598841ca9919d008b520114d8a4378c4ce4e40a1 ([S390] use gmap address spaces for kvm guest images) changed kvm to use a separate address space for kvm guests. This address space was switched in __vcpu_run In some cases (preemption, page fault) there is the possibility that this address space switch is lost. The typical symptom was a huge amount of validity intercepts or random guest addressing exceptions. Fix this by doing the switch in sie_loop and sie_exit and saving the address space in the gmap structure itself. Also use the preempt notifier.
Unified · Split
| @@ -658,12 +658,14 @@ static inline void pgste_set_pte(pte_t *ptep, pgste_t pgste) · 657 unmodified lines | ||
| 658 | 658 | * struct gmap_struct - guest address space |
| 659 | 659 | * @mm: pointer to the parent mm_struct |
| 660 | 660 | * @table: pointer to the page directory |
| 661 | + * @asce: address space control element for gmap page table | |
| 661 | 662 | * @crst_list: list of all crst tables used in the guest address space |
| 662 | 663 | */ |
| 663 | 664 | struct gmap { |
| 664 | 665 | struct list_head list; |
| 665 | 666 | struct mm_struct *mm; |
| 666 | 667 | unsigned long *table; |
| 668 | + unsigned long asce; | |
| 667 | 669 | struct list_head crst_list; |
| 668 | 670 | }; |
| 669 | 671 | |
| @@ -10,6 +10,7 @@ · 9 unmodified lines | ||
| 10 | 10 | #include <linux/sched.h> |
| 11 | 11 | #include <asm/vdso.h> |
| 12 | 12 | #include <asm/sigp.h> |
| 13 | +#include <asm/pgtable.h> | |
| 13 | 14 | |
| 14 | 15 | /* |
| 15 | 16 | * Make sure that the compiler is new enough. We want a compiler that |
| @@ -126,6 +127,7 @@ int main(void) · 110 unmodified lines | ||
| 126 | 127 | DEFINE(__LC_KERNEL_STACK, offsetof(struct _lowcore, kernel_stack)); |
| 127 | 128 | DEFINE(__LC_ASYNC_STACK, offsetof(struct _lowcore, async_stack)); |
| 128 | 129 | DEFINE(__LC_PANIC_STACK, offsetof(struct _lowcore, panic_stack)); |
| 130 | + DEFINE(__LC_USER_ASCE, offsetof(struct _lowcore, user_asce)); | |
| 129 | 131 | DEFINE(__LC_INT_CLOCK, offsetof(struct _lowcore, int_clock)); |
| 130 | 132 | DEFINE(__LC_MCCK_CLOCK, offsetof(struct _lowcore, mcck_clock)); |
| 131 | 133 | DEFINE(__LC_MACHINE_FLAGS, offsetof(struct _lowcore, machine_flags)); |
| @@ -151,6 +153,7 @@ int main(void) · 19 unmodified lines | ||
| 151 | 153 | DEFINE(__LC_VDSO_PER_CPU, offsetof(struct _lowcore, vdso_per_cpu_data)); |
| 152 | 154 | DEFINE(__LC_GMAP, offsetof(struct _lowcore, gmap)); |
| 153 | 155 | DEFINE(__LC_CMF_HPP, offsetof(struct _lowcore, cmf_hpp)); |
| 156 | + DEFINE(__GMAP_ASCE, offsetof(struct gmap, asce)); | |
| 154 | 157 | #endif /* CONFIG_32BIT */ |
| 155 | 158 | return 0; |
| 156 | 159 | } |
| @@ -1076,6 +1076,11 @@ sie_loop: · 1075 unmodified lines | ||
| 1076 | 1076 | lg %r14,__LC_THREAD_INFO # pointer thread_info struct |
| 1077 | 1077 | tm __TI_flags+7(%r14),_TIF_EXIT_SIE |
| 1078 | 1078 | jnz sie_exit |
| 1079 | + lg %r14,__LC_GMAP # get gmap pointer | |
| 1080 | + ltgr %r14,%r14 | |
| 1081 | + jz sie_gmap | |
| 1082 | + lctlg %c1,%c1,__GMAP_ASCE(%r14) # load primary asce | |
| 1083 | +sie_gmap: | |
| 1079 | 1084 | lg %r14,__SF_EMPTY(%r15) # get control block pointer |
| 1080 | 1085 | SPP __SF_EMPTY(%r15) # set guest id |
| 1081 | 1086 | sie 0(%r14) |
| @@ -1083,6 +1088,7 @@ sie_done: · 1 unmodified line | ||
| 1083 | 1088 | SPP __LC_CMF_HPP # set host id |
| 1084 | 1089 | lg %r14,__LC_THREAD_INFO # pointer thread_info struct |
| 1085 | 1090 | sie_exit: |
| 1091 | + lctlg %c1,%c1,__LC_USER_ASCE # load primary asce | |
| 1086 | 1092 | ni __TI_flags+6(%r14),255-(_TIF_SIE>>8) |
| 1087 | 1093 | lg %r14,__SF_EMPTY+8(%r15) # load guest register save area |
| 1088 | 1094 | stmg %r0,%r13,0(%r14) # save guest gprs 0-13 |
| @@ -263,10 +263,12 @@ void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu) · 262 unmodified lines | ||
| 263 | 263 | vcpu->arch.guest_fpregs.fpc &= FPC_VALID_MASK; |
| 264 | 264 | restore_fp_regs(&vcpu->arch.guest_fpregs); |
| 265 | 265 | restore_access_regs(vcpu->arch.guest_acrs); |
| 266 | + gmap_enable(vcpu->arch.gmap); | |
| 266 | 267 | } |
| 267 | 268 | |
| 268 | 269 | void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu) |
| 269 | 270 | { |
| 271 | + gmap_disable(vcpu->arch.gmap); | |
| 270 | 272 | save_fp_regs(&vcpu->arch.guest_fpregs); |
| 271 | 273 | save_access_regs(vcpu->arch.guest_acrs); |
| 272 | 274 | restore_fp_regs(&vcpu->arch.host_fpregs); |
| @@ -461,7 +463,6 @@ static void __vcpu_run(struct kvm_vcpu *vcpu) · 188 unmodified lines | ||
| 461 | 463 | local_irq_disable(); |
| 462 | 464 | kvm_guest_enter(); |
| 463 | 465 | local_irq_enable(); |
| 464 | − gmap_enable(vcpu->arch.gmap); | |
| 465 | 466 | VCPU_EVENT(vcpu, 6, "entering sie flags %x", |
| 466 | 467 | atomic_read(&vcpu->arch.sie_block->cpuflags)); |
| 467 | 468 | if (sie64a(vcpu->arch.sie_block, vcpu->arch.guest_gprs)) { |
| @@ -470,7 +471,6 @@ static void __vcpu_run(struct kvm_vcpu *vcpu) · 2 unmodified lines | ||
| 470 | 471 | } |
| 471 | 472 | VCPU_EVENT(vcpu, 6, "exit sie icptcode %d", |
| 472 | 473 | vcpu->arch.sie_block->icptcode); |
| 473 | − gmap_disable(vcpu->arch.gmap); | |
| 474 | 474 | local_irq_disable(); |
| 475 | 475 | kvm_guest_exit(); |
| 476 | 476 | local_irq_enable(); |
| @@ -160,6 +160,8 @@ struct gmap *gmap_alloc(struct mm_struct *mm) · 159 unmodified lines | ||
| 160 | 160 | table = (unsigned long *) page_to_phys(page); |
| 161 | 161 | crst_table_init(table, _REGION1_ENTRY_EMPTY); |
| 162 | 162 | gmap->table = table; |
| 163 | + gmap->asce = _ASCE_TYPE_REGION1 | _ASCE_TABLE_LENGTH | | |
| 164 | + _ASCE_USER_BITS | __pa(table); | |
| 163 | 165 | list_add(&gmap->list, &mm->context.gmap_list); |
| 164 | 166 | return gmap; |
| 165 | 167 | |
| @@ -240,10 +242,6 @@ EXPORT_SYMBOL_GPL(gmap_free); · 74 unmodified lines | ||
| 240 | 242 | */ |
| 241 | 243 | void gmap_enable(struct gmap *gmap) |
| 242 | 244 | { |
| 243 | − /* Load primary space page table origin. */ | |
| 244 | − S390_lowcore.user_asce = _ASCE_TYPE_REGION1 | _ASCE_TABLE_LENGTH | | |
| 245 | − _ASCE_USER_BITS | __pa(gmap->table); | |
| 246 | − asm volatile("lctlg 1,1,%0\n" : : "m" (S390_lowcore.user_asce) ); | |
| 247 | 245 | S390_lowcore.gmap = (unsigned long) gmap; |
| 248 | 246 | } |
| 249 | 247 | EXPORT_SYMBOL_GPL(gmap_enable); |
| @@ -254,10 +252,6 @@ EXPORT_SYMBOL_GPL(gmap_enable); · 4 unmodified lines | ||
| 254 | 252 | */ |
| 255 | 253 | void gmap_disable(struct gmap *gmap) |
| 256 | 254 | { |
| 257 | − /* Load primary space page table origin. */ | |
| 258 | − S390_lowcore.user_asce = | |
| 259 | − gmap->mm->context.asce_bits | __pa(gmap->mm->pgd); | |
| 260 | − asm volatile("lctlg 1,1,%0\n" : : "m" (S390_lowcore.user_asce) ); | |
| 261 | 255 | S390_lowcore.gmap = 0UL; |
| 262 | 256 | } |
| 263 | 257 | EXPORT_SYMBOL_GPL(gmap_disable); |