staging: comedi: addi_apci_3120: remove private data 'ai_running'
This member of the private data was used to determine if the interrupt
routine was handling data for the (*insn_read) or an async command. Now
that the (*insn_read) does not use interrupts this member is not needed.
Remove the member and refactor the code.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Now that the scanning and interrupt support have been removed from this
function it can be refactored to work correctly.
The comedi core expects (*insn_read) functions to read insn->n values
from the hardware and return the number of samples read. This function
currently just reads one sample but it returns insn->n.
Fix this function to work like the core expects.
Use comedi_timeout() to prevent a possible deadlock in the loop that
waits for the end-of-conversion.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
staging: comedi: addi_apci_3120: remove private data 'ui_AiReadData'
This member of the private data was used to return analog input samples that
were acquired for the (*insn_read) using interrupts. The interrupt support
code for the (*insn_read) has been removed. Remove this unused member from
the private data.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
staging: comedi: addi_apci_3120: remove interrupt support from ai (*insn_read)
(*insn_read) functions are supposed to do simple polled reads of a single
channel. This driver tries to be tricky and allow enabling interrupts in
the analog input (*insn_config) to allow the (*insn_read) to read samples
with the end-of-conversion interrupt.
The (*insn_config) doesn't follow the comedi API and this operation mode
is not part of the API.
Remove the interrupt support from the (*insn_read) as well as the support
code in the (*insn_config) and interrupt handler.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
staging: comedi: addi_apci_3120: remove private data 'ui_EocEosConversionTime'
This driver tries to be tricky and allow passing an analog input conversion time
for the (*insn_read) in the (*insn_config). The (*insn_config) doesn't follow
the comedi API and this programmable conversion time is not part of the API for
(*insn_read) operations.
Remove the member from the private data and use a fixed 10us (10000ns) conversion
time in the (*insn_read).
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
staging: comedi: addi_apci_3120: remove scanning from ai (*insn_read)
(*insn_read) functions are supposed to do simple polled reads of a single
channel. This driver tries to be tricky and allow passing a chanlist in
the analog input (*insn_config) to allow the (*insn_read) to do chanlist
scanning with or without interrupts.
The (*insn_config) doesn't follow the comedi API and this operation mode
is not part of the API.
Remove the scanning support from the (*insn_read) as well as the support
code in the (*insn_config) and interrupt handler.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The devpriv->mode is set to '0' at the start of apci3120_ai_insn_read() and
apci3120_cyclic_ai(). There is no need to clear any of the bits or do any
of the additional writes of those cleared bits.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
staging: comedi: addi_apci_3120: properly disable interrupts in apci3120_cancel()
The APCI3120_DISABLE_ALL_INTERRUPT define is a mask value used to clear bits
in devpriv->b_ModeSelectRegister to disable the interrupt sources. Writing this
value directly sets unintended bits.
Just set the devpriv->b_ModeSelectRegister to '0' and write that to the register.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
At the start of apci3120_ai_insn_read() the devpriv->b_ModeSelectRegister is set
to '0'. Remove the unnecessary masking to clear the interrupt enable bits.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Introduce a helper function to enable/disable a timer.
The disable of timers 0 and 1 in apci3120_interrupt() is probably not needed.
For now use the helper function to make sure the devpriv->ctrl bits are cleared
correctly to disable the timers.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
staging: comedi: addi_apci_3120: tidy up devpriv->ctrl use
The apci3120_setup_chan_list() function sets devpriv->ctrl to the scan length
(PR) and scan start (PA) values and writes the value to the register. There is
no need to mask the value first.
The apci3120_ai_insn_read() function calles apci3120_setup_chan_list(). There
is no need to clear devpriv->ctrl first or clear any additional bits and write
the register again. This also fixes an incorrect use of APCI3120_DISABLE_TIMER0
to disable the timer.
apci3120_cyclic_ai() also calls apci3120_setup_chan_list() so it does not need
to clear devpriv->ctrl or clear any addidional bits and write the register.
Update the comments in apci3120_reset() and apci3120_cancel().
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
staging: comedi: addi_apci_3120: fix counter and external interrupt disable
The APCI3120_DISABLE_ALL_TIMER define define is not needed, and it's used
incorrectly in apci3120_cancel(). The define is a mask of the bits needed
to disable the counters. Writing the value directly sets unintended bits.
Prior to writing the value, the 'devpriv->us_OutputRegister' is set to 0
and written to the register. This disables all the timers and the external
trigger.
Remove the unnecessary apci3120_exttrig_disable() call as well as the write
of APCI3120_DISABLE_ALL_TIMER.
In apci3120_interrupt(), remove the unnecessary mask and write to disable
all the timers. The COMEDI_CB_EOA event will cause the core to call the
(*cancel) operation which will disable the timers.
Remove the unused define.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
staging: comedi: addi_apci_3120: tidy up CTR0 register defines
Define the CTR0 register in the main driver source file and remove all
the old defines in hwrdv_apci3120.c.
For aesthetics, save the raw digital output state (devpriv->do_bits) in
the digital output (*insn_bits) function and use a macro to set them when
reading/writing a timer.
Use the CTR0 register define in the digital output (*insn_bits) function
and remove the current register define.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The timer divisor calculations in this driver are over complicated.
There are three timers on the board. They all use the same base clock
with a fixed prescaler for each timer. The base clock used depends on
the board version and type:
APCI-3120 Rev A boards OSC = 14.29MHz base clock (~70ns)
APCI-3120 Rev B boards OSC = 20MHz base clock (50ns)
APCI-3001 boards OSC = 20MHz base clock (50ns)
Daeseok Youn [Thu, 6 Nov 2014 10:27:59 +0000 (19:27 +0900)]
staging: dgap: introducing find_board_by_major()
use find_board_by_major function instead of getting a brd from static
arrary. Becasue tty's major number doesn't start zero and we can
find a brd from dgap_board[].
Daeseok Youn [Thu, 6 Nov 2014 10:27:33 +0000 (19:27 +0900)]
staging: dgap: remove useless variables for saving tty's major
The board_t has a tty_struct(serial_driver and print_driver)
that has a major number. When major number is compared in
dgap_tty_open(), just use brd->serial_driver{print_driver}->major.
Daeseok Youn [Thu, 6 Nov 2014 10:27:11 +0000 (19:27 +0900)]
staging: dgap: fix memory leak caused by double allocation of tty_structs
The tty_struct of serial_driver and print_driver were getting
allocated twice. One is allocated in tty_alloc_driver(), the other
is in dgap_tty_register(). So remove these in dgap_tty_register().
staging: comedi: amplc_pci230: use comedi_async 'scans_done' to detect AI EOA
Remove the private data member 'ai_scan_count' and use the comedi_async 'scans_done'
member to detect the analog input end-of-acquisition.
Use the comedi_nsamples_left() helper to work out the number of 'wake' samples in
pci230_ai_update_fifo_trigger_level() and the number of 'todo' samples actually
added to the async buffer in pci230_handle_ai().
Remove the unnecessary COMEDI_CB_OVERFLOW event for the hardware FIFO overflow
error. The COMEDI_CB_ERROR event will terminate the command.
comedi_buf_write_samples() can fail if the async buffer does not have room for
the sample. The it will set the COMEDI_CB_OVERFLOW event and return 0. Detect
this and quit trying to read and add more samples. The event will terminate the
command.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Ian Abbott [Fri, 31 Oct 2014 17:47:39 +0000 (17:47 +0000)]
staging: comedi: split out PCI support into new module
Setting the `CONFIG_COMEDI_PCI_DRIVERS` kernel configuration option
makes the main "comedi" module depend on the PCI support in the kernel.
That's not that big a deal since PCI support in the kernel is either
built into the kernel or is absent, and is not in a separate module.
Still, not all low-level Comedi drivers need PCI support, so we could
save a bit of space by not including it. The Comedi PCI support
functions are all in "comedi_pci.c". Turn it into a separate module so
the support code doesn't have to be loaded unnecessarily.
Signed-off-by: Ian Abbott <abbotti@mev.co.uk> Reviewed-by: H Hartley Sweeten <hsweeten@visionengravers.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Ian Abbott [Fri, 31 Oct 2014 17:47:38 +0000 (17:47 +0000)]
staging: comedi: split out USB support into new module
Setting the `CONFIG_COMEDI_USB_DRIVERS` kernel configuration option
makes the main "comedi" module depend on the "usbcore" module. But
perhaps the machine has no USB controllers (or they have been disabled),
in which case the "usbcore" module may have been pulled in
unnecessarily. Only a few low-level Comedi drivers require USB support.
The Comedi USB support functions are all in "comedi_usb.c". Turn it
into a separate module so we don't have to pull in the "usbcore" and
"usb_common" modules unnecessarily.
Signed-off-by: Ian Abbott <abbotti@mev.co.uk> Reviewed-by: H Hartley Sweeten <hsweeten@visionengravers.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Ian Abbott [Fri, 31 Oct 2014 17:47:37 +0000 (17:47 +0000)]
staging: comedi: split out PCMCIA support into new module
Setting the `CONFIG_COMEDI_PCMCIA_DRIVERS` kernel configuration option
makes the main "comedi" module depend on the "pcmcia" module, but many
machines don't have PCMCIA slots and only a few low-level Comedi drivers
need PCMCIA support. The Comedi PCMCIA support functions are all in
"comedi_pcmcia.c". Turn it into a separate module so we don't have to
pull in the other PCMCIA support modules unnecessarily.
Signed-off-by: Ian Abbott <abbotti@mev.co.uk> Reviewed-by: H Hartley Sweeten <hsweeten@visionengravers.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Ian Abbott [Fri, 31 Oct 2014 17:47:36 +0000 (17:47 +0000)]
staging: comedi: comedidev.h: remove some #ifdefs
Some declarations and macro definitions in "comedidev.h" are protected
by various `#ifdef`s for kernel configuration options, but the header
file compiles fine without the `#ifdef`s regardless of whether those
config option macros are defined or not. Remove the `#ifdef`s and
compile the affected code unconditionally.
Signed-off-by: Ian Abbott <abbotti@mev.co.uk> Reviewed-by: H Hartley Sweeten <hsweeten@visionengravers.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Ian Abbott [Fri, 31 Oct 2014 17:47:35 +0000 (17:47 +0000)]
staging: comedi: comedidev.h: remove dummy PCI support functions
Some low-level Comedi driver modules used to handle PCI devices and ISA
devices in the same module with some conditional compilation. To reduce
the amount of conditional compilation, some dummy inline versions of
`comedi_to_pci_dev()`, `comedi_pci_enable()`, `comedi_pci_disable()`,
and `comedi_pci_detach()` are defined in "comedidev.h" if
`CONFIG_COMEDI_PCI_DRIVERS` is undefined. Since those Comedi low-level
driver modules have since had PCI support split out into separate
modules, there is no need to keep the dummy inline functions around any
more, so remove them.
Signed-off-by: Ian Abbott <abbotti@mev.co.uk> Reviewed-by: H Hartley Sweeten <hsweeten@visionengravers.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Sudip Mukherjee [Fri, 7 Nov 2014 12:18:34 +0000 (17:48 +0530)]
staging: unisys: unneeded NULL check
the NULL check for memregion is not required as it has already been
checked for NULL after kzalloc. so we can reach this part of the code
only if memregion is not NULL.
staging: comedi: usbduxsigma: use comedi_async 'scans_done' to detect AI EOA
Remove the private data member 'ai_sample_count' and use the comedi_async
'scans_done' member to detect the analog input end-of-acquisition.
Move the EOA check so it happens after adding the samples from the current
urb to the async buffer. This prevents the unnecessary resubmit of the urb
when the EOA occurs.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com> Reviewed-by: Ian Abbott <abbotti@mev.co.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
projects / linux-beck.git / log