mm: Export split_page().

PATCH In review K. Y. Srinivasan 3 participants 7 messages Mar 4, 2013

Subsystems (2)

Memory management Page allocator

lore

Share & tools

lore

K. Y. Srinivasan

author

2013-03-04 01:58 UTC #0 lore raw

PATCH 1/1 mm: Export split_page().

The split_page() function will be very useful for balloon drivers. On Hyper-V, it will be very efficient to use 2M allocations in the guest as this (a) makes the ballooning protocol with the host that much more efficient and (b) moving memory in 2M chunks minimizes fragmentation in the host. Export the split_page() function to let the guest allocations be in 2M chunks while the host is free to return this memory at arbitrary granularity.

Signed-off-by K. Y. Srinivasan <[email protected]>

---
 mm/page_alloc.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

mm/page_alloc.c +1 -0

index 6cacfee..7e0ead6 100644
@@ -1404,6 +1404,7 @@ void split_page(struct page *page, unsigned int order)
 	for (i = 1; i < (1 << order); i++)
 		set_page_refcounted(page + i);
 }
+EXPORT_SYMBOL_GPL(split_page);
 
 static int __isolate_free_page(struct page *page, unsigned int order)
 {

Signature · 7 lines

1.7.4.1

--
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>

Greg KH

2013-03-04 02:07 UTC #1 lore raw

Quoted 18 lines · K. Y. Srinivasan

The split_page() function will be very useful for balloon drivers. On Hyper-V, it will be very efficient to use 2M allocations in the guest as this (a) makes the ballooning protocol with the host that much more efficient and (b) moving memory in 2M chunks minimizes fragmentation in the host. Export the split_page() function to let the guest allocations be in 2M chunks while the host is free to return this memory at arbitrary granularity.
Signed-off-by K. Y. Srinivasan <[email protected]>
---
 mm/page_alloc.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)
mm/page_alloc.c +0 -0
index 6cacfee..7e0ead6 100644

mm/page_alloc.c void split_page(struct page *page, unsigned int order) · line 1407

Quoted 4 lines

mm/page_alloc.c +1 -0

@@ -1404,6 +1404,7 @@ void split_page(struct page *page, unsigned int order)
 	for (i = 1; i < (1 << order); i++)
 		set_page_refcounted(page + i);
 }
+EXPORT_SYMBOL_GPL(split_page);

When you export a symbol, you also need to post the code that is going to use that symbol, otherwise people don't really know how to judge this request.

Can you just make this a part of your balloon driver update patch series instead?

thanks,

greg k-h

--
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>

KY Srinivasan

author

2013-03-04 02:14 UTC #2 lore raw

Quoted 40 lines

-----Original Message-----
From: Greg KH [mailto:[email protected]]
Sent: Sunday, March 03, 2013 9:08 PM
To: KY Srinivasan
Cc: [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; linux- [email protected]
Subject: Re: [PATCH 1/1] mm: Export split_page().
Quoted 4 lines · K. Y. Srinivasan
The split_page() function will be very useful for balloon drivers. On Hyper-V, it will be very efficient to use 2M allocations in the guest as this (a) makes the ballooning protocol with the host that much more efficient and (b) moving memory in 2M chunks minimizes fragmentation in the host. Export the
split_page()
Quoted 14 lines
function to let the guest allocations be in 2M chunks while the host is free to return this memory at arbitrary granularity.
Signed-off-by K. Y. Srinivasan <[email protected]>
---
 mm/page_alloc.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)
mm/page_alloc.c +0 -0
index 6cacfee..7e0ead6 100644
mm/page_alloc.c void split_page(struct page *page, unsigned int order) · line 1407
Quoted 4 lines
mm/page_alloc.c +1 -0
@@ -1404,6 +1404,7 @@ void split_page(struct page *page, unsigned int order)
 	for (i = 1; i < (1 << order); i++)
 		set_page_refcounted(page + i);
 }
+EXPORT_SYMBOL_GPL(split_page);
When you export a symbol, you also need to post the code that is going to use that symbol, otherwise people don't really know how to judge this request.

Quoted 2 lines

Can you just make this a part of your balloon driver update patch series instead?

Fair enough; I was hoping to see how inclined the mm folks were with regards to exporting this symbol before I went ahead and modified the balloon driver code to leverage this. Looking at the Windows guests on Hyper-V, I am convinced 2M balloon allocations in the Linux (Hyper-V) balloon driver will make significant difference. As you suggest, I will post this patch as part of the balloon driver changes that use this exported symbol. I am still hoping to get some feedback from the mm guys on this.

Regards,

K. Y

Greg KH

2013-03-04 02:25 UTC #3 lore raw

Quoted 43 lines · KY Srinivasan

Quoted 40 lines
-----Original Message-----
From: Greg KH [mailto:[email protected]]
Sent: Sunday, March 03, 2013 9:08 PM
To: KY Srinivasan
Cc: [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; linux- [email protected]
Subject: Re: [PATCH 1/1] mm: Export split_page().
Quoted 4 lines · K. Y. Srinivasan
The split_page() function will be very useful for balloon drivers. On Hyper-V, it will be very efficient to use 2M allocations in the guest as this (a) makes the ballooning protocol with the host that much more efficient and (b) moving memory in 2M chunks minimizes fragmentation in the host. Export the
split_page()
Quoted 14 lines
function to let the guest allocations be in 2M chunks while the host is free to return this memory at arbitrary granularity.
Signed-off-by K. Y. Srinivasan <[email protected]>
---
 mm/page_alloc.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)
mm/page_alloc.c +0 -0
index 6cacfee..7e0ead6 100644
mm/page_alloc.c void split_page(struct page *page, unsigned int order) · line 1407
Quoted 4 lines
mm/page_alloc.c +1 -0
@@ -1404,6 +1404,7 @@ void split_page(struct page *page, unsigned int order)
 	for (i = 1; i < (1 << order); i++)
 		set_page_refcounted(page + i);
 }
+EXPORT_SYMBOL_GPL(split_page);
When you export a symbol, you also need to post the code that is going to use that symbol, otherwise people don't really know how to judge this request.
Quoted 2 lines
Can you just make this a part of your balloon driver update patch series instead?

Quoted 6 lines

Fair enough; I was hoping to see how inclined the mm folks were with regards to exporting this symbol before I went ahead and modified the balloon driver code to leverage this. Looking at the Windows guests on Hyper-V, I am convinced 2M balloon allocations in the Linux (Hyper-V) balloon driver will make significant difference. As you suggest, I will post this patch as part of the balloon driver changes that use this exported symbol. I am still hoping to get some feedback from the mm guys on this.

I guess the most obvious question about exporting this symbol is, "Why doesn't any of the other hypervisor balloon drivers need this? What is so special about hyper-v?"

Or can those other drivers also need/use it as well, and they were just too chicken to be asking for the export? :)

thanks,

greg k-h

KY Srinivasan

author

2013-03-04 02:36 UTC #4 lore raw

Quoted 65 lines

-----Original Message-----
From: Greg KH [mailto:[email protected]]
Sent: Sunday, March 03, 2013 9:25 PM
To: KY Srinivasan
Cc: [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; linux- [email protected]
Subject: Re: [PATCH 1/1] mm: Export split_page().
Quoted 7 lines · KY Srinivasan
Quoted 7 lines
-----Original Message-----
From: Greg KH [mailto:[email protected]]
Sent: Sunday, March 03, 2013 9:08 PM
To: KY Srinivasan
Cc: [email protected]; [email protected];
[email protected];
Quoted 8 lines
Quoted 8 lines
[email protected]; [email protected]; [email protected]; linux- [email protected]
Subject: Re: [PATCH 1/1] mm: Export split_page().
Quoted 3 lines · K. Y. Srinivasan
The split_page() function will be very useful for balloon drivers. On Hyper-V, it will be very efficient to use 2M allocations in the guest as this (a) makes the ballooning protocol with the host that much more efficient and (b)
moving
Quoted 11 lines
Quoted 11 lines
Quoted 1 line
memory in 2M chunks minimizes fragmentation in the host. Export the
split_page()
Quoted 9 lines
function to let the guest allocations be in 2M chunks while the host is free to return this memory at arbitrary granularity.
Signed-off-by K. Y. Srinivasan <[email protected]>
---
 mm/page_alloc.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)
Quoted 5 lines
Quoted 5 lines
mm/page_alloc.c void split_page(struct page *page, unsigned int
Quoted 5 lines
mm/page_alloc.c +0 -0
index 6cacfee..7e0ead6 100644
@@ -1404,6 +1404,7 @@ void split_page(struct page *page, unsigned int
order)
Quoted 5 lines
Quoted 5 lines
Quoted 4 lines
mm/page_alloc.c +1 -0
 	for (i = 1; i < (1 << order); i++)
 		set_page_refcounted(page + i);
 }
+EXPORT_SYMBOL_GPL(split_page);
Quoted 8 lines
Quoted 6 lines
When you export a symbol, you also need to post the code that is going to use that symbol, otherwise people don't really know how to judge this request.
Can you just make this a part of your balloon driver update patch series instead?
Fair enough; I was hoping to see how inclined the mm folks were with regards
to
Quoted 1 line
exporting this symbol before I went ahead and modified the balloon driver
code to
Quoted 1 line
leverage this. Looking at the Windows guests on Hyper-V, I am convinced 2M
balloon
Quoted 1 line
allocations in the Linux (Hyper-V) balloon driver will make significant difference.

Quoted 8 lines

As you
Quoted 1 line
suggest, I will post this patch as part of the balloon driver changes that use this
exported
Quoted 1 line
symbol. I am still hoping to get some feedback from the mm guys on this.
I guess the most obvious question about exporting this symbol is, "Why doesn't any of the other hypervisor balloon drivers need this? What is so special about hyper-v?"

The balloon protocol that Hyper-V has specified is designed around the ability to move 2M pages. While the protocol can handle 4k allocations, it is going to be very chatty with 4K allocations. Furthermore, the Memory Balancer on the host is also designed to work best with memory moving around in 2M chunks. While I have not seen the code on the Windows host that does this memory balancing, looking at how Windows guests behave in this environment, (relative to Linux) I have to assume that the 2M allocations that Windows guests do are a big part of the difference we see.

Quoted 3 lines

Or can those other drivers also need/use it as well, and they were just too chicken to be asking for the export? :)

The 2M balloon allocations would make sense if the host is designed accordingly.

Regards,

K. Y

Dave Hansen

2013-03-04 17:06 UTC #5 lore raw

Quoted 7 lines · KY Srinivasan

Quoted 3 lines
I guess the most obvious question about exporting this symbol is, "Why doesn't any of the other hypervisor balloon drivers need this? What is so special about hyper-v?"
The balloon protocol that Hyper-V has specified is designed around the ability to move 2M pages. While the protocol can handle 4k allocations, it is going to be very chatty with 4K allocations.

What does "very chatty" mean? Do you think that there will be a noticeable performance difference ballooning 2M pages vs 4k?

Quoted 5 lines

Furthermore, the Memory Balancer on the host is also designed to work
best with memory moving around in 2M chunks. While I have not seen the code on the Windows host that does this memory balancing, looking at how Windows guests behave in this environment, (relative to Linux) I have to assume that the 2M allocations that Windows guests do are a big part of the difference we see.

You've been talking about differences. Could you elaborate on what the differences in behavior are that you are trying to rectify here?

Quoted 4 lines

Quoted 2 lines
Or can those other drivers also need/use it as well, and they were just too chicken to be asking for the export? :)
The 2M balloon allocations would make sense if the host is designed accordingly.

How does the guest decide which size pages to allocate? It seems like a relatively bad idea to be inflating the balloon with 2M pages from the guest in the case where the guest is under memory pressure _and_ fragmented.

KY Srinivasan

author

2013-03-04 18:13 UTC #6 lore raw

Quoted 16 lines

-----Original Message-----
From: Dave Hansen [mailto:[email protected]]
Sent: Monday, March 04, 2013 12:06 PM
To: KY Srinivasan
Cc: Greg KH; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; akpm@linux- foundation.org; [email protected]
Subject: Re: [PATCH 1/1] mm: Export split_page().
Quoted 4 lines · KY Srinivasan
Quoted 3 lines
I guess the most obvious question about exporting this symbol is, "Why doesn't any of the other hypervisor balloon drivers need this? What is so special about hyper-v?"

Quoted 8 lines

Quoted 1 line
The balloon protocol that Hyper-V has specified is designed around the ability
to
Quoted 1 line
move 2M pages. While the protocol can handle 4k allocations, it is going to be
very chatty
Quoted 1 line
with 4K allocations.
What does "very chatty" mean? Do you think that there will be a noticeable performance difference ballooning 2M pages vs 4k?

The balloon protocol that Hyper-V host specified allows you to specify page ranges - start_pfn: num_pfn. With 2M pages the number of messages that need to be exchanges is significantly fewer than with 4K page allocations.

Quoted 4 lines

Quoted 3 lines
Furthermore, the Memory Balancer on the host is also designed to work best with memory moving around in 2M chunks. While I have not seen the
code on the Windows

Quoted 8 lines

Quoted 1 line
host that does this memory balancing, looking at how Windows guests behave
in this environment,
Quoted 1 line
(relative to Linux) I have to assume that the 2M allocations that Windows
guests do are a big part of
Quoted 1 line
the difference we see.
You've been talking about differences. Could you elaborate on what the differences in behavior are that you are trying to rectify here?

As I look at how smoothly memory is balanced on Windows guests with changing load conditions in the guest relative to what I see with Linux, I see Linux taking more time to reach the steady state during a balancing operation. I will experiment with 2M allocations and report if this issue is addressed.

Quoted 11 lines

Quoted 5 lines
Quoted 3 lines
Or can those other drivers also need/use it as well, and they were just too chicken to be asking for the export? :)
The 2M balloon allocations would make sense if the host is designed
accordingly.
How does the guest decide which size pages to allocate? It seems like a relatively bad idea to be inflating the balloon with 2M pages from the guest in the case where the guest is under memory pressure _and_ fragmented.

I want to start with 2M allocations and if they fail, fall back onto lower order allocations. As I said, the host can support 4K allocations and that will be the final fallback position (that is what I have currently implemented). If the guest memory is fragmented, then obviously we will go in for lower order allocations.

Regards,

K. Y