[PATCH] mm/memremap.c: map FS_DAX device memory as decrypted

Pankaj Gupta posted 1 patch 3 years, 5 months ago
mm/memremap.c | 1 +
1 file changed, 1 insertion(+)
[PATCH] mm/memremap.c: map FS_DAX device memory as decrypted
Posted by Pankaj Gupta 3 years, 5 months ago
virtio_pmem use devm_memremap_pages() to map the device memory.
By default this memory is mapped as encrypted with SEV. Guest
reboot changes the current encryption key and guest no longer
properly decrypts the FSDAX device meta data.

Mark the corresponding device memory region for FSDAX devices
(mapped with memremap_pages) as decrypted to retain the persistent
memory property.

Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
---
 mm/memremap.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/mm/memremap.c b/mm/memremap.c
index 421bec3a29ee..08cbf54fe037 100644
--- a/mm/memremap.c
+++ b/mm/memremap.c
@@ -335,6 +335,7 @@ void *memremap_pages(struct dev_pagemap *pgmap, int nid)
 			WARN(1, "File system DAX not supported\n");
 			return ERR_PTR(-EINVAL);
 		}
+		params.pgprot = pgprot_decrypted(params.pgprot);
 		break;
 	case MEMORY_DEVICE_GENERIC:
 		break;
-- 
2.34.1
Re: [PATCH] mm/memremap.c: map FS_DAX device memory as decrypted
Posted by Andrew Morton 3 years, 5 months ago
On Wed, 2 Nov 2022 11:07:28 -0500 Pankaj Gupta <pankaj.gupta@amd.com> wrote:

> virtio_pmem use devm_memremap_pages() to map the device memory.
> By default this memory is mapped as encrypted with SEV. Guest
> reboot changes the current encryption key and guest no longer
> properly decrypts the FSDAX device meta data.
> 
> Mark the corresponding device memory region for FSDAX devices
> (mapped with memremap_pages) as decrypted to retain the persistent
> memory property.
> 
> ...
>
> --- a/mm/memremap.c
> +++ b/mm/memremap.c
> @@ -335,6 +335,7 @@ void *memremap_pages(struct dev_pagemap *pgmap, int nid)
>  			WARN(1, "File system DAX not supported\n");
>  			return ERR_PTR(-EINVAL);
>  		}
> +		params.pgprot = pgprot_decrypted(params.pgprot);
>  		break;
>  	case MEMORY_DEVICE_GENERIC:
>  		break;

Should this be backported into earlier kernels?

Is b7b3c01b19159 a suitable Fixes: target?
Re: [PATCH] mm/memremap.c: map FS_DAX device memory as decrypted
Posted by Gupta, Pankaj 3 years, 5 months ago
Hi Andrew,

>> virtio_pmem use devm_memremap_pages() to map the device memory.
>> By default this memory is mapped as encrypted with SEV. Guest
>> reboot changes the current encryption key and guest no longer
>> properly decrypts the FSDAX device meta data.
>>
>> Mark the corresponding device memory region for FSDAX devices
>> (mapped with memremap_pages) as decrypted to retain the persistent
>> memory property.
>>
>> ...
>>
>> --- a/mm/memremap.c
>> +++ b/mm/memremap.c
>> @@ -335,6 +335,7 @@ void *memremap_pages(struct dev_pagemap *pgmap, int nid)
>>   			WARN(1, "File system DAX not supported\n");
>>   			return ERR_PTR(-EINVAL);
>>   		}
>> +		params.pgprot = pgprot_decrypted(params.pgprot);
>>   		break;
>>   	case MEMORY_DEVICE_GENERIC:
>>   		break;
> 
> Should this be backported into earlier kernels?

Yes, kernel >= 5.10 for clean back-port as it contains the below commit.

> 
> Is b7b3c01b19159 a suitable Fixes: target?

Not very sure. But seems like the right commit for fixes tag as the 
target code is moved in this commit with multiple range support addition.

Should I send a v2 with fixes & stable kernel fixes tag?

Thanks,
Pankaj
Re: [PATCH] mm/memremap.c: map FS_DAX device memory as decrypted
Posted by Andrew Morton 3 years, 5 months ago
On Thu, 3 Nov 2022 08:55:40 +0100 "Gupta, Pankaj" <pankaj.gupta@amd.com> wrote:

> > Should this be backported into earlier kernels?
> 
> Yes, kernel >= 5.10 for clean back-port as it contains the below commit.
> 
> > 
> > Is b7b3c01b19159 a suitable Fixes: target?
> 
> Not very sure. But seems like the right commit for fixes tag as the 
> target code is moved in this commit with multiple range support addition.
> 
> Should I send a v2 with fixes & stable kernel fixes tag?

That's OK thanks - I made those changes to my copy.
Re: [PATCH] mm/memremap.c: map FS_DAX device memory as decrypted
Posted by Gupta, Pankaj 3 years, 5 months ago
+Cc Andrew [forgot to add earlier]

> virtio_pmem use devm_memremap_pages() to map the device memory.
> By default this memory is mapped as encrypted with SEV. Guest
> reboot changes the current encryption key and guest no longer
> properly decrypts the FSDAX device meta data.
> 
> Mark the corresponding device memory region for FSDAX devices
> (mapped with memremap_pages) as decrypted to retain the persistent
> memory property.
> 
> Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
> ---
>   mm/memremap.c | 1 +
>   1 file changed, 1 insertion(+)
> 
> diff --git a/mm/memremap.c b/mm/memremap.c
> index 421bec3a29ee..08cbf54fe037 100644
> --- a/mm/memremap.c
> +++ b/mm/memremap.c
> @@ -335,6 +335,7 @@ void *memremap_pages(struct dev_pagemap *pgmap, int nid)
>   			WARN(1, "File system DAX not supported\n");
>   			return ERR_PTR(-EINVAL);
>   		}
> +		params.pgprot = pgprot_decrypted(params.pgprot);
>   		break;
>   	case MEMORY_DEVICE_GENERIC:
>   		break;