[PATCH] KVM: SEV: Pin SEV guest memory out of CMA area

[PATCH] KVM: SEV: Pin SEV guest memory out of CMA area

[yangge1116]

From: yangge <yangge1116@126.com>

When pin_user_pages_fast() pins SEV guest memory without the
FOLL_LONGTERM flag, the pinned pages may inadvertently end up in the
CMA (Contiguous Memory Allocator) area. This can subsequently cause
cma_alloc() to fail in allocating these pages, due to the fact that
the pinned pages are not migratable.

To address the aforementioned problem, we propose adding the
FOLL_LONGTERM flag to the pin_user_pages_fast() function. By doing
so, we ensure that the pages allocated will not occupy space within
the CMA area, thereby preventing potential allocation failures.

Signed-off-by: yangge <yangge1116@126.com>
---
 arch/x86/kvm/svm/sev.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c
index 943bd07..35d0714 100644
--- a/arch/x86/kvm/svm/sev.c
+++ b/arch/x86/kvm/svm/sev.c
@@ -630,6 +630,7 @@ static struct page **sev_pin_memory(struct kvm *kvm, unsigned long uaddr,
 	unsigned long locked, lock_limit;
 	struct page **pages;
 	unsigned long first, last;
+	unsigned int flags = 0;
 	int ret;
 
 	lockdep_assert_held(&kvm->lock);
@@ -662,8 +663,10 @@ static struct page **sev_pin_memory(struct kvm *kvm, unsigned long uaddr,
 	if (!pages)
 		return ERR_PTR(-ENOMEM);
 
+	flags = write ? FOLL_WRITE : 0;
+
 	/* Pin the user virtual address. */
-	npinned = pin_user_pages_fast(uaddr, npages, write ? FOLL_WRITE : 0, pages);
+	npinned = pin_user_pages_fast(uaddr, npages, flags | FOLL_LONGTERM, pages);
 	if (npinned != npages) {
 		pr_err("SEV: Failure locking %lu pages.\n", npages);
 		ret = -ENOMEM;
-- 
2.7.4

Re: [PATCH] KVM: SEV: Pin SEV guest memory out of CMA area

[David Hildenbrand]

On 10.01.25 09:48, yangge1116@126.com wrote:
> From: yangge <yangge1116@126.com>

I would suggest the title to be something like

"KVM: SEV: fix wrong pinning of pages"

Then you can describe that without FOLL_LONGTERM, the pages will not get 
migrated out of MIGRATE_CMA/ZONE_MOVABLE, violating these mechanisms to 
avoid fragmentation with unmovable pages, for example making CMA 
allocations fail.

(CMA allocations failing is only one symptom of the missed usage of 
FOLL_LONGTERM)

> 
> When pin_user_pages_fast() pins SEV guest memory without the
> FOLL_LONGTERM flag, the pinned pages may inadvertently end up in the
> CMA (Contiguous Memory Allocator) area. This can subsequently cause
> cma_alloc() to fail in allocating these pages, due to the fact that
> the pinned pages are not migratable.
> 
> To address the aforementioned problem, we propose adding the
> FOLL_LONGTERM flag to the pin_user_pages_fast() function. By doing
> so, we ensure that the pages allocated will not occupy space within
> the CMA area, thereby preventing potential allocation failures.
> 
> Signed-off-by: yangge <yangge1116@126.com>
> ---
>   arch/x86/kvm/svm/sev.c | 5 ++++-
>   1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c
> index 943bd07..35d0714 100644
> --- a/arch/x86/kvm/svm/sev.c
> +++ b/arch/x86/kvm/svm/sev.c
> @@ -630,6 +630,7 @@ static struct page **sev_pin_memory(struct kvm *kvm, unsigned long uaddr,
>   	unsigned long locked, lock_limit;
>   	struct page **pages;
>   	unsigned long first, last;
> +	unsigned int flags = 0;

Why not set

unsigned int flags = FOLL_LONGTERM;

>   	int ret;
>   
>   	lockdep_assert_held(&kvm->lock);
> @@ -662,8 +663,10 @@ static struct page **sev_pin_memory(struct kvm *kvm, unsigned long uaddr,
>   	if (!pages)
>   		return ERR_PTR(-ENOMEM);
>   
> +	flags = write ? FOLL_WRITE : 0;


and here do

flags |= write ? FOLL_WRITE : 0;

> +
>   	/* Pin the user virtual address. */
> -	npinned = pin_user_pages_fast(uaddr, npages, write ? FOLL_WRITE : 0, pages);
> +	npinned = pin_user_pages_fast(uaddr, npages, flags | FOLL_LONGTERM, pages);
>   	if (npinned != npages) {
>   		pr_err("SEV: Failure locking %lu pages.\n", npages);
>   		ret = -ENOMEM;


-- 
Cheers,

David / dhildenb

Re: [PATCH] KVM: SEV: Pin SEV guest memory out of CMA area

[Ge Yang]

在 2025/1/10 16:58, David Hildenbrand 写道:
> On 10.01.25 09:48, yangge1116@126.com wrote:
>> From: yangge <yangge1116@126.com>
> 
> I would suggest the title to be something like
> 
> "KVM: SEV: fix wrong pinning of pages"
> 
> Then you can describe that without FOLL_LONGTERM, the pages will not get 
> migrated out of MIGRATE_CMA/ZONE_MOVABLE, violating these mechanisms to 
> avoid fragmentation with unmovable pages, for example making CMA 
> allocations fail.
> 
> (CMA allocations failing is only one symptom of the missed usage of 
> FOLL_LONGTERM)
> 
>>
>> When pin_user_pages_fast() pins SEV guest memory without the
>> FOLL_LONGTERM flag, the pinned pages may inadvertently end up in the
>> CMA (Contiguous Memory Allocator) area. This can subsequently cause
>> cma_alloc() to fail in allocating these pages, due to the fact that
>> the pinned pages are not migratable.
>>
>> To address the aforementioned problem, we propose adding the
>> FOLL_LONGTERM flag to the pin_user_pages_fast() function. By doing
>> so, we ensure that the pages allocated will not occupy space within
>> the CMA area, thereby preventing potential allocation failures.
>>
>> Signed-off-by: yangge <yangge1116@126.com>
>> ---
>>   arch/x86/kvm/svm/sev.c | 5 ++++-
>>   1 file changed, 4 insertions(+), 1 deletion(-)
>>
>> diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c
>> index 943bd07..35d0714 100644
>> --- a/arch/x86/kvm/svm/sev.c
>> +++ b/arch/x86/kvm/svm/sev.c
>> @@ -630,6 +630,7 @@ static struct page **sev_pin_memory(struct kvm 
>> *kvm, unsigned long uaddr,
>>       unsigned long locked, lock_limit;
>>       struct page **pages;
>>       unsigned long first, last;
>> +    unsigned int flags = 0;
> 
> Why not set
> 
> unsigned int flags = FOLL_LONGTERM;
> 
>>       int ret;
>>       lockdep_assert_held(&kvm->lock);
>> @@ -662,8 +663,10 @@ static struct page **sev_pin_memory(struct kvm 
>> *kvm, unsigned long uaddr,
>>       if (!pages)
>>           return ERR_PTR(-ENOMEM);
>> +    flags = write ? FOLL_WRITE : 0;
> 
> 
> and here do
> 
> flags |= write ? FOLL_WRITE : 0;
> 
>> +
>>       /* Pin the user virtual address. */
>> -    npinned = pin_user_pages_fast(uaddr, npages, write ? FOLL_WRITE : 
>> 0, pages);
>> +    npinned = pin_user_pages_fast(uaddr, npages, flags | 
>> FOLL_LONGTERM, pages);
>>       if (npinned != npages) {
>>           pr_err("SEV: Failure locking %lu pages.\n", npages);
>>           ret = -ENOMEM;
> 
> 
Ok, thanks.