Re: [PATCH v2 1/4] x86/hyper-v: move synic/stimer control structures definitions to hyperv-tlfs.h

From: Nadav Amit
Date: Tue Nov 27 2018 - 20:49:36 EST

> On Nov 27, 2018, at 10:48 AM, Roman Kagan <rkagan@xxxxxxxxxxxxx> wrote:
> On Tue, Nov 27, 2018 at 02:10:49PM +0100, Vitaly Kuznetsov wrote:
>> Roman Kagan <rkagan@xxxxxxxxxxxxx> writes:
>>> On Mon, Nov 26, 2018 at 04:47:29PM +0100, Vitaly Kuznetsov wrote:
>>> I personally tend to prefer masks over bitfields, so I'd rather do the
>>> consolidation in the opposite direction: use the definitions in
>>> hyperv-tlfs.h and replace those unions/bitfields elsewhere. (I vaguely
>>> remember posting such a patchset a couple of years ago but I lacked the
>>> motivation to complete it).
>> Are there any known advantages of using masks over bitfields or the
>> resulting binary code is the same?
> Strictly speaking bitwise ops are portable while bitfields are not, but
> I guess this is not much of an issue with gcc which is dependable to
> produce the right thing.
> I came to dislike the bitfields for the false feeling of atomicity of
> assignment while most of the time they are read-modify-write operations.
> And no, I don't feel strong about it, so if nobody backs me on this I
> give up :)

Last time I tried to push a change from bitmasks to bitfields I failed:

On a different note: how come all of the hyper-v structs are not marked
with the âpacked" attribute?