Re: [PATCH] LoongArch: mm: try VMA lock-based page fault handling first

[Wentao Guan]

Hello Huacai,

Thanks for your review, I will fixed in V2.

> > +       if (write) {
> > +               flags |= FAULT_FLAG_WRITE;
> > +               if (!(vma->vm_flags & VM_WRITE)) {
> > +                       vma_end_read(vma);
> > +                       si_code = SEGV_ACCERR;
> Need count_vm_vma_lock_event(VMA_LOCK_SUCCESS) here.
Ok, I forget it.
> > +                       goto bad_area_nosemaphore;
> > +               }
> > +       } else {
> > +               if (!(vma->vm_flags & VM_EXEC) && address == exception_era(regs)){
> > +                       vma_end_read(vma);
> > +                       si_code = SEGV_ACCERR;
> The same.
Ok.
> > +                       goto bad_area_nosemaphore;
> > +               }
> > +               if (!(vma->vm_flags & (VM_READ | VM_WRITE)) && address != exception_era(regs)){
> > +                       vma_end_read(vma);
> > +                       si_code = SEGV_ACCERR;
>The same.
Ok.

> >
> >         mmap_read_unlock(mm);
> > +done:
> > We need error handling here.
Ok.

Best Regards
Wentao Guan