On Wed, Mar 20, 2024 at 10:59 AM Paolo Bonzini <pbonzini@xxxxxxxxxx> wrote:
> I will now focus on reviewing patches 6-20. This way we can prepare a
> common tree for SEV_INIT2/SNP/TDX, for both vendors to build upon.
Ok, the attachment is the delta that I have. The only major change is
requiring discard (thus effectively blocking VFIO support for
SEV-SNP/TDX, at least for now).
I will push it shortly to the same sevinit2 branch, and will post the
patches sometime soon.
Xiaoyao, you can use that branch too (it's on
https://gitlab.com/bonzini/qemu) as the basis for your TDX work.
Paolo
diff --git a/accel/kvm/kvm-all.c b/accel/kvm/kvm-all.c
index bf0ae0c8adb..428468950d9 100644
--- a/accel/kvm/kvm-all.c
+++ b/accel/kvm/kvm-all.c
@@ -285,19 +285,8 @@ static int kvm_set_user_memory_region(KVMMemoryListener *kml, KVMSlot *slot, boo
{
KVMState *s = kvm_state;
struct kvm_userspace_memory_region2 mem;
- static int cap_user_memory2 = -1;
int ret;
- if (cap_user_memory2 == -1) {
- cap_user_memory2 = kvm_check_extension(s, KVM_CAP_USER_MEMORY2);
- }
-
- if (!cap_user_memory2 && slot->guest_memfd >= 0) {
- error_report("%s, KVM doesn't support KVM_CAP_USER_MEMORY2,"
- " which is required by guest memfd!", __func__);
- exit(1);
- }
-
mem.slot = slot->slot | (kml->as_id << 16);
mem.guest_phys_addr = slot->start_addr;
mem.userspace_addr = (unsigned long)slot->ram;
@@ -310,7 +299,7 @@ static int kvm_set_user_memory_region(KVMMemoryListener *kml, KVMSlot *slot, boo
* value. This is needed based on KVM commit 75d61fbc. */
mem.memory_size = 0;
- if (cap_user_memory2) {
+ if (kvm_guest_memfd_supported) {
ret = kvm_vm_ioctl(s, KVM_SET_USER_MEMORY_REGION2, &mem);
} else {
ret = kvm_vm_ioctl(s, KVM_SET_USER_MEMORY_REGION, &mem);
@@ -320,7 +309,7 @@ static int kvm_set_user_memory_region(KVMMemoryListener *kml, KVMSlot *slot, boo
}
}
mem.memory_size = slot->memory_size;
- if (cap_user_memory2) {
+ if (kvm_guest_memfd_supported) {
ret = kvm_vm_ioctl(s, KVM_SET_USER_MEMORY_REGION2, &mem);
} else {
ret = kvm_vm_ioctl(s, KVM_SET_USER_MEMORY_REGION, &mem);
@@ -332,7 +321,7 @@ err:
mem.userspace_addr, mem.guest_memfd,
mem.guest_memfd_offset, ret);
if (ret < 0) {
- if (cap_user_memory2) {
+ if (kvm_guest_memfd_supported) {
error_report("%s: KVM_SET_USER_MEMORY_REGION2 failed, slot=%d,"
" start=0x%" PRIx64 ", size=0x%" PRIx64 ","
" flags=0x%" PRIx32 ", guest_memfd=%" PRId32 ","
@@ -502,6 +491,7 @@ static int kvm_mem_flags(MemoryRegion *mr)
flags |= KVM_MEM_READONLY;
}
if (memory_region_has_guest_memfd(mr)) {
+ assert(kvm_guest_memfd_supported);
flags |= KVM_MEM_GUEST_MEMFD;
}
return flags;
@@ -1310,18 +1300,7 @@ static int kvm_set_memory_attributes(hwaddr start, hwaddr size, uint64_t attr)
struct kvm_memory_attributes attrs;
int r;
- if (kvm_supported_memory_attributes == 0) {
- error_report("No memory attribute supported by KVM\n");
- return -EINVAL;
- }
-
- if ((attr & kvm_supported_memory_attributes) != attr) {
- error_report("memory attribute 0x%lx not supported by KVM,"
- " supported bits are 0x%lx\n",
- attr, kvm_supported_memory_attributes);
- return -EINVAL;
- }
-
+ assert((attr & kvm_supported_memory_attributes) == attr);
attrs.attributes = attr;
attrs.address = start;
attrs.size = size;
@@ -2488,11 +2467,14 @@ static int kvm_init(MachineState *ms)
}
s->as = g_new0(struct KVMAs, s->nr_as);
- kvm_guest_memfd_supported = kvm_check_extension(s, KVM_CAP_GUEST_MEMFD);
-
ret = kvm_check_extension(s, KVM_CAP_MEMORY_ATTRIBUTES);
kvm_supported_memory_attributes = ret > 0 ? ret : 0;
+ kvm_guest_memfd_supported =
+ kvm_check_extension(s, KVM_CAP_GUEST_MEMFD) &&
+ kvm_check_extension(s, KVM_CAP_USER_MEMORY2) &&
+ (kvm_supported_memory_attributes & KVM_MEMORY_ATTRIBUTE_PRIVATE);
+
if (object_property_find(OBJECT(current_machine), "kvm-type")) {
g_autofree char *kvm_type = object_property_get_str(OBJECT(current_machine),
"kvm-type",
@@ -2962,14 +2944,10 @@ int kvm_convert_memory(hwaddr start, hwaddr size, bool to_private)
*/
return 0;
} else {
- ret = ram_block_discard_is_disabled()
- ? ram_block_discard_range(rb, offset, size)
- : 0;
+ ret = ram_block_discard_range(rb, offset, size);
}
} else {
- ret = ram_block_discard_is_disabled()
- ? ram_block_discard_guest_memfd_range(rb, offset, size)
- : 0;
+ ret = ram_block_discard_guest_memfd_range(rb, offset, size);
}
} else {
error_report("Convert non guest_memfd backed memory region "
diff --git a/system/physmem.c b/system/physmem.c
index 8be8053cf77..b839be58538 100644
--- a/system/physmem.c
+++ b/system/physmem.c
@@ -1810,6 +1810,7 @@ static void ram_block_add(RAMBlock *new_block, Error **errp)
const bool shared = qemu_ram_is_shared(new_block);
RAMBlock *block;
RAMBlock *last_block = NULL;
+ bool free_on_error = false;
ram_addr_t old_ram_size, new_ram_size;
Error *err = NULL;
@@ -1839,17 +1840,26 @@ static void ram_block_add(RAMBlock *new_block, Error **errp)
return;
}
memory_try_enable_merging(new_block->host, new_block->max_length);
+ free_on_error = true;
}
}
- if (kvm_enabled() && (new_block->flags & RAM_GUEST_MEMFD)) {
+ if (new_block->flags & RAM_GUEST_MEMFD) {
+ assert(kvm_enabled());
assert(new_block->guest_memfd < 0);
+ if (ram_block_discard_require(true) < 0) {
+ error_setg_errno(errp, errno,
+ "cannot set up private guest memory: discard currently blocked");
+ error_append_hint(errp, "Are you using assigned devices?\n");
+ goto out_free;
+ }
+
new_block->guest_memfd = kvm_create_guest_memfd(new_block->max_length,
0, errp);
if (new_block->guest_memfd < 0) {
qemu_mutex_unlock_ramlist();
- return;
+ goto out_free;
}
}
@@ -1901,6 +1911,13 @@ static void ram_block_add(RAMBlock *new_block, Error **errp)
ram_block_notify_add(new_block->host, new_block->used_length,
new_block->max_length);
}
+ return;
+
+out_free:
+ if (free_on_error) {
+ qemu_anon_ram_free(new_block->host, new_block->max_length);
+ new_block->host = NULL;
+ }
}
#ifdef CONFIG_POSIX
@@ -2032,7 +2049,7 @@ RAMBlock *qemu_ram_alloc_internal(ram_addr_t size, ram_addr_t max_size,
int align;
assert((ram_flags & ~(RAM_SHARED | RAM_RESIZEABLE | RAM_PREALLOC |
- RAM_NORESERVE| RAM_GUEST_MEMFD)) == 0);
+ RAM_NORESERVE | RAM_GUEST_MEMFD)) == 0);
assert(!host ^ (ram_flags & RAM_PREALLOC));
align = qemu_real_host_page_size();
@@ -2101,6 +2118,7 @@ static void reclaim_ramblock(RAMBlock *block)
if (block->guest_memfd >= 0) {
close(block->guest_memfd);
+ ram_block_discard_require(false);
}
g_free(block);
