2017-12-13 17:45 GMT+08:00 Guo Ren [off-list ref]:

Hello,

CPU team could improve the tlbop_*. Eg: Design a hardware
internal flag bit for SR_TLB_VPN, tlbop_* will invalid it and mtsr
SR_TLB_VPN will valid it.

So:
On Wed, Dec 13, 2017 at 05:03:33PM +0800, Greentime Hu wrote:

quoted

mtsr addr1 NDS32_SR_TLB_VPN
    interrupt coming
        mtsr addr2 NDS32_SR_TLB_VPN <- TLB_VPN has been set to addr2

        mtsr SR_TLB_VPN will valid the flag bit

quoted

        tlbop_rwr(*pte);

        tlbop_rwr will invalid SR_TLB_VPN flag bit

quoted

    interrupt finish
tlbop_rwr(*pte); <- it will use the wrong TLB_VPN

Because SR_TLB_VPN is in a invalid state, no operation happen on
tlbop_rwr.

Then they are atomic safe ,no spin_lock_irq need.
:)

Oh, I see. I may propose this idea to our ARCH colleagues for the next
version design.
Many thanks.
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html