| Message ID | 20180315020445.150604-4-djkurtz@chromium.org (mailing list archive) |
|---|---|
| State | Not Applicable, archived |
| Headers | show |
On Wed, 2018-03-14 at 20:04 -0600, Daniel Kurtz wrote: > The old_serial_port global array in 8250_core is supposed to hold an > entry > for each serial port on the system that cannot be discovered via a > standard enumeration mechanism (aka ACPI/PCI/DTS). The array is > populated > at compile-time from the value specified in the SERIAL_PORT_DFNS > macro. > This macro is defined in arch/serial.h. > > For x86, this macro is currently unconditionally initialized to supply > four ioport UARTs (0x3F8, 0x2F8, 0x3E8, 0x2E8). > > However, not all x86 CPUs have these four ioport UARTs. For example, > the > UARTs on AMD Carrizo and later are separate memory mapped Designware > IP > blocks. > > Fairly early in boot the console_initcall univ8250_console_init > iterates > over this array and installs these old UARTs into the global array > serial8250_ports. Further, it attempts to register them for use as > the console. In other words, if, for example, the kernel commandline > has > console=ttyS0, the console will be switched over to one of these > non-existent UARTs. Only later, when the real UART drivers are probed > and their devices are instantiated will the console switch back over > to > the proper UART. > > This is noticeable when using earlycon, since part of the serial > console > log will appear to disappear (when the bogus old takes over) and then > re-appear (when the real UART finally gets registered for the > console). > > The problem is even more noticable when *not* using earlycon, since in > this case the entire console output is missing, having been > incorrectly > played back to the non-existing serial port. > > Create a global variable to allow skipping old serial port > initialization > and wire it up to the AMDCZ ACPI SPCR quirk and the special amdcz > earlycon > setup handler. I don't like this approach at all. But unfortunately I have nothing to propose. Just felt like I have to share my opinion on this.
On Thu, Mar 15, 2018 at 7:04 AM Andy Shevchenko < andriy.shevchenko@linux.intel.com> wrote: > On Wed, 2018-03-14 at 20:04 -0600, Daniel Kurtz wrote: > > The old_serial_port global array in 8250_core is supposed to hold an > > entry > > for each serial port on the system that cannot be discovered via a > > standard enumeration mechanism (aka ACPI/PCI/DTS). The array is > > populated > > at compile-time from the value specified in the SERIAL_PORT_DFNS > > macro. > > This macro is defined in arch/serial.h. > > > > For x86, this macro is currently unconditionally initialized to supply > > four ioport UARTs (0x3F8, 0x2F8, 0x3E8, 0x2E8). > > > > However, not all x86 CPUs have these four ioport UARTs. For example, > > the > > UARTs on AMD Carrizo and later are separate memory mapped Designware > > IP > > blocks. > > > > Fairly early in boot the console_initcall univ8250_console_init > > iterates > > over this array and installs these old UARTs into the global array > > serial8250_ports. Further, it attempts to register them for use as > > the console. In other words, if, for example, the kernel commandline > > has > > console=ttyS0, the console will be switched over to one of these > > non-existent UARTs. Only later, when the real UART drivers are probed > > and their devices are instantiated will the console switch back over > > to > > the proper UART. > > > > This is noticeable when using earlycon, since part of the serial > > console > > log will appear to disappear (when the bogus old takes over) and then > > re-appear (when the real UART finally gets registered for the > > console). > > > > The problem is even more noticable when *not* using earlycon, since in > > this case the entire console output is missing, having been > > incorrectly > > played back to the non-existing serial port. > > > > Create a global variable to allow skipping old serial port > > initialization > > and wire it up to the AMDCZ ACPI SPCR quirk and the special amdcz > > earlycon > > setup handler. > I don't like this approach at all. > But unfortunately I have nothing to propose. > Just felt like I have to share my opinion on this. I'm not very proud of this either, but I don't know how else to set this flag. I proposed a Kconfig option earlier, but that also met resistance. Perhaps another kernel command line option? > -- > Andy Shevchenko <andriy.shevchenko@linux.intel.com> > Intel Finland Oy -- To unsubscribe from this list: send the line "unsubscribe linux-acpi" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi Daniel,
Thank you for the patch! Yet something to improve:
[auto build test ERROR on v4.16-rc4]
[also build test ERROR on next-20180316]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]
url: https://github.com/0day-ci/linux/commits/Daniel-Kurtz/Add-earlycon-support-for-AMD-Carrizo-Stoneyridge/20180318-024830
config: ia64-allmodconfig (attached as .config)
compiler: ia64-linux-gcc (GCC) 7.2.0
reproduce:
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
chmod +x ~/bin/make.cross
# save the attached .config to linux build tree
make.cross ARCH=ia64
All errors (new ones prefixed by >>):
drivers/acpi/spcr.o: In function `acpi_parse_spcr':
>> spcr.c:(.init.text+0x540): undefined reference to `serial8250_skip_old_ports'
spcr.c:(.init.text+0x560): undefined reference to `serial8250_skip_old_ports'
---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation
diff --git a/drivers/acpi/spcr.c b/drivers/acpi/spcr.c index 52d840d0e05b..d1affdf2bc37 100644 --- a/drivers/acpi/spcr.c +++ b/drivers/acpi/spcr.c @@ -14,6 +14,7 @@ #include <linux/acpi.h> #include <linux/console.h> #include <linux/kernel.h> +#include <linux/serial_8250.h> #include <linux/serial_core.h> /* @@ -208,6 +209,8 @@ int __init acpi_parse_spcr(bool enable_earlycon, bool enable_console) } if (amdcz_present(table)) { + if (IS_ENABLED(CONFIG_SERIAL_8250)) + serial8250_skip_old_ports = true; if (enable_earlycon) uart = "amdcz"; } diff --git a/drivers/tty/serial/8250/8250_core.c b/drivers/tty/serial/8250/8250_core.c index 9342fc2ee7df..a04da392a251 100644 --- a/drivers/tty/serial/8250/8250_core.c +++ b/drivers/tty/serial/8250/8250_core.c @@ -66,6 +66,9 @@ static unsigned int skip_txen_test; /* force skip of txen test at init time */ #define SERIAL_PORT_DFNS #endif +bool serial8250_skip_old_ports; +EXPORT_SYMBOL(serial8250_skip_old_ports); + static const struct old_serial_port old_serial_port[] = { SERIAL_PORT_DFNS /* defined in asm/serial.h */ }; @@ -537,6 +540,9 @@ static void __init serial8250_isa_init_ports(void) if (share_irqs) irqflag = IRQF_SHARED; + if (serial8250_skip_old_ports) + return; + for (i = 0, up = serial8250_ports; i < ARRAY_SIZE(old_serial_port) && i < nr_uarts; i++, up++) { diff --git a/drivers/tty/serial/8250/8250_early.c b/drivers/tty/serial/8250/8250_early.c index c6bf971a6038..288d2be82990 100644 --- a/drivers/tty/serial/8250/8250_early.c +++ b/drivers/tty/serial/8250/8250_early.c @@ -202,6 +202,7 @@ static int __init early_amdcz_setup(struct earlycon_device *dev, { struct uart_port *port = &dev->port; + serial8250_skip_old_ports = true; port->uartclk = 48000000; return early_serial8250_setup(dev, opt); diff --git a/include/linux/serial_8250.h b/include/linux/serial_8250.h index a27ef5f56431..ae739c34546f 100644 --- a/include/linux/serial_8250.h +++ b/include/linux/serial_8250.h @@ -136,6 +136,8 @@ struct uart_8250_port { struct uart_8250_em485 *em485; }; +extern bool serial8250_skip_old_ports; + static inline struct uart_8250_port *up_to_u8250p(struct uart_port *up) { return container_of(up, struct uart_8250_port, port);
The old_serial_port global array in 8250_core is supposed to hold an entry for each serial port on the system that cannot be discovered via a standard enumeration mechanism (aka ACPI/PCI/DTS). The array is populated at compile-time from the value specified in the SERIAL_PORT_DFNS macro. This macro is defined in arch/serial.h. For x86, this macro is currently unconditionally initialized to supply four ioport UARTs (0x3F8, 0x2F8, 0x3E8, 0x2E8). However, not all x86 CPUs have these four ioport UARTs. For example, the UARTs on AMD Carrizo and later are separate memory mapped Designware IP blocks. Fairly early in boot the console_initcall univ8250_console_init iterates over this array and installs these old UARTs into the global array serial8250_ports. Further, it attempts to register them for use as the console. In other words, if, for example, the kernel commandline has console=ttyS0, the console will be switched over to one of these non-existent UARTs. Only later, when the real UART drivers are probed and their devices are instantiated will the console switch back over to the proper UART. This is noticeable when using earlycon, since part of the serial console log will appear to disappear (when the bogus old takes over) and then re-appear (when the real UART finally gets registered for the console). The problem is even more noticable when *not* using earlycon, since in this case the entire console output is missing, having been incorrectly played back to the non-existing serial port. Create a global variable to allow skipping old serial port initialization and wire it up to the AMDCZ ACPI SPCR quirk and the special amdcz earlycon setup handler. Signed-off-by: Daniel Kurtz <djkurtz@chromium.org> --- Changes since v1: * Rename variable to serial8250_skip_old_ports * Also set variable in acpi/spcr to support no-earlycon case. Changes since v2: * Added IS_ENABLED(CONFIG_SERIAL_825) and removed the .h #ifdef as suggested by Kees Cook. drivers/acpi/spcr.c | 3 +++ drivers/tty/serial/8250/8250_core.c | 6 ++++++ drivers/tty/serial/8250/8250_early.c | 1 + include/linux/serial_8250.h | 2 ++ 4 files changed, 12 insertions(+)