[CRIU] [PATCH 00/11] cpuinfo: Detect compact frames and handle noxsaves

[Cyrill Gorcunov]

On Fri, Jun 15, 2018 at 02:20:29PM +0100, Dmitry Safonov wrote:
> 2018-06-14 18:50 GMT+01:00 Andrei Vagin <avagin at virtuozzo.com>:
> > Dima, could you revew this series? Thanks!
> >
> > On Thu, Jun 14, 2018 at 08:27:19PM +0300, Cyrill Gorcunov wrote:
> >> At moment we don't support yet compacted fpu frames but a user
> >> can boot kernel with noxsaves option poiting kernel to not use
> >> compact frames at all, thus until we implement xsaves support
> >> one can pass --cpu-cap='op=noxsave' on checkpoint/restore
> >> and process c/r as usual.
> >>
> >> Cyrill Gorcunov (11):
> >>   x86: Use uint_x types in rt_sigcontext
> >>   crit: Show cpuinfo image
> >>   x86: cpu -- Sync bits with kernel
> >>   x86: cpu -- Add new instruction bits to check
> >>   x86: cpu -- Show additional xsave info on init
> >>   x86: cpu -- Check for xsaves bit on cpu validation
> >>   x86: cpu -- Report when unsupported compact xsave frame detected
> >>   compel: cpu -- Add ability to clear features
> >>   x86: cpu -- Add ability to ignore xsaves
> >>   x86: cpu -- Use rt information since it might we filtered
> >>   x86: cpu -- Show which exactly features are failing in fpu capability mode
> 
> LGTM,
> except patch 8 where I believe we can share function over arches,
> Reviewed-by: Dmitry Safonov <0x7f454c46 at gmail.com>
> 
> P.S.
> *grunting* it would be also nice if the third patch would be spilt over
> adding new cpuinfo image and pure sync part.
> It makes it easier to review, so please, do this in the next patches series
> where you will sync and change something in sources.

They can't be splitted, because new bits increment size of bitmask
and I think they should be addressed in one patch for easier reverting ;)