Re: MSI/MSI-X implementation and interrupt handling on i386/amd64

["Mathew, Cherry G." <cherry%zyx.in@localhost>]

On December 28, 2018 10:27:27 PM GMT+05:30, "Jaromír Doleček" <jaromir.dolecek%gmail.com@localhost> wrote:
>Le mar. 25 déc. 2018 à 10:17, Cherry G.Mathew <cherry%zyx.in@localhost> a écrit :
>> Thank you - all done. There are no functional changes, but there's a
>> duplication for xen (under #ifdef XEN / #endif) in the cmpxchg paths
>and
>> a 32bit version for xen in amd64/lock_stubs.S
>>
>> This is the only tricky bit if you are modifying the bitmap to 64bit.
>
>OK, I'll take care of the various paths.
>
>> It's ok to leave the XEN version to do 32bit because only NIPL of the
>> 32 pending bits are currently used. Not sure if that helps you.
>
>Would it then make sense to explicitly have uint16_t instead of
>uint32_t for ci_xpending/ci_xunmask/ci_xmask variables then, so it's
>visible it follows different rules than the non-xen intr fields?
>
>Jaromir

That seems like a good idea. 

Thanks!

-- 
Sent from my Android device with K-9 Mail. Please excuse my brevity.