by Xunlei Pang

[permalink] [raw]

Subject: Re: [PATCH RFC 11/12] time: Convert xen_read_wallclock() to use timespec64

On 27 October 2014 23:35, Stefano Stabellini
<[email protected]> wrote:
> On Mon, 27 Oct 2014, pang.xunlei wrote:
>> The kernel uses 32-bit signed value(time_t) for seconds since 1970-01-01:00:00:00, so it
>> will overflow at 2038-01-19 03:14:08 on 32-bit systems. We call this "2038 safety" issue.
>>
>> As part of addressing 2038 safety for in-kernel uses, this patch creates no functional change
>> in existing users, converts xen_read_wallclock() to use timespec64 instead of timespec.
>>
>> Signed-off-by: pang.xunlei <[email protected]>
>
> Please CC [email protected] for Xen related patches.
Ok, apologies for missing this.
>
>
>> arch/x86/xen/time.c | 31 +++++++++++++++++++------------
>> 1 file changed, 19 insertions(+), 12 deletions(-)
>>
>> diff --git a/arch/x86/xen/time.c b/arch/x86/xen/time.c
>> index cf1b591..2ce064a 100644
>> --- a/arch/x86/xen/time.c
>> +++ b/arch/x86/xen/time.c
>> @@ -169,20 +169,33 @@ static cycle_t xen_clocksource_get_cycles(struct clocksource *cs)
>> return xen_clocksource_read();
>> }
>>
>> -static void xen_read_wallclock(struct timespec *ts)
>> +static void xen_read_wallclock(struct timespec64 *ts)
>> {
>> + struct timespec ts32;
>> struct shared_info *s = HYPERVISOR_shared_info;
>> struct pvclock_wall_clock *wall_clock = &(s->wc);
>> struct pvclock_vcpu_time_info *vcpu_time;
>>
>> vcpu_time = &get_cpu_var(xen_vcpu)->time;
>> - pvclock_read_wallclock(wall_clock, vcpu_time, ts);
>> + /*
>> + * TODO: [2038 safety] pvclock_read_wallclock() should be changed
>> + * to use timespec64 for 2038 safety as soon as possible.
>> + */
>> + pvclock_read_wallclock(wall_clock, vcpu_time, &ts32);
>> + *ts = timespec_to_timespec64(ts32);
>> put_cpu_var(xen_vcpu);
>> }
>>
>> +/*
>> + * TODO: [2038 safety] xen_get_wallclock() should be changed to use
>> + * timespec64 for 2038 safety as soon as possible.
>> + */
>> static void xen_get_wallclock(struct timespec *now)
>
> At this point shouldn't you convert xen_get_wallclock to timespec64 too?
This conversion is in the following patches. I'll send the version 2 tomorrow
which will demonstrate this in the first subset.
>
>
>> {
>> - xen_read_wallclock(now);
>> + struct timespec64 now64;
>> +
>> + xen_read_wallclock(&now64);
>> + *now = timespec64_to_timespec(now64);
>> }
>>
>> static int xen_set_wallclock(const struct timespec *now)
>> @@ -485,8 +498,7 @@ static const struct pv_time_ops xen_time_ops __initconst = {
>> static void __init xen_time_init(void)
>> {
>> int cpu = smp_processor_id();
>> - struct timespec tp;
>> - struct timespec64 tp64;
>> + struct timespec64 tp;
>>
>> clocksource_register_hz(&xen_clocksource, NSEC_PER_SEC);
>>
>> @@ -497,14 +509,9 @@ static void __init xen_time_init(void)
>> xen_clockevent = &xen_vcpuop_clockevent;
>> }
>>
>> - /*
>> - * Set initial system time with full resolution.
>> - * TODO: [2038 safety] xen_read_wallclock() should be changed to use
>> - * timespec64 for 2038 safety as soon as possible.
>> - */
>> + /* Set initial system time with full resolution. */
>> xen_read_wallclock(&tp);
>> - tp64 = timespec_to_timespec64(tp);
>> - do_settimeofday(&tp64);
>> + do_settimeofday(&tp);
>>
>> setup_force_cpu_cap(X86_FEATURE_TSC);
>>
>> --
>> 1.7.9.5
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
>> the body of a message to [email protected]
>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>> Please read the FAQ at http://secure-web.cisco.com/1R3n2I8-gDpn8akesS4B4JN0A1SlyiiKzIT0qgddW4Xm1xtBi5J7nJwhRaPMEkp0VmrMyOlBD5Lz9VdewgeobSUsSzhREJQNT2O0l1LUhfiaNzD2_oHWGA6fP4sD1KCgFB00xHD6jw1ggEqe03cpaKsGQmA-TwOpFD41O8IIqhwY/http%3A%2F%2Fwww.tux.org%2Flkml%2F
>>