Introduce a new memfd_create() flag indicating the content of the
created memfd is inaccessible from userspace. It does this by force
setting F_SEAL_INACCESSIBLE seal when the file is created. It also set
F_SEAL_SEAL to prevent future sealing, which means, it can not coexist
with MFD_ALLOW_SEALING.

The pages backed by such memfd will be used as guest private memory in
confidential computing environments such as Intel TDX/AMD SEV. Since
page migration/swapping is not yet supported for such usages so these
pages are currently marked as UNMOVABLE and UNEVICTABLE which makes
them behave like long-term pinned pages.

Signed-off-by: Chao Peng <chao.p.peng@linux.intel.com>
---
 include/uapi/linux/memfd.h |  1 +
 mm/memfd.c                 | 20 +++++++++++++++++++-
 2 files changed, 20 insertions(+), 1 deletion(-)

diff --git a/include/uapi/linux/memfd.h b/include/uapi/linux/memfd.h
index 7a8a26751c23..48750474b904 100644
--- a/include/uapi/linux/memfd.h
+++ b/include/uapi/linux/memfd.h
@@ -8,6 +8,7 @@
 #define MFD_CLOEXEC		0x0001U
 #define MFD_ALLOW_SEALING	0x0002U
 #define MFD_HUGETLB		0x0004U
+#define MFD_INACCESSIBLE	0x0008U
 
 /*
  * Huge page size encoding when MFD_HUGETLB is specified, and a huge page
diff --git a/mm/memfd.c b/mm/memfd.c
index 9f80f162791a..26998d96dc11 100644
--- a/mm/memfd.c
+++ b/mm/memfd.c
@@ -245,16 +245,19 @@ long memfd_fcntl(struct file *file, unsigned int cmd, unsigned long arg)
 #define MFD_NAME_PREFIX_LEN (sizeof(MFD_NAME_PREFIX) - 1)
 #define MFD_NAME_MAX_LEN (NAME_MAX - MFD_NAME_PREFIX_LEN)
 
-#define MFD_ALL_FLAGS (MFD_CLOEXEC | MFD_ALLOW_SEALING | MFD_HUGETLB)
+#define MFD_ALL_FLAGS (MFD_CLOEXEC | MFD_ALLOW_SEALING | MFD_HUGETLB | \
+		       MFD_INACCESSIBLE)
 
 SYSCALL_DEFINE2(memfd_create,
 		const char __user *, uname,
 		unsigned int, flags)
 {
+	struct address_space *mapping;
 	unsigned int *file_seals;
 	struct file *file;
 	int fd, error;
 	char *name;
+	gfp_t gfp;
 	long len;
 
 	if (!(flags & MFD_HUGETLB)) {
@@ -267,6 +270,10 @@ SYSCALL_DEFINE2(memfd_create,
 			return -EINVAL;
 	}
 
+	/* Disallow sealing when MFD_INACCESSIBLE is set. */
+	if (flags & MFD_INACCESSIBLE && flags & MFD_ALLOW_SEALING)
+		return -EINVAL;
+
 	/* length includes terminating zero */
 	len = strnlen_user(uname, MFD_NAME_MAX_LEN + 1);
 	if (len <= 0)
@@ -315,6 +322,17 @@ SYSCALL_DEFINE2(memfd_create,
 		*file_seals &= ~F_SEAL_SEAL;
 	}
 
+	if (flags & MFD_INACCESSIBLE) {
+		mapping = file_inode(file)->i_mapping;
+		gfp = mapping_gfp_mask(mapping);
+		gfp &= ~__GFP_MOVABLE;
+		mapping_set_gfp_mask(mapping, gfp);
+		mapping_set_unevictable(mapping);
+
+		file_seals = memfd_file_seals_ptr(file);
+		*file_seals &= F_SEAL_SEAL | F_SEAL_INACCESSIBLE;
+	}
+
 	fd_install(fd, file);
 	kfree(name);
 	return fd;
-- 
2.17.1

On 18/01/2022 13:21, Chao Peng wrote:
> Introduce a new memfd_create() flag indicating the content of the
> created memfd is inaccessible from userspace. It does this by force
> setting F_SEAL_INACCESSIBLE seal when the file is created. It also set
> F_SEAL_SEAL to prevent future sealing, which means, it can not coexist
> with MFD_ALLOW_SEALING.
> 
> The pages backed by such memfd will be used as guest private memory in
> confidential computing environments such as Intel TDX/AMD SEV. Since
> page migration/swapping is not yet supported for such usages so these
> pages are currently marked as UNMOVABLE and UNEVICTABLE which makes
> them behave like long-term pinned pages.
> 
> Signed-off-by: Chao Peng <chao.p.peng@linux.intel.com>
> ---
>  include/uapi/linux/memfd.h |  1 +
>  mm/memfd.c                 | 20 +++++++++++++++++++-
>  2 files changed, 20 insertions(+), 1 deletion(-)
> 
> diff --git a/include/uapi/linux/memfd.h b/include/uapi/linux/memfd.h
> index 7a8a26751c23..48750474b904 100644
> --- a/include/uapi/linux/memfd.h
> +++ b/include/uapi/linux/memfd.h
> @@ -8,6 +8,7 @@
>  #define MFD_CLOEXEC		0x0001U
>  #define MFD_ALLOW_SEALING	0x0002U
>  #define MFD_HUGETLB		0x0004U
> +#define MFD_INACCESSIBLE	0x0008U
>  
>  /*
>   * Huge page size encoding when MFD_HUGETLB is specified, and a huge page
> diff --git a/mm/memfd.c b/mm/memfd.c
> index 9f80f162791a..26998d96dc11 100644
> --- a/mm/memfd.c
> +++ b/mm/memfd.c
> @@ -245,16 +245,19 @@ long memfd_fcntl(struct file *file, unsigned int cmd, unsigned long arg)
>  #define MFD_NAME_PREFIX_LEN (sizeof(MFD_NAME_PREFIX) - 1)
>  #define MFD_NAME_MAX_LEN (NAME_MAX - MFD_NAME_PREFIX_LEN)
>  
> -#define MFD_ALL_FLAGS (MFD_CLOEXEC | MFD_ALLOW_SEALING | MFD_HUGETLB)
> +#define MFD_ALL_FLAGS (MFD_CLOEXEC | MFD_ALLOW_SEALING | MFD_HUGETLB | \
> +		       MFD_INACCESSIBLE)
>  
>  SYSCALL_DEFINE2(memfd_create,
>  		const char __user *, uname,
>  		unsigned int, flags)
>  {
> +	struct address_space *mapping;
>  	unsigned int *file_seals;
>  	struct file *file;
>  	int fd, error;
>  	char *name;
> +	gfp_t gfp;
>  	long len;
>  
>  	if (!(flags & MFD_HUGETLB)) {
> @@ -267,6 +270,10 @@ SYSCALL_DEFINE2(memfd_create,
>  			return -EINVAL;
>  	}
>  
> +	/* Disallow sealing when MFD_INACCESSIBLE is set. */
> +	if (flags & MFD_INACCESSIBLE && flags & MFD_ALLOW_SEALING)
> +		return -EINVAL;
> +
>  	/* length includes terminating zero */
>  	len = strnlen_user(uname, MFD_NAME_MAX_LEN + 1);
>  	if (len <= 0)
> @@ -315,6 +322,17 @@ SYSCALL_DEFINE2(memfd_create,
>  		*file_seals &= ~F_SEAL_SEAL;
>  	}
>  
> +	if (flags & MFD_INACCESSIBLE) {
> +		mapping = file_inode(file)->i_mapping;
> +		gfp = mapping_gfp_mask(mapping);
> +		gfp &= ~__GFP_MOVABLE;
> +		mapping_set_gfp_mask(mapping, gfp);
> +		mapping_set_unevictable(mapping);
> +
> +		file_seals = memfd_file_seals_ptr(file);
> +		*file_seals &= F_SEAL_SEAL | F_SEAL_INACCESSIBLE;

This looks backwards - the flags should be set on *file_seals, but here
you are unsetting all other flags.

Steve

> +	}
> +
>  	fd_install(fd, file);
>  	kfree(name);
>  	return fd;
> 

On Fri, Jan 21, 2022 at 03:50:55PM +0000, Steven Price wrote:
> On 18/01/2022 13:21, Chao Peng wrote:
> > Introduce a new memfd_create() flag indicating the content of the
> > created memfd is inaccessible from userspace. It does this by force
> > setting F_SEAL_INACCESSIBLE seal when the file is created. It also set
> > F_SEAL_SEAL to prevent future sealing, which means, it can not coexist
> > with MFD_ALLOW_SEALING.
> > 
> > The pages backed by such memfd will be used as guest private memory in
> > confidential computing environments such as Intel TDX/AMD SEV. Since
> > page migration/swapping is not yet supported for such usages so these
> > pages are currently marked as UNMOVABLE and UNEVICTABLE which makes
> > them behave like long-term pinned pages.
> > 
> > Signed-off-by: Chao Peng <chao.p.peng@linux.intel.com>
> > ---
> >  include/uapi/linux/memfd.h |  1 +
> >  mm/memfd.c                 | 20 +++++++++++++++++++-
> >  2 files changed, 20 insertions(+), 1 deletion(-)
> > 
> > diff --git a/include/uapi/linux/memfd.h b/include/uapi/linux/memfd.h
> > index 7a8a26751c23..48750474b904 100644
> > --- a/include/uapi/linux/memfd.h
> > +++ b/include/uapi/linux/memfd.h
> > @@ -8,6 +8,7 @@
> >  #define MFD_CLOEXEC		0x0001U
> >  #define MFD_ALLOW_SEALING	0x0002U
> >  #define MFD_HUGETLB		0x0004U
> > +#define MFD_INACCESSIBLE	0x0008U
> >  
> >  /*
> >   * Huge page size encoding when MFD_HUGETLB is specified, and a huge page
> > diff --git a/mm/memfd.c b/mm/memfd.c
> > index 9f80f162791a..26998d96dc11 100644
> > --- a/mm/memfd.c
> > +++ b/mm/memfd.c
> > @@ -245,16 +245,19 @@ long memfd_fcntl(struct file *file, unsigned int cmd, unsigned long arg)
> >  #define MFD_NAME_PREFIX_LEN (sizeof(MFD_NAME_PREFIX) - 1)
> >  #define MFD_NAME_MAX_LEN (NAME_MAX - MFD_NAME_PREFIX_LEN)
> >  
> > -#define MFD_ALL_FLAGS (MFD_CLOEXEC | MFD_ALLOW_SEALING | MFD_HUGETLB)
> > +#define MFD_ALL_FLAGS (MFD_CLOEXEC | MFD_ALLOW_SEALING | MFD_HUGETLB | \
> > +		       MFD_INACCESSIBLE)
> >  
> >  SYSCALL_DEFINE2(memfd_create,
> >  		const char __user *, uname,
> >  		unsigned int, flags)
> >  {
> > +	struct address_space *mapping;
> >  	unsigned int *file_seals;
> >  	struct file *file;
> >  	int fd, error;
> >  	char *name;
> > +	gfp_t gfp;
> >  	long len;
> >  
> >  	if (!(flags & MFD_HUGETLB)) {
> > @@ -267,6 +270,10 @@ SYSCALL_DEFINE2(memfd_create,
> >  			return -EINVAL;
> >  	}
> >  
> > +	/* Disallow sealing when MFD_INACCESSIBLE is set. */
> > +	if (flags & MFD_INACCESSIBLE && flags & MFD_ALLOW_SEALING)
> > +		return -EINVAL;
> > +
> >  	/* length includes terminating zero */
> >  	len = strnlen_user(uname, MFD_NAME_MAX_LEN + 1);
> >  	if (len <= 0)
> > @@ -315,6 +322,17 @@ SYSCALL_DEFINE2(memfd_create,
> >  		*file_seals &= ~F_SEAL_SEAL;
> >  	}
> >  
> > +	if (flags & MFD_INACCESSIBLE) {
> > +		mapping = file_inode(file)->i_mapping;
> > +		gfp = mapping_gfp_mask(mapping);
> > +		gfp &= ~__GFP_MOVABLE;
> > +		mapping_set_gfp_mask(mapping, gfp);
> > +		mapping_set_unevictable(mapping);
> > +
> > +		file_seals = memfd_file_seals_ptr(file);
> > +		*file_seals &= F_SEAL_SEAL | F_SEAL_INACCESSIBLE;
> 
> This looks backwards - the flags should be set on *file_seals, but here
> you are unsetting all other flags.

Thanks Steve. '|=' actually should be used here.

Chao
> 
> Steve
> 
> > +	}
> > +
> >  	fd_install(fd, file);
> >  	kfree(name);
> >  	return fd;
> >

On 1/18/22 14:21, Chao Peng wrote:
> Introduce a new memfd_create() flag indicating the content of the
> created memfd is inaccessible from userspace. It does this by force
> setting F_SEAL_INACCESSIBLE seal when the file is created. It also set
> F_SEAL_SEAL to prevent future sealing, which means, it can not coexist
> with MFD_ALLOW_SEALING.
> 
> The pages backed by such memfd will be used as guest private memory in
> confidential computing environments such as Intel TDX/AMD SEV. Since
> page migration/swapping is not yet supported for such usages so these
> pages are currently marked as UNMOVABLE and UNEVICTABLE which makes
> them behave like long-term pinned pages.

Shouldn't the amount of such memory allocations be restricted? E.g. similar
to secretmem_mmap() doing mlock_future_check().

> Signed-off-by: Chao Peng <chao.p.peng@linux.intel.com>
> ---
>  include/uapi/linux/memfd.h |  1 +
>  mm/memfd.c                 | 20 +++++++++++++++++++-
>  2 files changed, 20 insertions(+), 1 deletion(-)
> 
> diff --git a/include/uapi/linux/memfd.h b/include/uapi/linux/memfd.h
> index 7a8a26751c23..48750474b904 100644
> --- a/include/uapi/linux/memfd.h
> +++ b/include/uapi/linux/memfd.h
> @@ -8,6 +8,7 @@
>  #define MFD_CLOEXEC		0x0001U
>  #define MFD_ALLOW_SEALING	0x0002U
>  #define MFD_HUGETLB		0x0004U
> +#define MFD_INACCESSIBLE	0x0008U
>  
>  /*
>   * Huge page size encoding when MFD_HUGETLB is specified, and a huge page
> diff --git a/mm/memfd.c b/mm/memfd.c
> index 9f80f162791a..26998d96dc11 100644
> --- a/mm/memfd.c
> +++ b/mm/memfd.c
> @@ -245,16 +245,19 @@ long memfd_fcntl(struct file *file, unsigned int cmd, unsigned long arg)
>  #define MFD_NAME_PREFIX_LEN (sizeof(MFD_NAME_PREFIX) - 1)
>  #define MFD_NAME_MAX_LEN (NAME_MAX - MFD_NAME_PREFIX_LEN)
>  
> -#define MFD_ALL_FLAGS (MFD_CLOEXEC | MFD_ALLOW_SEALING | MFD_HUGETLB)
> +#define MFD_ALL_FLAGS (MFD_CLOEXEC | MFD_ALLOW_SEALING | MFD_HUGETLB | \
> +		       MFD_INACCESSIBLE)
>  
>  SYSCALL_DEFINE2(memfd_create,
>  		const char __user *, uname,
>  		unsigned int, flags)
>  {
> +	struct address_space *mapping;
>  	unsigned int *file_seals;
>  	struct file *file;
>  	int fd, error;
>  	char *name;
> +	gfp_t gfp;
>  	long len;
>  
>  	if (!(flags & MFD_HUGETLB)) {
> @@ -267,6 +270,10 @@ SYSCALL_DEFINE2(memfd_create,
>  			return -EINVAL;
>  	}
>  
> +	/* Disallow sealing when MFD_INACCESSIBLE is set. */
> +	if (flags & MFD_INACCESSIBLE && flags & MFD_ALLOW_SEALING)
> +		return -EINVAL;
> +
>  	/* length includes terminating zero */
>  	len = strnlen_user(uname, MFD_NAME_MAX_LEN + 1);
>  	if (len <= 0)
> @@ -315,6 +322,17 @@ SYSCALL_DEFINE2(memfd_create,
>  		*file_seals &= ~F_SEAL_SEAL;
>  	}
>  
> +	if (flags & MFD_INACCESSIBLE) {
> +		mapping = file_inode(file)->i_mapping;
> +		gfp = mapping_gfp_mask(mapping);
> +		gfp &= ~__GFP_MOVABLE;
> +		mapping_set_gfp_mask(mapping, gfp);
> +		mapping_set_unevictable(mapping);
> +
> +		file_seals = memfd_file_seals_ptr(file);
> +		*file_seals &= F_SEAL_SEAL | F_SEAL_INACCESSIBLE;
> +	}
> +
>  	fd_install(fd, file);
>  	kfree(name);
>  	return fd;

On 07.02.22 19:51, Vlastimil Babka wrote:
> On 1/18/22 14:21, Chao Peng wrote:
>> Introduce a new memfd_create() flag indicating the content of the
>> created memfd is inaccessible from userspace. It does this by force
>> setting F_SEAL_INACCESSIBLE seal when the file is created. It also set
>> F_SEAL_SEAL to prevent future sealing, which means, it can not coexist
>> with MFD_ALLOW_SEALING.
>>
>> The pages backed by such memfd will be used as guest private memory in
>> confidential computing environments such as Intel TDX/AMD SEV. Since
>> page migration/swapping is not yet supported for such usages so these
>> pages are currently marked as UNMOVABLE and UNEVICTABLE which makes
>> them behave like long-term pinned pages.
> 
> Shouldn't the amount of such memory allocations be restricted? E.g. similar
> to secretmem_mmap() doing mlock_future_check().

I've raised this already in the past and Kirill wanted to look into it [1].

We'll most certainly need a way to limit/control the amount of
unswappable + unmovable ("worse than mlock" memory) a user/process can
consume via this mechanism.


[1]
https://lkml.kernel.org/r/20211122135933.arjxpl7wyskkwvwv@box.shutemov.name


-- 
Thanks,

David / dhildenb

On Tue, Feb 08, 2022 at 09:49:35AM +0100, David Hildenbrand wrote:
> On 07.02.22 19:51, Vlastimil Babka wrote:
> > On 1/18/22 14:21, Chao Peng wrote:
> >> Introduce a new memfd_create() flag indicating the content of the
> >> created memfd is inaccessible from userspace. It does this by force
> >> setting F_SEAL_INACCESSIBLE seal when the file is created. It also set
> >> F_SEAL_SEAL to prevent future sealing, which means, it can not coexist
> >> with MFD_ALLOW_SEALING.
> >>
> >> The pages backed by such memfd will be used as guest private memory in
> >> confidential computing environments such as Intel TDX/AMD SEV. Since
> >> page migration/swapping is not yet supported for such usages so these
> >> pages are currently marked as UNMOVABLE and UNEVICTABLE which makes
> >> them behave like long-term pinned pages.
> > 
> > Shouldn't the amount of such memory allocations be restricted? E.g. similar
> > to secretmem_mmap() doing mlock_future_check().

Heh, for me it was easy, I had the VMA :)
 
> I've raised this already in the past and Kirill wanted to look into it [1].
> 
> We'll most certainly need a way to limit/control the amount of
> unswappable + unmovable ("worse than mlock" memory) a user/process can
> consume via this mechanism.

I think the accounting can be handled in notify_fallocate() and
notify_invalidate_page().

> [1] https://lkml.kernel.org/r/20211122135933.arjxpl7wyskkwvwv@box.shutemov.name
>
> -- 
> Thanks,
> 
> David / dhildenb

-- 
Sincerely yours,
Mike.