The following problem was found by Giovambattista Pulcini
<gpulcini@swintel.it>, who also provided a partial patch, and this has been
verified by our time guru Gabriel Paubert <paubert@iram.es>.
The problem is that in do_settimeofday() we always set time_state to
TIME_ERROR and except on two platforms, never re-set it.  This meant that
ntp_gettime() and ntp_adjtime() always returned TIME_ERROR, incorrectly.
Based on Gabriel's analysis, time_state is used for leap-second processing,
and ppc shouldn't be mucking with it.
Signed-off-by: Tom Rini <trini@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
 
        time_adjust = 0;                /* stop active adjtime() */
        time_status |= STA_UNSYNC;
-       time_state = TIME_ERROR;        /* p. 24, (a) */
        time_maxerror = NTP_PHASE_LIMIT;
        time_esterror = NTP_PHASE_LIMIT;
        write_sequnlock_irqrestore(&xtime_lock, flags);
 
        chrp_cmos_clock_write(save_control, RTC_CONTROL);
        chrp_cmos_clock_write(save_freq_select, RTC_FREQ_SELECT);
 
-       if ( (time_state == TIME_ERROR) || (time_state == TIME_BAD) )
-               time_state = TIME_OK;
        spin_unlock(&rtc_lock);
        return 0;
 }
 
        /* done writing */
        gemini_rtc_write(reg, M48T35_RTC_CONTROL);
 
-       if ((time_state == TIME_ERROR) || (time_state == TIME_BAD))
-               time_state = TIME_OK;
-
        return 0;
 }