Re: [PATCH v6 12/14] KVM: s390: Extend MEM_OP ioctl by storage key checked cmpxchg

On Thu, 2023-01-26 at 17:10 +0100, Janosch Frank wrote:

On 1/25/23 22:26, Janis Schoetterl-Glausch wrote:

...

User space can use the MEM_OP ioctl to make storage key checked reads and writes to the guest, however, it has no way of performing atomic, key checked, accesses to the guest. Extend the MEM_OP ioctl in order to allow for this, by adding a cmpxchg op. For now, support this op for absolute accesses only.

This op can be use, for example, to set the device-state-change

s/use/used/

...

indicator and the adapter-local-summary indicator atomically.

Signed-off-by: Janis Schoetterl-Glausch scgl@linux.ibm.com

[...]

...

+/**

• • cmpxchg_guest_abs_with_key() - Perform cmpxchg on guest absolute address.
• • @kvm: Virtual machine instance.
• • @gpa: Absolute guest address of the location to be changed.
• • @len: Operand length of the cmpxchg, required: 1 <= len <= 16. Providing a

• •   non power of two will result in failure.
    

• • @old_addr: Pointer to old value. If the location at @gpa contains this value,

• •        the exchange will succeed. After calling cmpxchg_guest_abs_with_key()
    

• •        *@old_addr contains the value at @gpa before the attempt to
    

• •        exchange the value.
    

• • @new: The value to place at @gpa.
• • @access_key: The access key to use for the guest access.
• • @success: output value indicating if an exchange occurred.
• • Atomically exchange the value at @gpa by @new, if it contains *@old.
• • Honors storage keys.
• • Return: * 0: successful exchange

• •     * a program interruption code indicating the reason cmpxchg could
    

• •       not be attempted
    

Nit:

...

0: a program interruption code...

...

• •     * -EINVAL: address misaligned or len not power of two
    

• •     * -EAGAIN: transient failure (len 1 or 2)
    

• •     * -EOPNOTSUPP: read-only memslot (should never occur)
    

• */

+int cmpxchg_guest_abs_with_key(struct kvm *kvm, gpa_t gpa, int len,

• 	       __uint128_t *old_addr, __uint128_t new,
  

• 	       u8 access_key, bool *success)
  

+{

• gfn_t gfn = gpa >> PAGE_SHIFT;

gpa_to_gfn()?

Yes.

...

• struct kvm_memory_slot *slot = gfn_to_memslot(kvm, gfn);
• bool writable;
• hva_t hva;
• int ret;
• if (!IS_ALIGNED(gpa, len))

• return -EINVAL;
  

• hva = gfn_to_hva_memslot_prot(slot, gfn, &writable);
• if (kvm_is_error_hva(hva))

• return PGM_ADDRESSING;
  

• /*

• * Check if it's a read-only memslot, even though that cannot occur
  

• * since those are unsupported.
  

• * Don't try to actually handle that case.
  

• */
  

• if (!writable)

• return -EOPNOTSUPP;
  

• hva += offset_in_page(gpa);

Hmm if we don't use a macro to generate these then I'd add an explanation:

cmpxchg_user_key() is a macro that is dependent on the type of "old" so there's no deduplication possible without further macros.

Can do. Btw. I could move the other two statements out of the switch by using a union of old values, memcmp and memcpy, but I think that would be less readable.

...

• switch (len) {
• case 1: {

• u8 old;
  

• ret = cmpxchg_user_key((u8 *)hva, &old, *old_addr, new, access_key);
  

• *success = !ret && old == *old_addr;
  

• *old_addr = old;
  

• break;
  

• }
• case 2: {

• u16 old;
  

• ret = cmpxchg_user_key((u16 *)hva, &old, *old_addr, new, access_key);
  

• *success = !ret && old == *old_addr;
  

• *old_addr = old;
  

• break;
  

• }
• case 4: {

• u32 old;
  

• ret = cmpxchg_user_key((u32 *)hva, &old, *old_addr, new, access_key);
  

• *success = !ret && old == *old_addr;
  

• *old_addr = old;
  

• break;
  

• }
• case 8: {

• u64 old;
  

• ret = cmpxchg_user_key((u64 *)hva, &old, *old_addr, new, access_key);
  

• *success = !ret && old == *old_addr;
  

• *old_addr = old;
  

• break;
  

• }
• case 16: {

• __uint128_t old;
  

• ret = cmpxchg_user_key((__uint128_t *)hva, &old, *old_addr, new, access_key);
  

• *success = !ret && old == *old_addr;
  

• *old_addr = old;
  

• break;
  

• }
• default:

• return -EINVAL;
  

• }
• mark_page_dirty_in_slot(kvm, slot, gfn);

Is that needed if we failed the store?

Indeed it isn't.

[...]