[U-Boot] [PATCH v2 3/6] dm: x86: Update timer_get_boot_us to work before DM is ready

[Simon Glass]

Hi Bin,

On 12 September 2017 at 07:47, Bin Meng <bmeng.cn at gmail.com> wrote:
> Hi Simon,
>
> On Wed, Sep 6, 2017 at 9:39 AM, Simon Glass <sjg at chromium.org> wrote:
>> Hi Bin,
>>
>> On 27 August 2017 at 23:18, Bin Meng <bmeng.cn at gmail.com> wrote:
>>> Hi Simon,
>>>
>>> On Sun, Aug 27, 2017 at 11:23 PM, Simon Glass <sjg at chromium.org> wrote:
>>>> Use the new separate init function so that we can make use of the timer
>>>> before driver model is started up.
>>>>
>>>> At some point we should consider adding the microsecond timer to the timer
>>>> uclass interface since it would reduce the amount of plumbing here
>>>> slightly.
>>>>
>>>> Signed-off-by: Simon Glass <sjg at chromium.org>
>>>> ---
>>>>
>>>> Changes in v2:
>>>> - Rebase on top of early timer code and simplify slightly
>>>>
>>>>  drivers/timer/tsc_timer.c | 19 ++++++++++++++-----
>>>>  1 file changed, 14 insertions(+), 5 deletions(-)
>>>>
>>>
>>> I don't think this patch is needed. The early timer support was
>>> already provided in the patch [2/6] in this series.
>>
>> The current version of timer_get_boot_us() does not init the timer if
>> it is not set up, so will return incorrect values at the start.
>>
>
> timer_get_boot_us() in tsc_timer.c calls timer_get_us():
>
> ulong notrace timer_get_us(void)
> {
> return get_ticks() / get_tbclk_mhz();
> }
>
> get_ticks() will call timer_early_get_count() if CONFIG_TIMER_EARLY is on.
>
> Am I missing anything? Isn't it still broken when CONFIG_TIMER_EARLY is on?
>
>> So I think this patch is needed.
>>

We don't subtract the base time in that version, but in fact we don't
correctly do that in timer_get_boot_us() either. So yes, let's drop
this patch from the series. It still works without it, and we are in
no worse position.

Regards,
Simon