From mboxrd@z Thu Jan  1 00:00:00 1970
From: Graeme Russ <graeme.russ@gmail.com>
Date: Wed, 25 May 2011 21:52:21 +1000
Subject: [U-Boot] [RFC] Review of U-Boot timer API
In-Reply-To: <20110525113738.E49DFCFB8A4@gemini.denx.de>
References: <4DD7B245.5000008@gmail.com>
	<BANLkTineVAT1b_nWV8zPyfjsVnRz=6DhPg@mail.gmail.com>
	<201105240024.26511.vapier@gentoo.org>
	<BANLkTi=ayrqMGzjT-EvU3Hq3ZJbo=_8oOw@mail.gmail.com>
	<4DDB42B1.2010206@emk-elektronik.de>
	<BANLkTini5tCMFjaKkeJaehqQ0AokEv5_0g@mail.gmail.com>
	<4DDBB29D.2050102@psyent.com>
	<20110524141950.2AB1CCF5DBF@gemini.denx.de>
	<4DDBE22D.6050806@gmail.com>
	<20110524191948.401CBCF5DBF@gemini.denx.de>
	<BANLkTinK+bxtZ+sumYZtvuHcd_Ft+jfyPw@mail.gmail.com>
	<4DDC6F46.4090809@comcast.net>
	<20110525052836.11A30CF5DBF@gemini.denx.de>
	<BANLkTinZ_q8_vqd9mzm7BY64Hi=zuyFZTQ@mail.gmail.com>
	<20110525080837.5925BCF5DBF@gemini.denx.de>
	<4DDCBFF4.40002@gmail.com>
	<20110525113738.E49DFCFB8A4@gemini.denx.de>
Message-ID: <4DDCED75.8030906@gmail.com>
List-Id: <u-boot.lists.denx.de>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
To: u-boot@lists.denx.de

Hi Wolfgang,

On 25/05/11 21:37, Wolfgang Denk wrote:
> Dear Graeme Russ,
> 
> In message <4DDCBFF4.40002@gmail.com> you wrote:
>>
>>> I don't get you.  In such a system, the interrupt would be the tick
>>> (see the PPC implementation).  If you miss interrupts, you miss ticks.
>>
>> Yes, you miss ticks, but if the hardware is keeping the tick counter
>> current in the background (i.e. without software or interrupts) then this
>> does not matter - The prescaler takes care of this...
> 
> I think you don't want to understand.
> 
> Look at the PPC implementation.  The decrementer causes an interrupt
> every millisecond.  This is our tick.  We never ever actually read the
> decrementer register itself.  We just use the interrupt.

x86 is the same - a 1ms interrupt which increments the timer

>> The key is the prescaler takes a 'tick delta' between the last time it was
>> called and now, uses the 'tick frequency' to calculate a corresponding
>> 'timer delta' which it adds to the current timer
> 
> The "tick frequency" in above example is 1,000 Hz, and determined by
> the frequency of the interrupts.  There is not any "timer count
> register" we are accessing here.
> 
>> This relies on the tick counter wrapping properly.
> 
> The "tick counter" is already in the higher level, i. e. implemented
> in software, without any hardware based registers.

And in the PPC case - No prescaler and life is easy

BUT - I just found something very disturbing in /board/zylonite/flash.c -
Interrupts are disables before the timer is used to detect a timeout!

Yes, this is in a board specific flash driver which is bad but who knows
where else this might (quite unintentionally) occur

No if you have:

	start = get_timer(0)

	while (hardware_busy) {
		if (get_timer(start) > TIMEOUT) {
			/* deal with timeout */
			break;
		}
		/* Do stuf (poll hardware etc) */
	}

Now every time get_timer() is called, the prescaler updates the timer from
the current tick count - This cannot fail if interrupts are inadvertently
disabled.

And if you do not care about the timer updating while you are not 'using
it' (i.e. checking for a timeout) - it doesn't really matter - You will get
a glitch, but this glitch is 'reset' by the call to get_timer() and your
timeout check starts a new timer increment afresh

Regards,

Graeme