| Message ID | 1448776863-13789-1-git-send-email-thomas@wytron.com.tw |
|---|---|
| State | Accepted |
| Delegated to: | Tom Rini |
| Headers | show |
On Sun, Nov 29, 2015 at 2:01 PM, Thomas Chou <thomas@wytron.com.tw> wrote: > Change reg-shift property default to zero. When the integer property > is missing, it should be taken as zero. This is consistent to Linux > drivers/tty/serial/of_serial.c. > > The x86 and most powerpc use reg-shift of 0. Most others use reg-shift > of 2. While reg-shift of 1 is rarely used. > > Signed-off-by: Thomas Chou <thomas@wytron.com.tw> > --- > drivers/serial/ns16550.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/drivers/serial/ns16550.c b/drivers/serial/ns16550.c > index d5bcbc3..5b64f7c 100644 > --- a/drivers/serial/ns16550.c > +++ b/drivers/serial/ns16550.c > @@ -403,7 +403,7 @@ int ns16550_serial_ofdata_to_platdata(struct udevice *dev) > > plat->base = addr; > plat->reg_shift = fdtdec_get_int(gd->fdt_blob, dev->of_offset, > - "reg-shift", 1); > + "reg-shift", 0); > plat->clock = fdtdec_get_int(gd->fdt_blob, dev->of_offset, > "clock-frequency", > CONFIG_SYS_NS16550_CLK); > -- Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
On Sunday 29 November 2015 11:31 AM, Thomas Chou wrote: > Change reg-shift property default to zero. When the integer property > is missing, it should be taken as zero. This is consistent to Linux > drivers/tty/serial/of_serial.c. > > The x86 and most powerpc use reg-shift of 0. Most others use reg-shift > of 2. While reg-shift of 1 is rarely used. > > Signed-off-by: Thomas Chou <thomas@wytron.com.tw> Reviewed-by: Mugunthan V N <mugunthanvnm@ti.com> Regards Mugunthan V N
On 29 November 2015 at 23:07, Mugunthan V N <mugunthanvnm@ti.com> wrote: > On Sunday 29 November 2015 11:31 AM, Thomas Chou wrote: >> Change reg-shift property default to zero. When the integer property >> is missing, it should be taken as zero. This is consistent to Linux >> drivers/tty/serial/of_serial.c. >> >> The x86 and most powerpc use reg-shift of 0. Most others use reg-shift >> of 2. While reg-shift of 1 is rarely used. >> >> Signed-off-by: Thomas Chou <thomas@wytron.com.tw> > > Reviewed-by: Mugunthan V N <mugunthanvnm@ti.com> Acked-by: Simon Glass <sjg@chromium.org>
On Sun, Nov 29, 2015 at 02:01:03PM +0800, Thomas Chou wrote: > Change reg-shift property default to zero. When the integer property > is missing, it should be taken as zero. This is consistent to Linux > drivers/tty/serial/of_serial.c. > > The x86 and most powerpc use reg-shift of 0. Most others use reg-shift > of 2. While reg-shift of 1 is rarely used. > > Signed-off-by: Thomas Chou <thomas@wytron.com.tw> > Reviewed-by: Bin Meng <bmeng.cn@gmail.com> > Reviewed-by: Mugunthan V N <mugunthanvnm@ti.com> > Acked-by: Simon Glass <sjg@chromium.org> Applied to u-boot/master, thanks!
On 12/06/2015 03:07 PM, Tom Rini wrote: > On Sun, Nov 29, 2015 at 02:01:03PM +0800, Thomas Chou wrote: > >> Change reg-shift property default to zero. When the integer property >> is missing, it should be taken as zero. This is consistent to Linux >> drivers/tty/serial/of_serial.c. It's not generally true that missing properties have value zero. Rather, whatever values was assumed by the semantics of the binding before that optional property was defined should be assumed. What the Linux kernel driver does also isn't justification for this change, since DT bindings define how they work, rather than a particular OS implementation forcing the hand of the binding. The DT binding documentation must state the default value/semantics for any optional property. Can you please make sure the DT binding documentation is updated to describe this case? (Note that I have no objection to this patch; the actual change seems fine. It's simply that the justifications given in the patch description for it aren't entirely robust.)
Hi Stephen, On 2015年12月08日 01:32, Stephen Warren wrote: > On 12/06/2015 03:07 PM, Tom Rini wrote: >> On Sun, Nov 29, 2015 at 02:01:03PM +0800, Thomas Chou wrote: >> >>> Change reg-shift property default to zero. When the integer property >>> is missing, it should be taken as zero. This is consistent to Linux >>> drivers/tty/serial/of_serial.c. > > It's not generally true that missing properties have value zero. Rather, > whatever values was assumed by the semantics of the binding before that > optional property was defined should be assumed. > > What the Linux kernel driver does also isn't justification for this > change, since DT bindings define how they work, rather than a particular > OS implementation forcing the hand of the binding. > > The DT binding documentation must state the default value/semantics for > any optional property. Can you please make sure the DT binding > documentation is updated to describe this case? > > (Note that I have no objection to this patch; the actual change seems > fine. It's simply that the justifications given in the patch description > for it aren't entirely robust.) > Thanks. Will send a follow-up patch for the DT binding. Best regards, Thomas
diff --git a/drivers/serial/ns16550.c b/drivers/serial/ns16550.c index d5bcbc3..5b64f7c 100644 --- a/drivers/serial/ns16550.c +++ b/drivers/serial/ns16550.c @@ -403,7 +403,7 @@ int ns16550_serial_ofdata_to_platdata(struct udevice *dev) plat->base = addr; plat->reg_shift = fdtdec_get_int(gd->fdt_blob, dev->of_offset, - "reg-shift", 1); + "reg-shift", 0); plat->clock = fdtdec_get_int(gd->fdt_blob, dev->of_offset, "clock-frequency", CONFIG_SYS_NS16550_CLK);
Change reg-shift property default to zero. When the integer property is missing, it should be taken as zero. This is consistent to Linux drivers/tty/serial/of_serial.c. The x86 and most powerpc use reg-shift of 0. Most others use reg-shift of 2. While reg-shift of 1 is rarely used. Signed-off-by: Thomas Chou <thomas@wytron.com.tw> --- drivers/serial/ns16550.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)