Re: [PATCH v6 46/76] x86/sev-es: Adjust #VC IST Stack on entering NMI handler
On Mon, Aug 24, 2020 at 10:54:41AM +0200, Joerg Roedel wrote:
> diff --git a/arch/x86/kernel/nmi.c b/arch/x86/kernel/nmi.c
> index 4fc9954a9560..951f098a4bf5 100644
> --- a/arch/x86/kernel/nmi.c
> +++ b/arch/x86/kernel/nmi.c
> @@ -33,6 +33,7 @@
> #include
> #include
> #include
> +#include
>
> #define CREATE_TRACE_POINTS
> #include
> @@ -488,6 +489,9 @@ DEFINE_IDTENTRY_RAW(exc_nmi)
> this_cpu_write(nmi_cr2, read_cr2());
> nmi_restart:
>
> + /* Needs to happen before DR7 is accessed */
... because? Let's explain why.
> + sev_es_ist_enter(regs);
> +
> this_cpu_write(nmi_dr7, local_db_save());
>
> irq_state = idtentry_enter_nmi(regs);
> @@ -501,6 +505,8 @@ DEFINE_IDTENTRY_RAW(exc_nmi)
>
> local_db_restore(this_cpu_read(nmi_dr7));
>
> + sev_es_ist_exit();
> +
> if (unlikely(this_cpu_read(nmi_cr2) != read_cr2()))
> write_cr2(this_cpu_read(nmi_cr2));
> if (this_cpu_dec_return(nmi_state))
> diff --git a/arch/x86/kernel/sev-es.c b/arch/x86/kernel/sev-es.c
> index 64002d86a237..95831d103418 100644
> --- a/arch/x86/kernel/sev-es.c
> +++ b/arch/x86/kernel/sev-es.c
> @@ -52,6 +52,9 @@ struct sev_es_runtime_data {
>
> static DEFINE_PER_CPU(struct sev_es_runtime_data*, runtime_data);
>
> +DEFINE_STATIC_KEY_FALSE(sev_es_enable_key);
> +EXPORT_SYMBOL_GPL(sev_es_enable_key);
So the GPL export is not needed. The key itself can be made static when
you uninline the sev_es_ist_enter/exit accessors as they're called only
in nmi.c but I guess Peter would object to that in high-NMI-load perf
scenarios...
The export looks unneeded tho.
> +
> static void __init sev_es_setup_vc_stacks(int cpu)
> {
> struct sev_es_runtime_data *data;
> @@ -73,6 +76,59 @@ static void __init sev_es_setup_vc_stacks(int cpu)
> cea_set_pte((void *)vaddr, pa, PAGE_KERNEL);
> }
>
> +static __always_inline bool on_vc_stack(unsigned long sp)
> +{
> + return ((sp >= __this_cpu_ist_bot_va(VC)) && (sp <
> __this_cpu_ist_top_va(VC)));
> +}
> +
> +/*
> + * This function handles the case when an NM is raised in the #VC exception
NMI
> + * handler entry code. In this case the IST entry for VC must be adjusted, so
"VC" or "#VC"? Choose one pls.
> + * that any subsequent VC exception will not overwrite the stack contents of
> the
> + * interrupted VC handler.
> + *
> + * The IST entry is adjusted unconditionally so that it can be also be
> + * unconditionally back-adjusted in sev_es_ist_exit(). Otherwise a nested
^
"adjusted back"
> + * sev_es_ist_exit() call may back-adjust the IST entry too early.
Ditto.
> + */
> +void noinstr __sev_es_ist_enter(struct pt_regs *regs)
> +{
> + unsigned long old_ist, new_ist;
> + unsigned long *p;
> +
> + /* Read old IST entry */
> + old_ist = __this_cpu_read(cpu_tss_rw.x86_tss.ist[IST_INDEX_VC]);
> +
> + /* Make room on the IST stack */
> + if (on_vc_stack(regs->sp))
> + new_ist = ALIGN_DOWN(regs->sp, 8) - sizeof(old_ist);
> + else
> + new_ist = old_ist - sizeof(old_ist);
> +
> + /* Store old IST entry */
> + p = (unsigned long *)new_ist;
> + *p = old_ist;
What's wrong with:
*(unsigned long *)new_ist = old_ist;
?
> +
> + /* Set new IST entry */
> + this_cpu_write(cpu_tss_rw.x86_tss.ist[IST_INDEX_VC], new_ist);
> +}
> +
> +void noinstr __sev_es_ist_exit(void)
> +{
> + unsigned long ist;
> + unsigned long *p;
> +
> + /* Read IST entry */
> + ist = __this_cpu_read(cpu_tss_rw.x86_tss.ist[IST_INDEX_VC]);
> +
> + if (WARN_ON(ist == __this_cpu_ist_top_va(VC)))
> + return;
> +
> + /* Read back old IST entry and write it to the TSS */
> + p = (unsigned long *)ist;
> + this_cpu_write(cpu_tss_rw.x86_tss.ist[IST_INDEX_VC], *p);
And
this_cpu_write(cpu_tss_rw.x86_tss.ist[IST_INDEX_VC], *(unsigned long
*)ist);
?
> +}
> +
> /* Needed in vc_early_forward_exception */
> void do_early_exception(struct pt_regs *regs, int trapnr);
>
> @@ -277,6 +333,9 @@ void __init sev_es_init_vc_handling(void)
> if (!sev_es_active())
> return;
>
> + /* Enable SEV-ES special handling */
> + static_branch_enable(_es_enable_key);
> +
> /* Initialize per-cpu GHCB pages */
> for_each_possible_cpu(cpu) {
> sev_es_alloc_runtime_data(cpu);
> diff --git a/arch/x86/kernel/traps.c b/arch/x86/kernel/traps.c
> index 74cfe6eb7ebb..030d882eaad1 100644
> --- a/arch/x86/kernel/traps.c
> +++ b/arch/x86/kernel/traps.c
> @@ -59,6 +59,7 @@
> #include
> #include
> #include
> +#include
>
> #ifdef CONFIG_X86_64
> #include
> @@ -731,6 +732,7 @@ static bool is_sysenter_singlestep(struct pt_regs *regs)
>
> static __always_inline void debug_enter(unsigned long *dr6, unsigned long
> *dr7)
> {
> +
> /*
>* Disable breakpoints during exception handling; recursive
[PATCH v6 46/76] x86/sev-es: Adjust #VC IST Stack on entering NMI handler
From: Joerg Roedel
When an NMI hits in the #VC handler entry code before it switched to
another stack, any subsequent #VC exception in the NMI code-path will
overwrite the interrupted #VC handlers stack.
Make sure this doesn't happen by explicitly adjusting the #VC IST entry
in the NMI handler for the time in can cause #VC exceptions.
Signed-off-by: Joerg Roedel
Link: https://lore.kernel.org/r/20200724160336.5435-46-j...@8bytes.org
---
arch/x86/include/asm/sev-es.h | 19 +++
arch/x86/kernel/nmi.c | 6
arch/x86/kernel/sev-es.c | 59 +++
arch/x86/kernel/traps.c | 2 ++
4 files changed, 86 insertions(+)
diff --git a/arch/x86/include/asm/sev-es.h b/arch/x86/include/asm/sev-es.h
index 824e9e6b067c..2dd19932a60d 100644
--- a/arch/x86/include/asm/sev-es.h
+++ b/arch/x86/include/asm/sev-es.h
@@ -77,4 +77,23 @@ static inline u64 lower_bits(u64 val, unsigned int bits)
extern void vc_no_ghcb(void);
extern bool handle_vc_boot_ghcb(struct pt_regs *regs);
+#ifdef CONFIG_AMD_MEM_ENCRYPT
+extern struct static_key_false sev_es_enable_key;
+extern void __sev_es_ist_enter(struct pt_regs *regs);
+extern void __sev_es_ist_exit(void);
+static __always_inline void sev_es_ist_enter(struct pt_regs *regs)
+{
+ if (static_branch_unlikely(_es_enable_key))
+ __sev_es_ist_enter(regs);
+}
+static __always_inline void sev_es_ist_exit(void)
+{
+ if (static_branch_unlikely(_es_enable_key))
+ __sev_es_ist_exit();
+}
+#else
+static inline void sev_es_ist_enter(struct pt_regs *regs) { }
+static inline void sev_es_ist_exit(void) { }
+#endif
+
#endif
diff --git a/arch/x86/kernel/nmi.c b/arch/x86/kernel/nmi.c
index 4fc9954a9560..951f098a4bf5 100644
--- a/arch/x86/kernel/nmi.c
+++ b/arch/x86/kernel/nmi.c
@@ -33,6 +33,7 @@
#include
#include
#include
+#include
#define CREATE_TRACE_POINTS
#include
@@ -488,6 +489,9 @@ DEFINE_IDTENTRY_RAW(exc_nmi)
this_cpu_write(nmi_cr2, read_cr2());
nmi_restart:
+ /* Needs to happen before DR7 is accessed */
+ sev_es_ist_enter(regs);
+
this_cpu_write(nmi_dr7, local_db_save());
irq_state = idtentry_enter_nmi(regs);
@@ -501,6 +505,8 @@ DEFINE_IDTENTRY_RAW(exc_nmi)
local_db_restore(this_cpu_read(nmi_dr7));
+ sev_es_ist_exit();
+
if (unlikely(this_cpu_read(nmi_cr2) != read_cr2()))
write_cr2(this_cpu_read(nmi_cr2));
if (this_cpu_dec_return(nmi_state))
diff --git a/arch/x86/kernel/sev-es.c b/arch/x86/kernel/sev-es.c
index 64002d86a237..95831d103418 100644
--- a/arch/x86/kernel/sev-es.c
+++ b/arch/x86/kernel/sev-es.c
@@ -52,6 +52,9 @@ struct sev_es_runtime_data {
static DEFINE_PER_CPU(struct sev_es_runtime_data*, runtime_data);
+DEFINE_STATIC_KEY_FALSE(sev_es_enable_key);
+EXPORT_SYMBOL_GPL(sev_es_enable_key);
+
static void __init sev_es_setup_vc_stacks(int cpu)
{
struct sev_es_runtime_data *data;
@@ -73,6 +76,59 @@ static void __init sev_es_setup_vc_stacks(int cpu)
cea_set_pte((void *)vaddr, pa, PAGE_KERNEL);
}
+static __always_inline bool on_vc_stack(unsigned long sp)
+{
+ return ((sp >= __this_cpu_ist_bot_va(VC)) && (sp <
__this_cpu_ist_top_va(VC)));
+}
+
+/*
+ * This function handles the case when an NM is raised in the #VC exception
+ * handler entry code. In this case the IST entry for VC must be adjusted, so
+ * that any subsequent VC exception will not overwrite the stack contents of
the
+ * interrupted VC handler.
+ *
+ * The IST entry is adjusted unconditionally so that it can be also be
+ * unconditionally back-adjusted in sev_es_ist_exit(). Otherwise a nested
+ * sev_es_ist_exit() call may back-adjust the IST entry too early.
+ */
+void noinstr __sev_es_ist_enter(struct pt_regs *regs)
+{
+ unsigned long old_ist, new_ist;
+ unsigned long *p;
+
+ /* Read old IST entry */
+ old_ist = __this_cpu_read(cpu_tss_rw.x86_tss.ist[IST_INDEX_VC]);
+
+ /* Make room on the IST stack */
+ if (on_vc_stack(regs->sp))
+ new_ist = ALIGN_DOWN(regs->sp, 8) - sizeof(old_ist);
+ else
+ new_ist = old_ist - sizeof(old_ist);
+
+ /* Store old IST entry */
+ p = (unsigned long *)new_ist;
+ *p = old_ist;
+
+ /* Set new IST entry */
+ this_cpu_write(cpu_tss_rw.x86_tss.ist[IST_INDEX_VC], new_ist);
+}
+
+void noinstr __sev_es_ist_exit(void)
+{
+ unsigned long ist;
+ unsigned long *p;
+
+ /* Read IST entry */
+ ist = __this_cpu_read(cpu_tss_rw.x86_tss.ist[IST_INDEX_VC]);
+
+ if (WARN_ON(ist == __this_cpu_ist_top_va(VC)))
+ return;
+
+ /* Read back old IST entry and write it to the TSS */
+ p = (unsigned long *)ist;
+ this_cpu_write(cpu_tss_rw.x86_tss.ist[IST_INDEX_VC], *p);
+}
+
/* Needed in vc_early_forward_exception */
void do_early_exception(struct pt_regs *regs, int trapnr);
@@ -277,6 +333,9 @@ void __init
