kvm: calculate correct gfn for small host pages which emulates large guest pages

RFC PATCH Applied Lai Jiangshan 2 participants 4 messages May 7, 2010 – May 26, 2010 2 messages not in the archive

Subsystems (3)

x86 KVM KVM x86

lore

Share & tools

lore

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

Lai Jiangshan

author

2010-05-07 08:52 UTC #2 lore raw

Quoted 41 lines · Lai Jiangshan

Quoted 9 lines · Lai Jiangshan
RFC, because maybe I missing something with the old code.
Frome: Lai Jiangshan <[email protected]>
In Document/kvm/mmu.txt:
gfn:
Either the guest page table containing the translations shadowed by this page, or the base page frame for linear translations. See role.direct.
Quoted 8 lines
But in function FNAME(fetch)(), sp->gfn is incorrect when one of following situations occurred:
1) guest is 32bit paging and guest uses pse-36 and the guest PDE maps a 4-MByte page(backed by 4k host pages) and bits 20:13 of the guest PDE is not equals to 0.
2) guest is long mode paging and the guest PDPTE maps a 1-GByte page (backed by 4k or 2M host pages)
Resend this patch with the changelog changed.
As Marcelo Tosatti and Gui Jianfeng points out, FNAME(fetch)() miss quadrant on 4mb large page emulation with shadow.
Subject: [PATCH] kvm: calculate correct gfn for small host pages which emulates large guest pages
In Document/kvm/mmu.txt:
gfn:
Either the guest page table containing the translations shadowed by this page, or the base page frame for linear translations. See role.direct.
But in function FNAME(fetch)(), sp->gfn is incorrect when one of following situations occurred:
1) guest is 32bit paging and the guest PDE maps a 4-MByte page (backed by 4k host pages), FNAME(fetch)() miss handling the quadrant.
And if guest use pse-36, "table_gfn = gpte_to_gfn(gw->ptes[level - delta]);" is incorrect.
2) guest is long mode paging and the guest PDPTE maps a 1-GByte page (backed by 4k or 2M host pages).

Quoted 8 lines

So we fix it to suit to the document and suit to the code which requires sp->gfn correct when sp->role.direct=1.
We use the goal mapping gfn(gw->gfn) to calculate the base page frame for linear translations, it is simple and easy to be understood.
Signed-off-by Lai Jiangshan <[email protected]>
---

Could you add these:

Reported-by Marcelo Tosatti <[email protected]>

Reported-by Gui Jianfeng <[email protected]>

Thanks.
Lai.

PS.
The whole patches includes:
[PATCH] kvm mmu: reduce 50% memory usage
[PATCH] kvm: calculate correct gfn for small host pages which emulates large guest pages [PATCH] kvm, tdp: calculate correct base gfn for non-DIR level
[PATCH] kvm: update document of gfns

Avi Kivity

2010-05-10 08:55 UTC #3 lore raw

In reply to a message not in the archive

Quoted 63 lines · Lai Jiangshan

Quoted 9 lines · Lai Jiangshan
RFC, because maybe I missing something with the old code.
Frome: Lai Jiangshan<[email protected]>
In Document/kvm/mmu.txt:
gfn:
Either the guest page table containing the translations shadowed by this page, or the base page frame for linear translations. See role.direct.
Quoted 9 lines
But in function FNAME(fetch)(), sp->gfn is incorrect when one of following situations occurred:
1) guest is 32bit paging and guest uses pse-36 and the guest PDE maps a 4-MByte page(backed by 4k host pages) and bits 20:13 of the guest PDE is not equals to 0.
2) guest is long mode paging and the guest PDPTE maps a 1-GByte page (backed by 4k or 2M host pages)
Resend this patch with the changelog changed.
As Marcelo Tosatti and Gui Jianfeng points out, FNAME(fetch)() miss quadrant on 4mb large page emulation with shadow.
Subject: [PATCH] kvm: calculate correct gfn for small host pages which emulates large guest pages
In Document/kvm/mmu.txt:
gfn:
Either the guest page table containing the translations shadowed by this page, or the base page frame for linear translations. See role.direct.
But in function FNAME(fetch)(), sp->gfn is incorrect when one of following situations occurred:
1) guest is 32bit paging and the guest PDE maps a 4-MByte page (backed by 4k host pages), FNAME(fetch)() miss handling the quadrant.
And if guest use pse-36, "table_gfn = gpte_to_gfn(gw->ptes[level - delta]);" is incorrect.
2) guest is long mode paging and the guest PDPTE maps a 1-GByte page (backed by 4k or 2M host pages).
So we fix it to suit to the document and suit to the code which requires sp->gfn correct when sp->role.direct=1.
We use the goal mapping gfn(gw->gfn) to calculate the base page frame for linear translations, it is simple and easy to be understood.
Signed-off-by Lai Jiangshan<[email protected]>
---
arch/x86/kvm/paging_tmpl.h +0 -4
index 702c016..958e9c6 100644
@@ -338,10 +338,13 @@ static u64 *FNAME(fetch)(struct kvm_vcpu *vcpu, gva_t addr,
  			direct = 1;
  			if (!is_dirty_gpte(gw->ptes[level - delta]))
  				access&= ~ACC_WRITE_MASK;
-			table_gfn = gpte_to_gfn(gw->ptes[level - delta]);
-			/* advance table_gfn when emulating 1gb pages with 4k */
-			if (delta == 0)
-				table_gfn += PT_INDEX(addr, level);

arch/x86/kvm/paging_tmpl.h static u64 *FNAME(fetch)(struct kvm_vcpu *vcpu, gva_t addr, · line 350

Quoted 10 lines

arch/x86/kvm/paging_tmpl.h +7 -0

+			/*
+			 * It is a large guest pages backed by small host pages,
+			 * So we set @direct(@shadow_page->role.direct)=1, and
+			 * set @table_gfn(@shadow_page->gfn)=the base page frame
+			 * for linear translations.
+			 */
+			table_gfn = gw->gfn&  ~(KVM_PAGES_PER_HPAGE(level) - 1);
  		} else {
  			direct = 0;
  			table_gfn = gw->table_gfn[level - 2];

Looks good, indeed it is a lot easier to understand than the original calculation (a minor issue is that the variable name is misleading, but that's a problem with kvm_mmu_page definition and not this patch).

Signature · 1 line

error compiling committee.c: too many arguments to function

Lai Jiangshan

author

2010-05-26 08:48 UTC #4 lore raw

PATCH RESEND 1/3 kvm: calculate correct gfn for small host pages which emulates large guest pages

In reply to a message not in the archive

In Document/kvm/mmu.txt:
gfn:
Either the guest page table containing the translations shadowed by this page, or the base page frame for linear translations. See role.direct.

But in function FNAME(fetch)(), sp->gfn is incorrect when one of following situations occurred:
1) guest is 32bit paging and the guest PDE maps a 4-MByte page (backed by 4k host pages), FNAME(fetch)() miss handling the quadrant.

And if guest use pse-36, "table_gfn = gpte_to_gfn(gw->ptes[level - delta]);" is incorrect.
2) guest is long mode paging and the guest PDPTE maps a 1-GByte page (backed by 4k or 2M host pages).

So we fix it to suit to the document and suit to the code which requires sp->gfn correct when sp->role.direct=1.

We use the goal mapping gfn(gw->gfn) to calculate the base page frame for linear translations, it is simple and easy to be understood.

Reported-by Marcelo Tosatti <[email protected]>

Reported-by Gui Jianfeng <[email protected]>

Signed-off-by Lai Jiangshan <[email protected]>

---

arch/x86/kvm/paging_tmpl.h +7 -4

index 702c016..958e9c6 100644
@@ -338,10 +338,13 @@ static u64 *FNAME(fetch)(struct kvm_vcpu *vcpu, gva_t addr,
 			direct = 1;
 			if (!is_dirty_gpte(gw->ptes[level - delta]))
 				access &= ~ACC_WRITE_MASK;
-			table_gfn = gpte_to_gfn(gw->ptes[level - delta]);
-			/* advance table_gfn when emulating 1gb pages with 4k */
-			if (delta == 0)
-				table_gfn += PT_INDEX(addr, level);
+			/*
+			 * It is a large guest pages backed by small host pages,
+			 * So we set @direct(@shadow_page->role.direct)=1, and
+			 * set @table_gfn(@shadow_page->gfn)=the base page frame
+			 * for linear translations.
+			 */
+			table_gfn = gw->gfn & ~(KVM_PAGES_PER_HPAGE(level) - 1);
 		} else {
 			direct = 0;
 			table_gfn = gw->table_gfn[level - 2];

Avi Kivity

Applied

2010-05-26 11:23 UTC #5 lore raw

Quoted 15 lines · Lai Jiangshan

In Document/kvm/mmu.txt:
gfn:
Either the guest page table containing the translations shadowed by this page, or the base page frame for linear translations. See role.direct.
But in function FNAME(fetch)(), sp->gfn is incorrect when one of following situations occurred:
1) guest is 32bit paging and the guest PDE maps a 4-MByte page (backed by 4k host pages), FNAME(fetch)() miss handling the quadrant.
And if guest use pse-36, "table_gfn = gpte_to_gfn(gw->ptes[level - delta]);" is incorrect.
2) guest is long mode paging and the guest PDPTE maps a 1-GByte page (backed by 4k or 2M host pages).

Quoted 6 lines

So we fix it to suit to the document and suit to the code which requires sp->gfn correct when sp->role.direct=1.
We use the goal mapping gfn(gw->gfn) to calculate the base page frame for linear translations, it is simple and easy to be understood.

Applied, thanks.

Signature · 1 line

error compiling committee.c: too many arguments to function