linux-toradex.git/drivers/mfd/tps65910.c, branch v3.10.76

mfd: tps65910: Fix possible invalid pointer dereference on regmap_add_irq_chip fail

2014-05-06T14:55:28+00:00

commit 483e2dfdbc94751430e41db9973985f5b054d322 upstream.

Fixes: 4aab3fadad32 ("mfd: tps65910: Move interrupt implementation code to mfd file")

tps65910_irq_init() sets 'tps65910->chip_irq' before calling
regmap_add_irq_chip(). If the regmap_add_irq_chip() call fails in
memory allocation of regmap_irq_chip_data members then:

1. The 'tps65910->chip_irq' will still hold some value
2. 'tps65910->irq_data' will be pointing to already freed memory
   (because regmap_add_irq_chip() will free it on error)

This results in invalid memory access during driver remove because the
tps65910_irq_exit() tests whether 'tps65910->chip_irq' is not zero.

Signed-off-by: Krzysztof Kozlowski 
Signed-off-by: Lee Jones 
Signed-off-by: Greg Kroah-Hartman

Merge tag 'mfd-3.8-1' of git://git.kernel.org/pub/scm/linux/kernel/git/sameo/mfd-2.6

2012-12-17T02:55:20+00:00

Pull MFS update from Samuel Ortiz:
 "This is the MFD patch set for the 3.8 merge window.

  We have several new drivers, most of the time coming with their sub
  devices drivers:

   - Austria Microsystem's AS3711
   - Nano River's viperboard
   - TI's TPS80031, AM335x TS/ADC,
   - Realtek's MMC/memstick card reader
   - Nokia's retu

  We also got some notable cleanups and improvements:

   - tps6586x got converted to IRQ domains.
   - tps65910 and tps65090 moved to the regmap IRQ API.
   - STMPE is now Device Tree aware.
   - A general twl6040 and twl-core cleanup, with moves to the regmap
     I/O and IRQ APIs and a conversion to the recently added PWM
     framework.
   - sta2x11 gained regmap support.

  Then the rest is mostly tiny cleanups and fixes, among which we have
  Mark's wm5xxx and wm8xxx patchset."

Far amount of annoying but largely trivial conflicts.  Many due to
__devinit/exit removal, others due to one or two of the new drivers also
having come in through another tree.

* tag 'mfd-3.8-1' of git://git.kernel.org/pub/scm/linux/kernel/git/sameo/mfd-2.6: (119 commits)
  mfd: tps6507x: Convert to devm_kzalloc
  mfd: stmpe: Update DT support for stmpe driver
  mfd: wm5102: Add readback of DSP status 3 register
  mfd: arizona: Log if we fail to create the primary IRQ domain
  mfd: tps80031: MFD_TPS80031 needs to select REGMAP_IRQ
  mfd: tps80031: Add terminating entry for tps80031_id_table
  mfd: sta2x11: Fix potential NULL pointer dereference in __sta2x11_mfd_mask()
  mfd: wm5102: Add tuning for revision B
  mfd: arizona: Defer patch initialistation until after first device boot
  mfd: tps65910: Fix wrong ack_base register
  mfd: tps65910: Remove unused data
  mfd: stmpe: Get rid of irq_invert_polarity
  mfd: ab8500-core: Fix invalid free of devm_ allocated data
  mfd: wm5102: Mark DSP memory regions as volatile
  mfd: wm5102: Correct default for LDO1_CONTROL_2
  mfd: arizona: Register haptics devices
  mfd: wm8994: Make current device behaviour the default
  mfd: tps65090: MFD_TPS65090 needs to select REGMAP_IRQ
  mfd: Fix stmpe.c build when OF is not enabled
  mfd: jz4740-adc: Use devm_kzalloc
  ...

mfd: tps65910: Fix wrong ack_base register

2012-11-30T11:20:48+00:00

The interrupt status registers of TPS65910/1 should be cleared
when the associated interrupt event occurs.
This work is done in the regmap irq thread - using 'ack_base' register.
The ACK registers should be fixed as status register, not mask registers.

This patch fixes the infinite interrupt event problem by clearing
interrupt status registers.

Tested on the AM3517 Craneboard.

Signed-off-by: Milo(Woogyom) Kim 
Acked-by: Laxman Dewangan 
Signed-off-by: Samuel Ortiz

mfd: tps65910: Remove unused data

2012-11-30T11:20:35+00:00

The 'io_mutex' is not used anywhere.
The regmap API supports the mutex internally, so no additional mutex
is required.

And 'domain' private data is unnecessary because the irq domain is
already registered by using regmap_add_irq_chip().

Signed-off-by: Milo(Woogyom) Kim 
Acked-by: Laxman Dewangan 
Signed-off-by: Samuel Ortiz

mfd: remove use of __devexit

2012-11-28T20:22:27+00:00

CONFIG_HOTPLUG is going away as an option so __devexit is no
longer needed.

Signed-off-by: Bill Pemberton 
Cc: Srinidhi Kasagar 
Cc: Peter Tyser 
Cc: Daniel Walker 
Cc: Bryan Huntsman 
Acked-by: David Brown 
Acked-by: Linus Walleij 
Acked-by: Mark Brown 
Signed-off-by: Greg Kroah-Hartman

mfd: remove use of __devinit

2012-11-28T20:20:19+00:00

CONFIG_HOTPLUG is going away as an option so __devinit is no longer
needed.

Signed-off-by: Bill Pemberton 
Cc: Srinidhi Kasagar 
Cc: Peter Tyser 
Cc: Daniel Walker 
Cc: Bryan Huntsman 
Acked-by: David Brown 
Acked-by: Mark Brown 
Signed-off-by: Greg Kroah-Hartman

mfd: remove use of __devexit_p

2012-11-28T20:16:42+00:00

CONFIG_HOTPLUG is going away as an option so __devexit_p is no longer
needed.

Signed-off-by: Bill Pemberton 
Cc: Srinidhi Kasagar 
Cc: Peter Tyser 
Cc: Daniel Walker 
Cc: Bryan Huntsman 
Acked-by: David Brown 
Acked-by: Linus Walleij 
Acked-by: Mark Brown 
Signed-off-by: Greg Kroah-Hartman

mfd: tps65910: Pass irq_domain when adding mfd sub devices

2012-11-20T11:21:10+00:00

When adding the sub device "tps65910-rtc", is it passed the
IO resource IRQ for the interrupt number. This interrupt needs
to map in the device irq domain. Pass the irq domain of device
in mfd_add_devices() so that proper irq mapping can be done when
adding the sub device RTC.

Signed-off-by: Laxman Dewangan 
Reviewed-by: Mark Brown 
Signed-off-by: Samuel Ortiz

mfd: tps65910: Move interrupt implementation code to mfd file

2012-11-20T11:21:09+00:00

In place of implementing the irq support in separate file,
moving implementation to main mfd file.
The irq files only contains the table and init steps only
and does not need extra file to have this only for this
purpose.

Signed-off-by: Laxman Dewangan 
Reviewed-by: Mark Brown 
Signed-off-by: Samuel Ortiz

mfd: tps65910: Initialize mfd devices after all initialization done

2012-11-20T11:21:08+00:00

Add sub devices of tps65910 after all initialization like interrupt,
clock etc. is done. This will make sure that require data gets
initialized properly before sub devices probe's get called.

Signed-off-by: Laxman Dewangan 
Reviewed-by: Mark Brown 
Signed-off-by: Samuel Ortiz