[RESEND][v2][PATCH] KVM: PPC: Book3S HV: Migrate pinned pages out of CMA

From: Balbir Singh <bsingharora@gmail.com>
Subject: [RESEND][v2][PATCH] KVM: PPC: Book3S HV: Migrate pinned pages out of CMA

When PCI Device pass-through is enabled via VFIO, KVM-PPC will
pin pages using get_user_pages_fast(). One of the downsides of
the pinning is that the page could be in CMA region. The CMA
region is used for other allocations like the hash page table.
Ideally we want the pinned pages to be from non CMA region.

This patch (currently only for KVM PPC with VFIO) forcefully
migrates the pages out (huge pages are omitted for the moment).
There are more efficient ways of doing this, but that might
be elaborate and might impact a larger audience beyond just
the kvm ppc implementation.

The magic is in new_iommu_non_cma_page() which allocates the
new page from a non CMA region.

I've tested the patches lightly at my end, but there might be bugs
For example if after lru_add_drain(), the page is not isolated
is this a BUG?

Previous discussion was at
http://permalink.gmane.org/gmane.linux.kernel.mm/136738

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Paul Mackerras <redacted>
Cc: Alexey Kardashevskiy <redacted>

Signed-off-by: Balbir Singh <bsingharora@gmail.com>
---
 arch/powerpc/include/asm/mmu_context.h |  1 +
 arch/powerpc/mm/mmu_context_iommu.c    | 80 ++++++++++++++++++++++++++++++++--
 2 files changed, 77 insertions(+), 4 deletions(-)

diff --git a/arch/powerpc/include/asm/mmu_context.h b/arch/powerpc/include/asm/mmu_context.h
index 9d2cd0c..475d1be 100644
--- a/arch/powerpc/include/asm/mmu_context.h
+++ b/arch/powerpc/include/asm/mmu_context.h

@@ -18,6 +18,7 @@ extern void destroy_context(struct mm_struct *mm);
 #ifdef CONFIG_SPAPR_TCE_IOMMU
 struct mm_iommu_table_group_mem_t;
 
+extern int isolate_lru_page(struct page *page);	/* from internal.h */
 extern bool mm_iommu_preregistered(void);
 extern long mm_iommu_get(unsigned long ua, unsigned long entries,
 		struct mm_iommu_table_group_mem_t **pmem);

diff --git a/arch/powerpc/mm/mmu_context_iommu.c b/arch/powerpc/mm/mmu_context_iommu.c
index da6a216..c18f742 100644
--- a/arch/powerpc/mm/mmu_context_iommu.c
+++ b/arch/powerpc/mm/mmu_context_iommu.c

@@ -15,6 +15,9 @@
 #include <linux/rculist.h>
 #include <linux/vmalloc.h>
 #include <linux/mutex.h>
+#include <linux/migrate.h>
+#include <linux/hugetlb.h>
+#include <linux/swap.h>
 #include <asm/mmu_context.h>
 
 static DEFINE_MUTEX(mem_list_mutex);

@@ -72,6 +75,54 @@ bool mm_iommu_preregistered(void)
 }
 EXPORT_SYMBOL_GPL(mm_iommu_preregistered);
 
+/*
+ * Taken from alloc_migrate_target with changes to remove CMA allocations
+ */
+struct page *new_iommu_non_cma_page(struct page *page, unsigned long private,
+					int **resultp)
+{
+	gfp_t gfp_mask = GFP_USER;
+	struct page *new_page;
+
+	if (PageHuge(page) || PageTransHuge(page) || PageCompound(page))
+		return NULL;
+
+	if (PageHighMem(page))
+		gfp_mask |= __GFP_HIGHMEM;
+
+	/*
+	 * We don't want the allocation to force an OOM if possibe
+	 */
+	new_page = alloc_page(gfp_mask | __GFP_NORETRY | __GFP_NOWARN);
+	return new_page;
+}
+
+static int mm_iommu_move_page_from_cma(struct page *page)
+{
+	int ret;
+	LIST_HEAD(cma_migrate_pages);
+
+	/* Ignore huge pages for now */
+	if (PageHuge(page) || PageTransHuge(page) || PageCompound(page))
+		return -EBUSY;
+
+	lru_add_drain();
+	ret = isolate_lru_page(page);
+	if (ret)
+		get_page(page); /* Potential BUG? */
+
+	list_add(&page->lru, &cma_migrate_pages);
+	put_page(page); /* Drop the gup reference */
+
+	ret = migrate_pages(&cma_migrate_pages, new_iommu_non_cma_page,
+				NULL, 0, MIGRATE_SYNC, MR_CMA);
+	if (ret) {
+		if (!list_empty(&cma_migrate_pages))
+			putback_movable_pages(&cma_migrate_pages);
+	}
+	return 0;
+}
+
 long mm_iommu_get(unsigned long ua, unsigned long entries,
 		struct mm_iommu_table_group_mem_t **pmem)
 {

@@ -124,15 +175,36 @@ long mm_iommu_get(unsigned long ua, unsigned long entries,
 	for (i = 0; i < entries; ++i) {
 		if (1 != get_user_pages_fast(ua + (i << PAGE_SHIFT),
 					1/* pages */, 1/* iswrite */, &page)) {
+			ret = -EFAULT;
 			for (j = 0; j < i; ++j)
-				put_page(pfn_to_page(
-						mem->hpas[j] >> PAGE_SHIFT));
+				put_page(pfn_to_page(mem->hpas[j] >>
+						PAGE_SHIFT));
 			vfree(mem->hpas);
 			kfree(mem);
-			ret = -EFAULT;
 			goto unlock_exit;
 		}
-
+		/*
+		 * If we get a page from the CMA zone, since we are going to
+		 * be pinning these entries, we might as well move them out
+		 * of the CMA zone if possible. NOTE: faulting in + migration
+		 * can be expensive. Batching can be considered later
+		 */
+		if (get_pageblock_migratetype(page) == MIGRATE_CMA) {
+			if (mm_iommu_move_page_from_cma(page))
+				goto populate;
+			if (1 != get_user_pages_fast(ua + (i << PAGE_SHIFT),
+						1/* pages */, 1/* iswrite */,
+						&page)) {
+				ret = -EFAULT;
+				for (j = 0; j < i; ++j)
+					put_page(pfn_to_page(mem->hpas[j] >>
+								PAGE_SHIFT));
+				vfree(mem->hpas);
+				kfree(mem);
+				goto unlock_exit;
+			}
+		}
+populate:
 		mem->hpas[i] = page_to_pfn(page) << PAGE_SHIFT;
 	}
 

-- 
2.5.5

On 14/07/16 14:25, Balbir Singh wrote:

From: Balbir Singh <bsingharora@gmail.com>
Subject: [RESEND][v2][PATCH] KVM: PPC: Book3S HV: Migrate pinned pages out of CMA

When PCI Device pass-through is enabled via VFIO, KVM-PPC will
pin pages using get_user_pages_fast(). One of the downsides of
the pinning is that the page could be in CMA region. The CMA
region is used for other allocations like the hash page table.
Ideally we want the pinned pages to be from non CMA region.

This patch (currently only for KVM PPC with VFIO) forcefully
migrates the pages out (huge pages are omitted for the moment).
There are more efficient ways of doing this, but that might
be elaborate and might impact a larger audience beyond just
the kvm ppc implementation.

The magic is in new_iommu_non_cma_page() which allocates the
new page from a non CMA region.

I've tested the patches lightly at my end, but there might be bugs
For example if after lru_add_drain(), the page is not isolated
is this a BUG?

Previous discussion was at
http://permalink.gmane.org/gmane.linux.kernel.mm/136738

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Paul Mackerras <redacted>
Cc: Alexey Kardashevskiy <redacted>

Signed-off-by: Balbir Singh <bsingharora@gmail.com>

Acked-by: Alexey Kardashevskiy <redacted>

-- 
Alexey

On 31/08/16 14:14, Alexey Kardashevskiy wrote:

On 14/07/16 14:25, Balbir Singh wrote:

quoted

From: Balbir Singh <bsingharora@gmail.com>
Subject: [RESEND][v2][PATCH] KVM: PPC: Book3S HV: Migrate pinned pages out of CMA

When PCI Device pass-through is enabled via VFIO, KVM-PPC will
pin pages using get_user_pages_fast(). One of the downsides of
the pinning is that the page could be in CMA region. The CMA
region is used for other allocations like the hash page table.
Ideally we want the pinned pages to be from non CMA region.

This patch (currently only for KVM PPC with VFIO) forcefully
migrates the pages out (huge pages are omitted for the moment).
There are more efficient ways of doing this, but that might
be elaborate and might impact a larger audience beyond just
the kvm ppc implementation.

The magic is in new_iommu_non_cma_page() which allocates the
new page from a non CMA region.

I've tested the patches lightly at my end, but there might be bugs
For example if after lru_add_drain(), the page is not isolated
is this a BUG?

Previous discussion was at
http://permalink.gmane.org/gmane.linux.kernel.mm/136738

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Paul Mackerras <redacted>
Cc: Alexey Kardashevskiy <redacted>

Signed-off-by: Balbir Singh <bsingharora@gmail.com>

Acked-by: Alexey Kardashevskiy <redacted>

Thanks! I tested this patch against latest mainline and here are the test results


System RAM - 64GB

VM instance 1 - size 55GB

Before patch - nr_free_cma after launch 8900
After patch - nr_free_cma after launch 39500

Balbir Singh.

Balbir Singh [off-list ref] writes:

Doesn't a PageCompound check cover all ?

+
+	if (PageHighMem(page))
+		gfp_mask |= __GFP_HIGHMEM;
+
+	/*
+	 * We don't want the allocation to force an OOM if possibe
+	 */
+	new_page = alloc_page(gfp_mask | __GFP_NORETRY | __GFP_NOWARN);
+	return new_page;
+}
+
+static int mm_iommu_move_page_from_cma(struct page *page)
+{
+	int ret;
+	LIST_HEAD(cma_migrate_pages);
+
+	/* Ignore huge pages for now */
+	if (PageHuge(page) || PageTransHuge(page) || PageCompound(page))
+		return -EBUSY;
+
+	lru_add_drain();

I guess I asked this last time. Shouldn't this be lru_add_drain_all() ?
What if the page is in other cpu's pagevec ?

+	ret = isolate_lru_page(page);
+	if (ret)
+		get_page(page); /* Potential BUG? */
+
+	list_add(&page->lru, &cma_migrate_pages);

Is that correct ? if we failed the isolate_lru_page(), can we be sure we
are not on lru at all ? ie, what if the page was on other cpu pagevec ?

+	put_page(page); /* Drop the gup reference */
+

Where is get user page (gup) here ? . I guess you mean drop the
reference taken above ?

+	ret = migrate_pages(&cma_migrate_pages, new_iommu_non_cma_page,
+				NULL, 0, MIGRATE_SYNC, MR_CMA);
+	if (ret) {
+		if (!list_empty(&cma_migrate_pages))
+			putback_movable_pages(&cma_migrate_pages);
+	}
+	return 0;
+}
+

I guess the plan was to not do it one page at a time and switch this to list
of pages which we need to migrate. Any reason why that is not tried ?

-aneesh

When PCI Device pass-through is enabled via VFIO, KVM-PPC will
pin pages using get_user_pages_fast(). One of the downsides of
the pinning is that the page could be in CMA region. The CMA
region is used for other allocations like the hash page table.
Ideally we want the pinned pages to be from non CMA region.

This patch (currently only for KVM PPC with VFIO) forcefully
migrates the pages out (huge pages are omitted for the moment).
There are more efficient ways of doing this, but that might
be elaborate and might impact a larger audience beyond just
the kvm ppc implementation.

The magic is in new_iommu_non_cma_page() which allocates the
new page from a non CMA region.

I've tested the patches lightly at my end. The full solution
requires migration of THP pages in the CMA region. That work
will be done incrementally on top of this.

Previous discussion was at
http://permalink.gmane.org/gmane.linux.kernel.mm/136738

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Paul Mackerras <redacted>
Cc: Alexey Kardashevskiy <redacted>

Signed-off-by: Balbir Singh <bsingharora@gmail.com>
Acked-by: Alexey Kardashevskiy <redacted>
---
 arch/powerpc/include/asm/mmu_context.h |  1 +
 arch/powerpc/mm/mmu_context_iommu.c    | 81 ++++++++++++++++++++++++++++++++--
 2 files changed, 78 insertions(+), 4 deletions(-)

diff --git a/arch/powerpc/include/asm/mmu_context.h b/arch/powerpc/include/asm/mmu_context.h
index 9d2cd0c..475d1be 100644
--- a/arch/powerpc/include/asm/mmu_context.h
+++ b/arch/powerpc/include/asm/mmu_context.h

@@ -18,6 +18,7 @@ extern void destroy_context(struct mm_struct *mm);
 #ifdef CONFIG_SPAPR_TCE_IOMMU
 struct mm_iommu_table_group_mem_t;
 
+extern int isolate_lru_page(struct page *page);	/* from internal.h */
 extern bool mm_iommu_preregistered(void);
 extern long mm_iommu_get(unsigned long ua, unsigned long entries,
 		struct mm_iommu_table_group_mem_t **pmem);

diff --git a/arch/powerpc/mm/mmu_context_iommu.c b/arch/powerpc/mm/mmu_context_iommu.c
index da6a216..e0f1c33 100644
--- a/arch/powerpc/mm/mmu_context_iommu.c
+++ b/arch/powerpc/mm/mmu_context_iommu.c

@@ -15,6 +15,9 @@
 #include <linux/rculist.h>
 #include <linux/vmalloc.h>
 #include <linux/mutex.h>
+#include <linux/migrate.h>
+#include <linux/hugetlb.h>
+#include <linux/swap.h>
 #include <asm/mmu_context.h>
 
 static DEFINE_MUTEX(mem_list_mutex);

@@ -72,6 +75,55 @@ bool mm_iommu_preregistered(void)
 }
 EXPORT_SYMBOL_GPL(mm_iommu_preregistered);
 
+/*
+ * Taken from alloc_migrate_target with changes to remove CMA allocations
+ */
+struct page *new_iommu_non_cma_page(struct page *page, unsigned long private,
+					int **resultp)
+{
+	gfp_t gfp_mask = GFP_USER;
+	struct page *new_page;
+
+	if (PageHuge(page) || PageTransHuge(page) || PageCompound(page))
+		return NULL;
+
+	if (PageHighMem(page))
+		gfp_mask |= __GFP_HIGHMEM;
+
+	/*
+	 * We don't want the allocation to force an OOM if possibe
+	 */
+	new_page = alloc_page(gfp_mask | __GFP_NORETRY | __GFP_NOWARN);
+	return new_page;
+}
+
+static int mm_iommu_move_page_from_cma(struct page *page)
+{
+	int ret = 0;
+	LIST_HEAD(cma_migrate_pages);
+
+	/* Ignore huge pages for now */
+	if (PageHuge(page) || PageTransHuge(page) || PageCompound(page))
+		return -EBUSY;
+
+	lru_add_drain();
+	ret = isolate_lru_page(page);
+	if (ret)
+		return ret;
+
+	list_add(&page->lru, &cma_migrate_pages);
+	put_page(page); /* Drop the gup reference */
+
+	ret = migrate_pages(&cma_migrate_pages, new_iommu_non_cma_page,
+				NULL, 0, MIGRATE_SYNC, MR_CMA);
+	if (ret) {
+		if (!list_empty(&cma_migrate_pages))
+			putback_movable_pages(&cma_migrate_pages);
+	}
+
+	return 0;
+}
+
 long mm_iommu_get(unsigned long ua, unsigned long entries,
 		struct mm_iommu_table_group_mem_t **pmem)
 {

@@ -124,15 +176,36 @@ long mm_iommu_get(unsigned long ua, unsigned long entries,
 	for (i = 0; i < entries; ++i) {
 		if (1 != get_user_pages_fast(ua + (i << PAGE_SHIFT),
 					1/* pages */, 1/* iswrite */, &page)) {
+			ret = -EFAULT;
 			for (j = 0; j < i; ++j)
-				put_page(pfn_to_page(
-						mem->hpas[j] >> PAGE_SHIFT));
+				put_page(pfn_to_page(mem->hpas[j] >>
+						PAGE_SHIFT));
 			vfree(mem->hpas);
 			kfree(mem);
-			ret = -EFAULT;
 			goto unlock_exit;
 		}
-
+		/*
+		 * If we get a page from the CMA zone, since we are going to
+		 * be pinning these entries, we might as well move them out
+		 * of the CMA zone if possible. NOTE: faulting in + migration
+		 * can be expensive. Batching can be considered later
+		 */
+		if (get_pageblock_migratetype(page) == MIGRATE_CMA) {
+			if (mm_iommu_move_page_from_cma(page))
+				goto populate;
+			if (1 != get_user_pages_fast(ua + (i << PAGE_SHIFT),
+						1/* pages */, 1/* iswrite */,
+						&page)) {
+				ret = -EFAULT;
+				for (j = 0; j < i; ++j)
+					put_page(pfn_to_page(mem->hpas[j] >>
+								PAGE_SHIFT));
+				vfree(mem->hpas);
+				kfree(mem);
+				goto unlock_exit;
+			}
+		}
+populate:
 		mem->hpas[i] = page_to_pfn(page) << PAGE_SHIFT;
 	}
 

-- 
2.5.5

On 06/09/16 15:49, Aneesh Kumar K.V wrote:

Balbir Singh [off-list ref] writes:

quoted

From: Balbir Singh <bsingharora@gmail.com>
Subject: [RESEND][v2][PATCH] KVM: PPC: Book3S HV: Migrate pinned pages out of CMA

When PCI Device pass-through is enabled via VFIO, KVM-PPC will
pin pages using get_user_pages_fast(). One of the downsides of
the pinning is that the page could be in CMA region. The CMA
region is used for other allocations like the hash page table.
Ideally we want the pinned pages to be from non CMA region.

This patch (currently only for KVM PPC with VFIO) forcefully
migrates the pages out (huge pages are omitted for the moment).
There are more efficient ways of doing this, but that might
be elaborate and might impact a larger audience beyond just
the kvm ppc implementation.

The magic is in new_iommu_non_cma_page() which allocates the
new page from a non CMA region.

I've tested the patches lightly at my end, but there might be bugs
For example if after lru_add_drain(), the page is not isolated
is this a BUG?

Previous discussion was at
http://permalink.gmane.org/gmane.linux.kernel.mm/136738

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Paul Mackerras <redacted>
Cc: Alexey Kardashevskiy <redacted>

Signed-off-by: Balbir Singh <bsingharora@gmail.com>
---
 arch/powerpc/include/asm/mmu_context.h |  1 +
 arch/powerpc/mm/mmu_context_iommu.c    | 80 ++++++++++++++++++++++++++++++++--
 2 files changed, 77 insertions(+), 4 deletions(-)

diff --git a/arch/powerpc/include/asm/mmu_context.h b/arch/powerpc/include/asm/mmu_context.h
index 9d2cd0c..475d1be 100644
--- a/arch/powerpc/include/asm/mmu_context.h
+++ b/arch/powerpc/include/asm/mmu_context.h

@@ -18,6 +18,7 @@ extern void destroy_context(struct mm_struct *mm);
 #ifdef CONFIG_SPAPR_TCE_IOMMU
 struct mm_iommu_table_group_mem_t;
 
+extern int isolate_lru_page(struct page *page);	/* from internal.h */
 extern bool mm_iommu_preregistered(void);
 extern long mm_iommu_get(unsigned long ua, unsigned long entries,
 		struct mm_iommu_table_group_mem_t **pmem);

diff --git a/arch/powerpc/mm/mmu_context_iommu.c b/arch/powerpc/mm/mmu_context_iommu.c
index da6a216..c18f742 100644
--- a/arch/powerpc/mm/mmu_context_iommu.c
+++ b/arch/powerpc/mm/mmu_context_iommu.c

@@ -15,6 +15,9 @@
 #include <linux/rculist.h>
 #include <linux/vmalloc.h>
 #include <linux/mutex.h>
+#include <linux/migrate.h>
+#include <linux/hugetlb.h>
+#include <linux/swap.h>
 #include <asm/mmu_context.h>
 
 static DEFINE_MUTEX(mem_list_mutex);

@@ -72,6 +75,54 @@ bool mm_iommu_preregistered(void)
 }
 EXPORT_SYMBOL_GPL(mm_iommu_preregistered);
 
+/*
+ * Taken from alloc_migrate_target with changes to remove CMA allocations
+ */
+struct page *new_iommu_non_cma_page(struct page *page, unsigned long private,
+					int **resultp)
+{
+	gfp_t gfp_mask = GFP_USER;
+	struct page *new_page;
+
+	if (PageHuge(page) || PageTransHuge(page) || PageCompound(page))
+		return NULL;

Doesn't a PageCompound check cover all ?

Yes, I was being overly conservative with the checks

quoted

+
+	if (PageHighMem(page))
+		gfp_mask |= __GFP_HIGHMEM;
+
+	/*
+	 * We don't want the allocation to force an OOM if possibe
+	 */
+	new_page = alloc_page(gfp_mask | __GFP_NORETRY | __GFP_NOWARN);
+	return new_page;
+}
+
+static int mm_iommu_move_page_from_cma(struct page *page)
+{
+	int ret;
+	LIST_HEAD(cma_migrate_pages);
+
+	/* Ignore huge pages for now */
+	if (PageHuge(page) || PageTransHuge(page) || PageCompound(page))
+		return -EBUSY;
+
+	lru_add_drain();

I guess I asked this last time. Shouldn't this be lru_add_drain_all() ?
What if the page is in other cpu's pagevec ?

lru_add_drain_all() is too expensive for a per-page migration. This is best
effort. If it is on the pagevec of another CPU, we skip it -- see v3

quoted

+	ret = isolate_lru_page(page);
+	if (ret)
+		get_page(page); /* Potential BUG? */
+
+	list_add(&page->lru, &cma_migrate_pages);

Is that correct ? if we failed the isolate_lru_page(), can we be sure we
are not on lru at all ? ie, what if the page was on other cpu pagevec ?

Fixed in v3

quoted

+	put_page(page); /* Drop the gup reference */
+

Where is get user page (gup) here ? . I guess you mean drop the
reference taken above ?

I say gup, because we'll do gup after this point if migration fails
and that we reacquire the reference lost here.

quoted

+	ret = migrate_pages(&cma_migrate_pages, new_iommu_non_cma_page,
+				NULL, 0, MIGRATE_SYNC, MR_CMA);
+	if (ret) {
+		if (!list_empty(&cma_migrate_pages))
+			putback_movable_pages(&cma_migrate_pages);
+	}
+	return 0;
+}
+

I guess the plan was to not do it one page at a time and switch this to list
of pages which we need to migrate. Any reason why that is not tried ?

Yes, it is a TODO. Here is my order of preference

1. get this in
2. Get THP migration in -- larger workset
3. Do page aggregation for both 1 and 2


Balbir Singh.

On 09/06/2016 11:57 AM, Balbir Singh wrote:

Small nit, cant we just add "mm/internal.h" header here with full path ?

So what guarantees that the new page too wont come from MIGRATE_CMA
page block ? Is absence of __GFP_MOVABLE flag enough. Also should not
we be checking that migrate type of the new allocated page is indeed
not MIGRATE_CMA ?

On 06/09/16 21:54, Anshuman Khandual wrote:

On 09/06/2016 11:57 AM, Balbir Singh wrote:

quoted

When PCI Device pass-through is enabled via VFIO, KVM-PPC will
pin pages using get_user_pages_fast(). One of the downsides of
the pinning is that the page could be in CMA region. The CMA
region is used for other allocations like the hash page table.
Ideally we want the pinned pages to be from non CMA region.

This patch (currently only for KVM PPC with VFIO) forcefully
migrates the pages out (huge pages are omitted for the moment).
There are more efficient ways of doing this, but that might
be elaborate and might impact a larger audience beyond just
the kvm ppc implementation.

The magic is in new_iommu_non_cma_page() which allocates the
new page from a non CMA region.

I've tested the patches lightly at my end. The full solution
requires migration of THP pages in the CMA region. That work
will be done incrementally on top of this.

Previous discussion was at
http://permalink.gmane.org/gmane.linux.kernel.mm/136738

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Paul Mackerras <redacted>
Cc: Alexey Kardashevskiy <redacted>

Signed-off-by: Balbir Singh <bsingharora@gmail.com>
Acked-by: Alexey Kardashevskiy <redacted>
---
 arch/powerpc/include/asm/mmu_context.h |  1 +
 arch/powerpc/mm/mmu_context_iommu.c    | 81 ++++++++++++++++++++++++++++++++--
 2 files changed, 78 insertions(+), 4 deletions(-)

diff --git a/arch/powerpc/include/asm/mmu_context.h b/arch/powerpc/include/asm/mmu_context.h
index 9d2cd0c..475d1be 100644
--- a/arch/powerpc/include/asm/mmu_context.h
+++ b/arch/powerpc/include/asm/mmu_context.h

@@ -18,6 +18,7 @@ extern void destroy_context(struct mm_struct *mm);
 #ifdef CONFIG_SPAPR_TCE_IOMMU
 struct mm_iommu_table_group_mem_t;

+extern int isolate_lru_page(struct page *page);	/* from internal.h */

Small nit, cant we just add "mm/internal.h" header here with full path ?

I did not think it was worth including it here

quoted

 extern bool mm_iommu_preregistered(void);
 extern long mm_iommu_get(unsigned long ua, unsigned long entries,
 		struct mm_iommu_table_group_mem_t **pmem);

diff --git a/arch/powerpc/mm/mmu_context_iommu.c b/arch/powerpc/mm/mmu_context_iommu.c
index da6a216..e0f1c33 100644
--- a/arch/powerpc/mm/mmu_context_iommu.c
+++ b/arch/powerpc/mm/mmu_context_iommu.c

@@ -15,6 +15,9 @@
 #include <linux/rculist.h>
 #include <linux/vmalloc.h>
 #include <linux/mutex.h>
+#include <linux/migrate.h>
+#include <linux/hugetlb.h>
+#include <linux/swap.h>
 #include <asm/mmu_context.h>

 static DEFINE_MUTEX(mem_list_mutex);

@@ -72,6 +75,55 @@ bool mm_iommu_preregistered(void)
 }
 EXPORT_SYMBOL_GPL(mm_iommu_preregistered);

+/*
+ * Taken from alloc_migrate_target with changes to remove CMA allocations
+ */
+struct page *new_iommu_non_cma_page(struct page *page, unsigned long private,
+					int **resultp)
+{
+	gfp_t gfp_mask = GFP_USER;
+	struct page *new_page;
+
+	if (PageHuge(page) || PageTransHuge(page) || PageCompound(page))
+		return NULL;
+
+	if (PageHighMem(page))
+		gfp_mask |= __GFP_HIGHMEM;
+
+	/*
+	 * We don't want the allocation to force an OOM if possibe
+	 */
+	new_page = alloc_page(gfp_mask | __GFP_NORETRY | __GFP_NOWARN);

So what guarantees that the new page too wont come from MIGRATE_CMA
page block ? Is absence of __GFP_MOVABLE flag enough.

I think so, that is what I am relying on and checked for

 Also should not

we be checking that migrate type of the new allocated page is indeed
not MIGRATE_CMA ?

I don't think that is required, may be I can do a VM_WARN_ON for debugging

Balbir Singh.

On Tue, 2016-06-09 at 06:27:31 UTC, Balbir Singh wrote:

When PCI Device pass-through is enabled via VFIO, KVM-PPC will
pin pages using get_user_pages_fast(). One of the downsides of
the pinning is that the page could be in CMA region. The CMA
region is used for other allocations like the hash page table.
Ideally we want the pinned pages to be from non CMA region.

This patch (currently only for KVM PPC with VFIO) forcefully
migrates the pages out (huge pages are omitted for the moment).
There are more efficient ways of doing this, but that might
be elaborate and might impact a larger audience beyond just
the kvm ppc implementation.

The magic is in new_iommu_non_cma_page() which allocates the
new page from a non CMA region.

I've tested the patches lightly at my end. The full solution
requires migration of THP pages in the CMA region. That work
will be done incrementally on top of this.

Signed-off-by: Balbir Singh <bsingharora@gmail.com>
Acked-by: Alexey Kardashevskiy <redacted>

Applied to powerpc next, thanks.

https://git.kernel.org/powerpc/c/2e5bbb5461f138cac631fe21b4ad95

cheers