -1;
-	if (nr_pages >= node_present_pages(node))
-		arg.status_change_nid = node;
-
-	ret = memory_notify(MEM_GOING_OFFLINE, &arg);
-	ret = notifier_to_errno(ret);
-	if (ret)
-		goto failed_removal;
+static int __ref offline_pages_migrate(unsigned long start_pfn,
+	unsigned long end_pfn, unsigned long timeout, long *offlined_pages)
+{
+	unsigned long pfn, expire;
+	int ret, drain, retry_max;
 
 	pfn = start_pfn;
 	expire = jiffies + timeout;
@@ -914,10 +894,10 @@ repeat:
 	/* start memory hot removal */
 	ret = -EAGAIN;
 	if (time_after(jiffies, expire))
-		goto failed_removal;
+		goto out;
 	ret = -EINTR;
 	if (signal_pending(current))
-		goto failed_removal;
+		goto out;
 	ret = 0;
 	if (drain) {
 		lru_add_drain_all();
@@ -934,7 +914,7 @@ repeat:
 		} else {
 			if (ret < 0)
 				if (--retry_max == 0)
-					goto failed_removal;
+					goto out;
 			yield();
 			drain = 1;
 			goto repeat;
@@ -946,11 +926,52 @@ repeat:
 	/* drain pcp pages , this is synchrouns. */
 	drain_all_pages();
 	/* check again */
-	offlined_pages = check_pages_isolated(start_pfn, end_pfn);
-	if (offlined_pages < 0) {
+	*offlined_pages = check_pages_isolated(start_pfn, end_pfn);
+	if (*offlined_pages < 0) {
 		ret = -EBUSY;
-		goto failed_removal;
+		goto out;
 	}
+out:
+	return ret;
+}
+
+static int __ref offline_pages(unsigned long start_pfn,
+		  unsigned long end_pfn, unsigned long timeout)
+{
+	unsigned long nr_pages;
+	long offlined_pages;
+	int ret, node;
+	struct zone *zone;
+	struct memory_notify arg;
+
+	if (offline_pages_check(start_pfn, end_pfn))
+		return -EINVAL;
+
+	lock_memory_hotplug();
+
+	zone = page_zone(pfn_to_page(start_pfn));
+	node = zone_to_nid(zone);
+	nr_pages = end_pfn - start_pfn;
+
+	/* set above range as isolated */
+	ret = start_isolate_page_range(start_pfn, end_pfn);
+	if (ret)
+		goto out;
+
+	arg.start_pfn = start_pfn;
+	arg.nr_pages = nr_pages;
+	arg.status_change_nid = -1;
+	if (nr_pages >= node_present_pages(node))
+		arg.status_change_nid = node;
+
+	ret = memory_notify(MEM_GOING_OFFLINE, &arg);
+	ret = notifier_to_errno(ret);
+	if (ret)
+		goto failed_removal;
+
+	ret = offline_pages_migrate(start_pfn, end_pfn, timeout, &offlined_pages);
+	if (ret)
+		goto failed_removal;
 	printk(KERN_INFO "Offlined Pages %ld\n", offlined_pages);
 	/* Ok, all of our target is islaoted.
 	   We cannot do rollback at this point. */
@@ -996,6 +1017,32 @@ int remove_memory(u64 start, u64 size)
 	end_pfn = start_pfn + PFN_DOWN(size);
 	return offline_pages(start_pfn, end_pfn, 120 * HZ);
 }
+
+int isolate_memory(u64 start, u64 size)
+{
+	unsigned long start_pfn, end_pfn;
+	long offlined_pages;
+	int ret;
+
+	start_pfn = PFN_DOWN(start);
+	end_pfn = start_pfn + PFN_DOWN(size);
+
+	if (offline_pages_check(start_pfn, end_pfn))
+		return -EINVAL;
+	lock_memory_hotplug();
+
+	/* set above range as isolated */
+	ret = start_isolate_page_range(start_pfn, end_pfn);
+	if (ret)
+		goto out;
+	ret = offline_pages_migrate(start_pfn, end_pfn,
+		120 * HZ, &offlined_pages);
+	if (ret)
+		undo_isolate_page_range(start_pfn, end_pfn);
+out:
+	unlock_memory_hotplug();
+	return ret;
+}
 #else
 int remove_memory(u64 start, u64 size)
 {

Re: questions about memory hotplug

[Larry Bassel]

On 03 Aug 11 15:55, Shaohua Li wrote:
> On Tue, 2011-08-02 at 09:09 +0800, Shaohua Li wrote:
> > On Tue, 2011-08-02 at 01:08 +0800, Larry Bassel wrote:
> > > 
> > > In use case #1 yes, maybe not in #2 (we can arrange it to be
> > > at the end of memory, but then might waste memory as it may
> > > not be aligned on a SPARSEMEM section boundary and so would
> > > need to be padded).
> > then maybe the new migrate type I suggested can help here for the
> > non-aligned memory. Anyway, let me do an experiment.
> so your problem is to offline memory in arbitrary address and size (eg,
> might not be at the end, and maybe smaller than a section)

Yes (and online it again). Also the decision to (attempt to)
on/offline must be done from userspace (as memory hotplug does already).

> 
> I had a hack. In my machine, I have DMA, DMA32, and NORMAL zone.
> At boot time, I mark 500M~600M ranges as MOVABLE_NOFB. the range is in
> DMA32 and not section size aligned.

A few questions:

* You still use SPARSEMEM though, correct?
* Would there be any problem using NORMAL memory as MOVABLE_NOFB?
* So you don't use ZONE_MOVABLE or kernelcore= or movablecore=
at all?
* Do you online/offline using the /sys/devices/system/memory files?
If so, does the kernel still attempt to on/offline the entire
section (as it does now) or only the MOVABLE_NOFB part?
If not, how do you on/offline memory?

> MOVABLE_NOFB is a new migrate type I
> added. That range memory is movable, but other type of allocation can't
> fallback into such ranges. so such range memory can only be used by
> userspace.

And the kernel will not reserve memory from it either (I had to put a
hack in an earlier version of what I'm doing to not allow reserving
from the movable zone because otherwise the existence of these
reserved pages would block logical hotremove), correct?

> I then run a memory stress test and do memory online/offline for the
> range at runtime, the offline always success.
> Does this meet your usage? If yes, I'll cook it up a little bit.

Yes, this looks very promising.
Do you see any reason this can't be backported to 2.6.38?

Thank you very much for your help here.

Larry

-- 
Sent by an employee of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.

Re: questions about memory hotplug

[Larry Bassel]

On 03 Aug 11 15:55, Shaohua Li wrote:
> On Tue, 2011-08-02 at 09:09 +0800, Shaohua Li wrote:
> > On Tue, 2011-08-02 at 01:08 +0800, Larry Bassel wrote:
> > > 
> > > In use case #1 yes, maybe not in #2 (we can arrange it to be
> > > at the end of memory, but then might waste memory as it may
> > > not be aligned on a SPARSEMEM section boundary and so would
> > > need to be padded).
> > then maybe the new migrate type I suggested can help here for the
> > non-aligned memory. Anyway, let me do an experiment.
> so your problem is to offline memory in arbitrary address and size (eg,
> might not be at the end, and maybe smaller than a section)

Yes (and online it again). Also the decision to (attempt to)
on/offline must be done from userspace (as memory hotplug does already).

> 
> I had a hack. In my machine, I have DMA, DMA32, and NORMAL zone.
> At boot time, I mark 500M~600M ranges as MOVABLE_NOFB. the range is in
> DMA32 and not section size aligned.

A few questions:

* You still use SPARSEMEM though, correct?
* Would there be any problem using NORMAL memory as MOVABLE_NOFB?
* So you don't use ZONE_MOVABLE or kernelcore= or movablecore=
at all?
* Do you online/offline using the /sys/devices/system/memory files?
If so, does the kernel still attempt to on/offline the entire
section (as it does now) or only the MOVABLE_NOFB part?
If not, how do you on/offline memory?

> MOVABLE_NOFB is a new migrate type I
> added. That range memory is movable, but other type of allocation can't
> fallback into such ranges. so such range memory can only be used by
> userspace.

And the kernel will not reserve memory from it either (I had to put a
hack in an earlier version of what I'm doing to not allow reserving
from the movable zone because otherwise the existence of these
reserved pages would block logical hotremove), correct?

> I then run a memory stress test and do memory online/offline for the
> range at runtime, the offline always success.
> Does this meet your usage? If yes, I'll cook it up a little bit.

Yes, this looks very promising.
Do you see any reason this can't be backported to 2.6.38?

Thank you very much for your help here.

Larry

-- 
Sent by an employee of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

questions about memory hotplug

[Larry Bassel]

On 03 Aug 11 15:55, Shaohua Li wrote:
> On Tue, 2011-08-02 at 09:09 +0800, Shaohua Li wrote:
> > On Tue, 2011-08-02 at 01:08 +0800, Larry Bassel wrote:
> > > 
> > > In use case #1 yes, maybe not in #2 (we can arrange it to be
> > > at the end of memory, but then might waste memory as it may
> > > not be aligned on a SPARSEMEM section boundary and so would
> > > need to be padded).
> > then maybe the new migrate type I suggested can help here for the
> > non-aligned memory. Anyway, let me do an experiment.
> so your problem is to offline memory in arbitrary address and size (eg,
> might not be at the end, and maybe smaller than a section)

Yes (and online it again). Also the decision to (attempt to)
on/offline must be done from userspace (as memory hotplug does already).

> 
> I had a hack. In my machine, I have DMA, DMA32, and NORMAL zone.
> At boot time, I mark 500M~600M ranges as MOVABLE_NOFB. the range is in
> DMA32 and not section size aligned.

A few questions:

* You still use SPARSEMEM though, correct?
* Would there be any problem using NORMAL memory as MOVABLE_NOFB?
* So you don't use ZONE_MOVABLE or kernelcore= or movablecore=
at all?
* Do you online/offline using the /sys/devices/system/memory files?
If so, does the kernel still attempt to on/offline the entire
section (as it does now) or only the MOVABLE_NOFB part?
If not, how do you on/offline memory?

> MOVABLE_NOFB is a new migrate type I
> added. That range memory is movable, but other type of allocation can't
> fallback into such ranges. so such range memory can only be used by
> userspace.

And the kernel will not reserve memory from it either (I had to put a
hack in an earlier version of what I'm doing to not allow reserving
from the movable zone because otherwise the existence of these
reserved pages would block logical hotremove), correct?

> I then run a memory stress test and do memory online/offline for the
> range at runtime, the offline always success.
> Does this meet your usage? If yes, I'll cook it up a little bit.

Yes, this looks very promising.
Do you see any reason this can't be backported to 2.6.38?

Thank you very much for your help here.

Larry

-- 
Sent by an employee of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.