From 415de44076640483648d6c0f6d645a9ee61328ad Mon Sep 17 00:00:00 2001 From: Jane Malalane Date: Thu, 21 Oct 2021 11:47:44 +0100 Subject: [PATCH] x86/cpu: Fix migration safety with X86_BUG_NULL_SEL Currently, Linux probes for X86_BUG_NULL_SEL unconditionally which makes it unsafe to migrate in a virtualised environment as the properties across the migration pool might differ. To be specific, the case which goes wrong is: 1. Zen1 (or earlier) and Zen2 (or later) in a migration pool 2. Linux boots on Zen2, probes and finds the absence of X86_BUG_NULL_SEL 3. Linux is then migrated to Zen1 Linux is now running on a X86_BUG_NULL_SEL-impacted CPU while believing that the bug is fixed. The only way to address the problem is to fully trust the "no longer affected" CPUID bit when virtualised, because in the above case it would be clear deliberately to indicate the fact "you might migrate to somewhere which has this behaviour". Zen3 adds the NullSelectorClearsBase CPUID bit to indicate that loading a NULL segment selector zeroes the base and limit fields, as well as just attributes. Zen2 also has this behaviour but doesn't have the NSCB bit. [ bp: Minor touchups. ] Signed-off-by: Jane Malalane Signed-off-by: Borislav Petkov CC: Link: https://lkml.kernel.org/r/20211021104744.24126-1-jane.malalane@citrix.com --- arch/x86/kernel/cpu/amd.c | 2 ++ arch/x86/kernel/cpu/common.c | 44 ++++++++++++++++++++++++++++++------ arch/x86/kernel/cpu/cpu.h | 1 + arch/x86/kernel/cpu/hygon.c | 2 ++ 4 files changed, 42 insertions(+), 7 deletions(-) diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c index 2131af9f2fa237..4edb6f0f628c20 100644 --- a/arch/x86/kernel/cpu/amd.c +++ b/arch/x86/kernel/cpu/amd.c @@ -989,6 +989,8 @@ static void init_amd(struct cpuinfo_x86 *c) if (cpu_has(c, X86_FEATURE_IRPERF) && !cpu_has_amd_erratum(c, amd_erratum_1054)) msr_set_bit(MSR_K7_HWCR, MSR_K7_HWCR_IRPERF_EN_BIT); + + check_null_seg_clears_base(c); } #ifdef CONFIG_X86_32 diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c index 325d6022599b24..1bfeb186452a66 100644 --- a/arch/x86/kernel/cpu/common.c +++ b/arch/x86/kernel/cpu/common.c @@ -1397,9 +1397,8 @@ void __init early_cpu_init(void) early_identify_cpu(&boot_cpu_data); } -static void detect_null_seg_behavior(struct cpuinfo_x86 *c) +static bool detect_null_seg_behavior(void) { -#ifdef CONFIG_X86_64 /* * Empirically, writing zero to a segment selector on AMD does * not clear the base, whereas writing zero to a segment @@ -1420,10 +1419,43 @@ static void detect_null_seg_behavior(struct cpuinfo_x86 *c) wrmsrl(MSR_FS_BASE, 1); loadsegment(fs, 0); rdmsrl(MSR_FS_BASE, tmp); - if (tmp != 0) - set_cpu_bug(c, X86_BUG_NULL_SEG); wrmsrl(MSR_FS_BASE, old_base); -#endif + return tmp == 0; +} + +void check_null_seg_clears_base(struct cpuinfo_x86 *c) +{ + /* BUG_NULL_SEG is only relevant with 64bit userspace */ + if (!IS_ENABLED(CONFIG_X86_64)) + return; + + /* Zen3 CPUs advertise Null Selector Clears Base in CPUID. */ + if (c->extended_cpuid_level >= 0x80000021 && + cpuid_eax(0x80000021) & BIT(6)) + return; + + /* + * CPUID bit above wasn't set. If this kernel is still running + * as a HV guest, then the HV has decided not to advertize + * that CPUID bit for whatever reason. For example, one + * member of the migration pool might be vulnerable. Which + * means, the bug is present: set the BUG flag and return. + */ + if (cpu_has(c, X86_FEATURE_HYPERVISOR)) { + set_cpu_bug(c, X86_BUG_NULL_SEG); + return; + } + + /* + * Zen2 CPUs also have this behaviour, but no CPUID bit. + * 0x18 is the respective family for Hygon. + */ + if ((c->x86 == 0x17 || c->x86 == 0x18) && + detect_null_seg_behavior()) + return; + + /* All the remaining ones are affected */ + set_cpu_bug(c, X86_BUG_NULL_SEG); } static void generic_identify(struct cpuinfo_x86 *c) @@ -1459,8 +1491,6 @@ static void generic_identify(struct cpuinfo_x86 *c) get_model_name(c); /* Default name */ - detect_null_seg_behavior(c); - /* * ESPFIX is a strange bug. All real CPUs have it. Paravirt * systems that run Linux at CPL > 0 may or may not have the diff --git a/arch/x86/kernel/cpu/cpu.h b/arch/x86/kernel/cpu/cpu.h index 95521302630d4c..ee6f23f7587d40 100644 --- a/arch/x86/kernel/cpu/cpu.h +++ b/arch/x86/kernel/cpu/cpu.h @@ -75,6 +75,7 @@ extern int detect_extended_topology_early(struct cpuinfo_x86 *c); extern int detect_extended_topology(struct cpuinfo_x86 *c); extern int detect_ht_early(struct cpuinfo_x86 *c); extern void detect_ht(struct cpuinfo_x86 *c); +extern void check_null_seg_clears_base(struct cpuinfo_x86 *c); unsigned int aperfmperf_get_khz(int cpu); diff --git a/arch/x86/kernel/cpu/hygon.c b/arch/x86/kernel/cpu/hygon.c index 6d50136f7ab983..3fcdda4c1e114f 100644 --- a/arch/x86/kernel/cpu/hygon.c +++ b/arch/x86/kernel/cpu/hygon.c @@ -335,6 +335,8 @@ static void init_hygon(struct cpuinfo_x86 *c) /* Hygon CPUs don't reset SS attributes on SYSRET, Xen does. */ if (!cpu_has(c, X86_FEATURE_XENPV)) set_cpu_bug(c, X86_BUG_SYSRET_SS_ATTRS); + + check_null_seg_clears_base(c); } static void cpu_detect_tlb_hygon(struct cpuinfo_x86 *c)