[PATCH 08/10] ARM: mxs: add ocotp read function

[Shawn Guo]

On Fri, Dec 31, 2010 at 05:11:41PM +0100, Uwe Kleine-König wrote:
> Hello Shawn,
> 
> On Fri, Dec 31, 2010 at 09:43:55AM +0800, Shawn Guo wrote:
> > On Thu, Dec 30, 2010 at 10:15:41AM +0100, Uwe Kleine-König wrote:
> > > Hello Shawn,
> > > 
> > > On Thu, Dec 30, 2010 at 01:50:12PM +0800, Shawn Guo wrote:
> > > > On Wed, Dec 29, 2010 at 12:22:08PM +0100, Uwe Kleine-König wrote:
> > > > > On Tue, Dec 28, 2010 at 10:55:53PM +0800, Shawn Guo wrote:
> > > > > > diff --git a/arch/arm/mach-mxs/ocotp.c b/arch/arm/mach-mxs/ocotp.c
> > > > > > new file mode 100644
> > > > > > index 0000000..24457d7
> > > > > > --- /dev/null
> > > > > > +++ b/arch/arm/mach-mxs/ocotp.c
> > > > > > @@ -0,0 +1,52 @@
> > > > > > +/*
> > > > > > + * Copyright 2010 Freescale Semiconductor, Inc. All Rights Reserved.
> > > > > > + *
> > > > > > + * This program is free software; you can redistribute it and/or modify
> > > > > > + * it under the terms of the GNU General Public License as published by
> > > > > > + * the Free Software Foundation; either version 2 of the License, or
> > > > > > + * (at your option) any later version.
> > > > > > + *
> > > > > > + * This program is distributed in the hope that it will be useful,
> > > > > > + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> > > > > > + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> > > > > > + * GNU General Public License for more details.
> > > > > > + */
> > > > > > +
> > > > > > +#include <linux/delay.h>
> > > > > > +#include <linux/err.h>
> > > > > > +
> > > > > > +#include <mach/mxs.h>
> > > > > > +
> > > > > > +#define BM_OCOTP_CTRL_RD_BANK_OPEN	(1 << 12)
> > > > > > +#define BM_OCOTP_CTRL_BUSY		(1 << 8)
> > > > > > +
> > > > > > +int mxs_read_ocotp(int offset, int count, u32 *values)
> > > > > > +{
> > > > > > +	void __iomem *ocotp_base = MXS_IO_ADDRESS(MXS_OCOTP_BASE_ADDR);
> > > > > > +	int i, timeout = 0x400;
> > > > > > +
> > > > > > +	/* open OCOTP banks for read */
> > > > > > +	__mxs_setl(BM_OCOTP_CTRL_RD_BANK_OPEN, ocotp_base);
> > > > > The reference manual specifies:
> > > > > 	1. Program the HCLK to a frequency up to the maximum allowable HCLK
> > > > > 	   frequency. [...]
> > > > > 	2. Check that HW_OCOTP_CTRL_BUSY and HW_OCOTP_CTRL_ERROR are clear.
> > > > > 	3. Set HW_OCOTP_CTRL_RD_BANK_OPEN. [...]
> > > > > 
> > > > > 1. isn't done (which is probably OK, or should it aquire a clk?)
> > > > > For 2. there is no check for HW_OCOTP_CTRL_ERROR which is not OK i
> > > > > guess?!
> > > > > 
> > > > Will add the check. Thanks.
> > > > 
> > > > > (Independant of that, shouldn't the list use BM_ prefixes as you did in
> > > > > the code?)
> > > > > 
> > > > Document description uses register name followed by bit-field name
> > > > to mean the bit.
> > > After rereading chapter 39.4 (Register Macro Usage/Naming Convention)
> > > I wonder if the BM_ values are shifted or not.  If they are not (as I
> > > think it is meant) your usage is inconsitent with the manual (but IMHO
> > > sane).  sigh.
> > > 
> > They are shifted. You can check regs-clkctrl-mx28.h as an example.
> > 
> > > > > > +	/* approximately wait 32 hclk cycles */
> > > > > > +	udelay(1);
> > > > > > +
> > > > > > +	/* poll BUSY bit becoming cleared */
> > > > > > +	while ((__raw_readl(ocotp_base) & BM_OCOTP_CTRL_BUSY) && --timeout)
> > > > > > +		/* nothing */;
> > > > > > +
> > > > > > +	if (unlikely(!timeout))
> > > > > > +		goto error;
> > > > > And I think you should recheck for BM_OCOTP_CTRL_RD_BANK_OPEN here (step
> > > > > 4 from the reference manual).
> > > > > 
> > > > The RM says:
> > > > 
> > > > Poll for HW_OCOTP_CTRL_BUSY clear. When HW_OCOTP_CTRL_BUSY is clear
> > > > and HW_OCOTP_CTRL_RD_BANK_OPEN is set, read the data from the
> > > > appropriate memory-mapped address.
> > > > 
> > > > My understanding is we only need to poll busy bit clear. And the
> > > > second sentence just tells that read operation should be working
> > > > when bit busy clear and rd_bank_open set, which has been done
> > > > in step 3. I also consulted the designer and was told that only
> > > > busy clear polling is needed.
> > > I expected to need this (shortend the names to improve readability):
> > > 
> > IMHO, BM_OCOTP_CTRL_BUSY is more readable than BUSY. Reader can
> > understand which bit it is more easily than BUSY.
> I only shortend it for the mail.  For a patch of course use the full
> name.
>  
> > > 	while (__raw_readl(ocotp_base) & BUSY) && --timeout);
> > > 
> > > 	if (unlikely((__raw_readl(ocotp_base) & (BUSY | RDBOPEN)) != RDBOPEN))
> > > 		goto error_unlock;
> > > 
> > I still did not get it. When we set bit RDBOPEN in step 3, it gets set.
> > That is it. Why do we still need to bother the unnecessary checking?
> That's how I interpret the hardware reference.  And note, registers are
> not memory.  Just because you write a value to it it doesn't mean it is
> set when you reread it.
> 
The note is right. That why I asked designer whether the bit will
definitely get set or not if software sets the bit implicitly.
The answer I got is yes.

> Best regards and happy new year (don't know if you have that? :-)
I have that holiday, but I still need to check your reply ;)

> Uwe
> 
> -- 
> Pengutronix e.K.                           | Uwe Kleine-König            |
> Industrial Linux Solutions                 | http://www.pengutronix.de/  |
> 

-- 
Regards,
Shawn