Needed to ensure the reference counting is correct.
Temporary fix whilst discussions are ongoing.
Signed-off-by: NJonathan Cameron <jic23@cam.ac.uk>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Smatch caught bug.
Signed-off-by: NJonathan Cameron <jic23@cam.ac.uk>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Introduced when I did the iio_chan_spec conversion. Detected by smatch.
Signed-off-by: NJonathan Cameron <jic23@cam.ac.uk>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Coccinelle found this one. I'm not certain what
the intent is, but this fix makes no functional changes.
Signed-off-by: NJonathan Cameron <jic23@cam.ac.uk>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Vivien pointed out that 99e5dc45
removed the IIO_DEVICE_ATTR_2 macro but didn't clean up the other
macros that used it.

As it turns out, no one is using val2 in tree, so lets scrap that
until it is needed.
Signed-off-by: NJonathan Cameron <jic23@cam.ac.uk>
Reported-by: NVivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

The unwinding index was j, not i and it is much simpler to do these as a single
loop unwinding those elements necessary in before jumping to the error handler.
Signed-off-by: NJonathan Cameron <jic23@cam.ac.uk>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

indio_dev->channels was accidentally removed thus preventing any sysfs
registrations occuring for direct reading.  That hid the other issue.

Unfortunately the incorrect handling leads to misspresentation of data
in sysfs reads.
Signed-off-by: NJonathan Cameron <jic23@cam.ac.uk>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Signed-off-by: NJonathan Cameron <jic23@cam.ac.uk>
Acked-by: NMichael Hennerich <michael.hennerich@analog.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Signed-off-by: NToon Schoenmakers <nighteyes1993@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Signed-off-by: NToon Schoenmakers <nighteyes1993@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

This patches nvec_kbd.c to the file that fixes up errors found by
the checkpath.pl tool.
Signed-off-by: NColin Brophy <colin@brophys.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

 Use kstrdup rather than duplicating its implementation

 The semantic patch that makes this output is available
 in scripts/coccinelle/api/kstrdup.cocci.

 More information about semantic patching is available at
 http://coccinelle.lip6.fr/Signed-off-by: NThomas Meyer <thomas@m3y3r.de>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Use kzalloc rather than kmalloc followed by memset with 0

This considers some simple cases that are common and easy to validate
Note in particular that there are no ...s in the rule, so all of the
matched code has to be contiguous

The semantic patch that makes this output is available
in scripts/coccinelle/api/alloc/kzalloc-simple.cocci.

More information about semantic patching is available at
http://coccinelle.lip6.fr/Signed-off-by: NThomas Meyer <thomas@m3y3r.de>
Acked-by: NRoland Vossen <rvossen@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

The brcmsmac driver supports different rates on 5GHz but this is not
taken into account when providing the rate index in the receive
status information passed to mac80211.
Reviewed-by: NRoland Vossen <rvossen@broadcom.com>
Reviewed-by: NPieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: NHenry Ptasinski <henryp@broadcom.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Message 'receive fifo overflow' appeared in the log. Root cause was an
invalid physical address being programmed into the DMA engine. This was
caused by an invalid pointer cast in the dma code. Issue was observed on
a Sparc (Sun Fire V120) machine but could theoretically also pop up
on other architectures.

Driver was tested to scan and ping on aforementioned machine.
Reviewed-by: NArend van Spriel <arend@broadcom.com>
Reviewed-by: NPieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Problem would pop up during driver load on a Sun Fire V120 and manifested
itself as an exception. This was caused by int* pointers provided to memcpy()
that were not aligned on an int boundary. The pointer type provided to
memcpy() is used by the compiler for optimization purposes. Fix was to cast
the int* pointers to void* pointers.

Bernhard R. Link and David S. Miller provided valuable feedback, thanks gents.
Reviewed-by: NArend van Spriel <arend@broadcom.com>
Reviewed-by: NPieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

When bringing down the netdevice interface a deadlock occurred
sporadically due to the rtnl_lock being held by a task that was
waiting for another task trying to get the lock. This patch fixes
that issue.
Reviewed-by: NFranky (Zhenhui) Lin <frankyl@broadcom.com>
Reviewed-by: NRoland Vossen <rvossen@broadcom.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Fix the __aligned checkpatch warning in wl_cfg80211.h
Reviewed-by: NRoland Vossen <rvossen@broadcom.com>
Reviewed-by: NHenry Ptasinski <henryp@broadcom.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

The cordic function which calculates cosine and sine values for given
angle is now provided in library module. The phy code now uses this
module function.
Reviewed-by: NHenry Ptasinski <henryp@broadcom.com>
Reviewed-by: NRoland Vossen <rvossen@broadcom.com>
Reviewed-by: NPieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

The crc8 function is now available as a library function provided
by kernel module crc8.ko. The function has been removed from brcmutil
module and srom.c now uses crc8 library function.
Reviewed-by: NHenry Ptasinski <henryp@broadcom.com>
Reviewed-by: NRoland Vossen <rvossen@broadcom.com>
Reviewed-by: NPieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

brcmsmac doesn't yet use bcma, but both drivers attempt to claim the same
device.  For now, turn of brcmsmac if bcma is enabled.
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Certain Mips based, big endian Broadcom products contain a memory window
for endianness swap purposes. However, this is too system specific for
this driver as there are more big endian platforms to support. Thus, this
'window' functionality is not to be used and can be removed.

The swap window would be a performance improvement and could be provided
with a dedicated mapping function in the platform specific code.

Tested on BCM63281.
Reviewed-by: NHenry Ptasinski <henryp@broadcom.com>
Reviewed-by: NArend van Spriel <arend@broadcom.com>
Tested-by: NJonas Gorski <jonas.gorski@gmail.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

The inline function dma_spin_for_len() was defined for MIPS platforms
but the problem only occurs with dma of the PCI core in bcm47xx chips.
This patch restricts the function further to BCM47XX platforms only.

Tested on BCM63281.
Reviewed-by: NPieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: NHenry Ptasinski <henryp@broadcom.com>
Reviewed-by: NRoland Vossen <rvossen@broadcom.com>
Tested-by: NJonas Gorski <jonas.gorski@gmail.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

The driver contained conditional code for resolving issue with
dma transaction reordering. This code was conditionalized using
__mips__ macro, but it actually is specific to bcm47xx chips.
This patch replaces it for the more speficic CONFIG_BCM47XX macro.

Tested on BCM63281.
Reviewed-by: NHenry Ptasinski <henryp@broadcom.com>
Reviewed-by: NRoland Vossen <rvossen@broadcom.com>
Tested-by: NJonas Gorski <jonas.gorski@gmail.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

The include <asm/addrspace.h> has been removed as it is not
needed. None of the definitions from it are used in the source
file.

Tested on BCM63281.
Reviewed-by: NHenry Ptasinski <henryp@broadcom.com>
Reviewed-by: NPieter-Paul Giesberts <pieterpg@broadcom.com>
Tested-by: NJonas Gorski <jonas.gorski@gmail.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

The firmware files are encoded as little-endian.  Do the appropriate swapping
for big-endian platforms.

Tested on Mac G5 PPC and BCM63281.
Signed-off-by: NHenry Ptasinski <henryp@broadcom.com>
Reviewed-by: NArend van Spriel <arend@broadcom.com>
Reviewed-by: NRoland Vossen <rvossen@broadcom.com>
Tested-by: NJonas Gorski <jonas.gorski@gmail.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Swapping the addresses is unnecessary, since the swaps are handled by the
underlying platform code (i.e. readb() etc. handle any necessary swapping).

Tested on Mac G5 PPC and BCM63281.
Signed-off-by: NHenry Ptasinski <henryp@broadcom.com>
Reviewed-by: NArend van Spriel <arend@broadcom.com>
Tested-by: NJonas Gorski <jonas.gorski@gmail.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

The original implementation iterated over the data in buf+1 through buf+size,
whereas it should have been over the data in buf through buf+size-1.

Tested on Mac G5 PPC and BCM63281.
Signed-off-by: NHenry Ptasinski <henryp@broadcom.com>
Reviewed-by: NArend van Spriel <arend@broadcom.com>
Reviewed-by: NRoland Vossen <rvossen@broadcom.com>
Tested-by: NJonas Gorski <jonas.gorski@gmail.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

The macro BUS_SWAP32(a) expanded to (a) which was fine for a little
endian system. For big endian platform this should do as the name
implies. As the driver is intended for PCI which is little-endian
the macro cpu_to_le32() can be used instead.
Reviewed-by: NPieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: NHenry Ptasinski <henryp@broadcom.com>
Reviewed-by: NRoland Vossen <rvossen@broadcom.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Four macros with PHYSADDR name prefix are not doing a lot and
complicate code readability. These have been expanded in the source
code and consequently removed the definitions.
Reviewed-by: NPieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: NHenry Ptasinski <henryp@broadcom.com>
Reviewed-by: NRoland Vossen <rvossen@broadcom.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

The macros W_SM and R_SM are not doing much conversion of the macro
parameters and complicate code readability without good cause. It
is more clear to remove usage of the macros expanding it in the source
code.
Reviewed-by: NPieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: NHenry Ptasinski <henryp@broadcom.com>
Reviewed-by: NRoland Vossen <rvossen@broadcom.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Remove improper usage of void * in dhd_linux.c and dhd_sdio.c
Reported-by: NJulian Calaby <julian.calaby@gmail.com>
Reviewed-by: NRoland Vossen <rvossen@broadcom.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Code cleanup. Replaced void pointers by less generic pointer types.
Reported-by: NJulian Calaby <julian.calaby@gmail.com>
Reviewed-by: NPieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Code cleanup. Otp is 'One Time Programmable' functionality. Replaced void
pointers by less generic pointer types.
Reported-by: NJulian Calaby <julian.calaby@gmail.com>
Reviewed-by: NPieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Code cleanup. Replaced by less generic pointer types.
Reported-by: NJulian Calaby <julian.calaby@gmail.com>
Reviewed-by: NPieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Code cleanup. Replace void * related to PCI functionality by less generic
pointer types.
Reported-by: NJulian Calaby <julian.calaby@gmail.com>
Reviewed-by: NPieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Get rid of some unused void *data parameter
Reviewed-by: NArend van Spriel <arend@broadcom.com>
Reviewed-by: NRoland Vossen <rvossen@broadcom.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Remove legacy dead code of SDTEST
Reviewed-by: NArend van Spriel <arend@broadcom.com>
Reviewed-by: NRoland Vossen <rvossen@broadcom.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Remove redundant complete function pointer in send/recv buffer interface.
Reviewed-by: NArend van Spriel <arend@broadcom.com>
Reviewed-by: NRoland Vossen <rvossen@broadcom.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Use brcmf_sdioh_request_byte directly instead of calling
brcmf_sdioh_cfg_read and brcmf_sdioh_cfg_write in brcmfmac
Reviewed-by: NArend van Spriel <arend@broadcom.com>
Reviewed-by: NRoland Vossen <rvossen@broadcom.com>
Signed-off-by: NArend van Spriel <arend@broadcom.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>