KVM: remove the usage of the mmap_sem for the protection of the memory slots.
This patch replaces the mmap_sem lock for the memory slots with a new
kvm private lock, it is needed beacuse untill now there were cases where
kvm accesses user memory while holding the mmap semaphore.
Signed-off-by: Izik Eidus <izike@qumranet.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
index 8efdcdb..2603710 100644
--- a/arch/x86/kvm/mmu.c
+++ b/arch/x86/kvm/mmu.c
@@ -876,11 +876,18 @@
struct page *gva_to_page(struct kvm_vcpu *vcpu, gva_t gva)
{
+ struct page *page;
+
gpa_t gpa = vcpu->arch.mmu.gva_to_gpa(vcpu, gva);
if (gpa == UNMAPPED_GVA)
return NULL;
- return gfn_to_page(vcpu->kvm, gpa >> PAGE_SHIFT);
+
+ down_read(¤t->mm->mmap_sem);
+ page = gfn_to_page(vcpu->kvm, gpa >> PAGE_SHIFT);
+ up_read(¤t->mm->mmap_sem);
+
+ return page;
}
static void mmu_set_spte(struct kvm_vcpu *vcpu, u64 *shadow_pte,
@@ -1020,15 +1027,18 @@
struct page *page;
+ down_read(&vcpu->kvm->slots_lock);
+
down_read(¤t->mm->mmap_sem);
page = gfn_to_page(vcpu->kvm, gfn);
+ up_read(¤t->mm->mmap_sem);
spin_lock(&vcpu->kvm->mmu_lock);
kvm_mmu_free_some_pages(vcpu);
r = __nonpaging_map(vcpu, v, write, gfn, page);
spin_unlock(&vcpu->kvm->mmu_lock);
- up_read(¤t->mm->mmap_sem);
+ up_read(&vcpu->kvm->slots_lock);
return r;
}
@@ -1362,6 +1372,7 @@
gfn_t gfn;
int r;
u64 gpte = 0;
+ struct page *page;
if (bytes != 4 && bytes != 8)
return;
@@ -1389,6 +1400,11 @@
if (!is_present_pte(gpte))
return;
gfn = (gpte & PT64_BASE_ADDR_MASK) >> PAGE_SHIFT;
+
+ down_read(¤t->mm->mmap_sem);
+ page = gfn_to_page(vcpu->kvm, gfn);
+ up_read(¤t->mm->mmap_sem);
+
vcpu->arch.update_pte.gfn = gfn;
vcpu->arch.update_pte.page = gfn_to_page(vcpu->kvm, gfn);
}
@@ -1496,9 +1512,9 @@
gpa_t gpa;
int r;
- down_read(¤t->mm->mmap_sem);
+ down_read(&vcpu->kvm->slots_lock);
gpa = vcpu->arch.mmu.gva_to_gpa(vcpu, gva);
- up_read(¤t->mm->mmap_sem);
+ up_read(&vcpu->kvm->slots_lock);
spin_lock(&vcpu->kvm->mmu_lock);
r = kvm_mmu_unprotect_page(vcpu->kvm, gpa >> PAGE_SHIFT);
diff --git a/arch/x86/kvm/paging_tmpl.h b/arch/x86/kvm/paging_tmpl.h
index 03ba860..2009c6e 100644
--- a/arch/x86/kvm/paging_tmpl.h
+++ b/arch/x86/kvm/paging_tmpl.h
@@ -91,7 +91,10 @@
pt_element_t *table;
struct page *page;
+ down_read(¤t->mm->mmap_sem);
page = gfn_to_page(kvm, table_gfn);
+ up_read(¤t->mm->mmap_sem);
+
table = kmap_atomic(page, KM_USER0);
ret = CMPXCHG(&table[index], orig_pte, new_pte);
@@ -378,7 +381,7 @@
if (r)
return r;
- down_read(¤t->mm->mmap_sem);
+ down_read(&vcpu->kvm->slots_lock);
/*
* Look up the shadow pte for the faulting address.
*/
@@ -392,11 +395,13 @@
pgprintk("%s: guest page fault\n", __FUNCTION__);
inject_page_fault(vcpu, addr, walker.error_code);
vcpu->arch.last_pt_write_count = 0; /* reset fork detector */
- up_read(¤t->mm->mmap_sem);
+ up_read(&vcpu->kvm->slots_lock);
return 0;
}
+ down_read(¤t->mm->mmap_sem);
page = gfn_to_page(vcpu->kvm, walker.gfn);
+ up_read(¤t->mm->mmap_sem);
spin_lock(&vcpu->kvm->mmu_lock);
kvm_mmu_free_some_pages(vcpu);
@@ -413,14 +418,14 @@
*/
if (shadow_pte && is_io_pte(*shadow_pte)) {
spin_unlock(&vcpu->kvm->mmu_lock);
- up_read(¤t->mm->mmap_sem);
+ up_read(&vcpu->kvm->slots_lock);
return 1;
}
++vcpu->stat.pf_fixed;
kvm_mmu_audit(vcpu, "post page fault (fixed)");
spin_unlock(&vcpu->kvm->mmu_lock);
- up_read(¤t->mm->mmap_sem);
+ up_read(&vcpu->kvm->slots_lock);
return write_pt;
}
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index ad36447..86f5bf1 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -1477,7 +1477,7 @@
struct kvm_userspace_memory_region kvm_userspace_mem;
int r = 0;
- down_write(¤t->mm->mmap_sem);
+ down_write(&kvm->slots_lock);
if (kvm->arch.apic_access_page)
goto out;
kvm_userspace_mem.slot = APIC_ACCESS_PAGE_PRIVATE_MEMSLOT;
@@ -1487,9 +1487,12 @@
r = __kvm_set_memory_region(kvm, &kvm_userspace_mem, 0);
if (r)
goto out;
+
+ down_read(¤t->mm->mmap_sem);
kvm->arch.apic_access_page = gfn_to_page(kvm, 0xfee00);
+ up_read(¤t->mm->mmap_sem);
out:
- up_write(¤t->mm->mmap_sem);
+ up_write(&kvm->slots_lock);
return r;
}
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 338764f..6b01552 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -184,7 +184,7 @@
int ret;
u64 pdpte[ARRAY_SIZE(vcpu->arch.pdptrs)];
- down_read(¤t->mm->mmap_sem);
+ down_read(&vcpu->kvm->slots_lock);
ret = kvm_read_guest_page(vcpu->kvm, pdpt_gfn, pdpte,
offset * sizeof(u64), sizeof(pdpte));
if (ret < 0) {
@@ -201,7 +201,7 @@
memcpy(vcpu->arch.pdptrs, pdpte, sizeof(vcpu->arch.pdptrs));
out:
- up_read(¤t->mm->mmap_sem);
+ up_read(&vcpu->kvm->slots_lock);
return ret;
}
@@ -215,13 +215,13 @@
if (is_long_mode(vcpu) || !is_pae(vcpu))
return false;
- down_read(¤t->mm->mmap_sem);
+ down_read(&vcpu->kvm->slots_lock);
r = kvm_read_guest(vcpu->kvm, vcpu->arch.cr3 & ~31u, pdpte, sizeof(pdpte));
if (r < 0)
goto out;
changed = memcmp(pdpte, vcpu->arch.pdptrs, sizeof(pdpte)) != 0;
out:
- up_read(¤t->mm->mmap_sem);
+ up_read(&vcpu->kvm->slots_lock);
return changed;
}
@@ -359,7 +359,7 @@
*/
}
- down_read(¤t->mm->mmap_sem);
+ down_read(&vcpu->kvm->slots_lock);
/*
* Does the new cr3 value map to physical memory? (Note, we
* catch an invalid cr3 even in real-mode, because it would
@@ -375,7 +375,7 @@
vcpu->arch.cr3 = cr3;
vcpu->arch.mmu.new_cr3(vcpu);
}
- up_read(¤t->mm->mmap_sem);
+ up_read(&vcpu->kvm->slots_lock);
}
EXPORT_SYMBOL_GPL(set_cr3);
@@ -1232,12 +1232,12 @@
if (kvm_nr_mmu_pages < KVM_MIN_ALLOC_MMU_PAGES)
return -EINVAL;
- down_write(¤t->mm->mmap_sem);
+ down_write(&kvm->slots_lock);
kvm_mmu_change_mmu_pages(kvm, kvm_nr_mmu_pages);
kvm->arch.n_requested_mmu_pages = kvm_nr_mmu_pages;
- up_write(¤t->mm->mmap_sem);
+ up_write(&kvm->slots_lock);
return 0;
}
@@ -1286,7 +1286,7 @@
< alias->target_phys_addr)
goto out;
- down_write(¤t->mm->mmap_sem);
+ down_write(&kvm->slots_lock);
p = &kvm->arch.aliases[alias->slot];
p->base_gfn = alias->guest_phys_addr >> PAGE_SHIFT;
@@ -1300,7 +1300,7 @@
kvm_mmu_zap_all(kvm);
- up_write(¤t->mm->mmap_sem);
+ up_write(&kvm->slots_lock);
return 0;
@@ -1376,7 +1376,7 @@
struct kvm_memory_slot *memslot;
int is_dirty = 0;
- down_write(¤t->mm->mmap_sem);
+ down_write(&kvm->slots_lock);
r = kvm_get_dirty_log(kvm, log, &is_dirty);
if (r)
@@ -1392,7 +1392,7 @@
}
r = 0;
out:
- up_write(¤t->mm->mmap_sem);
+ up_write(&kvm->slots_lock);
return r;
}
@@ -1570,7 +1570,7 @@
void *data = val;
int r = X86EMUL_CONTINUE;
- down_read(¤t->mm->mmap_sem);
+ down_read(&vcpu->kvm->slots_lock);
while (bytes) {
gpa_t gpa = vcpu->arch.mmu.gva_to_gpa(vcpu, addr);
unsigned offset = addr & (PAGE_SIZE-1);
@@ -1592,7 +1592,7 @@
addr += tocopy;
}
out:
- up_read(¤t->mm->mmap_sem);
+ up_read(&vcpu->kvm->slots_lock);
return r;
}
EXPORT_SYMBOL_GPL(emulator_read_std);
@@ -1611,9 +1611,9 @@
return X86EMUL_CONTINUE;
}
- down_read(¤t->mm->mmap_sem);
+ down_read(&vcpu->kvm->slots_lock);
gpa = vcpu->arch.mmu.gva_to_gpa(vcpu, addr);
- up_read(¤t->mm->mmap_sem);
+ up_read(&vcpu->kvm->slots_lock);
/* For APIC access vmexit */
if ((gpa & PAGE_MASK) == APIC_DEFAULT_PHYS_BASE)
@@ -1651,14 +1651,14 @@
{
int ret;
- down_read(¤t->mm->mmap_sem);
+ down_read(&vcpu->kvm->slots_lock);
ret = kvm_write_guest(vcpu->kvm, gpa, val, bytes);
if (ret < 0) {
- up_read(¤t->mm->mmap_sem);
+ up_read(&vcpu->kvm->slots_lock);
return 0;
}
kvm_mmu_pte_write(vcpu, gpa, val, bytes);
- up_read(¤t->mm->mmap_sem);
+ up_read(&vcpu->kvm->slots_lock);
return 1;
}
@@ -1670,9 +1670,9 @@
struct kvm_io_device *mmio_dev;
gpa_t gpa;
- down_read(¤t->mm->mmap_sem);
+ down_read(&vcpu->kvm->slots_lock);
gpa = vcpu->arch.mmu.gva_to_gpa(vcpu, addr);
- up_read(¤t->mm->mmap_sem);
+ up_read(&vcpu->kvm->slots_lock);
if (gpa == UNMAPPED_GVA) {
kvm_inject_page_fault(vcpu, addr, 2);
@@ -1749,7 +1749,7 @@
char *kaddr;
u64 val;
- down_read(¤t->mm->mmap_sem);
+ down_read(&vcpu->kvm->slots_lock);
gpa = vcpu->arch.mmu.gva_to_gpa(vcpu, addr);
if (gpa == UNMAPPED_GVA ||
@@ -1760,13 +1760,17 @@
goto emul_write;
val = *(u64 *)new;
+
+ down_read(¤t->mm->mmap_sem);
page = gfn_to_page(vcpu->kvm, gpa >> PAGE_SHIFT);
+ up_read(¤t->mm->mmap_sem);
+
kaddr = kmap_atomic(page, KM_USER0);
set_64bit((u64 *)(kaddr + offset_in_page(gpa)), val);
kunmap_atomic(kaddr, KM_USER0);
kvm_release_page_dirty(page);
emul_write:
- up_read(¤t->mm->mmap_sem);
+ up_read(&vcpu->kvm->slots_lock);
}
#endif
@@ -2159,10 +2163,10 @@
kvm_x86_ops->skip_emulated_instruction(vcpu);
for (i = 0; i < nr_pages; ++i) {
- down_read(¤t->mm->mmap_sem);
+ down_read(&vcpu->kvm->slots_lock);
page = gva_to_page(vcpu, address + i * PAGE_SIZE);
vcpu->arch.pio.guest_pages[i] = page;
- up_read(¤t->mm->mmap_sem);
+ up_read(&vcpu->kvm->slots_lock);
if (!page) {
kvm_inject_gp(vcpu, 0);
free_pio_guest_pages(vcpu);
@@ -2485,8 +2489,9 @@
down_read(¤t->mm->mmap_sem);
page = gfn_to_page(vcpu->kvm, apic->vapic_addr >> PAGE_SHIFT);
- vcpu->arch.apic->vapic_page = page;
up_read(¤t->mm->mmap_sem);
+
+ vcpu->arch.apic->vapic_page = page;
}
static void vapic_exit(struct kvm_vcpu *vcpu)
@@ -2959,9 +2964,9 @@
gpa_t gpa;
vcpu_load(vcpu);
- down_read(¤t->mm->mmap_sem);
+ down_read(&vcpu->kvm->slots_lock);
gpa = vcpu->arch.mmu.gva_to_gpa(vcpu, vaddr);
- up_read(¤t->mm->mmap_sem);
+ up_read(&vcpu->kvm->slots_lock);
tr->physical_address = gpa;
tr->valid = gpa != UNMAPPED_GVA;
tr->writeable = 1;
@@ -3234,11 +3239,13 @@
*/
if (!user_alloc) {
if (npages && !old.rmap) {
+ down_write(¤t->mm->mmap_sem);
memslot->userspace_addr = do_mmap(NULL, 0,
npages * PAGE_SIZE,
PROT_READ | PROT_WRITE,
MAP_SHARED | MAP_ANONYMOUS,
0);
+ up_write(¤t->mm->mmap_sem);
if (IS_ERR((void *)memslot->userspace_addr))
return PTR_ERR((void *)memslot->userspace_addr);
@@ -3246,8 +3253,10 @@
if (!old.user_alloc && old.rmap) {
int ret;
+ down_write(¤t->mm->mmap_sem);
ret = do_munmap(current->mm, old.userspace_addr,
old.npages * PAGE_SIZE);
+ up_write(¤t->mm->mmap_sem);
if (ret < 0)
printk(KERN_WARNING
"kvm_vm_ioctl_set_memory_region: "