[PATCH] KVM: VMX: Fix repeatly allocate of apic access page

[PATCH] KVM: VMX: Fix repeatly allocate of apic access page

[Sheng Yang]

[-- Attachment #1: Type: text/plain, Size: 1526 bytes --]

From 05164d73994a88ae130a086b592cd127c15a617f Mon Sep 17 00:00:00 2001
From: Sheng Yang <sheng.yang-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Date: Tue, 6 Nov 2007 11:37:44 +0800
Subject: [PATCH] KVM: VMX: Fix repeatly allocate of apic access page

For SMP guest, alloc_apic_access_page() would be called more than once.
So only the last one works, causing SMP guest can't benifit from FlexPriority.

This patch fixed this issue.

Signed-off-by: Sheng Yang <sheng.yang-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 drivers/kvm/vmx.c |   14 ++++++++++----
 1 files changed, 10 insertions(+), 4 deletions(-)

diff --git a/drivers/kvm/vmx.c b/drivers/kvm/vmx.c
index 42e7fad..6670e2b 100644
--- a/drivers/kvm/vmx.c
+++ b/drivers/kvm/vmx.c
@@ -1463,18 +1463,24 @@ static void seg_setup(int seg)
 static int alloc_apic_access_page(struct kvm *kvm)
 {
 	struct kvm_userspace_memory_region kvm_userspace_mem;
-	int r;
+	int r = 0;
 
-	r = -EFAULT;
+	mutex_lock(&kvm->lock);
+	if (kvm->apic_access_page)
+		goto out;
+	mutex_unlock(&kvm->lock);
 	kvm_userspace_mem.slot = APIC_ACCESS_PAGE_PRIVATE_MEMSLOT;
 	kvm_userspace_mem.flags = 0;
 	kvm_userspace_mem.guest_phys_addr = 0xfee00000ULL;
 	kvm_userspace_mem.memory_size = PAGE_SIZE;
 	r = kvm_set_memory_region(kvm, &kvm_userspace_mem, 0);
 	if (r)
-		return r;
+		goto out;
+	mutex_lock(&kvm->lock);
 	kvm->apic_access_page = gfn_to_page(kvm, 0xfee00);
-	return 0;
+out:
+	mutex_unlock(&kvm->lock);
+	return r;
 }
 
 /*
-- 
1.5.2


[-- Attachment #2: 0001-KVM-VMX-Fix-repeatly-allocate-of-apic-access-page.patch --]
[-- Type: text/x-diff, Size: 1526 bytes --]

From 05164d73994a88ae130a086b592cd127c15a617f Mon Sep 17 00:00:00 2001
From: Sheng Yang <sheng.yang-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Date: Tue, 6 Nov 2007 11:37:44 +0800
Subject: [PATCH] KVM: VMX: Fix repeatly allocate of apic access page

For SMP guest, alloc_apic_access_page() would be called more than once.
So only the last one works, causing SMP guest can't benifit from FlexPriority.

This patch fixed this issue.

Signed-off-by: Sheng Yang <sheng.yang-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
 drivers/kvm/vmx.c |   14 ++++++++++----
 1 files changed, 10 insertions(+), 4 deletions(-)

diff --git a/drivers/kvm/vmx.c b/drivers/kvm/vmx.c
index 42e7fad..6670e2b 100644
--- a/drivers/kvm/vmx.c
+++ b/drivers/kvm/vmx.c
@@ -1463,18 +1463,24 @@ static void seg_setup(int seg)
 static int alloc_apic_access_page(struct kvm *kvm)
 {
 	struct kvm_userspace_memory_region kvm_userspace_mem;
-	int r;
+	int r = 0;
 
-	r = -EFAULT;
+	mutex_lock(&kvm->lock);
+	if (kvm->apic_access_page)
+		goto out;
+	mutex_unlock(&kvm->lock);
 	kvm_userspace_mem.slot = APIC_ACCESS_PAGE_PRIVATE_MEMSLOT;
 	kvm_userspace_mem.flags = 0;
 	kvm_userspace_mem.guest_phys_addr = 0xfee00000ULL;
 	kvm_userspace_mem.memory_size = PAGE_SIZE;
 	r = kvm_set_memory_region(kvm, &kvm_userspace_mem, 0);
 	if (r)
-		return r;
+		goto out;
+	mutex_lock(&kvm->lock);
 	kvm->apic_access_page = gfn_to_page(kvm, 0xfee00);
-	return 0;
+out:
+	mutex_unlock(&kvm->lock);
+	return r;
 }
 
 /*
-- 
1.5.2


[-- Attachment #3: Type: text/plain, Size: 314 bytes --]

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/

[-- Attachment #4: Type: text/plain, Size: 186 bytes --]

_______________________________________________
kvm-devel mailing list
kvm-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
https://lists.sourceforge.net/lists/listinfo/kvm-devel

Re: [PATCH] KVM: VMX: Fix repeatly allocate of apic access page

[Avi Kivity]

Sheng Yang wrote:
> From 05164d73994a88ae130a086b592cd127c15a617f Mon Sep 17 00:00:00 2001
> From: Sheng Yang <sheng.yang-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
> Date: Tue, 6 Nov 2007 11:37:44 +0800
> Subject: [PATCH] KVM: VMX: Fix repeatly allocate of apic access page
>
> For SMP guest, alloc_apic_access_page() would be called more than once.
> So only the last one works, causing SMP guest can't benifit from FlexPriority.
>
> This patch fixed this issue.
>
> Signed-off-by: Sheng Yang <sheng.yang-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
> ---
>  drivers/kvm/vmx.c |   14 ++++++++++----
>  1 files changed, 10 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/kvm/vmx.c b/drivers/kvm/vmx.c
> index 42e7fad..6670e2b 100644
> --- a/drivers/kvm/vmx.c
> +++ b/drivers/kvm/vmx.c
> @@ -1463,18 +1463,24 @@ static void seg_setup(int seg)
>  static int alloc_apic_access_page(struct kvm *kvm)
>  {
>  	struct kvm_userspace_memory_region kvm_userspace_mem;
> -	int r;
> +	int r = 0;
>  
> -	r = -EFAULT;
> +	mutex_lock(&kvm->lock);
> +	if (kvm->apic_access_page)
> +		goto out;
> +	mutex_unlock(&kvm->lock);
>   

Still racy.  I two threads create a vcpu at the same time, they can both 
reach this point simultaneously.

I suggest having a __kvm_set_memory_region() that does not take 
kvm->lock, and call that instead, so the whole thing can be done within 
a single lock.

>  	kvm_userspace_mem.slot = APIC_ACCESS_PAGE_PRIVATE_MEMSLOT;
>  	kvm_userspace_mem.flags = 0;
>  	kvm_userspace_mem.guest_phys_addr = 0xfee00000ULL;
>  	kvm_userspace_mem.memory_size = PAGE_SIZE;
>  	r = kvm_set_memory_region(kvm, &kvm_userspace_mem, 0);
>  	if (r)
> -		return r;
> +		goto out;
> +	mutex_lock(&kvm->lock);
>  	kvm->apic_access_page = gfn_to_page(kvm, 0xfee00);
> -	return 0;
> +out:
> +	mutex_unlock(&kvm->lock);
> +	return r;
>  }
>  
>   

-- 
error compiling committee.c: too many arguments to function


-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/

[PATCH] KVM: VMX: Fix repeatly allocate of apic access page

[Sheng Yang]

[-- Attachment #1: Type: text/plain, Size: 5039 bytes --]

From 617412380eec67503d45cd91860b756f15293e73 Mon Sep 17 00:00:00 2001
From: Sheng Yang <sheng.yang-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Date: Tue, 6 Nov 2007 11:37:44 +0800
Subject: [PATCH] KVM: VMX: Fix repeatly allocate of apic access page

For SMP guest, alloc_apic_access_page() would be called more than once.
So only the last one works, causing SMP guest can't benifit from FlexPriority.

This patch fixed this issue.
---
 drivers/kvm/kvm.h      |    3 +++
 drivers/kvm/kvm_main.c |   38 ++++++++++++++++++++++++--------------
 drivers/kvm/vmx.c      |   14 +++++++++-----
 3 files changed, 36 insertions(+), 19 deletions(-)

diff --git a/drivers/kvm/kvm.h b/drivers/kvm/kvm.h
index 6b3b5e9..3ebfb19 100644
--- a/drivers/kvm/kvm.h
+++ b/drivers/kvm/kvm.h
@@ -524,6 +524,9 @@ int is_error_page(struct page *page);
 int kvm_set_memory_region(struct kvm *kvm,
 			  struct kvm_userspace_memory_region *mem,
 			  int user_alloc);
+int __kvm_set_memory_region(struct kvm *kvm,
+			    struct kvm_userspace_memory_region *mem,
+			    int user_alloc);
 gfn_t unalias_gfn(struct kvm *kvm, gfn_t gfn);
 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
 void kvm_release_page(struct page *page);
diff --git a/drivers/kvm/kvm_main.c b/drivers/kvm/kvm_main.c
index af7090f..d332a77 100644
--- a/drivers/kvm/kvm_main.c
+++ b/drivers/kvm/kvm_main.c
@@ -277,10 +277,12 @@ static int kvm_vm_release(struct inode *inode, struct 
file *filp)
  * space.
  *
  * Discontiguous memory is allowed, mostly for framebuffers.
+ *
+ * Must be called holding kvm->lock.
  */
-int kvm_set_memory_region(struct kvm *kvm,
-			  struct kvm_userspace_memory_region *mem,
-			  int user_alloc)
+int __kvm_set_memory_region(struct kvm *kvm,
+			    struct kvm_userspace_memory_region *mem,
+			    int user_alloc)
 {
 	int r;
 	gfn_t base_gfn;
@@ -307,8 +309,6 @@ int kvm_set_memory_region(struct kvm *kvm,
 	if (!npages)
 		mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES;
 
-	mutex_lock(&kvm->lock);
-
 	new = old = *memslot;
 
 	new.base_gfn = base_gfn;
@@ -318,7 +318,7 @@ int kvm_set_memory_region(struct kvm *kvm,
 	/* Disallow changing a memory slot's size. */
 	r = -EINVAL;
 	if (npages && old.npages && npages != old.npages)
-		goto out_unlock;
+		goto out_free;
 
 	/* Check for overlaps */
 	r = -EEXIST;
@@ -329,7 +329,7 @@ int kvm_set_memory_region(struct kvm *kvm,
 			continue;
 		if (!((base_gfn + npages <= s->base_gfn) ||
 		      (base_gfn >= s->base_gfn + s->npages)))
-			goto out_unlock;
+			goto out_free;
 	}
 
 	/* Free page dirty bitmap if unneeded */
@@ -343,7 +343,7 @@ int kvm_set_memory_region(struct kvm *kvm,
 		new.rmap = vmalloc(npages * sizeof(struct page *));
 
 		if (!new.rmap)
-			goto out_unlock;
+			goto out_free;
 
 		memset(new.rmap, 0, npages * sizeof(*new.rmap));
 
@@ -360,7 +360,7 @@ int kvm_set_memory_region(struct kvm *kvm,
 			up_write(&current->mm->mmap_sem);
 
 			if (IS_ERR((void *)new.userspace_addr))
-				goto out_unlock;
+				goto out_free;
 		}
 	} else {
 		if (!old.user_alloc && old.rmap) {
@@ -383,7 +383,7 @@ int kvm_set_memory_region(struct kvm *kvm,
 
 		new.dirty_bitmap = vmalloc(dirty_bytes);
 		if (!new.dirty_bitmap)
-			goto out_unlock;
+			goto out_free;
 		memset(new.dirty_bitmap, 0, dirty_bytes);
 	}
 
@@ -413,18 +413,28 @@ int kvm_set_memory_region(struct kvm *kvm,
 	kvm_mmu_slot_remove_write_access(kvm, mem->slot);
 	kvm_flush_remote_tlbs(kvm);
 
-	mutex_unlock(&kvm->lock);
-
 	kvm_free_physmem_slot(&old, &new);
 	return 0;
 
-out_unlock:
-	mutex_unlock(&kvm->lock);
+out_free:
 	kvm_free_physmem_slot(&new, &old);
 out:
 	return r;
 
 }
+EXPORT_SYMBOL_GPL(__kvm_set_memory_region);
+
+int kvm_set_memory_region(struct kvm *kvm,
+			  struct kvm_userspace_memory_region *mem,
+			  int user_alloc)
+{
+	int r;
+
+	mutex_lock(&kvm->lock);
+	r = __kvm_set_memory_region(kvm, mem, user_alloc);
+	mutex_unlock(&kvm->lock);
+	return r;
+}
 EXPORT_SYMBOL_GPL(kvm_set_memory_region);
 
 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
diff --git a/drivers/kvm/vmx.c b/drivers/kvm/vmx.c
index 42e7fad..da3a339 100644
--- a/drivers/kvm/vmx.c
+++ b/drivers/kvm/vmx.c
@@ -1463,18 +1463,22 @@ static void seg_setup(int seg)
 static int alloc_apic_access_page(struct kvm *kvm)
 {
 	struct kvm_userspace_memory_region kvm_userspace_mem;
-	int r;
+	int r = 0;
 
-	r = -EFAULT;
+	mutex_lock(&kvm->lock);
+	if (kvm->apic_access_page)
+		goto out;
 	kvm_userspace_mem.slot = APIC_ACCESS_PAGE_PRIVATE_MEMSLOT;
 	kvm_userspace_mem.flags = 0;
 	kvm_userspace_mem.guest_phys_addr = 0xfee00000ULL;
 	kvm_userspace_mem.memory_size = PAGE_SIZE;
-	r = kvm_set_memory_region(kvm, &kvm_userspace_mem, 0);
+	r = __kvm_set_memory_region(kvm, &kvm_userspace_mem, 0);
 	if (r)
-		return r;
+		goto out;
 	kvm->apic_access_page = gfn_to_page(kvm, 0xfee00);
-	return 0;
+out:
+	mutex_unlock(&kvm->lock);
+	return r;
 }
 
 /*
-- 
1.5.2


[-- Attachment #2: 0001-KVM-VMX-Fix-repeatly-allocate-of-apic-access-page.patch --]
[-- Type: text/x-diff, Size: 5037 bytes --]

From 617412380eec67503d45cd91860b756f15293e73 Mon Sep 17 00:00:00 2001
From: Sheng Yang <sheng.yang-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Date: Tue, 6 Nov 2007 11:37:44 +0800
Subject: [PATCH] KVM: VMX: Fix repeatly allocate of apic access page

For SMP guest, alloc_apic_access_page() would be called more than once.
So only the last one works, causing SMP guest can't benifit from FlexPriority.

This patch fixed this issue.
---
 drivers/kvm/kvm.h      |    3 +++
 drivers/kvm/kvm_main.c |   38 ++++++++++++++++++++++++--------------
 drivers/kvm/vmx.c      |   14 +++++++++-----
 3 files changed, 36 insertions(+), 19 deletions(-)

diff --git a/drivers/kvm/kvm.h b/drivers/kvm/kvm.h
index 6b3b5e9..3ebfb19 100644
--- a/drivers/kvm/kvm.h
+++ b/drivers/kvm/kvm.h
@@ -524,6 +524,9 @@ int is_error_page(struct page *page);
 int kvm_set_memory_region(struct kvm *kvm,
 			  struct kvm_userspace_memory_region *mem,
 			  int user_alloc);
+int __kvm_set_memory_region(struct kvm *kvm,
+			    struct kvm_userspace_memory_region *mem,
+			    int user_alloc);
 gfn_t unalias_gfn(struct kvm *kvm, gfn_t gfn);
 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
 void kvm_release_page(struct page *page);
diff --git a/drivers/kvm/kvm_main.c b/drivers/kvm/kvm_main.c
index af7090f..d332a77 100644
--- a/drivers/kvm/kvm_main.c
+++ b/drivers/kvm/kvm_main.c
@@ -277,10 +277,12 @@ static int kvm_vm_release(struct inode *inode, struct file *filp)
  * space.
  *
  * Discontiguous memory is allowed, mostly for framebuffers.
+ *
+ * Must be called holding kvm->lock.
  */
-int kvm_set_memory_region(struct kvm *kvm,
-			  struct kvm_userspace_memory_region *mem,
-			  int user_alloc)
+int __kvm_set_memory_region(struct kvm *kvm,
+			    struct kvm_userspace_memory_region *mem,
+			    int user_alloc)
 {
 	int r;
 	gfn_t base_gfn;
@@ -307,8 +309,6 @@ int kvm_set_memory_region(struct kvm *kvm,
 	if (!npages)
 		mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES;
 
-	mutex_lock(&kvm->lock);
-
 	new = old = *memslot;
 
 	new.base_gfn = base_gfn;
@@ -318,7 +318,7 @@ int kvm_set_memory_region(struct kvm *kvm,
 	/* Disallow changing a memory slot's size. */
 	r = -EINVAL;
 	if (npages && old.npages && npages != old.npages)
-		goto out_unlock;
+		goto out_free;
 
 	/* Check for overlaps */
 	r = -EEXIST;
@@ -329,7 +329,7 @@ int kvm_set_memory_region(struct kvm *kvm,
 			continue;
 		if (!((base_gfn + npages <= s->base_gfn) ||
 		      (base_gfn >= s->base_gfn + s->npages)))
-			goto out_unlock;
+			goto out_free;
 	}
 
 	/* Free page dirty bitmap if unneeded */
@@ -343,7 +343,7 @@ int kvm_set_memory_region(struct kvm *kvm,
 		new.rmap = vmalloc(npages * sizeof(struct page *));
 
 		if (!new.rmap)
-			goto out_unlock;
+			goto out_free;
 
 		memset(new.rmap, 0, npages * sizeof(*new.rmap));
 
@@ -360,7 +360,7 @@ int kvm_set_memory_region(struct kvm *kvm,
 			up_write(&current->mm->mmap_sem);
 
 			if (IS_ERR((void *)new.userspace_addr))
-				goto out_unlock;
+				goto out_free;
 		}
 	} else {
 		if (!old.user_alloc && old.rmap) {
@@ -383,7 +383,7 @@ int kvm_set_memory_region(struct kvm *kvm,
 
 		new.dirty_bitmap = vmalloc(dirty_bytes);
 		if (!new.dirty_bitmap)
-			goto out_unlock;
+			goto out_free;
 		memset(new.dirty_bitmap, 0, dirty_bytes);
 	}
 
@@ -413,18 +413,28 @@ int kvm_set_memory_region(struct kvm *kvm,
 	kvm_mmu_slot_remove_write_access(kvm, mem->slot);
 	kvm_flush_remote_tlbs(kvm);
 
-	mutex_unlock(&kvm->lock);
-
 	kvm_free_physmem_slot(&old, &new);
 	return 0;
 
-out_unlock:
-	mutex_unlock(&kvm->lock);
+out_free:
 	kvm_free_physmem_slot(&new, &old);
 out:
 	return r;
 
 }
+EXPORT_SYMBOL_GPL(__kvm_set_memory_region);
+
+int kvm_set_memory_region(struct kvm *kvm,
+			  struct kvm_userspace_memory_region *mem,
+			  int user_alloc)
+{
+	int r;
+
+	mutex_lock(&kvm->lock);
+	r = __kvm_set_memory_region(kvm, mem, user_alloc);
+	mutex_unlock(&kvm->lock);
+	return r;
+}
 EXPORT_SYMBOL_GPL(kvm_set_memory_region);
 
 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
diff --git a/drivers/kvm/vmx.c b/drivers/kvm/vmx.c
index 42e7fad..da3a339 100644
--- a/drivers/kvm/vmx.c
+++ b/drivers/kvm/vmx.c
@@ -1463,18 +1463,22 @@ static void seg_setup(int seg)
 static int alloc_apic_access_page(struct kvm *kvm)
 {
 	struct kvm_userspace_memory_region kvm_userspace_mem;
-	int r;
+	int r = 0;
 
-	r = -EFAULT;
+	mutex_lock(&kvm->lock);
+	if (kvm->apic_access_page)
+		goto out;
 	kvm_userspace_mem.slot = APIC_ACCESS_PAGE_PRIVATE_MEMSLOT;
 	kvm_userspace_mem.flags = 0;
 	kvm_userspace_mem.guest_phys_addr = 0xfee00000ULL;
 	kvm_userspace_mem.memory_size = PAGE_SIZE;
-	r = kvm_set_memory_region(kvm, &kvm_userspace_mem, 0);
+	r = __kvm_set_memory_region(kvm, &kvm_userspace_mem, 0);
 	if (r)
-		return r;
+		goto out;
 	kvm->apic_access_page = gfn_to_page(kvm, 0xfee00);
-	return 0;
+out:
+	mutex_unlock(&kvm->lock);
+	return r;
 }
 
 /*
-- 
1.5.2


[-- Attachment #3: Type: text/plain, Size: 314 bytes --]

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/

[-- Attachment #4: Type: text/plain, Size: 186 bytes --]

_______________________________________________
kvm-devel mailing list
kvm-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
https://lists.sourceforge.net/lists/listinfo/kvm-devel

Re: [PATCH] KVM: VMX: Fix repeatly allocate of apic access page

[Avi Kivity]

Sheng Yang wrote:
> From 617412380eec67503d45cd91860b756f15293e73 Mon Sep 17 00:00:00 2001
> From: Sheng Yang <sheng.yang-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
> Date: Tue, 6 Nov 2007 11:37:44 +0800
> Subject: [PATCH] KVM: VMX: Fix repeatly allocate of apic access page
>
> For SMP guest, alloc_apic_access_page() would be called more than once.
> So only the last one works, causing SMP guest can't benifit from FlexPriority.
>
>   

Patch is good, but no signoff...

(no need to resend, just provide one in the reply, sorry about the red tape)


-- 
error compiling committee.c: too many arguments to function


-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/