| Message ID | 1446770267-26749-1-git-send-email-timur@codeaurora.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 11/05/2015 07:37 PM, Timur Tabi wrote: > Add the "sbsa32" command-line option to earlycon to specify that the > pl011 registers should be accessed with 32-bit accessors only. This is > SBSA platforms > > Similarly, also add the "zte" option for the ZTE UART. It's functionally > similar to the standard PL011, but some registers are at different > offsets. > > Normal console uses the "access_32b" field of the vendor data structure > to determine whether the registers needs 32-bit accessors. Early console > is used before the vendor data is available, so we need a different > mechanism to choose the accessor. > > Example: > > earlycon=pl011,0x3ced1000,sbsa32 Perhaps I wasn't clear enough in my previous comments regarding this approach; I see no benefit to re-using the pl011 earlycon definition. Just define a completely new earlycon for sbsa32 (and zte): EARLYCON_DECLARE(sbsa32, sbsa32_early_console_setup); Example: earlycon=sbsa32,0x3ced1000 Regards, Peter Hurley > Signed-off-by: Timur Tabi <timur@codeaurora.org> > --- > > This patch applies on top of Russell's 12-part amba-pl011.c patchset. > > drivers/tty/serial/amba-pl011.c | 46 +++++++++++++++++++++++++++++++++++++---- > 1 file changed, 42 insertions(+), 4 deletions(-) > > diff --git a/drivers/tty/serial/amba-pl011.c b/drivers/tty/serial/amba-pl011.c > index 373b152..58eda89 100644 > --- a/drivers/tty/serial/amba-pl011.c > +++ b/drivers/tty/serial/amba-pl011.c > @@ -2301,10 +2301,10 @@ static struct console amba_console = { > > static void pl011_putc(struct uart_port *port, int c) > { > - while (readl(port->membase + REG_FR) & UART01x_FR_TXFF) > + while (readl(port->membase + UART01x_FR) & UART01x_FR_TXFF) > ; > - writeb(c, port->membase + REG_DR); > - while (readl(port->membase + REG_FR) & UART01x_FR_BUSY) > + writeb(c, port->membase + UART01x_DR); > + while (readl(port->membase + UART01x_FR) & UART01x_FR_BUSY) > ; > } > > @@ -2315,13 +2315,51 @@ static void pl011_early_write(struct console *con, const char *s, unsigned n) > uart_console_write(&dev->port, s, n, pl011_putc); > } > > +static void pl011_putc_sbsa32(struct uart_port *port, int c) > +{ > + while (readl(port->membase + UART01x_FR) & UART01x_FR_TXFF) > + cpu_relax(); > + writel(c, port->membase + UART01x_DR); > + while (readl(port->membase + UART01x_FR) & UART01x_FR_BUSY) > + cpu_relax(); > +} > + > +static void pl011_early_write_sbsa32(struct console *con, const char *s, unsigned n) > +{ > + struct earlycon_device *dev = con->data; > + > + uart_console_write(&dev->port, s, n, pl011_putc_sbsa32); > +} > + > +static void pl011_putc_zte(struct uart_port *port, int c) > +{ > + while (readl(port->membase + ZX_UART011_FR) & UART01x_FR_TXFF) > + cpu_relax(); > + writel(c, port->membase + ZX_UART011_DR); > + while (readl(port->membase + ZX_UART011_FR) & UART01x_FR_BUSY) > + cpu_relax(); > +} > + > +static void pl011_early_write_zte(struct console *con, const char *s, unsigned n) > +{ > + struct earlycon_device *dev = con->data; > + > + uart_console_write(&dev->port, s, n, pl011_putc_zte); > +} > + > static int __init pl011_early_console_setup(struct earlycon_device *device, > const char *opt) > { > if (!device->port.membase) > return -ENODEV; > > - device->con->write = pl011_early_write; > + if (strcmp(device->options, "sbsa32") == 0) > + device->con->write = pl011_early_write_sbsa32; > + else if (strcmp(device->options, "zte") == 0) > + device->con->write = pl011_early_write_zte; > + else > + device->con->write = pl011_early_write; > + > return 0; > } > EARLYCON_DECLARE(pl011, pl011_early_console_setup); >
Peter Hurley wrote: > Perhaps I wasn't clear enough in my previous comments regarding this > approach; I see no benefit to re-using the pl011 earlycon definition. > Just define a completely new earlycon for sbsa32 (and zte): > > EARLYCON_DECLARE(sbsa32, sbsa32_early_console_setup); > > Example: > earlycon=sbsa32,0x3ced1000 Sounds like six of one, half-dozen of another. But I'm happy to modify my patch accordingly. It will avoid the strcmp, but I'm not sure it will reduce the actual number of lines of code.
On 11/05/2015 09:00 PM, Timur Tabi wrote: > Peter Hurley wrote: >> Perhaps I wasn't clear enough in my previous comments regarding this >> approach; I see no benefit to re-using the pl011 earlycon definition. >> Just define a completely new earlycon for sbsa32 (and zte): >> >> EARLYCON_DECLARE(sbsa32, sbsa32_early_console_setup); >> >> Example: >> earlycon=sbsa32,0x3ced1000 > > Sounds like six of one, half-dozen of another. But I'm happy to modify my patch accordingly. It will avoid the strcmp, but I'm not sure it will reduce the actual number of lines of code. Or better yet, for sbsa32 earlycon, just use the existing port type constructs. So, earlycon=pl011,mmio32,0x3ced1000 which will set port->iotype to UPIO_MEM32, which in turn can be used to select the correct i/o width in pl011_putc(). Regards, Peter Hurley
On 11/05/2015 08:44 PM, Peter Hurley wrote: > Or better yet, for sbsa32 earlycon, just use the existing port type constructs. > So, > > earlycon=pl011,mmio32,0x3ced1000 > > which will set port->iotype to UPIO_MEM32, which in turn can be used to select > the correct i/o width in pl011_putc(). That's a great idea. But why aren't we using it for pl011_write()? static void pl011_write(unsigned int val, const struct uart_amba_port *uap, unsigned int reg) { void __iomem *addr = uap->port.membase + pl011_reg_to_offset(uap, reg); if (uap->port.iotype == UPIO_MEM32) writel(val, addr); else writew(val, addr); }
On 11/06/2015 01:59 PM, Timur Tabi wrote: > On 11/05/2015 08:44 PM, Peter Hurley wrote: >> Or better yet, for sbsa32 earlycon, just use the existing port type constructs. >> So, >> >> earlycon=pl011,mmio32,0x3ced1000 >> >> which will set port->iotype to UPIO_MEM32, which in turn can be used to select >> the correct i/o width in pl011_putc(). > > That's a great idea. But why aren't we using it for pl011_write()? I'll note that when I review Russell's v2 series, if necessary. What is the plan to introduce the SBSA32 support? After the ACPI support is upstreamed? Regards, Peter Hurley > static void pl011_write(unsigned int val, const struct uart_amba_port *uap, > unsigned int reg) > { > void __iomem *addr = uap->port.membase + pl011_reg_to_offset(uap, reg); > > if (uap->port.iotype == UPIO_MEM32) > writel(val, addr); > else > writew(val, addr); > } > >
On Sat, Nov 07, 2015 at 05:36:34PM -0500, Peter Hurley wrote: > On 11/06/2015 01:59 PM, Timur Tabi wrote: > > On 11/05/2015 08:44 PM, Peter Hurley wrote: > >> Or better yet, for sbsa32 earlycon, just use the existing port type constructs. > >> So, > >> > >> earlycon=pl011,mmio32,0x3ced1000 > >> > >> which will set port->iotype to UPIO_MEM32, which in turn can be used to select > >> the correct i/o width in pl011_putc(). > > > > That's a great idea. But why aren't we using it for pl011_write()? > > I'll note that when I review Russell's v2 series, if necessary. And I'm not rushing to post them yet as the merge window is currently open.
Peter Hurley wrote: > I'll note that when I review Russell's v2 series, if necessary. > What is the plan to introduce the SBSA32 support? After the ACPI support is > upstreamed? There are some patches from Leif Lindholm that address that. However, they will probably need to be reworked on top of Russell's patch set.
diff --git a/drivers/tty/serial/amba-pl011.c b/drivers/tty/serial/amba-pl011.c index 373b152..58eda89 100644 --- a/drivers/tty/serial/amba-pl011.c +++ b/drivers/tty/serial/amba-pl011.c @@ -2301,10 +2301,10 @@ static struct console amba_console = { static void pl011_putc(struct uart_port *port, int c) { - while (readl(port->membase + REG_FR) & UART01x_FR_TXFF) + while (readl(port->membase + UART01x_FR) & UART01x_FR_TXFF) ; - writeb(c, port->membase + REG_DR); - while (readl(port->membase + REG_FR) & UART01x_FR_BUSY) + writeb(c, port->membase + UART01x_DR); + while (readl(port->membase + UART01x_FR) & UART01x_FR_BUSY) ; } @@ -2315,13 +2315,51 @@ static void pl011_early_write(struct console *con, const char *s, unsigned n) uart_console_write(&dev->port, s, n, pl011_putc); } +static void pl011_putc_sbsa32(struct uart_port *port, int c) +{ + while (readl(port->membase + UART01x_FR) & UART01x_FR_TXFF) + cpu_relax(); + writel(c, port->membase + UART01x_DR); + while (readl(port->membase + UART01x_FR) & UART01x_FR_BUSY) + cpu_relax(); +} + +static void pl011_early_write_sbsa32(struct console *con, const char *s, unsigned n) +{ + struct earlycon_device *dev = con->data; + + uart_console_write(&dev->port, s, n, pl011_putc_sbsa32); +} + +static void pl011_putc_zte(struct uart_port *port, int c) +{ + while (readl(port->membase + ZX_UART011_FR) & UART01x_FR_TXFF) + cpu_relax(); + writel(c, port->membase + ZX_UART011_DR); + while (readl(port->membase + ZX_UART011_FR) & UART01x_FR_BUSY) + cpu_relax(); +} + +static void pl011_early_write_zte(struct console *con, const char *s, unsigned n) +{ + struct earlycon_device *dev = con->data; + + uart_console_write(&dev->port, s, n, pl011_putc_zte); +} + static int __init pl011_early_console_setup(struct earlycon_device *device, const char *opt) { if (!device->port.membase) return -ENODEV; - device->con->write = pl011_early_write; + if (strcmp(device->options, "sbsa32") == 0) + device->con->write = pl011_early_write_sbsa32; + else if (strcmp(device->options, "zte") == 0) + device->con->write = pl011_early_write_zte; + else + device->con->write = pl011_early_write; + return 0; } EARLYCON_DECLARE(pl011, pl011_early_console_setup);
Add the "sbsa32" command-line option to earlycon to specify that the pl011 registers should be accessed with 32-bit accessors only. This is SBSA platforms Similarly, also add the "zte" option for the ZTE UART. It's functionally similar to the standard PL011, but some registers are at different offsets. Normal console uses the "access_32b" field of the vendor data structure to determine whether the registers needs 32-bit accessors. Early console is used before the vendor data is available, so we need a different mechanism to choose the accessor. Example: earlycon=pl011,0x3ced1000,sbsa32 Signed-off-by: Timur Tabi <timur@codeaurora.org> --- This patch applies on top of Russell's 12-part amba-pl011.c patchset. drivers/tty/serial/amba-pl011.c | 46 +++++++++++++++++++++++++++++++++++++---- 1 file changed, 42 insertions(+), 4 deletions(-)