2023-09-23 16:19:24

by Li, Xin3

Subject: [PATCH v11 35/37] x86/syscall: Split IDT syscall setup code into idt_syscall_init()

Because FRED uses the ring 3 FRED entrypoint for SYSCALL and SYSENTER and
ERETU is the only legit instruction to return to ring 3, there is NO need
to setup SYSCALL and SYSENTER MSRs for FRED, except the IA32_STAR MSR.

Split IDT syscall setup code into idt_syscall_init() to make it easy to
skip syscall setup code when FRED is enabled.

Suggested-by: Thomas Gleixner <[email protected]>
Tested-by: Shan Kang <[email protected]>
Signed-off-by: Xin Li <[email protected]>
---
arch/x86/kernel/cpu/common.c | 13 ++++++++++---
1 file changed, 10 insertions(+), 3 deletions(-)

diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index 20bbedbf6dcb..2ee4e7b597a3 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -2071,10 +2071,8 @@ static void wrmsrl_cstar(unsigned long val)
wrmsrl(MSR_CSTAR, val);
}

-/* May not be marked __init: used by software suspend */
-void syscall_init(void)
+static inline void idt_syscall_init(void)
{
- wrmsr(MSR_STAR, 0, (__USER32_CS << 16) | __KERNEL_CS);
wrmsrl(MSR_LSTAR, (unsigned long)entry_SYSCALL_64);

if (ia32_enabled()) {
@@ -2108,6 +2106,15 @@ void syscall_init(void)
X86_EFLAGS_AC|X86_EFLAGS_ID);
}

+/* May not be marked __init: used by software suspend */
+void syscall_init(void)
+{
+ /* The default user and kernel segments */
+ wrmsr(MSR_STAR, 0, (__USER32_CS << 16) | __KERNEL_CS);
+
+ idt_syscall_init();
+}
+
#else /* CONFIG_X86_64 */

#ifdef CONFIG_STACKPROTECTOR
--
2.34.1

2023-09-25 18:58:49

by Li, Xin3

Subject: RE: [PATCH v11 35/37] x86/syscall: Split IDT syscall setup code into idt_syscall_init()

> >diff --git a/arch/x86/kernel/cpu/common.c
> >b/arch/x86/kernel/cpu/common.c index 20bbedbf6dcb..2ee4e7b597a3 100644
> >--- a/arch/x86/kernel/cpu/common.c
> >+++ b/arch/x86/kernel/cpu/common.c
> >@@ -2071,10 +2071,8 @@ static void wrmsrl_cstar(unsigned long val)
> > wrmsrl(MSR_CSTAR, val);
> > }
> >
> >-/* May not be marked __init: used by software suspend */ -void
> >syscall_init(void)
> >+static inline void idt_syscall_init(void)
> > {
> >- wrmsr(MSR_STAR, 0, (__USER32_CS << 16) | __KERNEL_CS);
> > wrmsrl(MSR_LSTAR, (unsigned long)entry_SYSCALL_64);
> >
> > if (ia32_enabled()) {
> >@@ -2108,6 +2106,15 @@ void syscall_init(void)
> > X86_EFLAGS_AC|X86_EFLAGS_ID);
> > }
> >
> >+/* May not be marked __init: used by software suspend */ void
> >+syscall_init(void) {
> >+ /* The default user and kernel segments */
> >+ wrmsr(MSR_STAR, 0, (__USER32_CS << 16) | __KERNEL_CS);
> >+
> >+ idt_syscall_init();
> >+}
> >+
> > #else /* CONFIG_X86_64 */
> >
> > #ifdef CONFIG_STACKPROTECTOR
>
> Am I missing something, or is this patch a noop?

Yes, this is a noop, just a cleanup patch w/o functionality change.

2023-09-26 00:49:39

by H. Peter Anvin

Subject: Re: [PATCH v11 35/37] x86/syscall: Split IDT syscall setup code into idt_syscall_init()

On September 23, 2023 2:42:10 AM PDT, Xin Li <[email protected]> wrote:
>Because FRED uses the ring 3 FRED entrypoint for SYSCALL and SYSENTER and
>ERETU is the only legit instruction to return to ring 3, there is NO need
>to setup SYSCALL and SYSENTER MSRs for FRED, except the IA32_STAR MSR.
>
>Split IDT syscall setup code into idt_syscall_init() to make it easy to
>skip syscall setup code when FRED is enabled.
>
>Suggested-by: Thomas Gleixner <[email protected]>
>Tested-by: Shan Kang <[email protected]>
>Signed-off-by: Xin Li <[email protected]>
>---
> arch/x86/kernel/cpu/common.c | 13 ++++++++++---
> 1 file changed, 10 insertions(+), 3 deletions(-)
>
>diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
>index 20bbedbf6dcb..2ee4e7b597a3 100644
>--- a/arch/x86/kernel/cpu/common.c
>+++ b/arch/x86/kernel/cpu/common.c
>@@ -2071,10 +2071,8 @@ static void wrmsrl_cstar(unsigned long val)
> wrmsrl(MSR_CSTAR, val);
> }
>
>-/* May not be marked __init: used by software suspend */
>-void syscall_init(void)
>+static inline void idt_syscall_init(void)
> {
>- wrmsr(MSR_STAR, 0, (__USER32_CS << 16) | __KERNEL_CS);
> wrmsrl(MSR_LSTAR, (unsigned long)entry_SYSCALL_64);
>
> if (ia32_enabled()) {
>@@ -2108,6 +2106,15 @@ void syscall_init(void)
> X86_EFLAGS_AC|X86_EFLAGS_ID);
> }
>
>+/* May not be marked __init: used by software suspend */
>+void syscall_init(void)
>+{
>+ /* The default user and kernel segments */
>+ wrmsr(MSR_STAR, 0, (__USER32_CS << 16) | __KERNEL_CS);
>+
>+ idt_syscall_init();
>+}
>+
> #else /* CONFIG_X86_64 */
>
> #ifdef CONFIG_STACKPROTECTOR

Am I missing something, or is this patch a noop?

2023-09-26 06:56:33

by H. Peter Anvin

Subject: RE: [PATCH v11 35/37] x86/syscall: Split IDT syscall setup code into idt_syscall_init()

On September 25, 2023 10:56:44 AM PDT, "Li, Xin3" <[email protected]> wrote:
>> >diff --git a/arch/x86/kernel/cpu/common.c
>> >b/arch/x86/kernel/cpu/common.c index 20bbedbf6dcb..2ee4e7b597a3 100644
>> >--- a/arch/x86/kernel/cpu/common.c
>> >+++ b/arch/x86/kernel/cpu/common.c
>> >@@ -2071,10 +2071,8 @@ static void wrmsrl_cstar(unsigned long val)
>> > wrmsrl(MSR_CSTAR, val);
>> > }
>> >
>> >-/* May not be marked __init: used by software suspend */ -void
>> >syscall_init(void)
>> >+static inline void idt_syscall_init(void)
>> > {
>> >- wrmsr(MSR_STAR, 0, (__USER32_CS << 16) | __KERNEL_CS);
>> > wrmsrl(MSR_LSTAR, (unsigned long)entry_SYSCALL_64);
>> >
>> > if (ia32_enabled()) {
>> >@@ -2108,6 +2106,15 @@ void syscall_init(void)
>> > X86_EFLAGS_AC|X86_EFLAGS_ID);
>> > }
>> >
>> >+/* May not be marked __init: used by software suspend */ void
>> >+syscall_init(void) {
>> >+ /* The default user and kernel segments */
>> >+ wrmsr(MSR_STAR, 0, (__USER32_CS << 16) | __KERNEL_CS);
>> >+
>> >+ idt_syscall_init();
>> >+}
>> >+
>> > #else /* CONFIG_X86_64 */
>> >
>> > #ifdef CONFIG_STACKPROTECTOR
>>
>> Am I missing something, or is this patch a noop?
>
>Yes, this is a noop, just a cleanup patch w/o functionality change.
>
>

It just seems to be completely redundant. We can just drop it, no? If we aren't going to explicitly clobber the registers there is no harm in setting them up for IDT unconditionally.

2023-09-26 14:27:29

by Li, Xin3

Subject: RE: [PATCH v11 35/37] x86/syscall: Split IDT syscall setup code into idt_syscall_init()

> >Yes, this is a noop, just a cleanup patch w/o functionality change.
> >
> It just seems to be completely redundant. We can just drop it, no? If we aren't going
> to explicitly clobber the registers there is no harm in setting them up for IDT
> unconditionally.

If no objection, I can make the change, i.e., unconditionally set these
MSRs up for IDT.

2023-09-29 01:36:57

by Thomas Gleixner

Subject: Re: [PATCH v11 35/37] x86/syscall: Split IDT syscall setup code into idt_syscall_init()

On Mon, Sep 25 2023 at 09:07, H. Peter Anvin wrote:
> On September 23, 2023 2:42:10 AM PDT, Xin Li <[email protected]> wrote:
>>+/* May not be marked __init: used by software suspend */
>>+void syscall_init(void)
>>+{
>>+ /* The default user and kernel segments */
>>+ wrmsr(MSR_STAR, 0, (__USER32_CS << 16) | __KERNEL_CS);
>>+
>>+ idt_syscall_init();
>>+}
>>+
>> #else /* CONFIG_X86_64 */
>>
>> #ifdef CONFIG_STACKPROTECTOR
>
> Am I missing something, or is this patch a noop?

Yes. It's a noop at this point. Later on it gains a

if (!fred)
idt_syscall_init();

Sure we could do

if (!fred) {
write_msr(foo...);
...
}

too, but I prefer the separation. No strong opinion though.

Thanks,

tglx