| Message ID | 20201214131445.954822-1-f.suligoi@asem.it |
|---|---|
| State | New |
| Headers | show |
| Series | [v1] serial: 8250_fintek: Print Fintek chip name | expand |
Hi, Greg Kroah-Hartman 於 2020/12/14 下午 09:42 寫道: >> pdata->pid = chip; >> + >> + pr_info("%s%s%s Fintek %s\n", >> + uart->port.dev ? dev_name(uart->port.dev) : "", >> + uart->port.dev ? ": " : "", >> + uart->port.name, >> + chip_name); > > Drivers, if all goes well, should not print anything to the kernel log. > This isn't ok. > > And even if it was, dev_info() would be the correct thing to do... Maybe can transform pr_info() to dev_dbg() for debug usage ? -- With Best Regards, Peter Hong
On 15. 12. 20, 1:29, Ji-Ze Hong (Peter Hong) wrote: > Hi, > > Greg Kroah-Hartman 於 2020/12/14 下午 09:42 寫道: >>> pdata->pid = chip; >>> + >>> + pr_info("%s%s%s Fintek %s\n", >>> + uart->port.dev ? dev_name(uart->port.dev) : "", >>> + uart->port.dev ? ": " : "", >>> + uart->port.name, >>> + chip_name); >> >> Drivers, if all goes well, should not print anything to the kernel log. >> This isn't ok. >> >> And even if it was, dev_info() would be the correct thing to do... > > Maybe can transform pr_info() to dev_dbg() for debug usage ? And even then, the newly introduced fintek_8250.chip_name is unused. thanks, -- js
Hi Greg, > > > > switch (chip) { > > case CHIP_ID_F81865: > > + chip_name = "F81865"; > > + break; > > case CHIP_ID_F81866: > > + chip_name = "F81866"; > > + break; > > case CHIP_ID_F81966: > > + chip_name = "F81966"; > > + break; > > case CHIP_ID_F81216AD: > > + chip_name = "F81216AD"; > > + break; > > case CHIP_ID_F81216H: > > + chip_name = "F81216H"; > > + break; > > case CHIP_ID_F81216: > > + chip_name = "F81216"; > > break; > > default: > > return -ENODEV; > > } > > > > pdata->pid = chip; > > + > > + pr_info("%s%s%s Fintek %s\n", > > + uart->port.dev ? dev_name(uart->port.dev) : "", > > + uart->port.dev ? ": " : "", > > + uart->port.name, > > + chip_name); > > Drivers, if all goes well, should not print anything to the kernel log. > This isn't ok. > > And even if it was, dev_info() would be the correct thing to do... Ok, too many information in the driver. But what do you think about the possibility to introduce a new additional field, in "serial8250_config" structure, such as "extra_name" or something like this: struct serial8250_config { const char *name; const char *extra_name; unsigned short fifo_size; unsigned short tx_loadsz; unsigned char fcr; unsigned char rxtrig_bytes[UART_FCR_R_TRIG_MAX_STATE]; unsigned int flags; }; In this way, if required, each driver can fill this additional field, for example adding the name of the particular uart chip or other useful info. As result, for example, the "uart_report_port" function output could be something like this: 00:01: ttyS0 at I/O 0x3f8 (irq = 4, base_baud = 115200) is a 16550A - Fintek F81216AD 00:02: ttyS3 at I/O 0x2e8 (irq = 11, base_baud = 115200) is a 16550A - Fintek F81216AD where the "extra_name", if not empty, is printed at the end of the line. For practical space reasons, the "extra_name" length can be limited to 16 chars. > > thanks, > > greg k-h Thanks and best regards, Flavio
On Tue, Dec 15, 2020 at 01:35:31PM +0000, Flavio Suligoi wrote: > Hi Greg, > > > > > > > switch (chip) { > > > case CHIP_ID_F81865: > > > + chip_name = "F81865"; > > > + break; > > > case CHIP_ID_F81866: > > > + chip_name = "F81866"; > > > + break; > > > case CHIP_ID_F81966: > > > + chip_name = "F81966"; > > > + break; > > > case CHIP_ID_F81216AD: > > > + chip_name = "F81216AD"; > > > + break; > > > case CHIP_ID_F81216H: > > > + chip_name = "F81216H"; > > > + break; > > > case CHIP_ID_F81216: > > > + chip_name = "F81216"; > > > break; > > > default: > > > return -ENODEV; > > > } > > > > > > pdata->pid = chip; > > > + > > > + pr_info("%s%s%s Fintek %s\n", > > > + uart->port.dev ? dev_name(uart->port.dev) : "", > > > + uart->port.dev ? ": " : "", > > > + uart->port.name, > > > + chip_name); > > > > Drivers, if all goes well, should not print anything to the kernel log. > > This isn't ok. > > > > And even if it was, dev_info() would be the correct thing to do... > > Ok, too many information in the driver. > > But what do you think about the possibility to introduce > a new additional field, in "serial8250_config" structure, > such as "extra_name" or something like this: > > struct serial8250_config { > const char *name; > const char *extra_name; > unsigned short fifo_size; > unsigned short tx_loadsz; > unsigned char fcr; > unsigned char rxtrig_bytes[UART_FCR_R_TRIG_MAX_STATE]; > unsigned int flags; > }; > > In this way, if required, each driver can fill this > additional field, for example adding the name of > the particular uart chip or other useful info. > > As result, for example, the "uart_report_port" function output > could be something like this: > > 00:01: ttyS0 at I/O 0x3f8 (irq = 4, base_baud = 115200) is a 16550A - Fintek F81216AD > 00:02: ttyS3 at I/O 0x2e8 (irq = 11, base_baud = 115200) is a 16550A - Fintek F81216AD > > where the "extra_name", if not empty, is printed > at the end of the line. > For practical space reasons, the "extra_name" length > can be limited to 16 chars. Why? What tool will use this, and why would userspace care about it? What problem are you trying to solve here? thanks, greg k-h
Hi Greg, > > > > + chip_name = "F81216H"; > > > > + break; > > > > case CHIP_ID_F81216: > > > > + chip_name = "F81216"; > > > > break; > > > > default: > > > > return -ENODEV; > > > > } > > > > > > > > pdata->pid = chip; > > > > + > > > > + pr_info("%s%s%s Fintek %s\n", > > > > + uart->port.dev ? dev_name(uart->port.dev) : "", > > > > + uart->port.dev ? ": " : "", > > > > + uart->port.name, > > > > + chip_name); > > > > > > Drivers, if all goes well, should not print anything to the kernel > log. > > > This isn't ok. > > > > > > And even if it was, dev_info() would be the correct thing to do... > > > > Ok, too many information in the driver. > > > > But what do you think about the possibility to introduce > > a new additional field, in "serial8250_config" structure, > > such as "extra_name" or something like this: > > > > struct serial8250_config { > > const char *name; > > const char *extra_name; > > unsigned short fifo_size; > > unsigned short tx_loadsz; > > unsigned char fcr; > > unsigned char rxtrig_bytes[UART_FCR_R_TRIG_MAX_STATE]; > > unsigned int flags; > > }; > > > > In this way, if required, each driver can fill this > > additional field, for example adding the name of > > the particular uart chip or other useful info. > > > > As result, for example, the "uart_report_port" function output > > could be something like this: > > > > 00:01: ttyS0 at I/O 0x3f8 (irq = 4, base_baud = 115200) is a 16550A - > Fintek F81216AD > > 00:02: ttyS3 at I/O 0x2e8 (irq = 11, base_baud = 115200) is a 16550A - > Fintek F81216AD > > > > where the "extra_name", if not empty, is printed > > at the end of the line. > > For practical space reasons, the "extra_name" length > > can be limited to 16 chars. > > Why? What tool will use this, and why would userspace care about it? > > What problem are you trying to solve here? I try to explain my requirement: we produce some x86 boards with multistandard RS232/422/485 ports and, to have this feature, in some of these boards, we use a Fintek uart or superIO. So this additional info "extra_name" can be useful for a quick check if the serial ports are multistandard or not, without any other investigations, but using only a simple command like: dmesg| grep ttyS > > thanks, > > greg k-h Thanks and best regards, Flavio
On Tue, Dec 15, 2020 at 02:06:09PM +0000, Flavio Suligoi wrote: > Hi Greg, > > > > > > > + chip_name = "F81216H"; > > > > > + break; > > > > > case CHIP_ID_F81216: > > > > > + chip_name = "F81216"; > > > > > break; > > > > > default: > > > > > return -ENODEV; > > > > > } > > > > > > > > > > pdata->pid = chip; > > > > > + > > > > > + pr_info("%s%s%s Fintek %s\n", > > > > > + uart->port.dev ? dev_name(uart->port.dev) : "", > > > > > + uart->port.dev ? ": " : "", > > > > > + uart->port.name, > > > > > + chip_name); > > > > > > > > Drivers, if all goes well, should not print anything to the kernel > > log. > > > > This isn't ok. > > > > > > > > And even if it was, dev_info() would be the correct thing to do... > > > > > > Ok, too many information in the driver. > > > > > > But what do you think about the possibility to introduce > > > a new additional field, in "serial8250_config" structure, > > > such as "extra_name" or something like this: > > > > > > struct serial8250_config { > > > const char *name; > > > const char *extra_name; > > > unsigned short fifo_size; > > > unsigned short tx_loadsz; > > > unsigned char fcr; > > > unsigned char rxtrig_bytes[UART_FCR_R_TRIG_MAX_STATE]; > > > unsigned int flags; > > > }; > > > > > > In this way, if required, each driver can fill this > > > additional field, for example adding the name of > > > the particular uart chip or other useful info. > > > > > > As result, for example, the "uart_report_port" function output > > > could be something like this: > > > > > > 00:01: ttyS0 at I/O 0x3f8 (irq = 4, base_baud = 115200) is a 16550A - > > Fintek F81216AD > > > 00:02: ttyS3 at I/O 0x2e8 (irq = 11, base_baud = 115200) is a 16550A - > > Fintek F81216AD > > > > > > where the "extra_name", if not empty, is printed > > > at the end of the line. > > > For practical space reasons, the "extra_name" length > > > can be limited to 16 chars. > > > > Why? What tool will use this, and why would userspace care about it? > > > > What problem are you trying to solve here? > > I try to explain my requirement: > > we produce some x86 boards with multistandard RS232/422/485 ports > and, to have this feature, in some of these boards, we use a > Fintek uart or superIO. > So this additional info "extra_name" can be useful for > a quick check if the serial ports are multistandard or not, > without any other investigations, but using only a simple command > like: > > dmesg| grep ttyS But as they work the same, why does it matter? Userspace should not care here. Isn't there some other id you can read/query for a hardware database tool to determine this? Printing a random string to the kernel log is not a good way to do hardware descriptions in a format that everyone can easily parse them :) thanks, greg k-h
Hi Greg, > > > Fintek F81216AD > > > > 00:02: ttyS3 at I/O 0x2e8 (irq = 11, base_baud = 115200) is a 16550A > - > > > Fintek F81216AD > > > > > > > > where the "extra_name", if not empty, is printed > > > > at the end of the line. > > > > For practical space reasons, the "extra_name" length > > > > can be limited to 16 chars. > > > > > > Why? What tool will use this, and why would userspace care about it? > > > > > > What problem are you trying to solve here? > > > > I try to explain my requirement: > > > > we produce some x86 boards with multistandard RS232/422/485 ports > > and, to have this feature, in some of these boards, we use a > > Fintek uart or superIO. > > So this additional info "extra_name" can be useful for > > a quick check if the serial ports are multistandard or not, > > without any other investigations, but using only a simple command > > like: > > > > dmesg| grep ttyS > > But as they work the same, why does it matter? Yes you are right, by the user point of view, they are the same. > > Userspace should not care here. Isn't there some other id you can > read/query for a hardware database tool to determine this? Yes, there is. I can add this info in the BIOS, in SMBIOS table type 8 for example, or I can read the board name and then search in a custom database tool what peripherals are present in the board. > > Printing a random string to the kernel log is not a good way to do > hardware descriptions in a format that everyone can easily parse them :) Ok, right! Thanks very much for your time Greg! 😊 > thanks, > > greg k-h Best regards, Flavio
Hi, Flavio Suligoi 於 2020/12/15 下午 11:06 寫道: >>> we produce some x86 boards with multistandard RS232/422/485 ports >>> and, to have this feature, in some of these boards, we use a >>> Fintek uart or superIO. >>> So this additional info "extra_name" can be useful for >>> a quick check if the serial ports are multistandard or not, >>> without any other investigations, but using only a simple command >>> like: >>> >>> dmesg| grep ttyS >> >> But as they work the same, why does it matter? > > Yes you are right, by the user point of view, they are the same. > >> >> Userspace should not care here. Isn't there some other id you can >> read/query for a hardware database tool to determine this? As Greg mentions, The userspace don't care what IC they are using. We can use Linux RS485 API to control or check the serial port. https://www.kernel.org/doc/html/latest/driver-api/serial/serial-rs485.html -- With Best Regards, Peter Hong
diff --git a/drivers/tty/serial/8250/8250_fintek.c b/drivers/tty/serial/8250/8250_fintek.c index 31c9e83ea3cb..ef2303cb5176 100644 --- a/drivers/tty/serial/8250/8250_fintek.c +++ b/drivers/tty/serial/8250/8250_fintek.c @@ -97,6 +97,7 @@ struct fintek_8250 { u16 base_port; u8 index; u8 key; + const char *chip_name; }; static u8 sio_read_reg(struct fintek_8250 *pdata, u8 reg) @@ -140,9 +141,11 @@ static void fintek_8250_exit_key(u16 base_port) release_region(base_port + ADDR_PORT, 2); } -static int fintek_8250_check_id(struct fintek_8250 *pdata) +static int fintek_8250_check_id(struct fintek_8250 *pdata, + struct uart_8250_port *uart) { u16 chip; + const char *chip_name; if (sio_read_reg(pdata, VENDOR_ID1) != VENDOR_ID1_VAL) return -ENODEV; @@ -155,17 +158,35 @@ static int fintek_8250_check_id(struct fintek_8250 *pdata) switch (chip) { case CHIP_ID_F81865: + chip_name = "F81865"; + break; case CHIP_ID_F81866: + chip_name = "F81866"; + break; case CHIP_ID_F81966: + chip_name = "F81966"; + break; case CHIP_ID_F81216AD: + chip_name = "F81216AD"; + break; case CHIP_ID_F81216H: + chip_name = "F81216H"; + break; case CHIP_ID_F81216: + chip_name = "F81216"; break; default: return -ENODEV; } pdata->pid = chip; + + pr_info("%s%s%s Fintek %s\n", + uart->port.dev ? dev_name(uart->port.dev) : "", + uart->port.dev ? ": " : "", + uart->port.name, + chip_name); + return 0; } @@ -406,7 +427,7 @@ static int probe_setup_port(struct fintek_8250 *pdata, if (fintek_8250_enter_key(addr[i], keys[j])) continue; - if (fintek_8250_check_id(pdata) || + if (fintek_8250_check_id(pdata, uart) || fintek_8250_get_ldn_range(pdata, &min, &max)) { fintek_8250_exit_key(addr[i]); continue;
At the moment, if a Fintek UART is detected, there is no printed information about this. The ttyS port is declared as a simple 16550A port, but, especially when we want to use the RS485 mode, it's very important understand if the Fintek UART is correctly detected as expected. Signed-off-by: Flavio Suligoi <f.suligoi@asem.it> --- drivers/tty/serial/8250/8250_fintek.c | 25 +++++++++++++++++++++++-- 1 file changed, 23 insertions(+), 2 deletions(-)