Upon a wakeup from S4, the restore kernel starts and initializes the FRED MSRs as needed from its perspective. It then loads a hibernation image, including the image kernel, and attempts to load image pages directly into their original page frames used before hibernation unless those frames are currently in use. Once all pages are moved to their original locations, it jumps to a "trampoline" page in the image kernel.
At this point, the image kernel takes control, but the FRED MSRs still contain values set by the restore kernel, which may differ from those set by the image kernel before hibernation. Therefore, the image kernel must ensure the FRED MSRs have the same values as before hibernation. Since these values depend only on the location of the kernel text and data, they can be recomputed from scratch.
Reported-by: Xi Pardee xi.pardee@intel.com Reported-by: Todd Brandt todd.e.brandt@intel.com Suggested-by: H. Peter Anvin (Intel) hpa@zytor.com Signed-off-by: Xin Li (Intel) xin@zytor.com Tested-by: Todd Brandt todd.e.brandt@intel.com Cc: Andy Lutomirski luto@kernel.org Cc: Brian Gerst brgerst@gmail.com Cc: Juergen Gross jgross@suse.com Cc: Linus Torvalds torvalds@linux-foundation.org Cc: stable@vger.kernel.org # 6.9+ ---
Change in v2: * Rewrite the change log and in-code comments based on Rafael's feedback. --- arch/x86/power/cpu.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+)
diff --git a/arch/x86/power/cpu.c b/arch/x86/power/cpu.c index 63230ff8cf4f..08e76a5ca155 100644 --- a/arch/x86/power/cpu.c +++ b/arch/x86/power/cpu.c @@ -27,6 +27,7 @@ #include <asm/mmu_context.h> #include <asm/cpu_device_id.h> #include <asm/microcode.h> +#include <asm/fred.h>
#ifdef CONFIG_X86_32 __visible unsigned long saved_context_ebx; @@ -231,6 +232,19 @@ static void notrace __restore_processor_state(struct saved_context *ctxt) */ #ifdef CONFIG_X86_64 wrmsrl(MSR_GS_BASE, ctxt->kernelmode_gs_base); + + /* + * Reinitialize FRED to ensure the FRED MSRs contain the same values + * as before hibernation. + * + * Note, the setup of FRED RSPs requires access to percpu data + * structures. Therefore, FRED reinitialization can only occur after + * the percpu access pointer (i.e., MSR_GS_BASE) is restored. + */ + if (ctxt->cr4 & X86_CR4_FRED) { + cpu_init_fred_exceptions(); + cpu_init_fred_rsps(); + } #else loadsegment(fs, __KERNEL_PERCPU); #endif
base-commit: 535bd326c5657fe570f41b1f76941e449d9e2062
On Tue, Apr 1, 2025 at 9:57 AM Xin Li (Intel) xin@zytor.com wrote:
Upon a wakeup from S4, the restore kernel starts and initializes the FRED MSRs as needed from its perspective. It then loads a hibernation image, including the image kernel, and attempts to load image pages directly into their original page frames used before hibernation unless those frames are currently in use. Once all pages are moved to their original locations, it jumps to a "trampoline" page in the image kernel.
At this point, the image kernel takes control, but the FRED MSRs still contain values set by the restore kernel, which may differ from those set by the image kernel before hibernation. Therefore, the image kernel must ensure the FRED MSRs have the same values as before hibernation. Since these values depend only on the location of the kernel text and data, they can be recomputed from scratch.
Reported-by: Xi Pardee xi.pardee@intel.com Reported-by: Todd Brandt todd.e.brandt@intel.com Suggested-by: H. Peter Anvin (Intel) hpa@zytor.com Signed-off-by: Xin Li (Intel) xin@zytor.com Tested-by: Todd Brandt todd.e.brandt@intel.com Cc: Andy Lutomirski luto@kernel.org Cc: Brian Gerst brgerst@gmail.com Cc: Juergen Gross jgross@suse.com Cc: Linus Torvalds torvalds@linux-foundation.org Cc: stable@vger.kernel.org # 6.9+
Reviewed-by: Rafael J. Wysocki rafael.j.wysocki@intel.com
Change in v2:
- Rewrite the change log and in-code comments based on Rafael's feedback.
Thanks!
arch/x86/power/cpu.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+)
diff --git a/arch/x86/power/cpu.c b/arch/x86/power/cpu.c index 63230ff8cf4f..08e76a5ca155 100644 --- a/arch/x86/power/cpu.c +++ b/arch/x86/power/cpu.c @@ -27,6 +27,7 @@ #include <asm/mmu_context.h> #include <asm/cpu_device_id.h> #include <asm/microcode.h> +#include <asm/fred.h>
#ifdef CONFIG_X86_32 __visible unsigned long saved_context_ebx; @@ -231,6 +232,19 @@ static void notrace __restore_processor_state(struct saved_context *ctxt) */ #ifdef CONFIG_X86_64 wrmsrl(MSR_GS_BASE, ctxt->kernelmode_gs_base);
/** Reinitialize FRED to ensure the FRED MSRs contain the same values* as before hibernation.** Note, the setup of FRED RSPs requires access to percpu data* structures. Therefore, FRED reinitialization can only occur after* the percpu access pointer (i.e., MSR_GS_BASE) is restored.*/if (ctxt->cr4 & X86_CR4_FRED) {cpu_init_fred_exceptions();cpu_init_fred_rsps();}#else loadsegment(fs, __KERNEL_PERCPU); #endif
base-commit: 535bd326c5657fe570f41b1f76941e449d9e2062
2.49.0
On April 1, 2025 12:57:27 AM PDT, "Xin Li (Intel)" xin@zytor.com wrote:
Upon a wakeup from S4, the restore kernel starts and initializes the FRED MSRs as needed from its perspective. It then loads a hibernation image, including the image kernel, and attempts to load image pages directly into their original page frames used before hibernation unless those frames are currently in use. Once all pages are moved to their original locations, it jumps to a "trampoline" page in the image kernel.
At this point, the image kernel takes control, but the FRED MSRs still contain values set by the restore kernel, which may differ from those set by the image kernel before hibernation. Therefore, the image kernel must ensure the FRED MSRs have the same values as before hibernation. Since these values depend only on the location of the kernel text and data, they can be recomputed from scratch.
Reported-by: Xi Pardee xi.pardee@intel.com Reported-by: Todd Brandt todd.e.brandt@intel.com Suggested-by: H. Peter Anvin (Intel) hpa@zytor.com Signed-off-by: Xin Li (Intel) xin@zytor.com Tested-by: Todd Brandt todd.e.brandt@intel.com Cc: Andy Lutomirski luto@kernel.org Cc: Brian Gerst brgerst@gmail.com Cc: Juergen Gross jgross@suse.com Cc: Linus Torvalds torvalds@linux-foundation.org Cc: stable@vger.kernel.org # 6.9+
Change in v2:
- Rewrite the change log and in-code comments based on Rafael's feedback.
arch/x86/power/cpu.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+)
diff --git a/arch/x86/power/cpu.c b/arch/x86/power/cpu.c index 63230ff8cf4f..08e76a5ca155 100644 --- a/arch/x86/power/cpu.c +++ b/arch/x86/power/cpu.c @@ -27,6 +27,7 @@ #include <asm/mmu_context.h> #include <asm/cpu_device_id.h> #include <asm/microcode.h> +#include <asm/fred.h>
#ifdef CONFIG_X86_32 __visible unsigned long saved_context_ebx; @@ -231,6 +232,19 @@ static void notrace __restore_processor_state(struct saved_context *ctxt) */ #ifdef CONFIG_X86_64 wrmsrl(MSR_GS_BASE, ctxt->kernelmode_gs_base);
- /*
* Reinitialize FRED to ensure the FRED MSRs contain the same values* as before hibernation.** Note, the setup of FRED RSPs requires access to percpu data* structures. Therefore, FRED reinitialization can only occur after* the percpu access pointer (i.e., MSR_GS_BASE) is restored.*/- if (ctxt->cr4 & X86_CR4_FRED) {
cpu_init_fred_exceptions();cpu_init_fred_rsps();- }
#else loadsegment(fs, __KERNEL_PERCPU); #endif
base-commit: 535bd326c5657fe570f41b1f76941e449d9e2062
Reviewed-by: H. Peter Anvin (Intel) hpa@zytor.com
linux-stable-mirror@lists.linaro.org
-
H. Peter Anvin -
Rafael J. Wysocki -
Xin Li (Intel)