On Tue, Sep 28, 2021 at 11:50:37PM +0000, Fenghua Yu wrote:
> If xfeatures's feature bit is 0, xsaves will not write its init value to the
> memory due to init optimization. So the xsaves will do nothing and the
> state is not initialized and may have random data.

> Setting TIF_NEED_FPU_LOAD cannot guaranteed to execute XRSTORS on exiting
> to user. In fpregs_restore_userregs():
>       if (!fpregs_state_valid(fpu, cpu)) {
>               ...
>                 __restore_fpregs_from_fpstate(&fpu->state, mask);
>               ...
>       }
> 
> fpregs state should be invalid to get the XRSTROS executed.
> 
> So setting TIF_NEED_FPU_LOAD may get the FPU register unchanged on exiting
> to user.

Does this help?
Changed lines marked with //<<<<<

-Tony

void *begin_update_one_xsave_feature(struct task_struct *tsk,
                                     enum xfeature xfeature, bool full)
{
        struct xregs_state *xsave = &tsk->thread.fpu.state.xsave;
        struct xregs_state *xinit = &init_fpstate.xsave;
        u64 fmask = 1ull << xfeature;
        void *addr;

        BUG_ON(!(xsave->header.xcomp_bv & fmask));

        fpregs_lock();

        addr = __raw_xsave_addr(xsave, xfeature);

        if (full || tsk != current) {
                memcpy(addr, __raw_xsave_addr(xinit, xfeature), 
xstate_sizes[xfeature]);
                goto out;
        }

        if (!(xsave->header.xfeatures & fmask)) {
                xsave->header.xfeatures |= fmask;       //<<<<<
                xsaves(xsave, fmask);
        }

out:
        xsave->header.xfeatures |= fmask;
        return addr;
}

void finish_update_one_xsave_feature(struct task_struct *tsk)
{
        set_ti_thread_flag(task_thread_info(tsk), TIF_NEED_FPU_LOAD);
        if (tsk == current)                             //<<<<<
                __cpu_invalidate_fpregs_state();        //<<<<<
        fpregs_unlock();
}
_______________________________________________
iommu mailing list
iommu@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/iommu

Reply via email to