| Message ID | 1c917278029b206317a2155fb78e63ed14b621e5.1566176127.git.mkow@invisiblethingslab.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [v1] x86: Restore IA32_MISC_ENABLE on wakeup | expand |
On 19/08/2019 03:23, Michał Kowalczyk wrote: > diff --git a/xen/arch/x86/boot/trampoline.S b/xen/arch/x86/boot/trampoline.S > index 7c6a2328d2..fcaa3eeaf1 100644 > --- a/xen/arch/x86/boot/trampoline.S > +++ b/xen/arch/x86/boot/trampoline.S > @@ -85,7 +85,7 @@ trampoline_gdt: > .long trampoline_gdt + BOOT_PSEUDORM_DS + 2 - . > .popsection > > -GLOBAL(trampoline_misc_enable_off) > +GLOBAL(misc_enable_off) The overall change is fine, but why have you renamed this variable? Without the rename, the patch would be just the single hunk in wakeup.S and therefore easier to backport. ~Andrew
On 8/19/19 11:04 AM, Andrew Cooper wrote: > On 19/08/2019 03:23, Michał Kowalczyk wrote: >> diff --git a/xen/arch/x86/boot/trampoline.S b/xen/arch/x86/boot/trampoline.S >> index 7c6a2328d2..fcaa3eeaf1 100644 >> --- a/xen/arch/x86/boot/trampoline.S >> +++ b/xen/arch/x86/boot/trampoline.S >> @@ -85,7 +85,7 @@ trampoline_gdt: >> .long trampoline_gdt + BOOT_PSEUDORM_DS + 2 - . >> .popsection >> >> -GLOBAL(trampoline_misc_enable_off) >> +GLOBAL(misc_enable_off) > The overall change is fine, but why have you renamed this variable? The old name had "trampoline_" prefix because the only place where it was used was trampoline_protmode_entry in arch/x86/boot/trampoline.S. Now it's also used in the wakeup code, so I removed the prefix which could be (IMO) misleading. > Without the rename, the patch would be just the single hunk in wakeup.S > and therefore easier to backport. True. Anyway, the decision is on your side, I can leave the old name if you prefer.
On 19/08/2019 14:50, Michał Kowalczyk wrote: > On 8/19/19 11:04 AM, Andrew Cooper wrote: >> On 19/08/2019 03:23, Michał Kowalczyk wrote: >>> diff --git a/xen/arch/x86/boot/trampoline.S b/xen/arch/x86/boot/trampoline.S >>> index 7c6a2328d2..fcaa3eeaf1 100644 >>> --- a/xen/arch/x86/boot/trampoline.S >>> +++ b/xen/arch/x86/boot/trampoline.S >>> @@ -85,7 +85,7 @@ trampoline_gdt: >>> .long trampoline_gdt + BOOT_PSEUDORM_DS + 2 - . >>> .popsection >>> >>> -GLOBAL(trampoline_misc_enable_off) >>> +GLOBAL(misc_enable_off) >> The overall change is fine, but why have you renamed this variable? > The old name had "trampoline_" prefix because the only place where it > was used was trampoline_protmode_entry in arch/x86/boot/trampoline.S. > Now it's also used in the wakeup code, so I removed the prefix which > could be (IMO) misleading. >> Without the rename, the patch would be just the single hunk in wakeup.S >> and therefore easier to backport. > True. Anyway, the decision is on your side, I can leave the old name if > you prefer. The trampoline_ prefix indicates where the data lives, which is in the 16 bit trampoline which contains both the AP boot path, and wakeup path. If you're happy with this, I can adjust on commit to avoid you sending a second time. ~Andrew
On 8/19/19 3:52 PM, Andrew Cooper wrote: > On 19/08/2019 14:50, Michał Kowalczyk wrote: >> On 8/19/19 11:04 AM, Andrew Cooper wrote: >>> On 19/08/2019 03:23, Michał Kowalczyk wrote: >>>> diff --git a/xen/arch/x86/boot/trampoline.S b/xen/arch/x86/boot/trampoline.S >>>> index 7c6a2328d2..fcaa3eeaf1 100644 >>>> --- a/xen/arch/x86/boot/trampoline.S >>>> +++ b/xen/arch/x86/boot/trampoline.S >>>> @@ -85,7 +85,7 @@ trampoline_gdt: >>>> .long trampoline_gdt + BOOT_PSEUDORM_DS + 2 - . >>>> .popsection >>>> >>>> -GLOBAL(trampoline_misc_enable_off) >>>> +GLOBAL(misc_enable_off) >>> The overall change is fine, but why have you renamed this variable? >> The old name had "trampoline_" prefix because the only place where it >> was used was trampoline_protmode_entry in arch/x86/boot/trampoline.S. >> Now it's also used in the wakeup code, so I removed the prefix which >> could be (IMO) misleading. >>> Without the rename, the patch would be just the single hunk in wakeup.S >>> and therefore easier to backport. >> True. Anyway, the decision is on your side, I can leave the old name if >> you prefer. > The trampoline_ prefix indicates where the data lives, which is in the > 16 bit trampoline which contains both the AP boot path, and wakeup path. Ah, if this is the convention you use then we should leave the old name. > If you're happy with this, I can adjust on commit to avoid you sending a > second time. Would be great, thanks!
On 19/08/2019 14:56, Michał Kowalczyk wrote: > On 8/19/19 3:52 PM, Andrew Cooper wrote: >> On 19/08/2019 14:50, Michał Kowalczyk wrote: >>> On 8/19/19 11:04 AM, Andrew Cooper wrote: >>>> On 19/08/2019 03:23, Michał Kowalczyk wrote: >>>>> diff --git a/xen/arch/x86/boot/trampoline.S b/xen/arch/x86/boot/trampoline.S >>>>> index 7c6a2328d2..fcaa3eeaf1 100644 >>>>> --- a/xen/arch/x86/boot/trampoline.S >>>>> +++ b/xen/arch/x86/boot/trampoline.S >>>>> @@ -85,7 +85,7 @@ trampoline_gdt: >>>>> .long trampoline_gdt + BOOT_PSEUDORM_DS + 2 - . >>>>> .popsection >>>>> >>>>> -GLOBAL(trampoline_misc_enable_off) >>>>> +GLOBAL(misc_enable_off) >>>> The overall change is fine, but why have you renamed this variable? >>> The old name had "trampoline_" prefix because the only place where it >>> was used was trampoline_protmode_entry in arch/x86/boot/trampoline.S. >>> Now it's also used in the wakeup code, so I removed the prefix which >>> could be (IMO) misleading. >>>> Without the rename, the patch would be just the single hunk in wakeup.S >>>> and therefore easier to backport. >>> True. Anyway, the decision is on your side, I can leave the old name if >>> you prefer. >> The trampoline_ prefix indicates where the data lives, which is in the >> 16 bit trampoline which contains both the AP boot path, and wakeup path. > Ah, if this is the convention you use then we should leave the old name. >> If you're happy with this, I can adjust on commit to avoid you sending a >> second time. > Would be great, thanks! Done. https://xenbits.xen.org/gitweb/?p=xen.git;a=commitdiff;h=c3cfa5b3084d71bccd8360d044bea813688b587c ~Andrew
On 8/19/19 7:28 PM, Andrew Cooper wrote: > On 19/08/2019 14:56, Michał Kowalczyk wrote: >> On 8/19/19 3:52 PM, Andrew Cooper wrote: >>> On 19/08/2019 14:50, Michał Kowalczyk wrote: >>>> On 8/19/19 11:04 AM, Andrew Cooper wrote: >>>>> On 19/08/2019 03:23, Michał Kowalczyk wrote: >>>>>> diff --git a/xen/arch/x86/boot/trampoline.S b/xen/arch/x86/boot/trampoline.S >>>>>> index 7c6a2328d2..fcaa3eeaf1 100644 >>>>>> --- a/xen/arch/x86/boot/trampoline.S >>>>>> +++ b/xen/arch/x86/boot/trampoline.S >>>>>> @@ -85,7 +85,7 @@ trampoline_gdt: >>>>>> .long trampoline_gdt + BOOT_PSEUDORM_DS + 2 - . >>>>>> .popsection >>>>>> >>>>>> -GLOBAL(trampoline_misc_enable_off) >>>>>> +GLOBAL(misc_enable_off) >>>>> The overall change is fine, but why have you renamed this variable? >>>> The old name had "trampoline_" prefix because the only place where it >>>> was used was trampoline_protmode_entry in arch/x86/boot/trampoline.S. >>>> Now it's also used in the wakeup code, so I removed the prefix which >>>> could be (IMO) misleading. >>>>> Without the rename, the patch would be just the single hunk in wakeup.S >>>>> and therefore easier to backport. >>>> True. Anyway, the decision is on your side, I can leave the old name if >>>> you prefer. >>> The trampoline_ prefix indicates where the data lives, which is in the >>> 16 bit trampoline which contains both the AP boot path, and wakeup path. >> Ah, if this is the convention you use then we should leave the old name. >>> If you're happy with this, I can adjust on commit to avoid you sending a >>> second time. >> Would be great, thanks! > Done. > > https://xenbits.xen.org/gitweb/?p=xen.git;a=commitdiff;h=c3cfa5b3084d71bccd8360d044bea813688b587c Looks good.
diff --git a/xen/arch/x86/boot/trampoline.S b/xen/arch/x86/boot/trampoline.S index 7c6a2328d2..fcaa3eeaf1 100644 --- a/xen/arch/x86/boot/trampoline.S +++ b/xen/arch/x86/boot/trampoline.S @@ -85,7 +85,7 @@ trampoline_gdt: .long trampoline_gdt + BOOT_PSEUDORM_DS + 2 - . .popsection -GLOBAL(trampoline_misc_enable_off) +GLOBAL(misc_enable_off) .quad 0 GLOBAL(cpuid_ext_features) @@ -117,8 +117,8 @@ trampoline_protmode_entry: mov %eax,%cr3 /* Adjust IA32_MISC_ENABLE if needed (for NX enabling below). */ - mov bootsym_rel(trampoline_misc_enable_off,4,%esi) - mov bootsym_rel(trampoline_misc_enable_off+4,4,%edi) + mov bootsym_rel(misc_enable_off,4,%esi) + mov bootsym_rel(misc_enable_off+4,4,%edi) mov %esi,%eax or %edi,%eax jz 1f diff --git a/xen/arch/x86/boot/wakeup.S b/xen/arch/x86/boot/wakeup.S index e3cb9e033a..b5f825e983 100644 --- a/xen/arch/x86/boot/wakeup.S +++ b/xen/arch/x86/boot/wakeup.S @@ -138,6 +138,21 @@ wakeup_32: add bootsym_rel(trampoline_xen_phys_start,4,%eax) mov %eax,%cr3 + /* Reapply IA32_MISC_ENABLE modifications from early_init_intel(). */ + mov bootsym_rel(misc_enable_off, 4, %esi) + mov bootsym_rel(misc_enable_off+4, 4, %edi) + mov %esi, %eax + or %edi, %eax + jz 1f + mov $MSR_IA32_MISC_ENABLE, %ecx + rdmsr + not %esi + not %edi + and %esi, %eax + and %edi, %edx + wrmsr +1: + /* Will cpuid feature change after resume? */ /* Set up EFER (Extended Feature Enable Register). */ mov bootsym_rel(cpuid_ext_features,4,%edi) diff --git a/xen/arch/x86/cpu/intel.c b/xen/arch/x86/cpu/intel.c index 5356a6ae10..a01e519281 100644 --- a/xen/arch/x86/cpu/intel.c +++ b/xen/arch/x86/cpu/intel.c @@ -269,7 +269,7 @@ static void early_init_intel(struct cpuinfo_x86 *c) MSR_IA32_MISC_ENABLE_XD_DISABLE); if (disable) { wrmsrl(MSR_IA32_MISC_ENABLE, misc_enable & ~disable); - bootsym(trampoline_misc_enable_off) |= disable; + bootsym(misc_enable_off) |= disable; } if (disable & MSR_IA32_MISC_ENABLE_LIMIT_CPUID) diff --git a/xen/include/asm-x86/processor.h b/xen/include/asm-x86/processor.h index 2862321eee..b325e4b0df 100644 --- a/xen/include/asm-x86/processor.h +++ b/xen/include/asm-x86/processor.h @@ -152,7 +152,7 @@ extern void (*ctxt_switch_masking)(const struct vcpu *next); extern bool_t opt_cpu_info; extern u32 cpuid_ext_features; -extern u64 trampoline_misc_enable_off; +extern u64 misc_enable_off; /* Maximum width of physical addresses supported by the hardware. */ extern unsigned int paddr_bits;
Code in intel.c:early_init_intel() modifies IA32_MISC_ENABLE MSR. Those modifications must be restored after resuming from S3 (see e.g. Linux wakeup code), otherwise bad things may happen (e.g. wakeup code may cause #GP when trying to set IA32_EFER.NXE [1]). This bug was noticed on a ThinkPad x230 with NX disabled in the BIOS: Xen could correctly boot, but crashed when resuming from suspend. Applying this patch fixed the problem. [1] Intel SDM vol 3: "If the execute-disable capability is not available, a write to set IA32_EFER.NXE produces a #GP exception." Signed-off-by: Michał Kowalczyk <mkow@invisiblethingslab.com> --- xen/arch/x86/boot/trampoline.S | 6 +++--- xen/arch/x86/boot/wakeup.S | 15 +++++++++++++++ xen/arch/x86/cpu/intel.c | 2 +- xen/include/asm-x86/processor.h | 2 +- 4 files changed, 20 insertions(+), 5 deletions(-)