On 30.10.2025 13:26, Grygorii Strashko wrote:
> On 30.10.25 13:23, Jan Beulich wrote:
>> On 30.10.2025 00:54, Grygorii Strashko wrote:
>>> --- a/xen/arch/x86/include/asm/hvm/vmx/vmcs.h
>>> +++ b/xen/arch/x86/include/asm/hvm/vmx/vmcs.h
>>> @@ -46,7 +46,9 @@ struct ept_data {
>>>
>>> #define _VMX_DOMAIN_PML_ENABLED 0
>>> #define VMX_DOMAIN_PML_ENABLED (1ul << _VMX_DOMAIN_PML_ENABLED)
>>> +
>>> struct vmx_domain {
>>> + spinlock_t uc_lock;
>>> mfn_t apic_access_mfn;
>>> /* VMX_DOMAIN_* */
>>> unsigned int status;
>>
>> Any reason to make this the very first field of the struct? It might better
>> live adjacent to the other field you move; there's going to be some padding
>> anyway, afaict.
>
> I've tried to put fields in holes and checked with pahole.
>
> With current change it is:
> struct vmx_domain {
> spinlock_t uc_lock; /* 0 8 */
> mfn_t apic_access_mfn; /* 8 8 */
> unsigned int status; /* 16 4 */
> _Bool exec_sp; /* 20 1 */
> _Bool is_in_uc_mode; /* 21 1 */
>
> /* size: 24, cachelines: 1, members: 5 */
> /* padding: 2 */
> /* last cacheline: 24 bytes */
> };
>
> It seems can be grouped like below?:
> struct vmx_domain {
> mfn_t apic_access_mfn; /* 0 8 */
> unsigned int status; /* 8 4 */
> _Bool exec_sp; /* 12 1 */
> _Bool is_in_uc_mode; /* 13 1 */
>
> /* XXX 2 bytes hole, try to pack */
>
> spinlock_t uc_lock; /* 16 8 */
>
> /* size: 24, cachelines: 1, members: 5 */
> /* sum members: 22, holes: 1, sum holes: 2 */
> /* last cacheline: 24 bytes */
> };
Yes, this is what I was thinking of.
Jan
