On 19.11.2021 18:09, Ian Jackson wrote:
> Andrew Cooper writes ("Re: [PATCH for-4.16 v3] efi: fix alignment of function
> parameters in compat mode"):
>> Some hardtabs appear to have slipped in.
>
> Thanks. Fixed.
>
>> Jan gave a conditional R-by which permitted a change along these lines,
>> but he's left the office now too, so you'll have to take him up on that
>> offer if you want this committing before Monday.
>
> Right.
>
> FTAOD I think I have done what Jan approved, and Andy double checked
> that. So I am about to push this. For the record, final version
> below.
Right, thanks. Nevertheless I think for 4.17 we will want to ...
> --- a/xen/common/efi/runtime.c
> +++ b/xen/common/efi/runtime.c
> @@ -607,6 +607,9 @@ int efi_runtime_call(struct xenpf_efi_runtime_call *op)
> break;
>
> case XEN_EFI_query_variable_info:
> + {
> + uint64_t max_store_size, remain_store_size, max_size;
> +
> if ( op->misc & ~XEN_EFI_VARINFO_BOOT_SNAPSHOT )
> return -EINVAL;
>
> @@ -638,16 +641,36 @@ int efi_runtime_call(struct xenpf_efi_runtime_call *op)
>
> if ( !efi_enabled(EFI_RS) || (efi_rs->Hdr.Revision >> 16) < 2 )
> return -EOPNOTSUPP;
> +
> + /*
> + * Bounce the variables onto the stack to make them 8 byte aligned
> when
> + * called from the compat handler, as their placement in
> + * compat_pf_efi_runtime_call will make them 4 byte aligned instead
> and
> + * and compilers may validly complain.
> + *
> + * Note that while the function parameters are OUT only, copy the
> + * values here anyway just in case. This is done regardless of
> whether
> + * called from the compat handler or not, as it's not worth the extra
> + * logic to differentiate.
> + */
> + max_store_size = op->u.query_variable_info.max_store_size;
> + remain_store_size = op->u.query_variable_info.remain_store_size;
> + max_size = op->u.query_variable_info.max_size;
... eliminate these.
Jan
