USB: serial: add port statistics
Subsystems (2)
PATCH 1/1 USB: serial: add port statistics
From: Yegor Yefremov <[email protected]>
Add additional port statistics like received and transmitted bytes the way /proc/tty/driver/serial does.
As usbserial driver already provides USB related information and this line is longer than 100 characters, this patch adds an additional line with the same port number:
0: module:ftdi_sio name:"FTDI USB Serial Device" vendor:0403 ... 0: tx:112 rx:0
--- drivers/usb/serial/usb-serial.c | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+)
drivers/usb/serial/usb-serial.c +22 -0
@@ -42,6 +42,9 @@#define USB_SERIAL_TTY_MAJOR 188#define USB_SERIAL_TTY_MINORS 512/* should be enough for a while */ +static int serial_get_icount (struct tty_struct *tty, +struct serial_icounter_struct *icount); +/* There is no MODULE_DEVICE_TABLE for usbserial.c. Instead the MODULE_DEVICE_TABLE declarations in each serial driver cause the "hotplug" program to pull in whatever module is necessary
@@ -471,10 +474,13 @@ static int serial_proc_show(struct seq_file *m, void *v)seq_puts (m,"usbserinfo:1.0 driver:2.0\n" );for (i =0 ; i <USB_SERIAL_TTY_MINORS ; ++i) { +struct tty_struct *tty; +struct serial_icounter_struct icount; port =usb_serial_port_get_by_minor (i);if (port ==NULL )continue ; serial = port->serial; + tty = port->port.tty;seq_printf (m,"%d:" , i);if (serial->type->driver.owner)
@@ -491,6 +497,22 @@ static int serial_proc_show(struct seq_file *m, void *v)seq_printf (m," path:%s" , tmp);seq_putc (m,'\n' ); +if (!serial_get_icount (tty, &icount)) { +seq_printf (m,"%d:" , i); +seq_printf (m," tx:%d rx:%d" , + icount.tx, icount.rx); +if (icount.frame) +seq_printf (m," fe:%d" , icount.frame); +if (icount.parity) +seq_printf (m," pe:%d" , icount.parity); +if (icount.brk) +seq_printf (m," brk:%d" , icount.brk); +if (icount.overrun) +seq_printf (m," oe:%d" , icount.overrun); +if (icount.buf_overrun) +seq_printf (m," bo:%d" , icount.buf_overrun); +seq_putc (m,'\n' ); + }usb_serial_put (serial);mutex_unlock (&serial->disc_mutex); }
Signature · 1 line
2.17.0
Greg KH
maintainerQuoted 9 lines · On Wed, Sep 18, 2019 at 11:14:15AM +0200, [email protected] wrote:
From: Yegor Yefremov <[email protected]>
Add additional port statistics like received and transmitted bytes the way
/proc/tty/driver/serialdoes.As usbserial driver already provides USB related information and this line is longer than 100 characters, this patch adds an additional line with the same port number:
Quoted 7 lines
0: module:ftdi_sio name:"FTDI USB Serial Device" vendor:0403 ... 0: tx:112 rx:0
Signed-off-by Yegor Yefremov <[email protected]>--- drivers/usb/serial/usb-serial.c | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+)
You can't change existing proc files without having the chance that userspace tools will break.
Have you tried this and seen what dies a horrible death?
thanks,
greg k-h
Quoted 19 lines · Greg KH
Quoted 9 lines · On Wed, Sep 18, 2019 at 11:14:15AM +0200, [email protected] wrote:
From: Yegor Yefremov <[email protected]>
Add additional port statistics like received and transmitted bytes the way
/proc/tty/driver/serialdoes.As usbserial driver already provides USB related information and this line is longer than 100 characters, this patch adds an additional line with the same port number:
Quoted 7 lines
0: module:ftdi_sio name:"FTDI USB Serial Device" vendor:0403 ... 0: tx:112 rx:0
Signed-off-by Yegor Yefremov <[email protected]>--- drivers/usb/serial/usb-serial.c | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+)
Quoted 4 lines
You can't change existing proc files without having the chance that userspace tools will break.
Have you tried this and seen what dies a horrible death?
This patch is more a proof of concept (forgot to add RFC keyword). I find statistics provdes by the 8250 driver very useful for debugging purposes. What would be the best way to implemnt this feature for usbserial driver?
a) extend current line:
0: module:ftdi_sio name:"FTDI USB Serial Device" vendor:0403 ...tx:112 rx:0
though this still can break parsing
b) creating special entries for FTDI and other UARTs? Though it would be greate to have all usbserial UART handled the same way in the same file
Best regards,
Yegor
Greg KH
maintainerQuoted 32 lines · Yegor Yefremov
Quoted 19 lines · Greg KH
Quoted 9 lines · On Wed, Sep 18, 2019 at 11:14:15AM +0200, [email protected] wrote:
From: Yegor Yefremov <[email protected]>
Add additional port statistics like received and transmitted bytes the way
/proc/tty/driver/serialdoes.As usbserial driver already provides USB related information and this line is longer than 100 characters, this patch adds an additional line with the same port number:
Quoted 7 lines
0: module:ftdi_sio name:"FTDI USB Serial Device" vendor:0403 ... 0: tx:112 rx:0
Signed-off-by Yegor Yefremov <[email protected]>--- drivers/usb/serial/usb-serial.c | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+)Quoted 4 lines
You can't change existing proc files without having the chance that userspace tools will break.
Have you tried this and seen what dies a horrible death?
This patch is more a proof of concept (forgot to add RFC keyword). I find statistics provdes by the 8250 driver very useful for debugging purposes. What would be the best way to implemnt this feature for usbserial driver?
a) extend current line:
Quoted 7 lines
0: module:ftdi_sio name:"FTDI USB Serial Device" vendor:0403 ...tx:112 rx:0
though this still can break parsing
b) creating special entries for FTDI and other UARTs? Though it would be greate to have all usbserial UART handled the same way in the same file
Why is any of this needed at all? Also, be very aware of the security issues involved here, we had to disable access of these values by "normal" users for other tty devices, so please don't break that by offering it up here again.
What is going to use this information?
thanks,
greg k-h
Quoted 42 lines · Greg KH
Quoted 32 lines · Yegor Yefremov
Quoted 19 lines · Greg KH
Quoted 9 lines · On Wed, Sep 18, 2019 at 11:14:15AM +0200, [email protected] wrote:
From: Yegor Yefremov <[email protected]>
Add additional port statistics like received and transmitted bytes the way
/proc/tty/driver/serialdoes.As usbserial driver already provides USB related information and this line is longer than 100 characters, this patch adds an additional line with the same port number:
Quoted 7 lines
0: module:ftdi_sio name:"FTDI USB Serial Device" vendor:0403 ... 0: tx:112 rx:0
Signed-off-by Yegor Yefremov <[email protected]>--- drivers/usb/serial/usb-serial.c | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+)Quoted 4 lines
You can't change existing proc files without having the chance that userspace tools will break.
Have you tried this and seen what dies a horrible death?
This patch is more a proof of concept (forgot to add RFC keyword). I find statistics provdes by the 8250 driver very useful for debugging purposes. What would be the best way to implemnt this feature for usbserial driver?
a) extend current line:
Quoted 7 lines
0: module:ftdi_sio name:"FTDI USB Serial Device" vendor:0403 ...tx:112 rx:0
though this still can break parsing
b) creating special entries for FTDI and other UARTs? Though it would be greate to have all usbserial UART handled the same way in the same file
Quoted 6 lines
Why is any of this needed at all? Also, be very aware of the security issues involved here, we had to disable access of these values by "normal" users for other tty devices, so please don't break that by offering it up here again.
What is going to use this information?
This feature is not a "must have" one but it is convenient to see transferred/received bytes and error flags from user space. If some serial software is not working like expected and doesn't provide enough debugging information one can quickly look at port statistics from the console in order to check whether and how many bytes were transferred or whether the were some communication errors.
Yegor
Greg KH
maintainerQuoted 48 lines · Yegor Yefremov
Quoted 47 lines · Greg KH
Quoted 32 lines · Yegor Yefremov
Quoted 19 lines · Greg KH
Quoted 9 lines · On Wed, Sep 18, 2019 at 11:14:15AM +0200, [email protected] wrote:
From: Yegor Yefremov <[email protected]>
Add additional port statistics like received and transmitted bytes the way
/proc/tty/driver/serialdoes.As usbserial driver already provides USB related information and this line is longer than 100 characters, this patch adds an additional line with the same port number:
Quoted 7 lines
0: module:ftdi_sio name:"FTDI USB Serial Device" vendor:0403 ... 0: tx:112 rx:0
Signed-off-by Yegor Yefremov <[email protected]>--- drivers/usb/serial/usb-serial.c | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+)Quoted 4 lines
You can't change existing proc files without having the chance that userspace tools will break.
Have you tried this and seen what dies a horrible death?
This patch is more a proof of concept (forgot to add RFC keyword). I find statistics provdes by the 8250 driver very useful for debugging purposes. What would be the best way to implemnt this feature for usbserial driver?
a) extend current line:
Quoted 7 lines
0: module:ftdi_sio name:"FTDI USB Serial Device" vendor:0403 ...tx:112 rx:0
though this still can break parsing
b) creating special entries for FTDI and other UARTs? Though it would be greate to have all usbserial UART handled the same way in the same file
Why is any of this needed at all? Also, be very aware of the security issues involved here, we had to disable access of these values by "normal" users for other tty devices, so please don't break that by offering it up here again.
Quoted 8 lines
Quoted 1 line
What is going to use this information?
This feature is not a "must have" one but it is convenient to see transferred/received bytes and error flags from user space. If some serial software is not working like expected and doesn't provide enough debugging information one can quickly look at port statistics from the console in order to check whether and how many bytes were transferred or whether the were some communication errors.
Again, it's a security issue, so be careful about this. If you _REALLY_ need it, make it a debugfs file, readble by root only.
Or a tracepoint, and then you can have a userspace read the data using ebpf :)
thanks,
greg k-h