Skip to contentSkip to conversation

drivers: xen-selfballoon: consider slab pages

PATCH In review Bob Liu 4 participants 9 messages Nov 4, 2013 – Nov 5, 2013 Xen hypervisor interface

lore
Share & tools
lore

Bob Liu

author
#0 lore raw

PATCH 1/1 drivers: xen-selfballoon: consider slab pages

Currently the goal_page in xen-selfballon doesn't consider much about pages used in kernel space.
A typical usage is slab pages, without consider slab pages the goal_page result may be too rough and lead extra memory pressure to guest os.

---
 drivers/xen/xen-selfballoon.c |    2 ++
 1 file changed, 2 insertions(+)
drivers/xen/xen-selfballoon.c +2 -0
index 21e18c1..4814759 100644
@@ -191,6 +191,8 @@ static void selfballoon_process(struct work_struct *work)
 		tgt_pages = cur_pages; /* default is no change */
 		goal_pages = vm_memory_committed() +
 				totalreserve_pages +
+				global_page_state(NR_SLAB_RECLAIMABLE) +
+				global_page_state(NR_SLAB_UNRECLAIMABLE) +
 				MB2PAGES(selfballoon_reserved_mb);
 #ifdef CONFIG_FRONTSWAP
 		/* allow space for frontswap pages to be repatriated */
Signature · 1 line
1.7.10.4
#1 lore raw
Quoted 4 lines · Bob Liu

Currently the goal_page in xen-selfballon doesn't consider much about pages used in kernel space.
A typical usage is slab pages, without consider slab pages the goal_page result may be too rough and lead extra memory pressure to guest os.

Does "lead extra memory pressure" mean OOM?

Quoted 22 lines
---
 drivers/xen/xen-selfballoon.c |    2 ++
 1 file changed, 2 insertions(+)
drivers/xen/xen-selfballoon.c +2 -0
index 21e18c1..4814759 100644
@@ -191,6 +191,8 @@ static void selfballoon_process(struct work_struct *work)
 		tgt_pages = cur_pages; /* default is no change */
 		goal_pages = vm_memory_committed() +
 				totalreserve_pages +
+				global_page_state(NR_SLAB_RECLAIMABLE) +
+				global_page_state(NR_SLAB_UNRECLAIMABLE) +
 				MB2PAGES(selfballoon_reserved_mb);
 #ifdef CONFIG_FRONTSWAP
 		/* allow space for frontswap pages to be repatriated */
Signature · 1 line
1.7.10.4
#2 lore raw
In reply to Bob Liu
Quoted 4 lines · Bob Liu

Currently the goal_page in xen-selfballon doesn't consider much about pages used in kernel space.
A typical usage is slab pages, without consider slab pages the goal_page result may be too rough and lead extra memory pressure to guest os.

Does "lead extra memory pressure" mean OOM?

Quoted 22 lines
---
 drivers/xen/xen-selfballoon.c |    2 ++
 1 file changed, 2 insertions(+)
drivers/xen/xen-selfballoon.c +2 -0
index 21e18c1..4814759 100644
@@ -191,6 +191,8 @@ static void selfballoon_process(struct work_struct *work)
 		tgt_pages = cur_pages; /* default is no change */
 		goal_pages = vm_memory_committed() +
 				totalreserve_pages +
+				global_page_state(NR_SLAB_RECLAIMABLE) +
+				global_page_state(NR_SLAB_UNRECLAIMABLE) +
 				MB2PAGES(selfballoon_reserved_mb);
 #ifdef CONFIG_FRONTSWAP
 		/* allow space for frontswap pages to be repatriated */
Signature · 1 line
1.7.10.4
#3 lore raw
In reply to Bob Liu
Quoted 4 lines · Bob Liu

Currently the goal_page in xen-selfballon doesn't consider much about pages used in kernel space.
A typical usage is slab pages, without consider slab pages the goal_page result may be too rough and lead extra memory pressure to guest os.

Can you provide some real world figures where the calculatation got it wrong? What was the resultant behavior? Swap death? OOM killer?

Quoted 10 lines
---
 drivers/xen/xen-selfballoon.c |    2 ++
 1 file changed, 2 insertions(+)
drivers/xen/xen-selfballoon.c +0 -0
index 21e18c1..4814759 100644

drivers/xen/xen-selfballoon.c static void selfballoon_process(struct work_struct *work) · line 194

Quoted 4 lines
drivers/xen/xen-selfballoon.c +1 -0
@@ -191,6 +191,8 @@ static void selfballoon_process(struct work_struct *work)
 		tgt_pages = cur_pages; /* default is no change */
 		goal_pages = vm_memory_committed() +
 				totalreserve_pages +
+				global_page_state(NR_SLAB_RECLAIMABLE) +

Does SLAB_RECLAIMABLE want to be included here? Unless I'm misunderstanding here, SLAB_RECLAIMABLE is effectively free.

drivers/xen/xen-selfballoon.c static void selfballoon_process(struct work_struct *work) · line 195

Quoted 1 line
drivers/xen/xen-selfballoon.c +1 -0
+				global_page_state(NR_SLAB_UNRECLAIMABLE) +

This bit looks fine to me.

drivers/xen/xen-selfballoon.c static void selfballoon_process(struct work_struct *work) · line 198

Quoted 3 lines
drivers/xen/xen-selfballoon.c +0 -0
 				MB2PAGES(selfballoon_reserved_mb);
 #ifdef CONFIG_FRONTSWAP
 		/* allow space for frontswap pages to be repatriated */

David

#4 lore raw
Quoted 9 lines · David Vrabel
Quoted 4 lines · Bob Liu

Currently the goal_page in xen-selfballon doesn't consider much about pages used in kernel space.
A typical usage is slab pages, without consider slab pages the goal_page result may be too rough and lead extra memory pressure to guest os.

Can you provide some real world figures where the calculatation got it wrong? What was the resultant behavior? Swap death? OOM killer?

Sorry, I didn't run any testing I just think it's unreasonable while reading the source code.

vm_memory_committed() only calculate pages which mapped to process address space, but the kernel itself(like block, fs and network subsystem) may occupy some memory. And it's possible that those subsystem may occupy a significant amount of memory in some situation.

I'm afraid if we don't consider those kernel memory while calculating goal_pages, guest memory will be set lower than guest really needs.

Quoted 5 lines
Quoted 5 lines
---
 drivers/xen/xen-selfballoon.c |    2 ++
 1 file changed, 2 insertions(+)
Quoted 13 lines

drivers/xen/xen-selfballoon.c static void selfballoon_process(struct work_struct *work) · line 194

Quoted 9 lines
drivers/xen/xen-selfballoon.c +1 -0
index 21e18c1..4814759 100644
@@ -191,6 +191,8 @@ static void selfballoon_process(struct work_struct *work)
 		tgt_pages = cur_pages; /* default is no change */
 		goal_pages = vm_memory_committed() +
 				totalreserve_pages +
+				global_page_state(NR_SLAB_RECLAIMABLE) +

Does SLAB_RECLAIMABLE want to be included here? Unless I'm misunderstanding here, SLAB_RECLAIMABLE is effectively free.

SLAB_RECLAIMABLE isn't effectively free, it means the slab page is in used but can be reclaimed(freed) during memory pressure.

Quoted 10 lines

drivers/xen/xen-selfballoon.c static void selfballoon_process(struct work_struct *work) · line 195

Quoted 1 line
drivers/xen/xen-selfballoon.c +1 -0
+				global_page_state(NR_SLAB_UNRECLAIMABLE) +

This bit looks fine to me.

drivers/xen/xen-selfballoon.c static void selfballoon_process(struct work_struct *work) · line 198

Quoted 3 lines
drivers/xen/xen-selfballoon.c +0 -0
 				MB2PAGES(selfballoon_reserved_mb);
 #ifdef CONFIG_FRONTSWAP
 		/* allow space for frontswap pages to be repatriated */

David

Thanks for your review.

Signature · 2 lines
Regards,
-Bob
#5 lore raw
Quoted 8 lines · Bob Liu
Quoted 6 lines · David Vrabel
Quoted 4 lines · Bob Liu

Currently the goal_page in xen-selfballon doesn't consider much about pages used in kernel space.
A typical usage is slab pages, without consider slab pages the goal_page result may be too rough and lead extra memory pressure to guest os.

Quoted 6 lines
Quoted 3 lines

Can you provide some real world figures where the calculatation got it wrong? What was the resultant behavior? Swap death? OOM killer?

Sorry, I didn't run any testing I just think it's unreasonable while reading the source code.

I'm not keen on changes to heuristics without testing with real workloads and a demonstration that it is better in practice.

Quoted 27 lines

vm_memory_committed() only calculate pages which mapped to process address space, but the kernel itself(like block, fs and network subsystem) may occupy some memory. And it's possible that those subsystem may occupy a significant amount of memory in some situation.

I'm afraid if we don't consider those kernel memory while calculating goal_pages, guest memory will be set lower than guest really needs.

Quoted 5 lines
Quoted 5 lines
---
 drivers/xen/xen-selfballoon.c |    2 ++
 1 file changed, 2 insertions(+)
Quoted 13 lines

drivers/xen/xen-selfballoon.c static void selfballoon_process(struct work_struct *work) · line 194

Quoted 9 lines
drivers/xen/xen-selfballoon.c +1 -0
index 21e18c1..4814759 100644
@@ -191,6 +191,8 @@ static void selfballoon_process(struct work_struct *work)
 		tgt_pages = cur_pages; /* default is no change */
 		goal_pages = vm_memory_committed() +
 				totalreserve_pages +
+				global_page_state(NR_SLAB_RECLAIMABLE) +

Does SLAB_RECLAIMABLE want to be included here? Unless I'm misunderstanding here, SLAB_RECLAIMABLE is effectively free.

Quoted 2 lines

SLAB_RECLAIMABLE isn't effectively free, it means the slab page is in used but can be reclaimed(freed) during memory pressure.

Similarly to the pages used for the page cache, I think the intention is for the selfballoon_reserved_mb value to include this.

Quoted 7 lines
Quoted 7 lines

drivers/xen/xen-selfballoon.c static void selfballoon_process(struct work_struct *work) · line 195

Quoted 1 line
drivers/xen/xen-selfballoon.c +1 -0
+				global_page_state(NR_SLAB_UNRECLAIMABLE) +

This bit looks fine to me.

Quoted 3 lines
drivers/xen/xen-selfballoon.c +0 -0
 				MB2PAGES(selfballoon_reserved_mb);
 #ifdef CONFIG_FRONTSWAP
 		/* allow space for frontswap pages to be repatriated */

David

#6 lore raw
In reply to Bob Liu
Quoted 8 lines · Bob Liu
Quoted 6 lines · David Vrabel
Quoted 4 lines · Bob Liu

Currently the goal_page in xen-selfballon doesn't consider much about pages used in kernel space.
A typical usage is slab pages, without consider slab pages the goal_page result may be too rough and lead extra memory pressure to guest os.

Quoted 6 lines
Quoted 3 lines

Can you provide some real world figures where the calculatation got it wrong? What was the resultant behavior? Swap death? OOM killer?

Sorry, I didn't run any testing I just think it's unreasonable while reading the source code.

I'm not keen on changes to heuristics without testing with real workloads and a demonstration that it is better in practice.

Quoted 27 lines

vm_memory_committed() only calculate pages which mapped to process address space, but the kernel itself(like block, fs and network subsystem) may occupy some memory. And it's possible that those subsystem may occupy a significant amount of memory in some situation.

I'm afraid if we don't consider those kernel memory while calculating goal_pages, guest memory will be set lower than guest really needs.

Quoted 5 lines
Quoted 5 lines
---
 drivers/xen/xen-selfballoon.c |    2 ++
 1 file changed, 2 insertions(+)
Quoted 13 lines

drivers/xen/xen-selfballoon.c static void selfballoon_process(struct work_struct *work) · line 194

Quoted 9 lines
drivers/xen/xen-selfballoon.c +1 -0
index 21e18c1..4814759 100644
@@ -191,6 +191,8 @@ static void selfballoon_process(struct work_struct *work)
 		tgt_pages = cur_pages; /* default is no change */
 		goal_pages = vm_memory_committed() +
 				totalreserve_pages +
+				global_page_state(NR_SLAB_RECLAIMABLE) +

Does SLAB_RECLAIMABLE want to be included here? Unless I'm misunderstanding here, SLAB_RECLAIMABLE is effectively free.

Quoted 2 lines

SLAB_RECLAIMABLE isn't effectively free, it means the slab page is in used but can be reclaimed(freed) during memory pressure.

Similarly to the pages used for the page cache, I think the intention is for the selfballoon_reserved_mb value to include this.

Quoted 7 lines
Quoted 7 lines

drivers/xen/xen-selfballoon.c static void selfballoon_process(struct work_struct *work) · line 195

Quoted 1 line
drivers/xen/xen-selfballoon.c +1 -0
+				global_page_state(NR_SLAB_UNRECLAIMABLE) +

This bit looks fine to me.

Quoted 3 lines
drivers/xen/xen-selfballoon.c +0 -0
 				MB2PAGES(selfballoon_reserved_mb);
 #ifdef CONFIG_FRONTSWAP
 		/* allow space for frontswap pages to be repatriated */

David

#7 lore raw
In reply to David Vrabel
Quoted 9 lines · David Vrabel
Quoted 4 lines · Bob Liu

Currently the goal_page in xen-selfballon doesn't consider much about pages used in kernel space.
A typical usage is slab pages, without consider slab pages the goal_page result may be too rough and lead extra memory pressure to guest os.

Can you provide some real world figures where the calculatation got it wrong? What was the resultant behavior? Swap death? OOM killer?

Sorry, I didn't run any testing I just think it's unreasonable while reading the source code.

vm_memory_committed() only calculate pages which mapped to process address space, but the kernel itself(like block, fs and network subsystem) may occupy some memory. And it's possible that those subsystem may occupy a significant amount of memory in some situation.

I'm afraid if we don't consider those kernel memory while calculating goal_pages, guest memory will be set lower than guest really needs.

Quoted 5 lines
Quoted 5 lines
---
 drivers/xen/xen-selfballoon.c |    2 ++
 1 file changed, 2 insertions(+)
Quoted 13 lines

drivers/xen/xen-selfballoon.c static void selfballoon_process(struct work_struct *work) · line 194

Quoted 9 lines
drivers/xen/xen-selfballoon.c +1 -0
index 21e18c1..4814759 100644
@@ -191,6 +191,8 @@ static void selfballoon_process(struct work_struct *work)
 		tgt_pages = cur_pages; /* default is no change */
 		goal_pages = vm_memory_committed() +
 				totalreserve_pages +
+				global_page_state(NR_SLAB_RECLAIMABLE) +

Does SLAB_RECLAIMABLE want to be included here? Unless I'm misunderstanding here, SLAB_RECLAIMABLE is effectively free.

SLAB_RECLAIMABLE isn't effectively free, it means the slab page is in used but can be reclaimed(freed) during memory pressure.

Quoted 10 lines

drivers/xen/xen-selfballoon.c static void selfballoon_process(struct work_struct *work) · line 195

Quoted 1 line
drivers/xen/xen-selfballoon.c +1 -0
+				global_page_state(NR_SLAB_UNRECLAIMABLE) +

This bit looks fine to me.

drivers/xen/xen-selfballoon.c static void selfballoon_process(struct work_struct *work) · line 198

Quoted 3 lines
drivers/xen/xen-selfballoon.c +0 -0
 				MB2PAGES(selfballoon_reserved_mb);
 #ifdef CONFIG_FRONTSWAP
 		/* allow space for frontswap pages to be repatriated */

David

Thanks for your review.

Signature · 2 lines
Regards,
-Bob
#8 lore raw
In reply to Bob Liu
Quoted 4 lines · Bob Liu

Currently the goal_page in xen-selfballon doesn't consider much about pages used in kernel space.
A typical usage is slab pages, without consider slab pages the goal_page result may be too rough and lead extra memory pressure to guest os.

Can you provide some real world figures where the calculatation got it wrong? What was the resultant behavior? Swap death? OOM killer?

Quoted 10 lines
---
 drivers/xen/xen-selfballoon.c |    2 ++
 1 file changed, 2 insertions(+)
drivers/xen/xen-selfballoon.c +0 -0
index 21e18c1..4814759 100644

drivers/xen/xen-selfballoon.c static void selfballoon_process(struct work_struct *work) · line 194

Quoted 4 lines
drivers/xen/xen-selfballoon.c +1 -0
@@ -191,6 +191,8 @@ static void selfballoon_process(struct work_struct *work)
 		tgt_pages = cur_pages; /* default is no change */
 		goal_pages = vm_memory_committed() +
 				totalreserve_pages +
+				global_page_state(NR_SLAB_RECLAIMABLE) +

Does SLAB_RECLAIMABLE want to be included here? Unless I'm misunderstanding here, SLAB_RECLAIMABLE is effectively free.

drivers/xen/xen-selfballoon.c static void selfballoon_process(struct work_struct *work) · line 195

Quoted 1 line
drivers/xen/xen-selfballoon.c +1 -0
+				global_page_state(NR_SLAB_UNRECLAIMABLE) +

This bit looks fine to me.

drivers/xen/xen-selfballoon.c static void selfballoon_process(struct work_struct *work) · line 198

Quoted 3 lines
drivers/xen/xen-selfballoon.c +0 -0
 				MB2PAGES(selfballoon_reserved_mb);
 #ifdef CONFIG_FRONTSWAP
 		/* allow space for frontswap pages to be repatriated */

David