Re: [patch V2 09/52] x86/fpu: Reject invalid MXCSR values in copy_kernel_to_xstate()

[Thomas Gleixner]

On Wed, Jun 16 2021 at 17:02, Borislav Petkov wrote:
> On Mon, Jun 14, 2021 at 05:44:17PM +0200, Thomas Gleixner wrote:
>> Instead of masking out reserved bits, check them and reject the provided
>> state as invalid if not zero.
>> 
>> Suggested-by: Andy Lutomirski <luto@xxxxxxxxxx>
>> Signed-off-by: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
>> ---
>> V2: New patch
>> ---
>>  arch/x86/kernel/fpu/xstate.c |   11 ++++++++---
>>  1 file changed, 8 insertions(+), 3 deletions(-)
>> 
>> --- a/arch/x86/kernel/fpu/xstate.c
>> +++ b/arch/x86/kernel/fpu/xstate.c
>> @@ -1166,6 +1166,14 @@ int copy_kernel_to_xstate(struct xregs_s
>>  	if (validate_user_xstate_header(&hdr))
>>  		return -EINVAL;
>>  
>> +	if (xfeatures_mxcsr_quirk(hdr.xfeatures)) {
>
> Since we're cleaning up this FPU stinking pile - that function needs to
> have a verb in the name, something like:
>
> 	if (xfeatures_mxcsr_quirk_needed(...))
>
> but that's unrelated to here and as a note to whoever gets to get to it
> first.
>
>> +		const u32 *mxcsr = kbuf + offsetof(struct fxregs_state, mxcsr);
>> +
>> +		/* Reserved bits in MXCSR must be zero. */
>> +		if (*mxcsr & ~mxcsr_feature_mask)
>> +			return -EINVAL;
>> +	}
>
> Btw, that function has another
>
> 	if (xfeatures_mxcsr_quirk(hdr.xfeatures)) {
>
> branch already below the loop.
>
> Should we merge both? Diff ontop of yours:

No, because the first usage is wrong. I found that while looking through
this stuff again. Sigh...