[v1] x86/sgx: use vm_flags_t for vm_prot_bits

[PATCH] x86/sgx: use vm_flags_t for vm_prot_bits

Posted by Wentong Tian 2 weeks, 2 days ago

The vm_flags_t type is the dedicated type for virtual memory flags.
Architecture and driver code should use this type instead of assuming
vm_flags is an unsigned long, as the underlying type may change in the
future.

This follows the cleanup in commit d75fa3c94750 ("mm: update
architecture and driver code to use vm_flags_t") by converting the
remaining vm_prot_bits usage in SGX code.

No functional change intended.

Signed-off-by: Wentong Tian <tianwentong2000@gmail.com>
---
 arch/x86/kernel/cpu/sgx/encl.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/x86/kernel/cpu/sgx/encl.c b/arch/x86/kernel/cpu/sgx/encl.c
index cf149b9f4916..814edcde225d 100644
--- a/arch/x86/kernel/cpu/sgx/encl.c
+++ b/arch/x86/kernel/cpu/sgx/encl.c
@@ -281,7 +281,7 @@ static struct sgx_encl_page *sgx_encl_load_page_in_vma(struct sgx_encl *encl,
 						       unsigned long addr,
 						       vm_flags_t vm_flags)
 {
-	unsigned long vm_prot_bits = vm_flags & VM_ACCESS_FLAGS;
+	vm_flags_t vm_prot_bits = vm_flags & VM_ACCESS_FLAGS;
 	struct sgx_encl_page *entry;
 
 	entry = xa_load(&encl->page_array, PFN_DOWN(addr));
-- 
2.34.1

Re: [PATCH] x86/sgx: use vm_flags_t for vm_prot_bits

Posted by Jarkko Sakkinen 1 week, 6 days ago

On Thu, Jan 22, 2026 at 01:54:35PM +0800, Wentong Tian wrote:
> The vm_flags_t type is the dedicated type for virtual memory flags.
> Architecture and driver code should use this type instead of assuming
> vm_flags is an unsigned long, as the underlying type may change in the
> future.
> 
> This follows the cleanup in commit d75fa3c94750 ("mm: update
> architecture and driver code to use vm_flags_t") by converting the
> remaining vm_prot_bits usage in SGX code.
> 
> No functional change intended.
> 
> Signed-off-by: Wentong Tian <tianwentong2000@gmail.com>
> ---
>  arch/x86/kernel/cpu/sgx/encl.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/arch/x86/kernel/cpu/sgx/encl.c b/arch/x86/kernel/cpu/sgx/encl.c
> index cf149b9f4916..814edcde225d 100644
> --- a/arch/x86/kernel/cpu/sgx/encl.c
> +++ b/arch/x86/kernel/cpu/sgx/encl.c
> @@ -281,7 +281,7 @@ static struct sgx_encl_page *sgx_encl_load_page_in_vma(struct sgx_encl *encl,
>  						       unsigned long addr,
>  						       vm_flags_t vm_flags)
>  {
> -	unsigned long vm_prot_bits = vm_flags & VM_ACCESS_FLAGS;
> +	vm_flags_t vm_prot_bits = vm_flags & VM_ACCESS_FLAGS;
>  	struct sgx_encl_page *entry;
>  
>  	entry = xa_load(&encl->page_array, PFN_DOWN(addr));
> -- 
> 2.34.1
> 

Reviewed-by: Jarkko Sakkinen <jarkko@kernel.org>

BR, Jarkko

Re: [PATCH] x86/sgx: use vm_flags_t for vm_prot_bits

Posted by Huang, Kai 2 weeks, 2 days ago

On Thu, 2026-01-22 at 13:54 +0800, Wentong Tian wrote:
> The vm_flags_t type is the dedicated type for virtual memory flags.
> Architecture and driver code should use this type instead of assuming
> vm_flags is an unsigned long, as the underlying type may change in the
> future.
> 
> This follows the cleanup in commit d75fa3c94750 ("mm: update
> architecture and driver code to use vm_flags_t") by converting the
> remaining vm_prot_bits usage in SGX code.
> 
> No functional change intended.
> 
> Signed-off-by: Wentong Tian <tianwentong2000@gmail.com>
> ---
>  arch/x86/kernel/cpu/sgx/encl.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/arch/x86/kernel/cpu/sgx/encl.c b/arch/x86/kernel/cpu/sgx/encl.c
> index cf149b9f4916..814edcde225d 100644
> --- a/arch/x86/kernel/cpu/sgx/encl.c
> +++ b/arch/x86/kernel/cpu/sgx/encl.c
> @@ -281,7 +281,7 @@ static struct sgx_encl_page *sgx_encl_load_page_in_vma(struct sgx_encl *encl,
>  						       unsigned long addr,
>  						       vm_flags_t vm_flags)
>  {
> -	unsigned long vm_prot_bits = vm_flags & VM_ACCESS_FLAGS;
> +	vm_flags_t vm_prot_bits = vm_flags & VM_ACCESS_FLAGS;
>  	struct sgx_encl_page *entry;
>  
>  	entry = xa_load(&encl->page_array, PFN_DOWN(addr));

Since commit d75fa3c94750 ("mm: update
architecture and driver code to use vm_flags_t") already converted one
'vm_prot_bits' to vm_flags_t in sgx_encl_may_map(), this looks good to me
too.

One thing though:

'vm_prot_bits' is later bit-ANDed with entry->vm_max_prot_bits:

	if ((entry->vm_max_prot_bits & vm_prot_bits) != vm_prot_bits)
                return ERR_PTR(-EFAULT);

But 'vm_max_prot_bits' a 8-bit field of 'unsigned long' in 'struct
sgx_encl_page':

	struct sgx_encl_page {
        	unsigned long desc;
       		unsigned long vm_max_prot_bits:8;
		...
	};

Not sure we should change that too:

	struct sgx_encl_page {
		unsigned long desc;
		vm_flags_t vm_max_prot_bits:8;
		...
	};

?

Re: [PATCH] x86/sgx: use vm_flags_t for vm_prot_bits

Posted by TwT pupupu 2 weeks, 1 day ago

On Wed, Jan 22, 2026 at 7:42 PM Huang, Kai <kai.huang@intel.com> wrote:
> Not sure we should change that too:
>
> struct sgx_encl_page {
> unsigned long desc;
> vm_flags_t vm_max_prot_bits:8;
> ...
> };
>
> ?
Hi, I'll include this in v2 for consistency. The :8
width is unchanged, so no functional or storage difference.

Huang, Kai <kai.huang@intel.com> 於 2026年1月22日週四 下午7:42寫道：
>
> On Thu, 2026-01-22 at 13:54 +0800, Wentong Tian wrote:
> > The vm_flags_t type is the dedicated type for virtual memory flags.
> > Architecture and driver code should use this type instead of assuming
> > vm_flags is an unsigned long, as the underlying type may change in the
> > future.
> >
> > This follows the cleanup in commit d75fa3c94750 ("mm: update
> > architecture and driver code to use vm_flags_t") by converting the
> > remaining vm_prot_bits usage in SGX code.
> >
> > No functional change intended.
> >
> > Signed-off-by: Wentong Tian <tianwentong2000@gmail.com>
> > ---
> >  arch/x86/kernel/cpu/sgx/encl.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/arch/x86/kernel/cpu/sgx/encl.c b/arch/x86/kernel/cpu/sgx/encl.c
> > index cf149b9f4916..814edcde225d 100644
> > --- a/arch/x86/kernel/cpu/sgx/encl.c
> > +++ b/arch/x86/kernel/cpu/sgx/encl.c
> > @@ -281,7 +281,7 @@ static struct sgx_encl_page *sgx_encl_load_page_in_vma(struct sgx_encl *encl,
> >                                                      unsigned long addr,
> >                                                      vm_flags_t vm_flags)
> >  {
> > -     unsigned long vm_prot_bits = vm_flags & VM_ACCESS_FLAGS;
> > +     vm_flags_t vm_prot_bits = vm_flags & VM_ACCESS_FLAGS;
> >       struct sgx_encl_page *entry;
> >
> >       entry = xa_load(&encl->page_array, PFN_DOWN(addr));
>
> Since commit d75fa3c94750 ("mm: update
> architecture and driver code to use vm_flags_t") already converted one
> 'vm_prot_bits' to vm_flags_t in sgx_encl_may_map(), this looks good to me
> too.
>
> One thing though:
>
> 'vm_prot_bits' is later bit-ANDed with entry->vm_max_prot_bits:
>
>         if ((entry->vm_max_prot_bits & vm_prot_bits) != vm_prot_bits)
>                 return ERR_PTR(-EFAULT);
>
> But 'vm_max_prot_bits' a 8-bit field of 'unsigned long' in 'struct
> sgx_encl_page':
>
>         struct sgx_encl_page {
>                 unsigned long desc;
>                 unsigned long vm_max_prot_bits:8;
>                 ...
>         };
>
> Not sure we should change that too:
>
>         struct sgx_encl_page {
>                 unsigned long desc;
>                 vm_flags_t vm_max_prot_bits:8;
>                 ...
>         };
>
> ?