Add UART_CAP_NOTEMT for UARTs that lack interrupt on TEMT but want to
use em485. Em485 framework needs to ensure not only FIFO is empty but
also that tx shift register is empty.
This approach uses Uwe Kleine-König's suggestion on simply
using/incrementing stop_tx timer rather than adding another timer. When
UART_CAP_NOTEMT is set and THRE is present w/o TEMT, stop tx timer is
reused to wait for the emptying of the shift register.
This change does not add the UART_CAP_NOTEMT define as it already exist
but is currently no-op. See 7a107b2c6b (Revert "serial: 8250: Handle
UART without interrupt on TEMT using em485") for further details.
Vicente Bergas reported that RTS is deasserted roughly one bit too
early losing stop bit tx. To address this problem, stop_delay now
accounts for one extra bit using rough formula /7 (assumes worst-case
of 2+5 bits). I suspect this glitch had to do with when THRE is getting
asserted. If FIFO is emptied already during the tx of the stop bit,
perhaps it leads to HW asserting THRE early for the normal frame time
formula to work accurately.
Suggested-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Cc: Eric Tremblay <etremblay@distech-controls.com>
Tested-by: Vicente Bergas <vicencb@gmail.com>
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Link: https://lore.kernel.org/r/20220425143410.12703-4-ilpo.jarvinen@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
8250 DMA tx complete path lacks calls to normal 8250 stop handling. It
does not use THRE to detect true completion of the tx and also doesn't
call __stop_tx. This leads to problems with em485 that needs to handle
RTS timing.
Instead of handling tx stop internally within 8250 dma code, enable
THRE when tx'able data runs out and tweak serial8250_handle_irq to call
only __stop_tx when uart is using DMA.
It also seems bit early to call serial8250_rpm_put_tx from there while
tx is still underway(?).
Tested-by: Vicente Bergas <vicencb@gmail.com>
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Link: https://lore.kernel.org/r/20220425143410.12703-3-ilpo.jarvinen@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Struct uart_port currently stores FIFO timeout. Having character timing
information readily available is useful. Even serial core itself
determines char_time from port->timeout using inverse calculation.
Store frame_time directly into uart_port. Character time is stored in
nanoseconds to have reasonable precision with high rates. To avoid
overflow, 64-bit math is necessary.
It might be possible to determine timeout from frame_time by
multiplying it with fifosize as needed but only part of the users seem
to be protected by a lock. Thus, this patch does not pursue storing
only frame_time in uart_port.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Link: https://lore.kernel.org/r/20220425143410.12703-2-ilpo.jarvinen@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
DW based controllers like the one on Renesas RZ/N1 must be programmed as
flow controllers when using DMA.
* Table 11.45 of the system manual, "Flow Control Combinations", states
that using UART with DMA requires setting the DMA in the peripheral
flow controller mode regardless of the direction.
* Chapter 11.6.1.3 of the system manual, "Basic Interface Definitions",
explains that the burst size in the above case must be configured in
the peripheral's register DEST/SRC_BURST_SIZE.
Experiments shown that upon Rx timeout, the DMA transaction needed to be
manually cleared as well.
Co-developed-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Phil Edworthy <phil.edworthy@renesas.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Link: https://lore.kernel.org/r/20220422180615.9098-9-miquel.raynal@bootlin.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Use device tree match data rather than multiple calls to
of_device_is_compatible() by introducing a platform data structure and
adding a quirks mask.
Provide a stub to the compatibles without quirks to simplify the
handling of the upcoming changes.
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Emil Renner Berthing <kernel@esmil.dk>
[<miquel.raynal@bootlin.com: Minor changes + creation of a real pdata structure]
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Link: https://lore.kernel.org/r/20220422180615.9098-3-miquel.raynal@bootlin.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Once kthread printing is available, console printing will no longer
occur in the context of the printk caller. However, there are some
special contexts where it is desirable for the printk caller to
directly print out kernel messages. Using pr_flush() to wait for
threaded printers is only possible if the caller is in a sleepable
context and the kthreads are active. That is not always the case.
Introduce printk_prefer_direct_enter() and printk_prefer_direct_exit()
functions to explicitly (and globally) activate/deactivate preferred
direct console printing. The term "direct console printing" refers to
printing to all enabled consoles from the context of the printk
caller. The term "prefer" is used because this type of printing is
only best effort. If the console is currently locked or other
printers are already actively printing, the printk caller will need
to rely on the other contexts to handle the printing.
This preferred direct printing is how all printing has been handled
until now (unless it was explicitly deferred).
When kthread printing is introduced, there may be some unanticipated
problems due to kthreads being unable to flush important messages.
In order to minimize such risks, preferred direct printing is
activated for the primary important messages when the system
experiences general types of major errors. These are:
- emergency reboot/shutdown
- cpu and rcu stalls
- hard and soft lockups
- hung tasks
- warn
- sysrq
Note that since kthread printing does not yet exist, no behavior
changes result from this commit. This is only implementing the
counter and marking the various places where preferred direct
printing is active.
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
Acked-by: Paul E. McKenney <paulmck@kernel.org> # for RCU
Signed-off-by: Petr Mladek <pmladek@suse.com>
Link: https://lore.kernel.org/r/20220421212250.565456-13-john.ogness@linutronix.de
The EndRun PTP/1588 dual serial port device is based on the Oxford
Semiconductor OXPCIe952 UART device with the PCI vendor:device ID set
for EndRun Technologies and is therefore driven by a fixed 62.5MHz clock
input derived from the 100MHz PCI Express clock. The clock rate is
divided by the oversampling rate of 16 as it is supplied to the baud
rate generator, yielding the baud base of 3906250.
Replace the incorrect baud base of 4000000 with the right value of
3906250 then, complementing commit 6cbe45d8ac ("serial: 8250: Correct
the clock for OxSemi PCIe devices").
Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Cc: stable <stable@kernel.org>
Fixes: 1bc8cde46a ("8250_pci: Added driver for Endrun Technologies PTP PCIe card.")
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Link: https://lore.kernel.org/r/alpine.DEB.2.21.2204181515270.9383@angie.orcam.me.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
n_gsm is based on the 3GPP 07.010 and its newer version is the 3GPP 27.010.
See https://portal.3gpp.org/desktopmodules/Specifications/SpecificationDetails.aspx?specificationId=1516
The changes from 07.010 to 27.010 are non-functional. Therefore, I refer to
the newer 27.010 here. Chapter 5.4.8.1 states that XON/XOFF characters
shall be used instead of Fcon/Fcoff command in advanced option mode to
handle flow control. Chapter 5.4.8.2 describes how XON/XOFF characters
shall be handled. Basic option mode only used Fcon/Fcoff commands and no
XON/XOFF characters. These are treated as data bytes here.
The current implementation uses the gsm_mux field 'constipated' to handle
flow control from the remote peer and the gsm_dlci field 'constipated' to
handle flow control from each DLCI. The later is unrelated to this patch.
The gsm_mux field is correctly set for Fcon/Fcoff commands in
gsm_control_message(). However, the same is not true for XON/XOFF
characters in gsm1_receive().
Disable software flow control handling in the tty to allow explicit
handling by n_gsm.
Add the missing handling in advanced option mode for gsm_mux in
gsm1_receive() to comply with the standard.
This patch depends on the following commit:
Commit 8838b2af23 ("tty: n_gsm: fix SW flow control encoding/handling")
Fixes: e1eaea46bb ("tty: n_gsm line discipline")
Cc: stable@vger.kernel.org
Signed-off-by: Daniel Starke <daniel.starke@siemens.com>
Link: https://lore.kernel.org/r/20220422071025.5490-3-daniel.starke@siemens.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
n_gsm is based on the 3GPP 07.010 and its newer version is the 3GPP 27.010.
See https://portal.3gpp.org/desktopmodules/Specifications/SpecificationDetails.aspx?specificationId=1516
The changes from 07.010 to 27.010 are non-functional. Therefore, I refer to
the newer 27.010 here. Chapter 5.4.6.3.7 states that the Modem Status
Command (MSC) shall only be used if the basic option was chosen.
The current implementation uses MSC frames even if advanced option was
chosen to inform the peer about modem line state updates. A standard
conform peer may choose to discard these frames in advanced option mode.
Furthermore, gsmtty_modem_update() is not part of the 'tty_operations'
functions despite its name.
Rename gsmtty_modem_update() to gsm_modem_update() to clarify this. Split
its function into gsm_modem_upd_via_data() and gsm_modem_upd_via_msc()
depending on the encoding and adaption. Introduce gsm_dlci_modem_output()
as adaption of gsm_dlci_data_output() to encode and queue empty frames in
advanced option mode. Use it in gsm_modem_upd_via_data().
gsm_modem_upd_via_msc() is based on the initial gsmtty_modem_update()
function which used only MSC frames to update modem states.
Fixes: e1eaea46bb ("tty: n_gsm line discipline")
Cc: stable@vger.kernel.org
Signed-off-by: Daniel Starke <daniel.starke@siemens.com>
Link: https://lore.kernel.org/r/20220422071025.5490-2-daniel.starke@siemens.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Dynamic virtual tty registration was introduced to allow the user to handle
these cases with uevent rules. The following commits relate to this:
Commit 5b87686e32 ("tty: n_gsm: Modify gsmtty driver register method when config requester")
Commit 0b91b53323 ("tty: n_gsm: Save dlci address open status when config requester")
Commit 46292622ad ("tty: n_gsm: clean up indenting in gsm_queue()")
However, the following behavior can be seen with this implementation:
- n_gsm ldisc is activated via ioctl
- all configuration parameters are set to their default value (initiator=0)
- the mux gets activated and attached and gsmtty0 is being registered in
in gsm_dlci_open() after DLCI 0 was established (DLCI 0 is the control
channel)
- the user configures n_gsm via ioctl GSMIOC_SETCONF as initiator
- this re-attaches the n_gsm mux
- no new gsmtty devices are registered in gsmld_attach_gsm() because the
mux is already active
- the initiator side registered only the control channel as gsmtty0
(which should never happen) and no user channel tty
The commits above make it impossible to operate the initiator side as no
user channel tty is or will be available.
On the other hand, this behavior will make it also impossible to allow DLCI
parameter negotiation on responder side in the future. The responder side
first needs to provide a device for the application before the application
can set its parameters of the associated DLCI via ioctl.
Note that the user application is still able to detect a link establishment
without relaying to uevent by waiting for DTR open on responder side. This
is the same behavior as on a physical serial interface. And on initiator
side a tty hangup can be detected if a link establishment request failed.
Revert the commits above completely to always register all user channels
and no control channel after mux attachment. No other changes are made.
Fixes: 5b87686e32 ("tty: n_gsm: Modify gsmtty driver register method when config requester")
Cc: stable@vger.kernel.org
Signed-off-by: Daniel Starke <daniel.starke@siemens.com>
Link: https://lore.kernel.org/r/20220422071025.5490-1-daniel.starke@siemens.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Do not set timeout to twice the approximate amount of time to send the
entire FIFO if CTS is enabled. If the caller requested no timeout, e.g.
when userspace program called tcdrain(), then wait without any timeout.
Premature return from tcdrain() was observed on imx based system which
has 32 character long transmitter FIFO with hardware CTS handling.
Simple userspace application that reproduces problem has to:
* Open tty device, enable hardware flow control (CRTSCTS)
* Write data, e.g. 26 bytes
* Call tcdrain() to wait for the transmitter
* Close tty device
The other side of serial connection has to:
* Receive some data, e.g. 10 bytes
* Set RTS output (CTS input from sender perspective) inactive for
at least twice the port timeout
* Try to receive remaining data
Without this patch, userspace application will finish without any error
while the other side of connection will never receive remaining data.
Signed-off-by: Tomasz Moń <tomasz.mon@camlingroup.com>
Link: https://lore.kernel.org/r/20220228054911.1420221-1-tomasz.mon@camlingroup.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Note: I am using a small test app + driver located at [0] for the
problem description. serco is a driver whose write function dispatches
to the serial controller. sertest is a user-mode app that writes n bytes
to the serial console using the serco driver.
While investigating a bug in the RHEL kernel, I noticed that the serial
console throughput is way below the configured speed of 115200 bps in
a HP Proliant DL380 Gen9. I was expecting something above 10KB/s, but
I got 2.5KB/s.
$ time ./sertest -n 2500 /tmp/serco
real 0m0.997s
user 0m0.000s
sys 0m0.997s
With the help of the function tracer, I then noticed the serial
controller was taking around 410us seconds to dispatch one single byte:
$ trace-cmd record -p function_graph -g serial8250_console_write \
./sertest -n 1 /tmp/serco
$ trace-cmd report
| serial8250_console_write() {
0.384 us | _raw_spin_lock_irqsave();
1.836 us | io_serial_in();
1.667 us | io_serial_out();
| uart_console_write() {
| serial8250_console_putchar() {
| wait_for_xmitr() {
1.870 us | io_serial_in();
2.238 us | }
1.737 us | io_serial_out();
4.318 us | }
4.675 us | }
| wait_for_xmitr() {
1.635 us | io_serial_in();
| __const_udelay() {
1.125 us | delay_tsc();
1.429 us | }
...
...
...
1.683 us | io_serial_in();
| __const_udelay() {
1.248 us | delay_tsc();
1.486 us | }
1.671 us | io_serial_in();
411.342 us | }
In another machine, I measured a throughput of 11.5KB/s, with the serial
controller taking between 80-90us to send each byte. That matches the
expected throughput for a configuration of 115200 bps.
This patch changes the serial8250_console_write to use the 16550 fifo
if available. In my benchmarks I got around 25% improvement in the slow
machine, and no performance penalty in the fast machine.
Signed-off-by: Wander Lairson Costa <wander@redhat.com>
Link: https://lore.kernel.org/r/20220411174841.34936-2-wander@redhat.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There is a deadlock in sa1100_set_termios(), which is shown
below:
(Thread 1) | (Thread 2)
| sa1100_enable_ms()
sa1100_set_termios() | mod_timer()
spin_lock_irqsave() //(1) | (wait a time)
... | sa1100_timeout()
del_timer_sync() | spin_lock_irqsave() //(2)
(wait timer to stop) | ...
We hold sport->port.lock in position (1) of thread 1 and
use del_timer_sync() to wait timer to stop, but timer handler
also need sport->port.lock in position (2) of thread 2. As a result,
sa1100_set_termios() will block forever.
This patch moves del_timer_sync() before spin_lock_irqsave()
in order to prevent the deadlock.
Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
Link: https://lore.kernel.org/r/20220417111626.7802-1-duoming@zju.edu.cn
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Some more serial drivers can be compile-tested under certain
circumstances (when building a specific architecture). So allow for
that.
This reduces the need of zillion mach/subarch-specific configs. And
since the 0day bot has only allmodconfig's for some archs, this
increases build coverage there too.
Note that cpm needs a minor update in the header, so that it drags in
at least some defines (CPM2 ones).
Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Link: https://lore.kernel.org/r/20220421101708.5640-8-jslaby@suse.cz
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
pic32_uart contains this:
#ifdef CONFIG_SERIAL_PIC32_CONSOLE
...
console_initcall(pic32_console_init);
...
core_initcall(pic32_late_console_init);
...
#endif
...
arch_initcall(pic32_uart_init);
When the driver is built as module, all three above become
module_init(). So if SERIAL_PIC32_CONSOLE is set while SERIAL_PIC32=m,
it results in the following build error:
In file included from include/linux/device/driver.h:21,
from include/linux/device.h:32,
from include/linux/platform_device.h:13,
from drivers/tty/serial/pic32_uart.c:12:
include/linux/module.h:131:49: error: redefinition of '__inittest'
So make sure SERIAL_PIC32_CONSOLE can be set only when SERIAL_PIC32=y --
similar as for other drivers.
Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Link: https://lore.kernel.org/r/20220421101708.5640-7-jslaby@suse.cz
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Use list_for_each_entry() helper instead of explicit combo of
list_for_each() and list_entry().
Note that pos is used as a reference point in list_add_tail() in
icom_alloc_adapter(). This functionality remains as with an empty list,
cur_adapter_entry->icom_adapter_entry is still the list head.
This simplifies the code a bit.
Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Link: https://lore.kernel.org/r/20220421085808.24152-9-jslaby@suse.cz
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
