[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Xen-devel] Re: [GIT PULL RFC] pvclock cleanups and pvclock vsyscall support

On 10/15/2009 04:28 AM, Jeremy Fitzhardinge wrote:
Hi all,

This series contains several things:

  - Unify the separate vdso and vsyscall implementations of vgettimeofday and
    vgetcpu.  There's at least one bugfix which was only applied to one copy
    (ignore tcache in vgetcpu was only applied to the vdso version); this
    should avoid such skews in future.

  - Bug fixes for the Xen and KVM clocksource.read functions to make sure
    the returned time doesn't regress compared to clocksource.cycle_last.
    (Probably stable material.)

  - Make sure the pvclock rdtsc is surrounded by appropriate barriers so
    that it doesn't get speculated to the wrong place with respect to reading
    the time parameters.  (Probably stable material.)

  - General cleanups of the pvclock algorithm (there's no need to make a local
    copy of the time parameters before use).

  - Add a new CONFIG_X86_VSYSCALL to control the compilation of vsyscall-related
    code rather than just using CONFIG_X86_64 - we may want to implement 32-bit
    vsyscall at some point, and this will make it easier.

  - Add the sched notifier for task migration between CPUs, for use by
    pvclock vread.

  - Implement a pvclock vread function, so that pvclock-using clocksources can 
    used by vsyscall/vdso vgettimeofday and vclock_gettime.

  - Use pvclock vread in the Xen clocksource.

Looks good to me.

Acked-by etc.

I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.

Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.