On many new Intel SoCs the UART has an integrated DMA engine
(iDMA). In order to use it a special filter function is needed.
Signed-off-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The function clk_prepare_enable() may fail, and in that case it
does not make sense to proceed. Let's check its return code and
error out if it is a negative value.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The probe method of this driver calls clk_get(), but clk_put() is
missing from the remove callback.
Using the managed clk function is easier than fixing other parts.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Support command line parameters of the form:
earlycon=<name>,io|mmio|mmio32|mmio32be,<addr>,<options>
This commit seem to be needed even after commit:
serial: 8250: Add support for big-endian MMIO accesses
c627f2ceb6
Signed-off-by: Noam Camus <noamc@ezchip.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
when the serial works as a bluetooth sink, due to audio realtime
requirement, the driver should have something similar with ALSA:
1. one big DMA buffer to easy the schedule jitter
2. split this big DMA buffer to multiple small periods, for each
period, we get a DMA interrupt, then push the data to userspace.
the small periods will easy the audio latency.
so ALSA generally uses a cyclic chained DMA.
but for sirfsoc, the dma hardware has the limitation: we have
only two loops in the cyclic mode, so we can only support two
small periods to switch. if we make the DMA buffer too big, we
get long latency, if we make the DMA buffer too little, we get
miss in scheduling for audio realtime.
so this patch moves to use a hrtimer to simulate the cyclic
DMA, then we can have a big buffer, and also have a timely
data push to users as the hrtimer can generate in small period
then actual HW interrupts.
with this patch, we also delete a lot of complex codes to handle
loop buffers, and RX timeout interrupt since the RX work can be
completely handled from hrtimer interrupt.
tests show using this way will make our bad audio streaming be-
come smooth.
Signed-off-by: Qipan Li <Qipan.Li@csr.com>
Signed-off-by: Barry Song <Baohua.Song@csr.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Qipan Li <Qipan.Li@csr.com>
Signed-off-by: Barry Song <Baohua.Song@csr.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
A corner case exists in the current driver. if an app opens the console
device, and before writing to console device, and there are huge kernel
ogs to print out, system will hang on
sirfsoc_uart_console_putchar:
while (rd_regl(port, ureg->sirfsoc_tx_fifo_status) &
ufifo_st->ff_full(port->line))
cpu_relax();
as in sirfsoc_uart_startup(), the driver assigns tx_fifo_op to 0 will stop
TX FIFO, this loop will be endless.
Signed-off-by: Qipan Li <Qipan.Li@csr.com>
Signed-off-by: Barry Song <Baohua.Song@csr.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
serial8250_set_mctrl() is a void type function. Returning something
does not look nice.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
RS-485 configuration is also done under the spinlock
so no blocking I/O allowed.
Signed-off-by: Jakub Kicinski <kubakici@wp.pl>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
.stop_rx/tx() are called in atomic context, we cannot use
blocking I/O. While at it correct the name of RX bit and
'*' placement in pointer declarations.
Signed-off-by: Jakub Kicinski <kubakici@wp.pl>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Convert md_proc into general async reconfiguration procedure.
Signed-off-by: Jakub Kicinski <kubakici@wp.pl>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Instead of spinning under port->lock let's just sleep
inside the kthread. It should be equivalent as TX cannot
proceed when thread is blocked.
Signed-off-by: Jakub Kicinski <kubakici@wp.pl>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Convert workqueue usage to a real-time kworker. The problem
with workqueues is that we cannot set real-time priorities on
our work and asynchronous reconfiguration can be blocked by
less important tasks.
We need kthread for the interrupt anyway and because we will
now be using single kthread for all TX-related operations we
can get rid of the port mutex.
Signed-off-by: Jakub Kicinski <kubakici@wp.pl>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
LSR_TEMT_BIT (LSR bit 6) provides us exactly the information
we need to determine if transmission is finished - FIFO level
and shift register empty. We can save ourselves reading FIFO
level explicitly if we use this bit.
Signed-off-by: Jakub Kicinski <kubakici@wp.pl>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Without matching bus-specific strings driver will not be loaded
automatically.
Signed-off-by: Jakub Kicinski <kubakici@wp.pl>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Users of RS-485 can request via ioctl that RTS signals should
be activated selected number of milliseconds before the actual
data transmission or delay reception certain number of milli-
seconds after the transmission is finished. In sc16is7xx,
however, RTS signalling is handled by the hardware and driver
has no way of providing this feature.
We still try to provide .delay_rts_before_send by delaying
transmission but without actual effect on the RTS line.
Note: this change will make the driver return -EINVAL when the
feature is requested (.delay_rts_after_send is set).
Signed-off-by: Jakub Kicinski <kubakici@wp.pl>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
spi interface for sc16is7xx is added along with Kconfig flag
to enable spi or i2c, thus in a instance we can have either
spi or i2c or both, in sync to the hw.
Signed-off-by: Rama Kiran Kumar Indrakanti <indrakanti_ram@hotmail.com>
Signed-off-by: Jakub Kicinski <kubakici@wp.pl>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Add the necessary driver boilerplate to let the driver be used when
the respective ACPI table is discovered by the ACPI subsystem.
[Andre: change table name, add MODULE_DEVICE_TABLE entry and improve
commit message]
Reviewed-by: Hanjun Guo <hanjun.guo@linaro.org>
Signed-off-by: Graeme Gregory <graeme.gregory@linaro.org>
Tested-by: Mark Langsdorf <mlangsdo@redhat.com>
Tested-by: Naresh Bhat <nbhat@cavium.com>
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The ARM Server Base System Architecture[1] document describes a
generic UART which is a subset of the PL011 UART.
It lacks DMA support, baud rate control and modem status line
control, among other things.
The idea is to move the UART initialization and setup into the
firmware (which does this job today already) and let the kernel just
use the UART for sending and receiving characters.
We use the recent refactoring to build a new struct uart_ops
variable which points to some new functions avoiding access to the
missing registers. We reuse as much existing PL011 code as possible.
In contrast to the PL011 the SBSA UART does not define any AMBA or
PrimeCell relations, so we go with a pretty generic probe function
which only uses platform device functions.
A DT binding is provided with this patch, ACPI support is added in a
separate one.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Tested-by: Mark Langsdorf <mlangsdo@redhat.com>
Tested-by: Naresh Bhat <nbhat@cavium.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The SBSA UART has a fixed baud rate and flow control setting, which
cannot be changed or queried by software.
Add a vendor specific property to always return fixed values when
trying to read the console options.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Tested-by: Mark Langsdorf <mlangsdo@redhat.com>
Tested-by: Naresh Bhat <nbhat@cavium.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The SBSA UART should not be enabled or disabled (it is always on),
and consequently the spec lacks the UART_CR register.
Add a vendor specific property to skip disabling or enabling of the
UART. This will be used later by the SBSA UART support.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Tested-by: Mark Langsdorf <mlangsdo@redhat.com>
Tested-by: Naresh Bhat <nbhat@cavium.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
To avoid lines with more than 80 characters and to make the
pl011_int() function more readable, move the workaround out into a
separate function.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Tested-by: Mark Langsdorf <mlangsdo@redhat.com>
Tested-by: Naresh Bhat <nbhat@cavium.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The PL011 register UART_MIS is actually a bitwise AND of the
UART_RIS and the UART_MISC register.
Since the SBSA UART does not include the _MIS register, use the
two separate registers to get the same behaviour. Since we are
inside the spinlock and we read the _IMSC register only once, there
should be no race issue.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Tested-by: Mark Langsdorf <mlangsdo@redhat.com>
Tested-by: Naresh Bhat <nbhat@cavium.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently the pl011_probe() function is relying on some AMBA IDs
and a device tree node to initialize the driver and a port.
Both features are not necessarily required for the driver:
- we lack AMBA IDs in the ARM SBSA generic UART and
- we lack a DT node in ACPI systems.
So lets refactor the function to ease later reuse.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Tested-by: Mark Langsdorf <mlangsdo@redhat.com>
Tested-by: Naresh Bhat <nbhat@cavium.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Split the pl011_set_termios() function into smaller chunks to allow
easier reuse later when adding SBSA support.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Tested-by: Mark Langsdorf <mlangsdo@redhat.com>
Tested-by: Naresh Bhat <nbhat@cavium.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Split the pl011_shutdown() function into smaller chunks to allow
easier reuse later when adding SBSA support.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Tested-by: Mark Langsdorf <mlangsdo@redhat.com>
Tested-by: Naresh Bhat <nbhat@cavium.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Split the pl011_startup() function into smaller chunks to allow
easier reuse later when adding SBSA support.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Tested-by: Mark Langsdorf <mlangsdo@redhat.com>
Tested-by: Naresh Bhat <nbhat@cavium.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Although we care about not unregistering the driver if there are
still ports connected during the .remove callback, we do miss this
check in the pl011_probe function. So if the current port allocation
fails, but there are other ports already registered, we will kill
those.
So factor out the port removal into a separate function and use that
in the probe function, too.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Tested-by: Mark Langsdorf <mlangsdo@redhat.com>
Tested-by: Naresh Bhat <nbhat@cavium.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If the mediatek serial port driver is built-in, but serial
console is disabled in Kconfig (e.g. when the serial driver
itself is a loadable module), we get this build error:
drivers/built-in.o: In function `early_mtk8250_setup':
undefined reference to `early_serial8250_setup'
To avoid that problem, this patch encloses the early_mtk8250_setup
function in #ifdef CONFIG_SERIAL_8250_CONSOLE, the same symbol
that guards the early_serial8250_setup function.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Eddie Huang <eddie.huang@mediatek.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
A configuration that enables earlycon but not the core console
code causes a link error:
drivers/built-in.o: In function `setup_earlycon':
drivers/tty/serial/earlycon.c:70: undefined reference to `uart_parse_earlycon'
That error can be triggered by the newly added samsung earlycon support,
which is missing a 'select' statement.
As suggested by Peter Hurley, solves the problem by moving the
'select SERIAL_EARLYCON' statement to the samsung console driver
option, as it is done by all other console drivers.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: b94ba0328d ("serial: samsung: Add support for early console")
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The platform_sysrq_reset_seq code was intended as a way for an embedded
platform to provide its own sysrq sequence at compile time. After over
two years, nobody has started using it in an upstream kernel, and
the platforms that were interested in it have moved on to devicetree,
which can be used to configure the sequence without requiring kernel
changes. The method is also incompatible with the way that most
architectures build support for multiple platforms into a single
kernel.
Now the code is producing warnings when built with gcc-5.1:
drivers/tty/sysrq.c: In function 'sysrq_init':
drivers/tty/sysrq.c:959:33: warning: array subscript is above array bounds [-Warray-bounds]
key = platform_sysrq_reset_seq[i];
We could fix this, but it seems unlikely that it will ever be used,
so let's just remove the code instead. We still have the option to
pass the sequence either in DT, using the kernel command line,
or using the /sys/module/sysrq/parameters/reset_seq file.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: 154b7a489a ("Input: sysrq - allow specifying alternate reset sequence")
----
v2: moved sysrq_reset_downtime_ms variable to avoid introducing a compile
warning when CONFIG_INPUT is disabled
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The function mctrl_gpio_init returns failure if the assignment to any
member of the gpio array results in an error pointer. So there is no
need to check for such error values in the other functions.
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The currently in-use port->startup and port->shutdown are "okay". The
startup part for instance does the tiny omap extra part and invokes
serial8250_do_startup() for the remaining pieces. The workflow in
serial8250_do_startup() is okay except for the part where UART_RX is
read without a check if there is something to read. I tried to
workaround it in commit 0aa525d118 ("tty: serial: 8250_core: read only
RX if there is something in the FIFO") but then reverted it later in
commit ca8bb4aefb ("serial: 8250: Revert "tty: serial: 8250_core: read
only RX if there is something in the FIFO"").
This is the second attempt to get it to work on older OMAPs without
breaking other chips this time
Peter Hurley suggested to pull in the few needed lines from
serial8250_do_startup() and drop everything else that is not required
including making it simpler like using just request_irq() instead the
chain handler like it is doing now.
So lets try that.
Fixes: ca8bb4aefb ("serial: 8250: Revert "tty: serial: 8250_core:
read only RX if there is something in the FIFO"")
Tested-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If the call to dmaengine_slave_config() fails, then the DMA buffer will
not be freed/unmapped. Fix this by moving the code that stores the
address of the buffer in the tegra_uart_port structure to before the
call to dmaengine_slave_config().
Reported-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Reviewed-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
it seems this is a more typical behaviour from reviewing other console
drivers.
Signed-off-by: Qipan Li <Qipan.Li@csr.com>
Signed-off-by: Barry Song <Baohua.Song@csr.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Qipan Li <Qipan.Li@csr.com>
Signed-off-by: Barry Song <Baohua.Song@csr.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
add overrun error's flag mark and parity's counter, we can show the
statistic from procfs node.
BTW, let the indentation of stick bits configuration look better.
Signed-off-by: Qipan Li <Qipan.Li@csr.com>
Signed-off-by: Barry Song <Baohua.Song@csr.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In special condition, when cpu schedule into rx_tmo_process_tl or
rx_dma_complete_tl and all the receive dma tasks have done, it will
go into endless loop because no dma task cookie status be changed.
Signed-off-by: Qipan Li <Qipan.Li@csr.com>
Signed-off-by: Barry Song <Baohua.Song@csr.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
>From HW spec, when rxfifo's data is less than AFC_RX_THD(RX threshhold), RTS
signal is active. otherwise, RTS signal is inactive.
Crrently the RX threshhold is set as zero, so RTS has no chance to be
active. This patch replaces the default 0 by a positive number.
Signed-off-by: Qipan Li <Qipan.Li@csr.com>
Signed-off-by: Barry Song <Baohua.Song@csr.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
differentiate difference port types by re-defining the status MARCO
or putting HW differences into private data of the related ports.
Signed-off-by: Qipan Li <Qipan.Li@csr.com>
Signed-off-by: Barry Song <Baohua.Song@csr.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This comment does not reflect the actual code. It should be 57600,
not 56000.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In RS485 mode, we may want to set the delay_rts_after_send value to 0.
In the datasheet, the 0 value is said to "disable" the Transmitter Timeguard but
this is exactly the expected behavior if we want no delay...
Moreover, if the value was set to non-zero value by device-tree or earlier
ioctl command, it was impossible to change it back to zero.
Reported-by: Sami Pietikäinen <Sami.Pietikainen@wapice.com>
Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Cc: stable@vger.kernel.org # 3.2+
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In commit 02730d3c05
(Merge 4.1-rc4 into tty-next), git mismerged some lines,
reintroducing a reference to the removed field
uart_amba_port.tx_irq_seen. This causes a build failure.
This patch removes the mismerged lines, restoring the code to what
was in tty-next (which was the intention).
Signed-off-by: Dave Martin <Dave.Martin@arm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Here's some TTY and serial driver fixes for reported issues. All of
these have been in linux-next successfully.
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2
iEYEABECAAYFAlVX4JMACgkQMUfUDdst+ymy/QCfSx/npI/WfRNlKBMHI20xwOaE
szQAoJKRxeF0d+2GCJ56OVbmqqjnG4IN
=FUtJ
-----END PGP SIGNATURE-----
Merge tag 'tty-4.1-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
Pull tty/serial fixes from Greg KH:
"Here's some TTY and serial driver fixes for reported issues.
All of these have been in linux-next successfully"
* tag 'tty-4.1-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
pty: Fix input race when closing
tty/n_gsm.c: fix a memory leak when gsmtty is removed
Revert "serial/amba-pl011: Leave the TX IRQ alone when the UART is not open"
serial: omap: Fix error handling in probe
earlycon: Revert log warnings
This was introduced in
commit 6db4063c5b
Author: Antonino A. Daplas <adaplas@gmail.com>
Date: Mon Jun 26 00:27:12 2006 -0700
[PATCH] VT binding: Add sysfs control to the VT layer
with the justification
"In addition, if any of the consoles are in KD_GRAPHICS mode, binding and
unbinding will not succeed. KD_GRAPHICS mode usually indicates that the
underlying console hardware is used for other purposes other than displaying
text (ie X). This feature should prevent binding/unbinding from interfering
with a graphics application using the VT."
I think we should lift this artificial restriction though:
- KD_GRAPHICS doesn't get cleaned up automatically, which means it's
easy to have terminals stuck in KD_GRAPHICS when hacking around on
X.
- X doesn't really care, especially with drm where kms already blocks
fbdev (and hence fbcon) when there's an active compositor.
- This is a root-only interface with a separate .config option and
it's possible to hang your machine already anyway if you
unload/reload drivers and don't know what you're doing.
With this patch i915.ko module reloading works again reliably,
something in the recent fedora upgrades broke things.
Cc: Antonino A. Daplas <adaplas@gmail.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Acked-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently there is a lock order problem between the console lock and the
kernfs s_active lock of the console driver's bind sysfs entry. When
writing to the sysfs entry the lock order is first s_active then console
lock, when unregistering the console driver via
do_unregister_con_driver() the order is the opposite. See the below
bugzilla reference for one instance of a lockdep backtrace.
Fix this by unregistering the console driver from a deferred work, where
we can safely drop the console lock while unregistering the device and
corresponding sysfs entries (which in turn acquire s_active). Note that
we have to keep the console driver slot in the registered_con_driver
array reserved for the driver that's being unregistered until it's fully
removed. Otherwise a concurrent call to do_register_con_driver could
try to reuse the same slot and fail when registering the corresponding
device with a minor index that's still in use.
Note that the referenced bug report contains two dmesg logs with two
distinct lockdep reports: [1] is about a locking scenario involving
s_active, console_lock and the fb_notifier list lock, while [2] is
about a locking scenario involving only s_active and console_lock.
In [1] locking fb_notifier triggers the lockdep warning only because
of its dependence on console_lock, otherwise case [1] is the same
s_active<->console_lock dependency problem fixed by this patch.
Before this change we have the following locking scenarios involving
the 3 locks:
a) via do_unregister_framebuffer()->...->do_unregister_con_driver():
1. console lock 2. fb_notifier lock 3. s_active lock
b) for example via give_up_console()->do_unregister_con_driver():
1. console lock 2. s_active lock
c) via vt_bind()/vt_unbind():
1. s_active lock 2. console lock
Since c) is the console bind sysfs entry's write code path we can't
change the locking order there. We can only fix this issue by removing
s_active's dependence on the other two locks in a) and b). We can do
this only in the vt code which owns the corresponding sysfs entry, so
that after the change we have:
a) 1. console lock 2. fb_notifier lock
b) 1. console lock
c) 1. s_active lock 2. console lock
d) in the new con_driver_unregister_callback():
1. s_active lock
[1] https://bugs.freedesktop.org/attachment.cgi?id=87716
[2] https://bugs.freedesktop.org/attachment.cgi?id=107602
v2:
- get console_lock earlier in con_driver_unregister_callback(), so we
protect the following console driver field assignments there
- add code coment explaining the reason for deferring the sysfs entry
removal
- add a third paragraph to the commit message explaining why there are
two distinct lockdep reports and that this issue is independent of
fb/fbcon. (Peter Hurley)
v3:
- clarify further the third paragraph
v4:
- rebased on v4 of patch 1/3
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
A read() from a pty master may mistakenly indicate EOF (errno == -EIO)
after the pty slave has closed, even though input data remains to be read.
For example,
pty slave | input worker | pty master
| |
| | n_tty_read()
pty_write() | | input avail? no
add data | | sleep
schedule worker --->| | .
|---> flush_to_ldisc() | .
pty_close() | fill read buffer | .
wait for worker | wakeup reader --->| .
| read buffer full? |---> input avail ? yes
|<--- yes - exit worker | copy 4096 bytes to user
TTY_OTHER_CLOSED <---| |<--- kick worker
| |
**** New read() before worker starts ****
| | n_tty_read()
| | input avail? no
| | TTY_OTHER_CLOSED? yes
| | return -EIO
Several conditions are required to trigger this race:
1. the ldisc read buffer must become full so the input worker exits
2. the read() count parameter must be >= 4096 so the ldisc read buffer
is empty
3. the subsequent read() occurs before the kicked worker has processed
more input
However, the underlying cause of the race is that data is pipelined, while
tty state is not; ie., data already written by the pty slave end is not
yet visible to the pty master end, but state changes by the pty slave end
are visible to the pty master end immediately.
Pipeline the TTY_OTHER_CLOSED state through input worker to the reader.
1. Introduce TTY_OTHER_DONE which is set by the input worker when
TTY_OTHER_CLOSED is set and either the input buffers are flushed or
input processing has completed. Readers/polls are woken when
TTY_OTHER_DONE is set.
2. Reader/poll checks TTY_OTHER_DONE instead of TTY_OTHER_CLOSED.
3. A new input worker is started from pty_close() after setting
TTY_OTHER_CLOSED, which ensures the TTY_OTHER_DONE state will be
set if the last input worker is already finished (or just about to
exit).
Remove tty_flush_to_ldisc(); no in-tree callers.
Fixes: 52bce7f8d4 ("pty, n_tty: Simplify input processing on final close")
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=96311
BugLink: http://bugs.launchpad.net/bugs/1429756
Cc: <stable@vger.kernel.org> # 3.19+
Reported-by: Andy Whitcroft <apw@canonical.com>
Reported-by: H.J. Lu <hjl.tools@gmail.com>
Signed-off-by: Peter Hurley <peter@hurleysoftware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
when gsmtty_remove put dlci, it will cause memory leak if dlci->port's refcount is zero.
So we do the cleanup work in .cleanup callback instead.
dlci will be last put in two call chains.
1) gsmld_close -> gsm_cleanup_mux -> gsm_dlci_release -> dlci_put
2) gsmld_remove -> dlci_put
so there is a race. the memory leak depends on the race.
In call chain 2. we hit the memory leak. below comment tells.
release_tty -> tty_driver_remove_tty -> gsmtty_remove -> dlci_put -> tty_port_destructor (WARN_ON(port->itty) and return directly)
|
tty->port->itty = NULL;
|
tty_kref_put ---> release_one_tty -> gsmtty_cleanup (added by our patch)
So our patch fix the memory leak by doing the cleanup work after tty core did.
Signed-off-by: Pan Xinhui <xinhuix.pan@intel.com>
Fixes: dfabf7ffa3
Cc: stable <stable@vger.kernel.org> # 3.14+
Acked-by: Jiri Slaby <jslaby@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
