usb-serial: Fix usb serial console open/close regression
Subsystems (4)
PATCH 1/1 usb-serial: Fix usb serial console open/close regression
Commit e1108a63e10d344284011cccc06328b2cd3e5da3 ("usb_serial: Use the shutdown() operation") breaks the ability to use a usb console starting in 2.6.32. This was observed when using console=ttyUSB0,115200 as a boot argument with an FTDI device. The error is:
ftdi_sio ttyUSB0: ftdi_submit_read_urb - failed submitting read urb, error -22
The handling of the ASYNCB_INITIALIZED changed in 2.6.32 such that in tty_port_shutdown() it always clears the flag if it is set. As a work around the usb console must reset this flag, in order for the hw to stay open.
--- drivers/usb/serial/usb-serial.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)
drivers/usb/serial/usb-serial.c +5 -2
@@ -287,10 +287,13 @@ static void serial_down(struct tty_portstruct usb_serial_driver *drv = port->serial->type;/* * The console is magical. Do not hang up the console hardware -* or there will be tears. +* or there will be tears. If this is a console the initialized +* flag is reset. */ -if (port->console) +if (port->console) { +set_bit (ASYNCB_INITIALIZED , &port->port.flags);return ; + }if (drv->close) drv->close (port); }
Quoted 34 lines · Jason Wessel
Commit
e1108a63e10d344284011cccc06328b2cd3e5da3("usb_serial: Use theshutdown()operation") breaks the ability to use a usb console starting in 2.6.32. This was observed when using console=ttyUSB0,115200 as a boot argument with an FTDI device. The error is:
ftdi_siottyUSB0:ftdi_submit_read_urb- failed submitting read urb, error -22The handling of the
ASYNCB_INITIALIZEDchanged in 2.6.32 such that intty_port_shutdown()it always clears the flag if it is set. As a work around the usb console must reset this flag, in order for the hw to stay open.CC Greg Kroah-Hartman <[email protected]>CC Alan Cox <[email protected]>CC Alan Stern <[email protected]>CC Oliver Neukum <[email protected]>CC Andrew Morton <[email protected]>Signed-off-by Jason Wessel <[email protected]>--- drivers/usb/serial/usb-serial.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)drivers/usb/serial/usb-serial.c +1
-1@@ -287,10 +287,13 @@ static void serial_down(struct tty_portstruct usb_serial_driver *drv = port->serial->type;/* * The console is magical. Do not hang up the console hardware -* or there will be tears. +* or there will be tears. If this is a console the initialized
drivers/usb/serial/usb-serial.c static void serial_down(struct tty_port · line 299
Quoted 10 lines
drivers/usb/serial/usb-serial.c +4
-1+* flag is reset. */ -if (port->console) +if (port->console) { +set_bit (ASYNCB_INITIALIZED , &port->port.flags);return ; + }if (drv->close) drv->close (port); }
This is a little unfortunate. It would be better to prevent tty_port_shutdown() from clearing ASYNCB_INITIALIZED in the first place. The problem is that the tty core doesn't know when the port is being used as a console. There ought to be a way to tell it.
Alan Stern
Quoted 45 lines · Alan Stern
Quoted 42 lines · Jason Wessel
Commit
e1108a63e10d344284011cccc06328b2cd3e5da3("usb_serial: Use theshutdown()operation") breaks the ability to use a usb console starting in 2.6.32. This was observed when using console=ttyUSB0,115200 as a boot argument with an FTDI device. The error is:
ftdi_siottyUSB0:ftdi_submit_read_urb- failed submitting read urb, error -22The handling of the
ASYNCB_INITIALIZEDchanged in 2.6.32 such that intty_port_shutdown()it always clears the flag if it is set. As a work around the usb console must reset this flag, in order for the hw to stay open.CC Greg Kroah-Hartman <[email protected]>CC Alan Cox <[email protected]>CC Alan Stern <[email protected]>CC Oliver Neukum <[email protected]>CC Andrew Morton <[email protected]>Signed-off-by Jason Wessel <[email protected]>--- drivers/usb/serial/usb-serial.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)drivers/usb/serial/usb-serial.c +5
-2@@ -287,10 +287,13 @@ static void serial_down(struct tty_portstruct usb_serial_driver *drv = port->serial->type;/* * The console is magical. Do not hang up the console hardware -* or there will be tears. +* or there will be tears. If this is a console the initialized +* flag is reset. */ -if (port->console) +if (port->console) { +set_bit (ASYNCB_INITIALIZED , &port->port.flags);return ; + }if (drv->close)
Quoted 9 lines
drivers/usb/serial/usb-serial.c static void serial_down(struct tty_port · line 299
Quoted 3 lines
drivers/usb/serial/usb-serial.c +0
-0drv->close (port); }This is a little unfortunate. It would be better to prevent
tty_port_shutdown()from clearingASYNCB_INITIALIZEDin the first place. The problem is that the tty core doesn't know when the port is being used as a console. There ought to be a way to tell it.
I agree, but presently there is no way to do so. Up until 2.6.33 theASYNCB_INITIALIZED was being used to track this, but now it is used a bit differently.
We still also have the same sort of issue for the passing in the initial baud. I don't know if you want to go for a short term approach this way, or try to implement something different right now. When you do consider something longer term, I would like it to incorporate the other serial settings as well, such that if the console initializes them first the get correctly inherited by the tty open().
Jason.
Quoted 10 lines · Jason Wessel
Quoted 5 lines
This is a little unfortunate. It would be better to prevent
tty_port_shutdown()from clearingASYNCB_INITIALIZEDin the first place. The problem is that the tty core doesn't know when the port is being used as a console. There ought to be a way to tell it.I agree, but presently there is no way to do so. Up until 2.6.33 the
ASYNCB_INITIALIZEDwas being used to track this, but now it is used a bit differently.
Quoted 6 lines
We still also have the same sort of issue for the passing in the initial baud. I don't know if you want to go for a short term approach this way, or try to implement something different right now. When you do consider something longer term, I would like it to incorporate the other serial settings as well, such that if the console initializes them first the get correctly inherited by the tty
open().
I don't want to make any changes before hearing from Alan Cox. Doing the right thing probably means switching all the tty drivers over to the tty_port model. I don't know which ones still need to be changed.
Alan Stern
Quoted 4 lines
This is a little unfortunate. It would be better to prevent
tty_port_shutdown()from clearingASYNCB_INITIALIZEDin the first place. The problem is that the tty core doesn't know when the port is being used as a console. There ought to be a way to tell it.
Agreed - there should probably be a port.console flag to push it up into the tty_port logic as well.
Alan
Quoted 8 lines · Alan Cox
Quoted 4 lines
This is a little unfortunate. It would be better to prevent
tty_port_shutdown()from clearingASYNCB_INITIALIZEDin the first place. The problem is that the tty core doesn't know when the port is being used as a console. There ought to be a way to tell it.Agreed - there should probably be a port.console flag to push it up into the
tty_portlogic as well.
Alan, are you thinking about something like the patch one listed below?
That led me to wonder if we additionally want to remove the ->console semi-private variable from the usb-serial port struct. I tested that and included a patch here as well as an RFC.
If we come to agreement I'll send new patches with appropriate patch headers.
Thanks,
Jason.
----- PATCH ONE STARTS HERE-----
--- drivers/char/tty_port.c | 2 +- drivers/usb/serial/console.c | 1 + include/linux/serial.h | 1 + 3 files changed, 3 insertions(+), 1 deletion(-)
drivers/char/tty_port.c +1 -1
@@ -119,7 +119,7 @@ EXPORT_SYMBOL(tty_port_tty_set);static void tty_port_shutdown (struct tty_port *port) {mutex_lock (&port->mutex); -if (port->ops->shutdown && +if (port->ops->shutdown && !test_bit (ASYNCB_CONSOLE , &port->flags) &&test_and_clear_bit (ASYNCB_INITIALIZED , &port->flags)) port->ops->shutdown (port);mutex_unlock (&port->mutex);
include/linux/serial.h +1 -0
@@ -132,6 +132,7 @@ struct serial_uart_config {#define ASYNCB_CONS_FLOW 23/* flow control for console */ #define ASYNCB_BOOT_ONLYMCA 22/* Probe only if MCA bus */ #define ASYNCB_FIRST_KERNEL 22 +#define ASYNCB_CONSOLE 21/* Serial port is console */ #define ASYNC_HUP_NOTIFY (1U << ASYNCB_HUP_NOTIFY)#define ASYNC_SUSPENDED (1U << ASYNCB_SUSPENDED)
drivers/usb/serial/console.c +1 -0
@@ -181,6 +181,7 @@ static int usb_console_setup(struct cons/* The console is special in terms of closing the device so * indicate this port is now acting as a system console. */ port->console =1 ; +set_bit (ASYNCB_CONSOLE , &port->port.flags);mutex_unlock (&serial->disc_mutex);return retval;
----- PART TWO STARTS HERE -----
Remove the console variable from the usb serial private data.
--- drivers/usb/serial/console.c | 5 ++--- drivers/usb/serial/ftdi_sio.c | 3 ++- drivers/usb/serial/generic.c | 4 ++-- drivers/usb/serial/pl2303.c | 3 ++- drivers/usb/serial/usb-serial.c | 4 ++-- include/linux/usb/serial.h | 2 -- 6 files changed, 10 insertions(+), 11 deletions(-)
include/linux/usb/serial.h +0 -2
@@ -66,7 +66,6 @@ enum port_dev_state {* @work: work queue entry for the line discipline waking up. * @throttled: nonzero if the read urb is inactive to throttle the device * @throttle_req: nonzero if the tty wants to throttle us -* @console: attached usb serial console * @dev: pointer to the serial device * * This structure is used by the usb-serial core and drivers for the specific
@@ -106,7 +105,6 @@ struct usb_serial_port {struct work_struct work;char throttled;char throttle_req; -char console;unsigned long sysrq;/* sysrq timeout */ struct device dev;enum port_dev_state dev_state;
drivers/usb/serial/console.c +2 -3
@@ -180,7 +180,6 @@ static int usb_console_setup(struct cons --port->port.count;/* The console is special in terms of closing the device so * indicate this port is now acting as a system console. */ - port->console =1 ;set_bit (ASYNCB_CONSOLE , &port->port.flags);mutex_unlock (&serial->disc_mutex);
@@ -217,7 +216,7 @@ static void usb_console_write(struct condbg ("%s - port %d, %d byte(s)" , __func__, port->number, count); -if (!port->console) { +if (!test_bit (ASYNCB_INITIALIZED , &port->port.flags)) {dbg ("%s - port not opened" , __func__);return ; }
@@ -313,7 +312,7 @@ void usb_serial_console_exit(void) {if (usbcons_info.port) {unregister_console (&usbcons); - usbcons_info.port->console =0 ; +clear_bit (ASYNCB_CONSOLE , &usbcons_info.port->port.flags); usbcons_info.port =NULL ; } }
drivers/usb/serial/ftdi_sio.c +2 -1
@@ -2073,7 +2073,8 @@ static int ftdi_process_packet(struct ttreturn 0 ;/* status only */ ch = packet +2 ; -if (!(port->console && port->sysrq) && flag ==TTY_NORMAL ) +if (!(test_bit (ASYNCB_INITIALIZED , &port->port.flags) && + port->sysrq) && flag ==TTY_NORMAL )tty_insert_flip_string (tty, ch, len);else {for (i =0 ; i < len; i++, ch++) {
drivers/usb/serial/generic.c +2 -2
@@ -437,7 +437,7 @@ static void flush_and_resubmit_read_urb(/* The per character mucking around with sysrq path it too slow for stuff like 3G modems, so shortcircuit it in the 99.9999999% of cases where the USB serial is not a console anyway */ -if (!port->console || !port->sysrq) +if (!test_bit (ASYNCB_INITIALIZED , &port->port.flags) || !port->sysrq)tty_insert_flip_string (tty, ch, urb->actual_length);else {/* Push data to tty */
@@ -556,7 +556,7 @@ void usb_serial_generic_unthrottle(strucint usb_serial_handle_sysrq_char (struct tty_struct *tty,struct usb_serial_port *port,unsigned int ch) { -if (port->sysrq && port->console) { +if (port->sysrq &&test_bit (ASYNCB_INITIALIZED , &port->port.flags)) {if (ch &&time_before (jiffies, port->sysrq)) {handle_sysrq (ch, tty); port->sysrq =0 ;
drivers/usb/serial/pl2303.c +2 -1
@@ -1056,7 +1056,8 @@ static void pl2303_push_data(struct tty_if (line_status &UART_OVERRUN_ERROR )tty_insert_flip_char (tty,0 ,TTY_OVERRUN ); -if (tty_flag ==TTY_NORMAL && !(port->console && port->sysrq)) +if (tty_flag ==TTY_NORMAL && + !(test_bit (ASYNCB_INITIALIZED , &port->port.flags) && port->sysrq))tty_insert_flip_string (tty, data, urb->actual_length);else {int i;
drivers/usb/serial/usb-serial.c +2 -2
@@ -289,7 +289,7 @@ static void serial_down(struct tty_port* The console is magical. Do not hang up the console hardware * or there will be tears. */ -if (port->console) +if (test_bit (ASYNCB_INITIALIZED , &port->port.flags))return ;if (drv->close) drv->close (port);
@@ -328,7 +328,7 @@ static void serial_cleanup(struct tty_st/* The console is magical. Do not hang up the console hardware * or there will be tears. */ -if (port->console) +if (test_bit (ASYNCB_INITIALIZED , &port->port.flags))return ;dbg ("%s - port %d" , __func__, port->number);
Quoted 25 lines · Jason Wessel
Alan, are you thinking about something like the patch one listed below?
That led me to wonder if we additionally want to remove the ->console semi-private variable from the usb-serial port struct. I tested that and included a patch here as well as an RFC.
If we come to agreement I'll send new patches with appropriate patch headers.
----- PART TWO STARTS HERE -----
Remove the console variable from the usb serial private data.
drivers/usb/serial/console.c +0
-1@@ -180,7 +180,6 @@ static int usb_console_setup(struct cons --port->port.count;/* The console is special in terms of closing the device so * indicate this port is now acting as a system console. */ - port->console =1 ;set_bit (ASYNCB_CONSOLE , &port->port.flags);mutex_unlock (&serial->disc_mutex);
drivers/usb/serial/console.c static void usb_console_write(struct con · line 219
Quoted 5 lines
drivers/usb/serial/console.c +1
-1@@ -217,7 +216,7 @@ static void usb_console_write(struct condbg ("%s - port %d, %d byte(s)" , __func__, port->number, count); -if (!port->console) { +if (!test_bit (ASYNCB_INITIALIZED , &port->port.flags)) {
Here and below you wrote INITIALIZED instead of CONSOLE.
Alan Stern