Stephen Boyd [Wed, 24 Jul 2013 18:37:31 +0000 (11:37 -0700)]
msm_serial: Send more than 1 character at a time on UARTDM
UARTDM cores have a TX fifo that can accept more than one
character per register write, but the msm_serial driver currently
only supports 1 character mode. Add support for this mode of operation
to speed up the transmit path on DM devices.
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org> Acked-by: David Brown <davidb@codeaurora.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Stephen Boyd [Wed, 24 Jul 2013 18:37:30 +0000 (11:37 -0700)]
msm_serial: Make baud_code detection more dynamic
Currently msm_set_baud_rate() assumes the uart clock rate is
1.8432 MHz. This is not always true, and limits our options to
program the baud rate. Instead of assuming the rate and
hard-coding the baud_code based on it, calculate the divider that
we want and try to find the closest baud_code that matches. This
allows us to support uarts with faster clock speeds.
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org> Acked-by: David Brown <davidb@codeaurora.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Stephen Boyd [Wed, 24 Jul 2013 18:37:29 +0000 (11:37 -0700)]
msm_serial: Fix sparse warnings
drivers/tty/serial/msm_serial.c:302:6: warning: symbol 'msm_set_mctrl' was
not declared. Should it be static?
drivers/tty/serial/msm_serial.c:597:17: warning: incorrect type in argument 2
(different address spaces)
drivers/tty/serial/msm_serial.c:597:17: expected void volatile [noderef] <asn:2>*addr
drivers/tty/serial/msm_serial.c:597:17: got unsigned int *
drivers/tty/serial/msm_serial.c:608:33: warning: incorrect type in argument 1
(different address spaces)
drivers/tty/serial/msm_serial.c:608:33: expected void volatile [noderef] <asn:2>*addr
drivers/tty/serial/msm_serial.c:608:33: got unsigned int *gsbi_base
drivers/tty/serial/msm_serial.c:648:37: warning: incorrect type in assignment
(different address spaces)
drivers/tty/serial/msm_serial.c:648:37: expected unsigned int *gsbi_base
drivers/tty/serial/msm_serial.c:648:37: got void [noderef] <asn:2>*
Mark the ioremapped memory as __iomem and use writel instead of
iowrite because we're not dealing with PCI devices. Also, mark
msm_set_mctrl() static because it isn't used outside this file.
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org> Acked-by: David Brown <davidb@codeaurora.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Stephen Boyd [Wed, 24 Jul 2013 18:37:28 +0000 (11:37 -0700)]
msm_serial: Fix NUL byte output on UARTDM
UARTDM serial devices require us to wait for the entire TX fifo
to drain before we can change the contents of the NCF_TX
register. Furthermore, if we write any characters to the TX fifo
within the same clock cycle of changing the NCF_TX register the
NCF_TX register won't latch properly.
To fix these issues we should read back the NCF_TX register to
delay any TX fifo accesses by a clock cycle and we should wait
for the TX fifo to drain (instead of just waiting for the fifo to
be ready to receive more characters). Failure to do so leads to
random NUL bytes interspersed in the output.
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org> Acked-by: David Brown <davidb@codeaurora.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Ian Abbott [Tue, 16 Jul 2013 15:14:41 +0000 (16:14 +0100)]
pci_ids.h: remove PCI_VENDOR_ID_ADDIDATA_OLD and PCI_DEVICE_ID_ADDIDATA_APCI7800
These two defines are no longer used. They were only used by the PCI
serial driver "8250_pci" to support the original ADDI-DATA APCI-7800
card. In that driver, PCI_VENDOR_ID_ADDIDATA_OLD has been replaced with
PCI_VENDOR_ID_AMCC which has the same value (0x10e8), and
PCI_DEVICE_ID_ADDIDATA_APCI7800 has been replaced with a local #define
PCI_DEVICE_ID_AMCC_ADDIDATA_APCI7800 with the same value (0x818e).
Ian Abbott [Tue, 16 Jul 2013 15:14:40 +0000 (16:14 +0100)]
serial: 8250_pci: use local device ID for ADDI-DATA APCI-7800
The quirks and PCI ID table entries for the original ADDI-DATA APCI-7800
(not the newer APCI-7800-3) use PCI_DEVICE_ID_ADDIDATA_APCI7800 from
<linux/pci_ids.h> but the device ID was actually assigned to ADDI-DATA
by Applied Micro Circuits Corporation (PCI_VENDOR_ID_AMCC). Replace it
locally with #define PCI_DEVICE_ID_AMCC_ADDIDATA_APCI7800.
Signed-off-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
PCI_VENDOR_ID_ADDIDATA_OLD has the same value (0x10e8) as
PCI_VENDOR_ID_AMCC in <linux/pci_ids.h>. The vender ID is actually
assigned to Applied Micro Circuits Corporation. The 8250_pci driver
uses PCI_VENDOR_ID_ADDIDATA_OLD in the lists of quirks and PCI IDs for
the ADDI-DATA APCI-7800 card. Change it to use the more accurate
PCI_VENDOR_ID_AMCC.
Signed-off-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Ian Abbott [Tue, 16 Jul 2013 15:14:38 +0000 (16:14 +0100)]
pci_ids.h: move PCI_VENDOR_ID_AMCC here
PCI_VENDOR_ID_AMCC is defined locally in
"drivers/staging/comedi/comedidev.h" for a few comedi hardware drivers,
namely "adl_pci9118", "addi_apci_1500" and "addi_apci_3120" (also
"addi_apci_1710" but that is not currently built and will probably be
removed soon). Move the define into "include/linux/pci_ids.h" as it is
shared by several drivers (albeit all comedi drivers currently).
PCI_VENDOR_ID_AMCC happens to have the same value (0x10e8) as
PCI_VENDOR_ID_ADDIDATA_OLD. The vendor ID is actually assigned to
Applied Micro Circuits Corporation and Addi-Data were using device IDs
assigned by AMCC on some of their earlier PCI boards. The
PCI_VENDOR_ID_ADDIDATA_OLD define is still being used by the "8250_pci"
PCI serial board driver.
This patch adds support to ASC (asynchronous serial controller)
driver, which is basically a standard serial driver. This IP is common
across all the ST parts for settop box platforms.
ASC is embedded in ST COMMS IP block. It supports Rx & Tx functionality.
It support all industry standard baud rates.
serial/efm32-uart: use COMPILE_TEST symbol to extend compile test coverage
The driver fails to build on x86 because writel_relaxed isn't available
there. That function exists on arm, arm64, avr32, hexagon, mips and sh,
but adding all these is overkill so stick to arm only.
Axel Lin [Mon, 15 Jul 2013 05:36:03 +0000 (13:36 +0800)]
serial: bfin_uart: Remove redundant testing for ifdef CONFIG_SERIAL_BFIN_MODULE
No idea why we have redundant testing for ifdef CONFIG_SERIAL_BFIN_MODULE,
check it once is enough.
Signed-off-by: Axel Lin <axel.lin@ingics.com> Acked-by: Sonic Zhang <sonic.zhang@analog.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Some serial ports will not use the standard base baud rate. Report
this after initialization so it might be discovered and used for early
console configuration.
serial: omap: enable PM runtime only when its fully configured
If earlyprintk is enabled and current UART is console port the platform
code can mark it as RPM_ACTIVE to sync real IP state with PM Runtime and
avoid resuming of already active device, but now, driver initialization
will be performed in the wrong way:
pm_runtime_enable(&pdev->dev);
<-- PM runtime alowed (device state RPM_ACTIVE)
if (omap_up_info->autosuspend_timeout == 0)
omap_up_info->autosuspend_timeout = -1;
device_init_wakeup(up->dev, true);
pm_runtime_use_autosuspend(&pdev->dev);
<-- update_autosuspend() will be called and it will disable device
(device state RPM_SUSPENDED)
pm_runtime_set_autosuspend_delay(&pdev->dev,
omap_up_info->autosuspend_timeout);
<-- update_autosuspend() will be called which will re-enable device
(device state RPM_ACTIVE), because autosuspend_timeout < 0
pm_runtime_irq_safe(&pdev->dev);
pm_runtime_get_sync(&pdev->dev);
<-- will do nothing
Such behavior isn't expected by OMAP serial drivers and causes
unpredictable calls of serial_omap_runtime_suspend() and
serial_omap_runtime_resume().
Hence, fix it by allowing PM runtime only after all its parameters are
configured.
CC: Tony Lindgren <tony@atomide.com> CC: Rajendra Nayak <rnayak@ti.com> CC: Felipe Balbi <balbi@ti.com> CC: Kevin Hilman <khilman@linaro.org> Tested-by: Mark Jackson <mpfj-list@newflow.co.uk> Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
serial: fsl_lpuart: restore UARTCR2 after watermark setup is done
Function lpuart_setup_watermark() clears some bits in register UARTCR2
before writing FIFO configuration registers as required by hardware.
But it should restore UARTCR2 after that. Otherwise, we end up changing
UARTCR2 register when setting up watermark, and that is not really
desirable. At least, when low-level debug and earlyprint is enabled,
serial console is broken due to it.
Fix the problem by restoring UARTCR2 register at the end of function
lpuart_setup_watermark().
The revision register is a 32 bit register. The serial_in() function reads
only the lower 16 bits of the register. This leads to an incorrect computation
of the Module revision.
Signed-off-by: Ruchika Kharwar <ruchika@ti.com>
[oleksandr.savchenko@ti.com: add some whitespaces] Signed-off-by: Alexander Savchenko <oleksandr.savchenko@ti.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Current logic results in interrupt storm since the fifo
is constantly below the threshold level. Change the logic
to fill all the available spaces in the fifo as long as
we have data to minimize the possibilty of underflow and
elimiate excessive interrupts.
wer has TX wakeup bit available enable the same
by populating the necessary tx wakeup flag for the
applicable module ip blocks and use the same
while configuaring wer reg.
Also wer is not context restored, restore wer when
context is lost.
Cc: Paul Walmsley <paul@pwsan.com> Cc: Felipe Balbi <balbi@ti.com> Cc: Kevin Hilman <kevin.hilman@linaro.org> Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(for drivers/tty changes) Signed-off-by: Govindraj.R <govindraj.raja@ti.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Alexander Shiyan [Sat, 29 Jun 2013 06:44:17 +0000 (10:44 +0400)]
serial: max310x: Driver rework
This patch rework max310x driver.
Major changes have been made:
- Prepare driver to support ICs with more than one UART.
- Prepare driver to support work with I2C-bus.
The patch changes almost every function and can not be divided into parts.
Signed-off-by: Alexander Shiyan <shc_work@mail.ru> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
serial: imx: distinguish the imx6q uart from the others
We will add the DMA support for the imx uart. For the firmware's limit,
only the imx6 serial chips (including the imx6q, imx6dl, imx6sl) can
support the DMA.
This patch adds the necessary macro and helper to distinguish the
imx6q uart from the other imx uart. Other chips can use the "fsl,imx6q-uart"
to enable the DMA support.
This patch also replaces the check "is_imx21_uart()" with "!is_imx1_uart()".
Jingoo Han [Tue, 25 Jun 2013 01:08:49 +0000 (10:08 +0900)]
serial: remove unnecessary platform_set_drvdata()
The driver core clears the driver data to NULL after device_release
or on probe failure, since commit 0998d0631001288a5974afc0b2a5f568bcdecb4d
(device-core: Ensure drvdata = NULL when no driver is bound).
Thus, it is not needed to manually clear the device driver data to NULL.
Signed-off-by: Jingoo Han <jg1.han@samsung.com> Acked-by: Barry Song <baohua.song@csr.com> Acked-by: Tony Prisk <linux@prisktech.co.nz> Acked-by: Shawn Guo <shawn.guo@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
tty/hvc_iucv: Disconnect IUCV connection when lowering DTR
Implement the dtr_rts() hvc console callback to improve control when to
disconnect the IUCV connection. Previously, the IUCV connection was
disconnected during the notifier_del() callback, i.e., when the last file
descriptor to the hvc terminal device was closed.
Recent changes in login programs caused undesired disconnects during the
login phase. To prevent these kind of disconnects, implement the dtr_rts
callback to implicitly handle the HUPCL termios control via the hvc_console
driver.
Signed-off-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
tty/hvc_console: Add DTR/RTS callback to handle HUPCL control
Introduce a new callback to explicitly handle the HUPCL termios control flag.
This prepares for a follow-up commit for the hvc_iucv device driver to
improve handling when to drop an established network connection.
The callback naming is based on the recently added tty_port interface to
facilitate a potential refactoring of the hvc_console to use tty_port
functions.
Signed-off-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Wolfram Sang [Wed, 10 Jul 2013 15:57:46 +0000 (16:57 +0100)]
drivers/tty/serial: don't use devm_pinctrl_get_select_default() in probe
Since commit ab78029 (drivers/pinctrl: grab default handles from device core),
we can rely on device core for setting the default pins. Compile tested only.
Acked-by: Linus Walleij <linus.walleij@linaro.org> (personally at LCE13) Signed-off-by: Wolfram Sang <wsa@the-dreams.de> Acked-by: Nicolas Ferre <nicolas.ferre@atmel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Specifically:
n_gsm.c:810: ERROR: space required before the open parenthesis '('
n_gsm.c:830: WARNING: line over 80 characters
n_gsm.c:971: ERROR: trailing whitespace
n_gsm.c:984: ERROR: code indent should use tabs where possible
n_gsm.c:984: WARNING: please, no space before tabs
n_gsm.c:984: WARNING: please, no spaces at the start of a line
n_gsm.c:1141: WARNING: space prohibited before semicolon
n_gsm.c:1743: ERROR: space required before the open brace '{'
n_gsm.c:1744: WARNING: line over 80 characters
n_gsm.c:1745: ERROR: code indent should use tabs where possible
n_gsm.c:1746: ERROR: code indent should use tabs where possible
n_gsm.c:2908: WARNING: line over 80 characters
n_gsm.c:2912: ERROR: trailing whitespace
The lockdep report [1] shows that n_tty_write() will attempt
to claim console_lock while holding the termios_rwsem, whereas
tty_do_resize() may already hold the console_lock while
claiming the termios_rwsem.
Since n_tty_write() and tty_do_resize() do not contend
over the same data -- the tty->winsize structure -- correct
the lock dependency by introducing a new lock which
specifically serializes access to tty->winsize only.
[1] Lockdep report
======================================================
[ INFO: possible circular locking dependency detected ]
3.10.0-0+tip-xeon+lockdep #0+tip Not tainted
-------------------------------------------------------
modprobe/277 is trying to acquire lock:
(&tty->termios_rwsem){++++..}, at: [<ffffffff81452656>] tty_do_resize+0x36/0xe0
but task is already holding lock:
((fb_notifier_list).rwsem){.+.+.+}, at: [<ffffffff8107aac6>] __blocking_notifier_call_chain+0x56/0xc0
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
Peter Hurley [Wed, 24 Jul 2013 12:29:57 +0000 (08:29 -0400)]
tty: Remove extra wakeup from pty write() path
Acquiring the write_wait queue spin lock now accounts for the largest
slice of cpu time on the tty write path. Two factors contribute to
this situation; a overly-pessimistic line discipline write loop which
_always_ sets up a wait loop even if i/o will immediately succeed, and
on ptys, a wakeup storm from reads and writes.
Writer wakeup does not need to be performed by the pty driver.
Firstly, since the actual i/o is performed within the write, the
line discipline write loop will continue while space remains in
the flip buffers. Secondly, when space becomes avail in the
line discipline receive buffer (and thus also in the flip buffers),
the pty unthrottle re-wakes the writer (non-flow-controlled line
disciplines unconditionally unthrottle the driver when data is
received). Thus, existing in-kernel i/o is guaranteed to advance.
Finally, writer wakeup occurs at the conclusion of the line discipline
write (in tty_write_unlock()). This guarantees that any user-space write
waiters are woken to continue additional i/o.
Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Peter Hurley [Wed, 24 Jul 2013 12:29:56 +0000 (08:29 -0400)]
n_tty: Factor LNEXT processing from per-char i/o path
LNEXT processing accounts for ~15% of total cpu time in end-to-end
tty i/o; factor the lnext test/clear from the per-char i/o path.
Instead, attempt to immediately handle the literal next char if not
at the end of this received buffer; otherwise, handle the first char
of the next received buffer as the literal next char, then continue
with normal i/o.
Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Peter Hurley [Wed, 24 Jul 2013 12:29:54 +0000 (08:29 -0400)]
n_tty: Remove overflow tests from receive_buf() path
Always pre-figure the space available in the read_buf and limit
the inbound receive request to that amount.
For compatibility reasons with the non-flow-controlled interface,
n_tty_receive_buf() will continue filling read_buf until all data
has been received or receive_room() returns 0.
Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Peter Hurley [Sat, 15 Jun 2013 14:21:17 +0000 (10:21 -0400)]
n_tty: Fix EOF push handling
In canonical mode, an EOF which is not the first character of the line
causes read() to complete and return the number of characters read so
far (commonly referred to as EOF push). However, if the previous read()
returned because the user buffer was full _and_ the next character
is an EOF not at the beginning of the line, read() must not return 0,
thus mistakenly indicating the end-of-file condition.
The TTY_PUSH flag is used to indicate an EOF was received which is not
at the beginning of the line. Because the EOF push condition is
evaluated by a thread other than the read(), multiple EOF pushes can
cause a premature end-of-file to be indicated.
Instead, discover the 'EOF push as first read character' condition
from the read() thread itself, and restart the i/o loop if detected.
Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Peter Hurley [Sat, 15 Jun 2013 14:04:26 +0000 (10:04 -0400)]
n_tty: Process echoes in blocks
Byte-by-byte echo output is painfully slow, requiring a lock/unlock
cycle for every input byte.
Instead, perform the echo output in blocks of 256 characters, and
at least once per flip buffer receive. Enough space is reserved in
the echo buffer to guarantee a full block can be saved without
overrunning the echo output. Overrun is prevented by discarding
the oldest echoes until enough space exists in the echo buffer
to receive at least a full block of new echoes.
Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Peter Hurley [Sat, 15 Jun 2013 14:04:24 +0000 (10:04 -0400)]
n_tty: Remove echo_lock
Adding data to echo_buf (via add_echo_byte()) is guaranteed to be
single-threaded, since all callers are from the n_tty_receive_buf()
path. Processing the echo_buf can be called from either the
n_tty_receive_buf() path or the n_tty_write() path; however, these
callers are already serialized by output_lock.
Publish cumulative echo_head changes to echo_commit; process echo_buf
from echo_tail to echo_commit; remove echo_lock.
On echo_buf overrun, claim output_lock to serialize changes to
echo_tail.
Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Peter Hurley [Sat, 15 Jun 2013 14:04:22 +0000 (10:04 -0400)]
n_tty: Use separate head and tail indices for echo_buf
Instead of using a single index to track the current echo_buf position,
use a head index when adding to the buffer and a tail index when
consuming from the buffer. Allow these head and tail indices to wrap
at max representable value; perform modulo reduction via helper
functions when accessing the buffer.
Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Peter Hurley [Sat, 15 Jun 2013 13:36:15 +0000 (09:36 -0400)]
tty: Fix unsafe vt paste_selection()
Convert the tty_buffer_flush() exclusion mechanism to a
public interface - tty_buffer_lock/unlock_exclusive() - and use
the interface to safely write the paste selection to the line
discipline.
Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Peter Hurley [Sat, 15 Jun 2013 13:36:11 +0000 (09:36 -0400)]
tty: Only perform flip buffer flush from tty_buffer_flush()
Now that dropping the buffer lock is not necessary (as result of
converting the spin lock to a mutex), the flip buffer flush no
longer needs to be handled by the buffer work.
Simply signal a flush is required; the buffer work will exit the
i/o loop, which allows tty_buffer_flush() to proceed.
Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Peter Hurley [Sat, 15 Jun 2013 13:36:09 +0000 (09:36 -0400)]
tty: Make driver-side flip buffers lockless
Driver-side flip buffer input is already single-threaded; 'publish'
the .next link as the last operation on the tail buffer so the
'consumer' sees the already-completed flip buffer.
The commit buffer index is already 'published' by driver-side functions.
Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Peter Hurley [Sat, 15 Jun 2013 13:36:08 +0000 (09:36 -0400)]
tty: Track flip buffer memory limit atomically
Lockless flip buffers require atomically updating the bytes-in-use
watermark.
The pty driver also peeks at the watermark value to limit
memory consumption to a much lower value than the default; query
the watermark with new fn, tty_buffer_space_avail().
Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Peter Hurley [Sat, 15 Jun 2013 13:36:07 +0000 (09:36 -0400)]
tty: Simplify flip buffer list with 0-sized sentinel
Use a 0-sized sentinel to avoid assigning the head ptr from
the driver side thread. This also eliminates testing head/tail
for NULL.
When the sentinel is first 'consumed' by the buffer work
(or by tty_buffer_flush()), it is detached from the list but not
freed nor added to the free list. Both buffer work and
tty_buffer_flush() continue to preserve at least 1 flip buffer
to which head & tail is pointed.
Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Peter Hurley [Sat, 15 Jun 2013 13:36:06 +0000 (09:36 -0400)]
tty: Use lockless flip buffer free list
In preparation for lockless flip buffers, make the flip buffer
free list lockless.
NB: using llist is not the optimal solution, as the driver and
buffer work may contend over the llist head unnecessarily. However,
test measurements indicate this contention is low.
Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Peter Hurley [Sat, 15 Jun 2013 13:36:04 +0000 (09:36 -0400)]
tty: Merge tty_buffer_find() into tty_buffer_alloc()
tty_buffer_find() implements a simple free list lookaside cache.
Merge this functionality into tty_buffer_alloc() to reflect the
more traditional alloc/free symmetry.
Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Peter Hurley [Sat, 15 Jun 2013 13:36:02 +0000 (09:36 -0400)]
tty: Fix flip buffer free list
Since flip buffers are size-aligned to 256 bytes and all flip
buffers 512-bytes or larger are not added to the free list, the
free list only contains 256-byte flip buffers.
Remove the list search when allocating a new flip buffer.
Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Peter Hurley [Sat, 15 Jun 2013 13:14:36 +0000 (09:14 -0400)]
n_tty: Queue buffer work on any available cpu
Scheduling buffer work on the same cpu as the read() thread
limits the parallelism now possible between the receive_buf path
and the n_tty_read() path.
Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Peter Hurley [Tue, 23 Jul 2013 12:47:30 +0000 (08:47 -0400)]
n_tty: Special case pty flow control
The pty driver forces ldisc flow control on, regardless of available
receive buffer space, so the writer can be woken whenever unthrottle
is called. However, this 'forced throttle' has performance
consequences, as multiple atomic operations are necessary to
unthrottle and perform the write wakeup for every input line (in
canonical mode).
Instead, short-circuit the unthrottle if the tty is a pty and perform
the write wakeup directly.
Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Peter Hurley [Sat, 15 Jun 2013 13:14:27 +0000 (09:14 -0400)]
n_tty: Reset lnext if canonical mode changes
lnext escapes the next input character as a literal, and must
be reset when canonical mode changes (to avoid misinterpreting
a special character as a literal if canonical mode is changed
back again).
lnext is specifically not reset on a buffer flush so as to avoid
misinterpreting the next input character as a special character.
Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Peter Hurley [Sat, 15 Jun 2013 13:14:26 +0000 (09:14 -0400)]
n_tty: Make N_TTY ldisc receive path lockless
n_tty has a single-producer/single-consumer input model;
use lockless publish instead.
Use termios_rwsem to exclude both consumer and producer while
changing or resetting buffer indices, eg., when flushing. Also,
claim exclusive termios_rwsem to safely retrieve the buffer
indices from a thread other than consumer or producer
(eg., TIOCINQ ioctl).
Note the read_tail is published _after_ clearing the newline
indicator in read_flags to avoid racing the producer.
Drop read_lock spinlock.
Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Peter Hurley [Sat, 15 Jun 2013 13:14:25 +0000 (09:14 -0400)]
n_tty: Replace canon_data with index comparison
canon_data represented the # of lines which had been copied
to the receive buffer but not yet copied to the user buffer.
The value was tested to determine if input was available in
canonical mode (and also to force input overrun if the
receive buffer was full but a newline had not been received).
However, the actual count was irrelevent; only whether it was
non-zero (meaning 'is there any input to transfer?'). This
shared count is unnecessary and unsafe with a lockless algorithm.
The same check is made by comparing canon_head with read_tail instead.
Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Peter Hurley [Sat, 15 Jun 2013 13:14:24 +0000 (09:14 -0400)]
n_tty: Access termios values safely
Use termios_rwsem to guarantee safe access to the termios values.
This is particularly important for N_TTY as changing certain termios
settings alters the mode of operation.
termios_rwsem must be dropped across throttle/unthrottle since
those functions claim the termios_rwsem exclusively (to guarantee
safe access to the termios and for mutual exclusion).
Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Peter Hurley [Sat, 15 Jun 2013 13:14:21 +0000 (09:14 -0400)]
n_tty: Don't wrap input buffer indices at buffer size
Wrap read_buf indices (read_head, read_tail, canon_head) at
max representable value, instead of at the N_TTY_BUF_SIZE. This step
is necessary to allow lockless reads of these shared variables
(by updating the variables atomically).
Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
projects / linux-beck.git / log