| Message ID | ae4583176ee5c9d8a8a2d53c000b86711eb66287.1389018985.git.michal.simek@xilinx.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Mon, Jan 6, 2014 at 6:36 AM, Michal Simek <michal.simek@xilinx.com> wrote: > From: Steffen Trumtrar <s.trumtrar@pengutronix.de> > > For later usage by zynq clk driver. > > Signed-off-by: Steffen Trumtrar <s.trumtrar@pengutronix.de> > Signed-off-by: Michal Simek <michal.simek@xilinx.com> > --- > arch/arm/mach-zynq/slcr.c | 2 ++ > 1 file changed, 2 insertions(+) > > diff --git a/arch/arm/mach-zynq/slcr.c b/arch/arm/mach-zynq/slcr.c > index 9746de7..2891dcf 100644 > --- a/arch/arm/mach-zynq/slcr.c > +++ b/arch/arm/mach-zynq/slcr.c > @@ -121,6 +121,8 @@ int __init zynq_early_slcr_init(void) > BUG(); > } > > + np->data = (__force void *)zynq_slcr_base; This is quite unusual, we normally remap again if needed. Is there a reason you can't just have the driver request the resource and map it like most other ones do? -Olof
On 03/11/2014 10:26 PM, Olof Johansson wrote: > On Mon, Jan 6, 2014 at 6:36 AM, Michal Simek <michal.simek@xilinx.com> wrote: >> From: Steffen Trumtrar <s.trumtrar@pengutronix.de> >> >> For later usage by zynq clk driver. >> >> Signed-off-by: Steffen Trumtrar <s.trumtrar@pengutronix.de> >> Signed-off-by: Michal Simek <michal.simek@xilinx.com> >> --- >> arch/arm/mach-zynq/slcr.c | 2 ++ >> 1 file changed, 2 insertions(+) >> >> diff --git a/arch/arm/mach-zynq/slcr.c b/arch/arm/mach-zynq/slcr.c >> index 9746de7..2891dcf 100644 >> --- a/arch/arm/mach-zynq/slcr.c >> +++ b/arch/arm/mach-zynq/slcr.c >> @@ -121,6 +121,8 @@ int __init zynq_early_slcr_init(void) >> BUG(); >> } >> >> + np->data = (__force void *)zynq_slcr_base; > > > This is quite unusual, we normally remap again if needed. Is there a > reason you can't just have the driver request the resource and map it > like most other ones do? I took this patch from Steffen because remapping slcr again in clock driver is another mapping for the same IP. Use one ioremap region seems to me better if there is a way. Of course if this is strictly prohibited we can do another mapping in clock driver. Here is the RFC I have sent after my discussion with Arnd that we should use early syscon initialization and regmap. This is just RFC to show slcr cleaning up. http://permalink.gmane.org/gmane.linux.kernel/1651213 (Here is broken CLK part because clk subsystem uses clk_readl/clk_writel - changing this is discussed in other threads) From v1 - regmap device separation was applied by Mark git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regmap.git tags/nodev From v2 - "mfd: syscon: Support early initialization" I expected to get any comment from Arnd but I will respin this based on Lee's comments To summarize this for future I would like to use access with regmap with one ioremap. Thanks, Michal
diff --git a/arch/arm/mach-zynq/slcr.c b/arch/arm/mach-zynq/slcr.c index 9746de7..2891dcf 100644 --- a/arch/arm/mach-zynq/slcr.c +++ b/arch/arm/mach-zynq/slcr.c @@ -121,6 +121,8 @@ int __init zynq_early_slcr_init(void) BUG(); } + np->data = (__force void *)zynq_slcr_base; + /* unlock the SLCR so that registers can be changed */ writel(SLCR_UNLOCK_MAGIC, zynq_slcr_base + SLCR_UNLOCK_OFFSET);