Re: [patch V3 3/6] x86/process: Check PF_KTHREAD and not current->mm for kernel threads

[Andy Lutomirski]

On 6/10/21 1:54 PM, Thomas Gleixner wrote:
> On Thu, Jun 10 2021 at 10:10, Andy Lutomirski wrote:
> 
>> On Tue, Jun 8, 2021, at 7:36 AM, Thomas Gleixner wrote:
>>> switch_fpu_finish() checks current->mm as indicator for kernel threads.
>>> That's wrong because kernel threads can temporarily use a mm of a user
>>> process via kthread_use_mm().
>>>
>>> Check the task flags for PF_KTHREAD instead.
>>>
>>> Fixes: 0cecca9d03c9 ("x86/fpu: Eager switch PKRU state")
>>> Signed-off-by: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
>>> Cc: Rik van Riel <riel@xxxxxxxxxxx>
>>> Cc: stable@xxxxxxxxxxxxxxx
>>> ---
>>>  arch/x86/include/asm/fpu/internal.h |    2 +-
>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> --- a/arch/x86/include/asm/fpu/internal.h
>>> +++ b/arch/x86/include/asm/fpu/internal.h
>>> @@ -578,7 +578,7 @@ static inline void switch_fpu_finish(str
>>>  	 * PKRU state is switched eagerly because it needs to be valid before we
>>>  	 * return to userland e.g. for a copy_to_user() operation.
>>>  	 */
>>> -	if (current->mm) {
>>> +	if (!(current->flags & PF_KTHREAD)) {
>>>  		pk = get_xsave_addr(&new_fpu->state.xsave, XFEATURE_PKRU);
>>>  		if (pk)
>>>  			pkru_val = pk->pkru;
>>>
>>>
>> Why are we checking this at all?  I actually tend to agree with the
>> ->mm check more than PF_anything. If we have a user address space,
>> then PKRU matters. If we don’t, then it doesn’t.
> 
> Which PKRU matters? A kernel thread has always the default PKRU no
> matter whether it uses a mm or not. It _cannot_ borrow the PKRU from the
> mm owning process. There is no way, so let's not pretend there would be.
> 

Hmm.  I guess PK_KTHREAD is consistent with switch_fpu_prepare() --
kernel threads have no FPU state.

It might be worth a loud comment here that kernel threads' PKRU is not
context switched and that, if anyone wants kthread_use_mm() users to use
anything other than the default PKRU, that they will need to change this.

So I guess your patch is okay.

--Andy