mmu:remove the usage of the private field by the rmap.

RFC PATCH In review Izik Eidus 4 participants 6 messages Sep 24, 2007 – Sep 30, 2007 6 messages not in the archive KVM

lore

Share & tools

lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

Izik Eidus

author

2007-09-24 18:59 UTC #1 lore raw

RFC PATCH 1/1 mmu:remove the usage of the private field by the rmap.

in order to be able to allocate the memory to kvm from userspace we have to remove
the private usage made by the rmap.

this patch add array to each memory slot sorted by gfn that hold rmap to each page.
to make the finding of gfn from spte address, we are adding another page to each kvm_mmu_page, that hold
to each spte - spt offset the gfn of it.

thanks

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/

_______________________________________________
kvm-devel mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/kvm-devel

A message that is not in the archive (sent off-list or not yet received): 64F9B87B6B770947A9F8391472E032160CBECFB7-yEcIvxbTEBqsx+V+t5oei8rau4O3wl8o3fe8/T/[email protected]

Anthony Liguori

2007-09-24 19:34 UTC #3 lore raw

I think you forgot to include the patch?

Regards,

Anthony Liguori

Quoted 22 lines · Izik Eidus

in order to be able to allocate the memory to kvm from userspace we have to remove
the private usage made by the rmap.
this patch add array to each memory slot sorted by gfn that hold rmap to each page.
to make the finding of gfn from spte address, we are adding another page to each kvm_mmu_page, that hold
to each spte - spt offset the gfn of it.
thanks
------------------------------------------------------------------------
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
------------------------------------------------------------------------

Quoted 5 lines

_______________________________________________
kvm-devel mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/kvm-devel

A message that is not in the archive (sent off-list or not yet received): [email protected]

Izik Eidus

2007-09-25 07:24 UTC #5 lore raw

Quoted 7 lines · Anthony Liguori

I think you forgot to include the patch?
Regards,
Anthony Liguori

yes i forgot :)
here it is...

commit 3ed7059496e6e7d802dfb46e604d0af4abc2b040

Author: Izik Eidus <izike-atKUWr5tajBWk0Htik3J/[email protected]>
Date:   Mon Sep 24 13:35:12 2007 +0200

Remove the usage of the private field by the rmap.
this patch add to each memory slot an array sorted by gfn that hold to each page its reverse mapping,

Signed-off-by Izik Eidus <izike-atKUWr5tajBWk0Htik3J/[email protected]>

drivers/kvm/kvm.h +5 -1

index b7cd276..e403441 100644
@@ -126,6 +126,8 @@ struct kvm_mmu_page {
 	union kvm_mmu_page_role role;
 
 	u64 *spt;
+	/* hold the gfn of each spte inside spt */
+	gfn_t *gfns;
 	unsigned long slot_bitmap; /* One bit set per slot which has memory
 				    * in this shadow page.
 				    */

@@ -157,7 +159,7 @@ struct kvm_mmu {
 	u64 *pae_root;
 };
 
-#define KVM_NR_MEM_OBJS 20
+#define KVM_NR_MEM_OBJS 40
 
 struct kvm_mmu_memory_cache {
 	int nobjs;

@@ -399,6 +401,7 @@ struct kvm_memory_slot {
 	unsigned long npages;
 	unsigned long flags;
 	struct page **phys_mem;
+	unsigned long *rmap;
 	unsigned long *dirty_bitmap;
 };

@@ -550,6 +553,7 @@ struct page *gva_to_page(struct kvm_vcpu *vcpu, gva_t gva);
 
 extern hpa_t bad_page_address;
 
+gfn_t unalias_gfn(struct kvm *kvm, gfn_t gfn);
 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);

drivers/kvm/kvm_main.c +9 -2

index bc2d32d..366e769 100644
@@ -329,6 +329,8 @@ static void kvm_free_physmem_slot(struct kvm_memory_slot *free,
 					__free_page(free->phys_mem[i]);
 			vfree(free->phys_mem);
 		}
+	if (!dont || free->rmap != dont->rmap)
+		vfree(free->rmap);
 
 	if (!dont || free->dirty_bitmap != dont->dirty_bitmap)
 		vfree(free->dirty_bitmap);

@@ -742,13 +744,18 @@ static int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
 		if (!new.phys_mem)
 			goto out_unlock;
 
+		new.rmap = vmalloc(npages * sizeof(struct page*));
+
+		if (!new.rmap)
+			goto out_unlock;
+
 		memset(new.phys_mem, 0, npages * sizeof(struct page *));
+		memset(new.rmap, 0, npages * sizeof(*new.rmap));
 		for (i = 0; i < npages; ++i) {
 			new.phys_mem[i] = alloc_page(GFP_HIGHUSER
 						     | __GFP_ZERO);
 			if (!new.phys_mem[i])
 				goto out_unlock;
-			set_page_private(new.phys_mem[i],0);
 		}
 	}

@@ -932,7 +939,7 @@ static int kvm_vm_ioctl_set_irqchip(struct kvm *kvm, struct kvm_irqchip *chip)
 	return r;
 }
 
-static gfn_t unalias_gfn(struct kvm *kvm, gfn_t gfn)
+gfn_t unalias_gfn(struct kvm *kvm, gfn_t gfn)
 {
 	int i;
 	struct kvm_mem_alias *alias;

drivers/kvm/mmu.c +75 ~~-52~~

269 lines not on this page · raw message

drivers/kvm/paging_tmpl.h +2 -1

index 6b094b4..e07ee3b 100644
@@ -289,7 +289,8 @@ unshadowed:
 	set_shadow_pte(shadow_pte, spte);
 	page_header_update_slot(vcpu->kvm, shadow_pte, gaddr);
 	if (!was_rmapped)
-		rmap_add(vcpu, shadow_pte);
+		rmap_add(vcpu, shadow_pte, (gaddr & PT64_BASE_ADDR_MASK)
+								>> PAGE_SHIFT);
 }
 
 static void FNAME(set_pte)(struct kvm_vcpu *vcpu, pt_element_t gpte,

_______________________________________________
kvm-devel mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/kvm-devel

A message that is not in the archive (sent off-list or not yet received): 46F8B7B2.9020104-atKUWr5tajBWk0Htik3J/[email protected]

Avi Kivity

2007-09-27 10:55 UTC #7 lore raw

Quoted 10 lines · Izik Eidus

Quoted 7 lines · Anthony Liguori
I think you forgot to include the patch?
Regards,
Anthony Liguori
yes i forgot :)
here it is...

Patch doesn't apply. Can you rebase to latest kvm.git?

Also, fix whitespace issues reported by 'git apply'.

Please repost with 'git format-patch' instead of 'git show'.

Signature · 7 lines

Do not meddle in the internals of kernels, for they are subtle and quick to panic.


-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2005.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/

A message that is not in the archive (sent off-list or not yet received): 46FB8C39.2060505-atKUWr5tajBWk0Htik3J/[email protected]

Izik Eidus

2007-09-27 12:22 UTC #9 lore raw

Quoted 17 lines · Avi Kivity

Quoted 4 lines · Izik Eidus
Quoted 2 lines · Anthony Liguori
I think you forgot to include the patch?
Quoted 9 lines
Quoted 6 lines
Regards,
Anthony Liguori
yes i forgot :)
here it is...

Quoted 7 lines

Patch doesn't apply. Can you rebase to latest kvm.git?
Also, fix whitespace issues reported by 'git apply'.
Please repost with 'git format-patch' instead of 'git show'.

avi, is it ok now?

Quoted 1 line

From 974f43b6b550a7b895d8374f43e9ea454d8636e3 Mon Sep 17 00:00:00 2001

From: Izik Eidus <izike-atKUWr5tajBWk0Htik3J/[email protected]>
Date: Thu, 27 Sep 2007 14:11:22 +0200
Subject: [PATCH] Remove the usage of the private field by the rmap. this patch add to each memory slot an array sorted by gfn that hold to each page its reverse mapping.

Signed-off-by Izik Eidus <izik-atKUWr5tajBWk0Htik3J/[email protected]>

---
 drivers/kvm/kvm.h         |    6 ++-
 drivers/kvm/kvm_main.c    |   11 +++-
 drivers/kvm/mmu.c         |  122 ++++++++++++++++++++++++++-------------------
 drivers/kvm/paging_tmpl.h |    3 +-
 4 files changed, 86 insertions(+), 56 deletions(-)

drivers/kvm/kvm.h +5 -1

index 051cdbe..8e58f3b 100644
@@ -126,6 +126,8 @@ struct kvm_mmu_page {
 	union kvm_mmu_page_role role;
 
 	u64 *spt;
+	/* hold the gfn of each spte inside spt */
+	gfn_t *gfns;
 	unsigned long slot_bitmap; /* One bit set per slot which has memory
 				    * in this shadow page.
 				    */

@@ -159,7 +161,7 @@ struct kvm_mmu {
 	u64 *pae_root;
 };
 
-#define KVM_NR_MEM_OBJS 20
+#define KVM_NR_MEM_OBJS 40
 
 struct kvm_mmu_memory_cache {
 	int nobjs;

@@ -402,6 +404,7 @@ struct kvm_memory_slot {
 	unsigned long npages;
 	unsigned long flags;
 	struct page **phys_mem;
+	unsigned long *rmap;
 	unsigned long *dirty_bitmap;
 };

@@ -554,6 +557,7 @@ struct page *gva_to_page(struct kvm_vcpu *vcpu, gva_t gva);
 
 extern hpa_t bad_page_address;
 
+gfn_t unalias_gfn(struct kvm *kvm, gfn_t gfn);
 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);

drivers/kvm/kvm_main.c +9 -2

index 60798e3..2c3986f 100644
@@ -329,6 +329,8 @@ static void kvm_free_physmem_slot(struct kvm_memory_slot *free,
 					__free_page(free->phys_mem[i]);
 			vfree(free->phys_mem);
 		}
+	if (!dont || free->rmap != dont->rmap)
+		vfree(free->rmap);
 
 	if (!dont || free->dirty_bitmap != dont->dirty_bitmap)
 		vfree(free->dirty_bitmap);

@@ -739,13 +741,18 @@ static int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
 		if (!new.phys_mem)
 			goto out_unlock;
 
+		new.rmap = vmalloc(npages * sizeof(struct page*));
+
+		if (!new.rmap)
+			goto out_unlock;
+
 		memset(new.phys_mem, 0, npages * sizeof(struct page *));
+		memset(new.rmap, 0, npages * sizeof(*new.rmap));
 		for (i = 0; i < npages; ++i) {
 			new.phys_mem[i] = alloc_page(GFP_HIGHUSER
 						     | __GFP_ZERO);
 			if (!new.phys_mem[i])
 				goto out_unlock;
-			set_page_private(new.phys_mem[i],0);
 		}
 	}

@@ -929,7 +936,7 @@ static int kvm_vm_ioctl_set_irqchip(struct kvm *kvm, struct kvm_irqchip *chip)
 	return r;
 }
 
-static gfn_t unalias_gfn(struct kvm *kvm, gfn_t gfn)
+gfn_t unalias_gfn(struct kvm *kvm, gfn_t gfn)
 {
 	int i;
 	struct kvm_mem_alias *alias;

drivers/kvm/mmu.c +70 ~~-52~~

264 lines not on this page · raw message

drivers/kvm/paging_tmpl.h +2 -1

index be0f852..fbe595f 100644
@@ -295,7 +295,8 @@ unshadowed:
 	set_shadow_pte(shadow_pte, spte);
 	page_header_update_slot(vcpu->kvm, shadow_pte, gaddr);
 	if (!was_rmapped)
-		rmap_add(vcpu, shadow_pte);
+		rmap_add(vcpu, shadow_pte, (gaddr & PT64_BASE_ADDR_MASK)
+			 >> PAGE_SHIFT);
 	if (!ptwrite || !*ptwrite)
 		vcpu->last_pte_updated = shadow_pte;
 }

Signature · 1 line

1.5.2.4

_______________________________________________
kvm-devel mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/kvm-devel

A message that is not in the archive (sent off-list or not yet received): 46FBA07F.8030402-atKUWr5tajBWk0Htik3J/[email protected]

Avi Kivity

2007-09-30 09:37 UTC #11 lore raw

Quoted 17 lines · Izik Eidus

Quoted 16 lines · Avi Kivity
Quoted 12 lines · Izik Eidus
Quoted 7 lines · Anthony Liguori
I think you forgot to include the patch?
Regards,
Anthony Liguori
yes i forgot :)
here it is...

Quoted 9 lines

Quoted 7 lines
Patch doesn't apply. Can you rebase to latest kvm.git?
Also, fix whitespace issues reported by 'git apply'.
Please repost with 'git format-patch' instead of 'git show'.
avi, is it ok now?

Sure; applied.

Signature · 7 lines

error compiling committee.c: too many arguments to function


-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2005.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/