2023-11-20 15:36:27

by Uros Bizjak

[permalink] [raw]
Subject: [PATCH] x86/mm/encrypt: Use %a asm operand modifier to obtain %rip-relative address

The "a" asm operand modifier substitutes a memory reference, with the
actual operand treated as address. For x86_64, when a symbol is
provided, the "a" modifier emits "sym(%rip)" instead of "sym".

Clang allows only "i" and "r" operand constraints with an "a" modifier,
so the patch normalizes the modifier/constraint pair to "a"/"i"
which is consistent between both compilers.

No functional change intended.

Suggested-by: Linus Torvalds <[email protected]>
Cc: Dave Hansen <[email protected]>
Cc: Andy Lutomirski <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: "H. Peter Anvin" <[email protected]>
Signed-off-by: Uros Bizjak <[email protected]>
---
arch/x86/mm/mem_encrypt_identity.c | 16 ++++------------
1 file changed, 4 insertions(+), 12 deletions(-)

diff --git a/arch/x86/mm/mem_encrypt_identity.c b/arch/x86/mm/mem_encrypt_identity.c
index d73aeb16417f..6a351fdd1b39 100644
--- a/arch/x86/mm/mem_encrypt_identity.c
+++ b/arch/x86/mm/mem_encrypt_identity.c
@@ -346,9 +346,7 @@ void __init sme_encrypt_kernel(struct boot_params *bp)
* We're running identity mapped, so we must obtain the address to the
* SME encryption workarea using rip-relative addressing.
*/
- asm ("lea sme_workarea(%%rip), %0"
- : "=r" (workarea_start)
- : "p" (sme_workarea));
+ asm ("lea %a1, %0" : "=r" (workarea_start) : "i" (sme_workarea));

/*
* Calculate required number of workarea bytes needed:
@@ -582,15 +580,9 @@ void __init sme_enable(struct boot_params *bp)
* identity mapped, so we must obtain the address to the SME command
* line argument data using rip-relative addressing.
*/
- asm ("lea sme_cmdline_arg(%%rip), %0"
- : "=r" (cmdline_arg)
- : "p" (sme_cmdline_arg));
- asm ("lea sme_cmdline_on(%%rip), %0"
- : "=r" (cmdline_on)
- : "p" (sme_cmdline_on));
- asm ("lea sme_cmdline_off(%%rip), %0"
- : "=r" (cmdline_off)
- : "p" (sme_cmdline_off));
+ asm ("lea %a1, %0" : "=r" (cmdline_arg) : "i" (sme_cmdline_arg));
+ asm ("lea %a1, %0" : "=r" (cmdline_on) : "i" (sme_cmdline_on));
+ asm ("lea %a1, %0" : "=r" (cmdline_off) : "i" (sme_cmdline_off));

if (IS_ENABLED(CONFIG_AMD_MEM_ENCRYPT_ACTIVE_BY_DEFAULT))
active_by_default = true;
--
2.41.0


2023-11-20 17:16:49

by Borislav Petkov

[permalink] [raw]
Subject: Re: [PATCH] x86/mm/encrypt: Use %a asm operand modifier to obtain %rip-relative address

On Mon, Nov 20, 2023 at 04:33:50PM +0100, Uros Bizjak wrote:
> The "a" asm operand modifier substitutes a memory reference, with the
> actual operand treated as address. For x86_64, when a symbol is
> provided, the "a" modifier emits "sym(%rip)" instead of "sym".
>
> Clang allows only "i" and "r" operand constraints with an "a" modifier,
> so the patch normalizes the modifier/constraint pair to "a"/"i"

s/the patch normalizes/normalize/

> which is consistent between both compilers.
>
> No functional change intended.
>
> Suggested-by: Linus Torvalds <[email protected]>
> Cc: Dave Hansen <[email protected]>
> Cc: Andy Lutomirski <[email protected]>
> Cc: Peter Zijlstra <[email protected]>
> Cc: Thomas Gleixner <[email protected]>
> Cc: Ingo Molnar <[email protected]>
> Cc: Borislav Petkov <[email protected]>
> Cc: "H. Peter Anvin" <[email protected]>
> Signed-off-by: Uros Bizjak <[email protected]>
> ---
> arch/x86/mm/mem_encrypt_identity.c | 16 ++++------------
> 1 file changed, 4 insertions(+), 12 deletions(-)
>
> diff --git a/arch/x86/mm/mem_encrypt_identity.c b/arch/x86/mm/mem_encrypt_identity.c
> index d73aeb16417f..6a351fdd1b39 100644
> --- a/arch/x86/mm/mem_encrypt_identity.c
> +++ b/arch/x86/mm/mem_encrypt_identity.c
> @@ -346,9 +346,7 @@ void __init sme_encrypt_kernel(struct boot_params *bp)
> * We're running identity mapped, so we must obtain the address to the
> * SME encryption workarea using rip-relative addressing.
> */
> - asm ("lea sme_workarea(%%rip), %0"
> - : "=r" (workarea_start)
> - : "p" (sme_workarea));
> + asm ("lea %a1, %0" : "=r" (workarea_start) : "i" (sme_workarea));

Yeah, I saw that particular subthread today.

Are you sure this "a" modifier DTRT with all gcc version we support?

I.e., from 5.1 onwards...

Just making sure.

Thx.

--
Regards/Gruss,
Boris.

https://people.kernel.org/tglx/notes-about-netiquette

2023-11-20 19:59:15

by Uros Bizjak

[permalink] [raw]
Subject: Re: [PATCH] x86/mm/encrypt: Use %a asm operand modifier to obtain %rip-relative address

On Mon, Nov 20, 2023 at 6:15 PM Borislav Petkov <[email protected]> wrote:
>
> On Mon, Nov 20, 2023 at 04:33:50PM +0100, Uros Bizjak wrote:
> > The "a" asm operand modifier substitutes a memory reference, with the
> > actual operand treated as address. For x86_64, when a symbol is
> > provided, the "a" modifier emits "sym(%rip)" instead of "sym".
> >
> > Clang allows only "i" and "r" operand constraints with an "a" modifier,
> > so the patch normalizes the modifier/constraint pair to "a"/"i"
>
> s/the patch normalizes/normalize/
>
> > which is consistent between both compilers.
> >
> > No functional change intended.
> >
> > Suggested-by: Linus Torvalds <[email protected]>
> > Cc: Dave Hansen <[email protected]>
> > Cc: Andy Lutomirski <[email protected]>
> > Cc: Peter Zijlstra <[email protected]>
> > Cc: Thomas Gleixner <[email protected]>
> > Cc: Ingo Molnar <[email protected]>
> > Cc: Borislav Petkov <[email protected]>
> > Cc: "H. Peter Anvin" <[email protected]>
> > Signed-off-by: Uros Bizjak <[email protected]>
> > ---
> > arch/x86/mm/mem_encrypt_identity.c | 16 ++++------------
> > 1 file changed, 4 insertions(+), 12 deletions(-)
> >
> > diff --git a/arch/x86/mm/mem_encrypt_identity.c b/arch/x86/mm/mem_encrypt_identity.c
> > index d73aeb16417f..6a351fdd1b39 100644
> > --- a/arch/x86/mm/mem_encrypt_identity.c
> > +++ b/arch/x86/mm/mem_encrypt_identity.c
> > @@ -346,9 +346,7 @@ void __init sme_encrypt_kernel(struct boot_params *bp)
> > * We're running identity mapped, so we must obtain the address to the
> > * SME encryption workarea using rip-relative addressing.
> > */
> > - asm ("lea sme_workarea(%%rip), %0"
> > - : "=r" (workarea_start)
> > - : "p" (sme_workarea));
> > + asm ("lea %a1, %0" : "=r" (workarea_start) : "i" (sme_workarea));
>
> Yeah, I saw that particular subthread today.
>
> Are you sure this "a" modifier DTRT with all gcc version we support?
>
> I.e., from 5.1 onwards...

Yes [1].

[1] https://godbolt.org/z/aWe7b16rT

Thanks,
Uros.

2023-11-21 09:01:13

by Kirill A. Shutemov

[permalink] [raw]
Subject: Re: [PATCH] x86/mm/encrypt: Use %a asm operand modifier to obtain %rip-relative address

On Mon, Nov 20, 2023 at 08:58:29PM +0100, Uros Bizjak wrote:
> On Mon, Nov 20, 2023 at 6:15 PM Borislav Petkov <[email protected]> wrote:
> >
> > On Mon, Nov 20, 2023 at 04:33:50PM +0100, Uros Bizjak wrote:
> > > The "a" asm operand modifier substitutes a memory reference, with the
> > > actual operand treated as address. For x86_64, when a symbol is
> > > provided, the "a" modifier emits "sym(%rip)" instead of "sym".
> > >
> > > Clang allows only "i" and "r" operand constraints with an "a" modifier,
> > > so the patch normalizes the modifier/constraint pair to "a"/"i"
> >
> > s/the patch normalizes/normalize/
> >
> > > which is consistent between both compilers.
> > >
> > > No functional change intended.
> > >
> > > Suggested-by: Linus Torvalds <[email protected]>
> > > Cc: Dave Hansen <[email protected]>
> > > Cc: Andy Lutomirski <[email protected]>
> > > Cc: Peter Zijlstra <[email protected]>
> > > Cc: Thomas Gleixner <[email protected]>
> > > Cc: Ingo Molnar <[email protected]>
> > > Cc: Borislav Petkov <[email protected]>
> > > Cc: "H. Peter Anvin" <[email protected]>
> > > Signed-off-by: Uros Bizjak <[email protected]>
> > > ---
> > > arch/x86/mm/mem_encrypt_identity.c | 16 ++++------------
> > > 1 file changed, 4 insertions(+), 12 deletions(-)
> > >
> > > diff --git a/arch/x86/mm/mem_encrypt_identity.c b/arch/x86/mm/mem_encrypt_identity.c
> > > index d73aeb16417f..6a351fdd1b39 100644
> > > --- a/arch/x86/mm/mem_encrypt_identity.c
> > > +++ b/arch/x86/mm/mem_encrypt_identity.c
> > > @@ -346,9 +346,7 @@ void __init sme_encrypt_kernel(struct boot_params *bp)
> > > * We're running identity mapped, so we must obtain the address to the
> > > * SME encryption workarea using rip-relative addressing.
> > > */
> > > - asm ("lea sme_workarea(%%rip), %0"
> > > - : "=r" (workarea_start)
> > > - : "p" (sme_workarea));
> > > + asm ("lea %a1, %0" : "=r" (workarea_start) : "i" (sme_workarea));
> >
> > Yeah, I saw that particular subthread today.
> >
> > Are you sure this "a" modifier DTRT with all gcc version we support?
> >
> > I.e., from 5.1 onwards...
>
> Yes [1].
>
> [1] https://godbolt.org/z/aWe7b16rT

But Clang generates RIP-relative load only starting from version 15.
Clange 11 claimed to be supported.

--
Kiryl Shutsemau / Kirill A. Shutemov

2023-11-21 09:30:04

by Uros Bizjak

[permalink] [raw]
Subject: Re: [PATCH] x86/mm/encrypt: Use %a asm operand modifier to obtain %rip-relative address

On Tue, Nov 21, 2023 at 10:05 AM Kirill A. Shutemov
<[email protected]> wrote:
>
> On Mon, Nov 20, 2023 at 08:58:29PM +0100, Uros Bizjak wrote:
> > On Mon, Nov 20, 2023 at 6:15 PM Borislav Petkov <[email protected]> wrote:
> > >
> > > On Mon, Nov 20, 2023 at 04:33:50PM +0100, Uros Bizjak wrote:
> > > > The "a" asm operand modifier substitutes a memory reference, with the
> > > > actual operand treated as address. For x86_64, when a symbol is
> > > > provided, the "a" modifier emits "sym(%rip)" instead of "sym".
> > > >
> > > > Clang allows only "i" and "r" operand constraints with an "a" modifier,
> > > > so the patch normalizes the modifier/constraint pair to "a"/"i"
> > >
> > > s/the patch normalizes/normalize/
> > >
> > > > which is consistent between both compilers.
> > > >
> > > > No functional change intended.
> > > >
> > > > Suggested-by: Linus Torvalds <[email protected]>
> > > > Cc: Dave Hansen <[email protected]>
> > > > Cc: Andy Lutomirski <[email protected]>
> > > > Cc: Peter Zijlstra <[email protected]>
> > > > Cc: Thomas Gleixner <[email protected]>
> > > > Cc: Ingo Molnar <[email protected]>
> > > > Cc: Borislav Petkov <[email protected]>
> > > > Cc: "H. Peter Anvin" <[email protected]>
> > > > Signed-off-by: Uros Bizjak <[email protected]>
> > > > ---
> > > > arch/x86/mm/mem_encrypt_identity.c | 16 ++++------------
> > > > 1 file changed, 4 insertions(+), 12 deletions(-)
> > > >
> > > > diff --git a/arch/x86/mm/mem_encrypt_identity.c b/arch/x86/mm/mem_encrypt_identity.c
> > > > index d73aeb16417f..6a351fdd1b39 100644
> > > > --- a/arch/x86/mm/mem_encrypt_identity.c
> > > > +++ b/arch/x86/mm/mem_encrypt_identity.c
> > > > @@ -346,9 +346,7 @@ void __init sme_encrypt_kernel(struct boot_params *bp)
> > > > * We're running identity mapped, so we must obtain the address to the
> > > > * SME encryption workarea using rip-relative addressing.
> > > > */
> > > > - asm ("lea sme_workarea(%%rip), %0"
> > > > - : "=r" (workarea_start)
> > > > - : "p" (sme_workarea));
> > > > + asm ("lea %a1, %0" : "=r" (workarea_start) : "i" (sme_workarea));
> > >
> > > Yeah, I saw that particular subthread today.
> > >
> > > Are you sure this "a" modifier DTRT with all gcc version we support?
> > >
> > > I.e., from 5.1 onwards...
> >
> > Yes [1].
> >
> > [1] https://godbolt.org/z/aWe7b16rT
>
> But Clang generates RIP-relative load only starting from version 15.
> Clange 11 claimed to be supported.

Huh, what a bummer...

The patch is retracted then.

Thanks,
Uros.