Re: [RFC PATCH 1/3] x86/mm: fix LAM cr3 mask inconsistency during context switch

[Yosry Ahmed <yosryahmed@google.com>]

On Thu, Mar 07, 2024 at 07:22:36PM +0200, Kirill A. Shutemov wrote:
> On Thu, Mar 07, 2024 at 01:39:14PM +0000, Yosry Ahmed wrote:
> > In switch_mm_irqs_off(), we read the 'mm->context.lam_cr3_mask' into
> > 'new_lam', which is later passed to load_new_mm_cr3(). However, there is
> > a call to set_tlbstate_lam_mode() in between which will read
> > 'mm->context.lam_cr3_mask' again and set 'cpu_tlbstate.lam' accordingly.
> > If we race with another thread updating 'mm->context.lam_cr3_mask', the
> > value in 'cpu_tlbstate.lam' could end up being different from CR3.
> 
> What other thread? LAM can only be enabled when the process has single
> thread. And cannot be disabled. See MM_CONTEXT_LOCK_LAM.

Right, but a kthread may run with that single-threaded process's mm
IIUC. I think this can happen via kthread_use_mm() or if we context
switch directly from the user process to the kthread (context_switch()
doesn't seem to update the mm in this case).

> 
> > While we are at it, remove the misguiding comment that states that
> > 'new_lam' may not match tlbstate_lam_cr3_mask() if a race occurs.
> 
> The comment is indeed misguiding, but for different reason. It is leftover
> from the earlier version of LAM patchset.