platform/kernel/linux-rpi.git
5 years agostaging: bcm2835-camera: Handle empty EOS buffers whilst streaming
Dave Stevenson [Sat, 29 Jun 2019 12:48:23 +0000 (14:48 +0200)]
staging: bcm2835-camera: Handle empty EOS buffers whilst streaming

The change to mapping V4L2 to MMAL buffers 1:1 didn't handle
the condition we get with raw pixel buffers (eg YUV and RGB)
direct from the camera's stills port. That sends the pixel buffer
and then an empty buffer with the EOS flag set. The EOS buffer
wasn't handled and returned an error up the stack.

Handle the condition correctly by returning it to the component
if streaming, or returning with an error if stopping streaming.

Fixes: 938416707071 ("staging: bcm2835-camera: Remove V4L2/MMAL buffer remapping")
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Stefan Wahren <wahrenst@gmx.net>
Acked-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Acked-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: bcm2835-camera: Remove check of the number of buffers supplied
Dave Stevenson [Sat, 29 Jun 2019 12:13:30 +0000 (14:13 +0200)]
staging: bcm2835-camera: Remove check of the number of buffers supplied

Before commit "staging: bcm2835-camera: Remove V4L2/MMAL buffer remapping"
there was a need to ensure that there were sufficient buffers supplied from
the user to cover those being sent to the VPU (always 1).

Now the buffers are linked 1:1 between MMAL and V4L2,
therefore there is no need for that check, and indeed it is wrong
as there is no need to submit all the buffers before starting streaming.

Fixes: 938416707071 ("staging: bcm2835-camera: Remove V4L2/MMAL buffer remapping")
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Stefan Wahren <wahrenst@gmx.net>
Acked-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Acked-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: bcm2835-camera: Ensure all buffers are returned on disable
Dave Stevenson [Sat, 29 Jun 2019 12:13:29 +0000 (14:13 +0200)]
staging: bcm2835-camera: Ensure all buffers are returned on disable

With the recent change to match MMAL and V4L2 buffers there
is a need to wait for all MMAL buffers to be returned during
stop_streaming.

Fixes: 938416707071 ("staging: bcm2835-camera: Remove V4L2/MMAL buffer remapping")
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Stefan Wahren <wahrenst@gmx.net>
Acked-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Acked-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: bcm2835-camera: Fix open parenthesis alignment
Dave Stevenson [Sat, 29 Jun 2019 12:13:28 +0000 (14:13 +0200)]
staging: bcm2835-camera: Fix open parenthesis alignment

Fix checkpatch "Alignment should match open parenthesis"
errors.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Stefan Wahren <wahrenst@gmx.net>
Acked-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Acked-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: bcm2835-camera: Fix missing lines between items
Dave Stevenson [Sat, 29 Jun 2019 12:13:27 +0000 (14:13 +0200)]
staging: bcm2835-camera: Fix missing lines between items

Fix checkpatch errors for missing blank lines after variable
or structure declarations.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Stefan Wahren <wahrenst@gmx.net>
Acked-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Acked-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: bcm2835-camera: Fix brace style issues.
Dave Stevenson [Sat, 29 Jun 2019 12:13:26 +0000 (14:13 +0200)]
staging: bcm2835-camera: Fix brace style issues.

Fix mismatched or missing brace issues flagged by checkpatch.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Stefan Wahren <wahrenst@gmx.net>
Acked-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Acked-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: bcm2835-camera: Fix multiple line dereference errors
Dave Stevenson [Sat, 29 Jun 2019 12:13:25 +0000 (14:13 +0200)]
staging: bcm2835-camera: Fix multiple line dereference errors

Fix checkpatch errors "Avoid multiple line dereference"

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Stefan Wahren <wahrenst@gmx.net>
Acked-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Acked-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: bcm2835-camera: Reduce length of enum names
Dave Stevenson [Sat, 29 Jun 2019 12:13:24 +0000 (14:13 +0200)]
staging: bcm2835-camera: Reduce length of enum names

We have numerous lines over 80 chars, or oddly split. Many
of these are due to using long enum names such as
MMAL_COMPONENT_CAMERA.
Reduce the length of these enum names.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Stefan Wahren <wahrenst@gmx.net>
Acked-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Acked-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: bcm2835-camera: Fix spacing around operators
Dave Stevenson [Sat, 29 Jun 2019 12:13:23 +0000 (14:13 +0200)]
staging: bcm2835-camera: Fix spacing around operators

Fix checkpatch warnings over spaces around operators.
Many were around operations that can be replaced with the
BIT(x) macro, so replace with that where appropriate.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Stefan Wahren <wahrenst@gmx.net>
Acked-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Acked-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: bcm2835-camera: Fix comment style violations.
Dave Stevenson [Sat, 29 Jun 2019 12:13:22 +0000 (14:13 +0200)]
staging: bcm2835-camera: Fix comment style violations.

Fix comment style violations in the header files.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Stefan Wahren <wahrenst@gmx.net>
Acked-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Acked-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: bcm2835-camera: Remove dead email addresses
Dave Stevenson [Sat, 29 Jun 2019 12:13:21 +0000 (14:13 +0200)]
staging: bcm2835-camera: Remove dead email addresses

None of the listed author email addresses were valid.
Keep list of authors and the companies they represented.
Update my email address.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Stefan Wahren <wahrenst@gmx.net>
Acked-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Acked-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: bcm2835-camera: Return early on errors
Dave Stevenson [Sat, 29 Jun 2019 12:13:20 +0000 (14:13 +0200)]
staging: bcm2835-camera: Return early on errors

Fix several instances where it is easier to return
early on error conditions than handle it as an else
clause. As requested by Mauro.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Stefan Wahren <wahrenst@gmx.net>
Acked-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Acked-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: bcm2835-camera: Correctly denote key frames in encoded data
Dave Stevenson [Sat, 29 Jun 2019 12:13:19 +0000 (14:13 +0200)]
staging: bcm2835-camera: Correctly denote key frames in encoded data

Forward MMAL key frame flags to the V4L2 buffers.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Stefan Wahren <wahrenst@gmx.net>
Acked-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Acked-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: bcm2835-camera: Do not bulk receive from service thread
Dave Stevenson [Sat, 29 Jun 2019 12:13:18 +0000 (14:13 +0200)]
staging: bcm2835-camera: Do not bulk receive from service thread

vchi_bulk_queue_receive will queue up to a default of 4
bulk receives on a connection before blocking.
If called from the VCHI service_callback thread, then
that thread is unable to service the VCHI_CALLBACK_BULK_RECEIVED
events that would enable the queue call to succeed.

Add a workqueue to schedule the call vchi_bulk_queue_receive
in an alternate context to avoid the lock up.

Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Stefan Wahren <wahrenst@gmx.net>
Acked-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Acked-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: bcm2835-camera: Replace spinlock protecting context_map with mutex
Dave Stevenson [Sat, 29 Jun 2019 12:13:17 +0000 (14:13 +0200)]
staging: bcm2835-camera: Replace spinlock protecting context_map with mutex

The commit "staging: bcm2835-camera: Replace open-coded idr with a struct idr."
replaced an internal implementation of an idr with the standard functions
and a spinlock. idr_alloc(GFP_KERNEL) can sleep whilst calling kmem_cache_alloc
to allocate the new node, but this is not valid whilst in an atomic context
due to the spinlock.

There is no need for this to be a spinlock as a standard mutex is
sufficient.

Fixes: 950fd867c635 ("staging: bcm2835-camera: Replace open-coded idr with a struct idr.")
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Stefan Wahren <wahrenst@gmx.net>
Acked-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Acked-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: bcm2835-camera: Check the error for REPEAT_SEQ_HEADER
Dave Stevenson [Sat, 29 Jun 2019 12:13:16 +0000 (14:13 +0200)]
staging: bcm2835-camera: Check the error for REPEAT_SEQ_HEADER

When handling for V4L2_CID_MPEG_VIDEO_REPEAT_SEQ_HEADER was added
the firmware would reject the setting if H264 hadn't already been
selected. This was fixed in the firmware at that point, but to
enable backwards compatibility the returned error was ignored.

That was Dec 2013, so the chances of having a firmware that still
has that issue is so close to zero that the workaround can be
removed.

Link: https://github.com/raspberrypi/linux/pull/2782/
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Stefan Wahren <wahrenst@gmx.net>
Acked-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Acked-by: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging/rtl8188eu/os_dep: Remove unneeded variable ret
Hariprasad Kelam [Sat, 29 Jun 2019 18:26:50 +0000 (23:56 +0530)]
staging/rtl8188eu/os_dep: Remove unneeded variable ret

Below list of functions returns 0 in success and -EINVAL in failure. So
directly return 0 on Success.

Signed-off-by: Hariprasad Kelam <hariprasad.kelam@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8188eu: Remove declarations of unused functions
Nishka Dasgupta [Wed, 26 Jun 2019 06:50:57 +0000 (12:20 +0530)]
staging: rtl8188eu: Remove declarations of unused functions

Remove the declarations of the following unused functions from
rtw_eeprom.h:
- eeprom_write16
- eeprom_read16
- eeprom_read_sz
- read_eeprom_content
- read_eeprom_content_by_attrib.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8188eu: hal: Replace function ODM_TXPowerTrackingCheck()
Nishka Dasgupta [Wed, 26 Jun 2019 06:50:56 +0000 (12:20 +0530)]
staging: rtl8188eu: hal: Replace function ODM_TXPowerTrackingCheck()

Remove function ODM_TXPowerTrackingCheck as all it does is call
odm_TXPowerTrackingCheckCE.
Rename odm_TXPowerTrackingCheckCE to ODM_TXPowerTrackingCheck for
compatibility with call sites.
Issue found with Coccinelle.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8188eu: hal: Replace function odm_TXPowerTrackingInit()
Nishka Dasgupta [Wed, 26 Jun 2019 06:50:55 +0000 (12:20 +0530)]
staging: rtl8188eu: hal: Replace function odm_TXPowerTrackingInit()

Remove function odm_TXPowerTrackingInit as all it does is call
odm_TXPowerTrackingThermalMeterInit.
Rename odm_TXPowerTrackingThermalMeterInit to odm_TXPowerTrackingInit
for compatibility with call sites.
Issue found with Coccinelle.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8188eu: os_dep: Remove return variable
Nishka Dasgupta [Wed, 26 Jun 2019 06:50:54 +0000 (12:20 +0530)]
staging: rtl8188eu: os_dep: Remove return variable

Remove return variable as its value is not altered between
initialisation and return.
Issue found with Coccinelle.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: netlogic: Change GFP_ATOMIC to GFP_KERNEL
Hariprasad Kelam [Sat, 29 Jun 2019 19:10:35 +0000 (00:40 +0530)]
staging: netlogic: Change GFP_ATOMIC to GFP_KERNEL

Below is data path of xlr_config_spill
xlr_net_probe
  -->xlr_config_fifo_spill_area
  --->xlr_config_spill

We can use GFP_KERNEL as this function is getting called from
xlr_net_probe and there are no locks.

Signed-off-by: Hariprasad Kelam <hariprasad.kelam@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rts5208: Simplify boolean expression to improve code style
Tobias Nießen [Wed, 26 Jun 2019 14:28:57 +0000 (16:28 +0200)]
staging: rts5208: Simplify boolean expression to improve code style

This bitwisen / boolean expression can be made more readable while
reducing the line lengths at the same time. This commit uses the
fact that

    a & (b | c) == (b | c)

evaluates to true if and only if

    (a & b) && (a & c)

is true. Since b and c are constants with relatively long names,
using the second form makes the code much more readable and shorter.

Signed-off-by: Tobias Nießen <tobias.niessen@stud.uni-hannover.de>
Signed-off-by: Sabrina Gaube <sabrina-gaube@web.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rts5208: Rewrite redundant if statement to improve code style
Tobias Nießen [Wed, 26 Jun 2019 14:28:56 +0000 (16:28 +0200)]
staging: rts5208: Rewrite redundant if statement to improve code style

This commit uses the fact that

    if (a) {
            if (b) {
                    ...
            }
    }

can instead be written as

    if (a && b) {
            ...
    }

without any change in behavior, allowing to decrease the indentation
of the contained code block and thus reducing the average line length.

Signed-off-by: Tobias Nießen <tobias.niessen@stud.uni-hannover.de>
Signed-off-by: Sabrina Gaube <sabrina-gaube@web.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: wilc1000: rename 'host_interface' source and header
Ajay Singh [Wed, 26 Jun 2019 12:41:06 +0000 (12:41 +0000)]
staging: wilc1000: rename 'host_interface' source and header

Rename 'host_interface' source and header file to include the 'wilc_'
prefix in its name.

Signed-off-by: Ajay Singh <ajay.kathat@microchip.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: wilc1000: remove extra argument passing to wilc_send_config_pkt()
Ajay Singh [Wed, 26 Jun 2019 12:41:04 +0000 (12:41 +0000)]
staging: wilc1000: remove extra argument passing to wilc_send_config_pkt()

Cleanup patch to remove the passing of driver handler, get the 'idx'
value inside the called function.

Signed-off-by: Ajay Singh <ajay.kathat@microchip.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: wilc1000: remove use of 'src_addr' element in 'wilc_vif' struct
Ajay Singh [Wed, 26 Jun 2019 12:41:01 +0000 (12:41 +0000)]
staging: wilc1000: remove use of 'src_addr' element in 'wilc_vif' struct

Remove use of 'src_addr' element in wilc_vif, as the same information
already copied to net_device->dev_addr.

Signed-off-by: Ajay Singh <ajay.kathat@microchip.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: wilc1000: remove unnecessary loop to traverse vif interfaces
Ajay Singh [Wed, 26 Jun 2019 12:40:58 +0000 (12:40 +0000)]
staging: wilc1000: remove unnecessary loop to traverse vif interfaces

Cleanup patch to avoid loop to traverse the interfaces instead make use
of vif received from net_device priv data.

Signed-off-by: Ajay Singh <ajay.kathat@microchip.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: wilc1000: remove use of driver_handler_id & ifc_id
Ajay Singh [Wed, 26 Jun 2019 12:40:55 +0000 (12:40 +0000)]
staging: wilc1000: remove use of driver_handler_id & ifc_id

Removed the 'driver_handler_id' & 'ifc_id' elements and used 'idx' to
identify the handler.

Signed-off-by: Ajay Singh <ajay.kathat@microchip.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: wilc1000: added support to dynamically add/remove interfaces
Ajay Singh [Wed, 26 Jun 2019 12:40:52 +0000 (12:40 +0000)]
staging: wilc1000: added support to dynamically add/remove interfaces

Removed the use of two hardcoded interfaces and added support to
add/remove the network interfaces dynamically.
Now the driver will have single default interface with name 'wlan0' and
later other interface can be added from user space application e.g
using 'iw add' command.
Also taken care to maintain 'wilc_vif' as part of 'net_device'
private data and 'wilc' struct as 'wiphy' private data.

Signed-off-by: Ajay Singh <ajay.kathat@microchip.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: wilc1000: fix error path cleanup in wilc_wlan_initialize()
Ajay Singh [Wed, 26 Jun 2019 12:40:48 +0000 (12:40 +0000)]
staging: wilc1000: fix error path cleanup in wilc_wlan_initialize()

For the error path in wilc_wlan_initialize(), the resources are not
cleanup in the correct order. Reverted the previous changes and use the
correct order to free during error condition.

Fixes: b46d68825c2d ("staging: wilc1000: remove COMPLEMENT_BOOT")
Cc: <stable@vger.kernel.org>
Signed-off-by: Ajay Singh <ajay.kathat@microchip.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: wilc1000: handle p2p operations in caller context
Ajay Singh [Wed, 26 Jun 2019 12:40:45 +0000 (12:40 +0000)]
staging: wilc1000: handle p2p operations in caller context

Moved the handling of p2p related operation in the caller context instead
of using workqueue.

Signed-off-by: Ajay Singh <ajay.kathat@microchip.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8188eu: Add 'rtl8188eufw.bin' to MODULE_FIRMWARE list
Alexander Dahl [Sun, 30 Jun 2019 08:05:40 +0000 (10:05 +0200)]
staging: rtl8188eu: Add 'rtl8188eufw.bin' to MODULE_FIRMWARE list

This is the file loaded by the code anyway, but now you can use
'modinfo' to determine the needed firmware file for this module.

Spotted when packaging firmware files for the fli4l Linux router
distribution, where a script uses the information from 'modinfo' to
collect all needed firmware files to package.

Cc: Christoph Schulz <mail@kristov.de>
Signed-off-by: Alexander Dahl <post@lespocky.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: mt7621-pci: fix PCIE_FTS_NUM_LO macro
Sergio Paracuellos [Wed, 26 Jun 2019 12:43:18 +0000 (14:43 +0200)]
staging: mt7621-pci: fix PCIE_FTS_NUM_LO macro

Add missing parenthesis to PCIE_FTS_NUM_LO macro to do the
same it was being done in original code.

Fixes: a4b2eb912bb1 ("staging: mt7621-pci: rewrite RC FTS configuration")
Signed-off-by: Sergio Paracuellos <sergio.paracuellos@gmail.com>
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: bcm2835-camera: Restore return behavior of ctrl_set_bitrate()
Stefan Wahren [Wed, 26 Jun 2019 15:48:11 +0000 (17:48 +0200)]
staging: bcm2835-camera: Restore return behavior of ctrl_set_bitrate()

The commit 52c4dfcead49 ("Staging: vc04_services: Cleanup in
ctrl_set_bitrate()") changed the return behavior of ctrl_set_bitrate().
We cannot do this because of a bug in the firmware, which breaks probing
of bcm2835-camera:

    bcm2835-v4l2: mmal_init: failed to set all camera controls: -3
    Cleanup: Destroy video encoder
    Cleanup: Destroy image encoder
    Cleanup: Destroy video render
    Cleanup: Destroy camera
    bcm2835-v4l2: bcm2835_mmal_probe: mmal init failed: -3
    bcm2835-camera: probe of bcm2835-camera failed with error -3

So restore the old behavior, add an explaining comment and a debug message
to verify that the bug has been fixed in firmware.

Fixes: 52c4dfcead49 ("Staging: vc04_services: Cleanup in ctrl_set_bitrate()")
Signed-off-by: Stefan Wahren <wahrenst@gmx.net>
Cc: stable <stable@vger.kernel.org>
Acked-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: greybus: tools: Remove function log_csv_error()
Nishka Dasgupta [Wed, 26 Jun 2019 06:44:50 +0000 (12:14 +0530)]
staging: greybus: tools: Remove function log_csv_error()

Remove unused function log_csv_error.
Issue found with Coccinelle.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Reviewed-by: Bryan O'Donoghue <pure.logic@nexus-software.ie>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: comedi: dt282x: fix a null pointer deref on interrupt
Ian Abbott [Wed, 26 Jun 2019 13:18:04 +0000 (14:18 +0100)]
staging: comedi: dt282x: fix a null pointer deref on interrupt

The interrupt handler `dt282x_interrupt()` causes a null pointer
dereference for those supported boards that have no analog output
support.  For these boards, `dev->write_subdev` will be `NULL` and
therefore the `s_ao` subdevice pointer variable will be `NULL`.  In that
case, the following call near the end of the interrupt handler results
in a null pointer dereference:

comedi_handle_events(dev, s_ao);

Fix it by only calling the above function if `s_ao` is valid.

(There are other uses of `s_ao` by the interrupt handler that may or may
not be reached depending on values of hardware registers.  Trust that
they are reliable for now.)

Note:
commit 4f6f009b204f ("staging: comedi: dt282x: use comedi_handle_events()")
propagates an earlier error from
commit f21c74fa4cfe ("staging: comedi: dt282x: use cfc_handle_events()").

Fixes: 4f6f009b204f ("staging: comedi: dt282x: use comedi_handle_events()")
Cc: <stable@vger.kernel.org> # v3.19+
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: comedi: amplc_pci230: fix null pointer deref on interrupt
Ian Abbott [Wed, 26 Jun 2019 13:17:39 +0000 (14:17 +0100)]
staging: comedi: amplc_pci230: fix null pointer deref on interrupt

The interrupt handler `pci230_interrupt()` causes a null pointer
dereference for a PCI260 card.  There is no analog output subdevice for
a PCI260.  The `dev->write_subdev` subdevice pointer and therefore the
`s_ao` subdevice pointer variable will be `NULL` for a PCI260.  The
following call near the end of the interrupt handler results in the null
pointer dereference for a PCI260:

comedi_handle_events(dev, s_ao);

Fix it by only calling the above function if `s_ao` is valid.

Note that the other uses of `s_ao` in the calls
`pci230_handle_ao_nofifo(dev, s_ao);` and `pci230_handle_ao_fifo(dev,
s_ao);` will never be reached for a PCI260, so they are safe.

Fixes: 39064f23284c ("staging: comedi: amplc_pci230: use comedi_handle_events()")
Cc: <stable@vger.kernel.org> # v3.19+
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging/most/video: set device_caps in struct video_device
Hans Verkuil [Wed, 26 Jun 2019 08:30:17 +0000 (10:30 +0200)]
staging/most/video: set device_caps in struct video_device

Instead of filling in the struct v4l2_capability device_caps
field, fill in the struct video_device device_caps field.

That way the V4L2 core knows what the capabilities of the
video device are.

Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: erofs: Replace kzalloc(struct ..) with kzalloc(*ptr)
Shobhit Kukreti [Thu, 27 Jun 2019 05:31:18 +0000 (22:31 -0700)]
staging: erofs: Replace kzalloc(struct ..) with kzalloc(*ptr)

Resolve checkpatch warning:
Prefer kzalloc(sizeof(*ptr)...) over kzalloc(sizeof(struct ..)

Signed-off-by: Shobhit Kukreti <shobhitkukreti@gmail.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Reviewed-by: Gao Xiang <gaoxiang25@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: erofs: don't check special inode layout
Yue Hu [Fri, 28 Jun 2019 03:42:34 +0000 (11:42 +0800)]
staging: erofs: don't check special inode layout

Currently, we will check if inode layout is compression or inline if
the inode is special in fill_inode(). Also set ->i_mapping->a_ops for
it. That is pointless since the both modes won't be set for special
inode when creating EROFS filesystem image. So, let's avoid it.

Signed-off-by: Yue Hu <huyue2@yulong.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: erofs: return the error value if fill_inline_data() fails
Yue Hu [Thu, 27 Jun 2019 09:46:15 +0000 (17:46 +0800)]
staging: erofs: return the error value if fill_inline_data() fails

We should consider the error returned by fill_inline_data() when filling
last page in fill_inode(). If not getting inode will be successful even
though last page is bad. That is illogical. Also change -EAGAIN to 0 in
fill_inline_data() to stand for successful filling.

Signed-off-by: Yue Hu <huyue2@yulong.com>
Reviewed-by: Gao Xiang <gaoxiang25@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8712: reduce stack usage, again
Arnd Bergmann [Fri, 28 Jun 2019 12:37:48 +0000 (14:37 +0200)]
staging: rtl8712: reduce stack usage, again

An earlier patch I sent reduced the stack usage enough to get
below the warning limit, and I could show this was safe, but with
GCC_PLUGIN_STRUCTLEAK_BYREF_ALL, it gets worse again because large stack
variables in the same function no longer overlap:

drivers/staging/rtl8712/rtl871x_ioctl_linux.c: In function 'translate_scan.isra.2':
drivers/staging/rtl8712/rtl871x_ioctl_linux.c:322:1: error: the frame size of 1200 bytes is larger than 1024 bytes [-Werror=frame-larger-than=]

Split out the largest two blocks in the affected function into two
separate functions and mark those noinline_for_stack.

Fixes: 8c5af16f7953 ("staging: rtl8712: reduce stack usage")
Fixes: 81a56f6dcd20 ("gcc-plugins: structleak: Generalize to all variable types")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8712: Change return values of r8712_setrfreg_cmd()
Nishka Dasgupta [Mon, 24 Jun 2019 05:49:03 +0000 (11:19 +0530)]
staging: rtl8712: Change return values of r8712_setrfreg_cmd()

Change return values of function r8712_setrfreg_cmd from _SUCCESS/_FAIL
to 0/-ENOMEM respectively. Modify call site accordingly.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8712: Remove r8712_setassocsta_cmd()
Nishka Dasgupta [Mon, 24 Jun 2019 05:49:02 +0000 (11:19 +0530)]
staging: rtl8712: Remove r8712_setassocsta_cmd()

Remove function r8712_setassocsta_cmd as it is only called by
mp_start_joinbss, which was removed in the previous patch.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8712: Remove mp_start_joinbss()
Nishka Dasgupta [Mon, 24 Jun 2019 05:49:01 +0000 (11:19 +0530)]
staging: rtl8712: Remove mp_start_joinbss()

Remove unused function mp_start_joinbss.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8712: Replace r8712_find_network()
Nishka Dasgupta [Mon, 24 Jun 2019 05:49:00 +0000 (11:19 +0530)]
staging: rtl8712: Replace r8712_find_network()

Remove function r8712_find_network as all it does is call
_r8712_find_network.
Rename _r8712_find_network to r8712_find_network for compatibility with
call sites.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8712: Remove r8712_setrttbl_cmd()
Nishka Dasgupta [Mon, 24 Jun 2019 05:48:59 +0000 (11:18 +0530)]
staging: rtl8712: Remove r8712_setrttbl_cmd()

Remove unused function r8712_setrttbl_cmd.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8712: Replace r8712_free_evt_priv()
Nishka Dasgupta [Mon, 24 Jun 2019 05:48:58 +0000 (11:18 +0530)]
staging: rtl8712: Replace r8712_free_evt_priv()

Remove function r8712_free_evt_priv as all it does is call
_free_evt_priv.
Rename _free_evt_priv to r8712_free_evt_priv to maintain compatibility
with call sites.
Change type of new r8712_free_evt_priv from static to non-static to
match old definition.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8712: Replace r8712_free_cmd_priv()
Nishka Dasgupta [Mon, 24 Jun 2019 05:48:57 +0000 (11:18 +0530)]
staging: rtl8712: Replace r8712_free_cmd_priv()

Remove function r8712_free_cmd_priv as all it does is call
_free_cmd_priv.
Change type of new r8712_free_cmd_priv from static to non-static to
match definition of original r8712_free_cmd_priv.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8712: Change return values of r8712_init_evt_priv()
Nishka Dasgupta [Mon, 24 Jun 2019 05:48:56 +0000 (11:18 +0530)]
staging: rtl8712: Change return values of r8712_init_evt_priv()

Change return values of the function r8712_init_evt_priv from
_SUCCESS/_FAIL to 0/-ENOMEM. Modify call site accordingly.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8712: Replace function r8712_init_evt_priv()
Nishka Dasgupta [Mon, 24 Jun 2019 05:48:55 +0000 (11:18 +0530)]
staging: rtl8712: Replace function r8712_init_evt_priv()

Remove function r8712_init_evt_priv as all it does is call
_init_evt_priv.
Rename _init_evt_priv to r8712_init_evt_priv to maintain compatibility
with call sites.
Change type of new r8712_init_evt_priv from static to non-static as
original r8712_init_evt_priv was non-static.
Change return type of new r8712_init_evt_priv to int as original had
return type u32 but new (formerly _init_evt_priv) had return type sint.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8712: Change return values in r8712_init_cmd_priv()
Nishka Dasgupta [Mon, 24 Jun 2019 05:48:54 +0000 (11:18 +0530)]
staging: rtl8712: Change return values in r8712_init_cmd_priv()

Change return values in r8712_init_cmd_priv from _SUCCESS/_FAIL to
0/-ENOMEM. Modify call site accordingly.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: vc04_services: Remove function block_resume()
Nishka Dasgupta [Tue, 25 Jun 2019 18:17:09 +0000 (23:47 +0530)]
staging: vc04_services: Remove function block_resume()

Remove function block_resume as it was only called by
vchiq_arm_force_suspend, which was removed in a previous patch.
Issue found with Coccinelle.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: vc04_services: Remove function output_timeout_error()
Nishka Dasgupta [Tue, 25 Jun 2019 18:17:08 +0000 (23:47 +0530)]
staging: vc04_services: Remove function output_timeout_error()

Remove function output_timeout_error as it was only called by
vchiq_arm_force_suspend, which was deleted in a previous patch.
Issue found with Coccinelle.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: vc04_services: Remove vchiq_send_remote_release()
Nishka Dasgupta [Tue, 25 Jun 2019 18:17:07 +0000 (23:47 +0530)]
staging: vc04_services: Remove vchiq_send_remote_release()

Remove unused function vchiq_send_remote_release.
Issue found with Coccinelle.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: vc04_services: Remove vchiq_use_service_no_resume()
Nishka Dasgupta [Tue, 25 Jun 2019 18:17:06 +0000 (23:47 +0530)]
staging: vc04_services: Remove vchiq_use_service_no_resume()

Remove unused function vchiq_use_service_no_resume.
Issue found with Coccinelle.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: vc04_services: Remove vchiq_resume_internal()
Nishka Dasgupta [Tue, 25 Jun 2019 18:17:05 +0000 (23:47 +0530)]
staging: vc04_services: Remove vchiq_resume_internal()

Remove unused function vchiq_resume_internal.
Issue found with Coccinelle.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: vc04_services: Remove vchiq_pause_internal()
Nishka Dasgupta [Tue, 25 Jun 2019 18:17:04 +0000 (23:47 +0530)]
staging: vc04_services: Remove vchiq_pause_internal()

Remove unused function vchiq_pause_internal.
Issue found with Coccinelle.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: vc04_services: Remove vchiq_arm_force_suspend()
Nishka Dasgupta [Tue, 25 Jun 2019 18:17:03 +0000 (23:47 +0530)]
staging: vc04_services: Remove vchiq_arm_force_suspend()

Remove unused function vchiq_arm_force_suspend.
Issue found with Coccinelle.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: vc04_services: Remove function vchiq_arm_allow_resume()
Nishka Dasgupta [Tue, 25 Jun 2019 18:17:02 +0000 (23:47 +0530)]
staging: vc04_services: Remove function vchiq_arm_allow_resume()

Remove unused function vchiq_arm_allow_resume.
Issue found with Coccinelle.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: vc04_services: Remove function vchiu_queue_is_full()
Nishka Dasgupta [Tue, 25 Jun 2019 18:17:01 +0000 (23:47 +0530)]
staging: vc04_services: Remove function vchiu_queue_is_full()

Remove unused function vchiu_queue_is_full.
Issue found with Coccinelle.

Signed-off-by: Nishka Dasgupta <nishkadg.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: kpc2000: add missing spaces in kpc2000_spi.c
Simon Sandström [Tue, 25 Jun 2019 08:41:28 +0000 (10:41 +0200)]
staging: kpc2000: add missing spaces in kpc2000_spi.c

Fixes checkpatch errors:
- spaces required around that '=' (ctx:VxV)
- space required before the open parenthesis '('
- spaces preferred around that '-' (ctx:VxV)
- space required before the open brace '{'

Signed-off-by: Simon Sandström <simon@nikanor.nu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: kpc2000: add missing spaces in kpc2000_i2c.c
Simon Sandström [Tue, 25 Jun 2019 08:41:27 +0000 (10:41 +0200)]
staging: kpc2000: add missing spaces in kpc2000_i2c.c

Fixes checkpatch "CHECK: spaces preferred around that '+' (ctx:VxV)".

Signed-off-by: Simon Sandström <simon@nikanor.nu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8188eu: remove unused function is_ap_in_wep()
Michael Straube [Sun, 23 Jun 2019 17:59:57 +0000 (19:59 +0200)]
staging: rtl8188eu: remove unused function is_ap_in_wep()

Function is_ap_in_wep() is not used in the driver code, so remove it.

Signed-off-by: Michael Straube <straube.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8188eu: remove unused function get_bsstype()
Michael Straube [Sun, 23 Jun 2019 17:59:56 +0000 (19:59 +0200)]
staging: rtl8188eu: remove unused function get_bsstype()

Function get_bsstype() is not used in the driver code, so remove it.

Signed-off-by: Michael Straube <straube.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8723bs: os_dep: Change return type of rtw_init_default_value() to void
Shobhit Kukreti [Sat, 22 Jun 2019 16:40:42 +0000 (09:40 -0700)]
staging: rtl8723bs: os_dep: Change return type of rtw_init_default_value() to void

rtw_init_default_value() func always returns a value (u8)_SUCCESS.
Modified return type to void to resolve coccicheck warnings
of unneeded variable.

Signed-off-by: Shobhit Kukreti <shobhitkukreti@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8723bs: os_dep: modified return type of function rtw_suspend_wow() to...
Shobhit Kukreti [Sat, 22 Jun 2019 16:40:41 +0000 (09:40 -0700)]
staging: rtl8723bs: os_dep: modified return type of function rtw_suspend_wow() to void

Changed return type of function rtw_suspend_wow() to void.
The function always return _SUCCESS and the value is never
checked in the calling function.

Resolves coccicheck Unneeded variable "ret" warning.

Signed-off-by: Shobhit Kukreti <shobhitkukreti@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8723bs: os_dep: Change return type of function rtw_suspend_normal() to...
Shobhit Kukreti [Sat, 22 Jun 2019 16:40:40 +0000 (09:40 -0700)]
staging: rtl8723bs: os_dep: Change return type of function rtw_suspend_normal() to void

Coccicheck issues Unneeded variable "ret" warning.
The return value of function rtw_suspend_normal() is set to _SUCCESS.
The return value is never never checked by the calling function.
Modified return type to void to remove the coccicheck warning..

Signed-off-by: Shobhit Kukreti <shobhitkukreti@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8723bs: os_dep: Modify return type of function rtw_reset_drv_sw() to...
Shobhit Kukreti [Sat, 22 Jun 2019 15:23:08 +0000 (08:23 -0700)]
staging: rtl8723bs: os_dep: Modify return type of function rtw_reset_drv_sw() to void.

The function rtw_reset_drv_sw() return value is set to _SUCCESS.
The return value is never checked when the function is called.
Modified the return value to void to remove "Unneeded Variable warning
of coccicheck.

Signed-off-by: Shobhit Kukreti <shobhitkukreti@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8723bs: os_dep: Modify return type of function loadparam(..) to void
Shobhit Kukreti [Sat, 22 Jun 2019 15:23:07 +0000 (08:23 -0700)]
staging: rtl8723bs: os_dep: Modify return type of function loadparam(..) to void

The function static uint loadparam(struct adapter *padapter, _nic_hdl
pnetdev) return type is modified to void.
The initial return value was always returning _SUCCESS and the return value
is never checked when the function is called.
This resolves coccicheck warnings of unneeded variables.

Signed-off-by: Shobhit Kukreti <shobhitkukreti@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8188eu: remove hal_init_macaddr()
Michael Straube [Sat, 22 Jun 2019 15:14:49 +0000 (17:14 +0200)]
staging: rtl8188eu: remove hal_init_macaddr()

Function hal_init_macaddr() just calls rtw_hal_set_hwreg().
Use rtw_hal_set_hwreg() directly and remove hal_init_macaddr().

Signed-off-by: Michael Straube <straube.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: rtl8188eu: cleanup lines ending with a '('
Michael Straube [Sat, 22 Jun 2019 15:14:48 +0000 (17:14 +0200)]
staging: rtl8188eu: cleanup lines ending with a '('

Cleanup checkpatch issues in usb_halinit.c.
CHECK: Lines should not end with a '('

Signed-off-by: Michael Straube <straube.linux@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: erofs: integrate decompression inplace
Gao Xiang [Mon, 24 Jun 2019 07:22:58 +0000 (15:22 +0800)]
staging: erofs: integrate decompression inplace

Decompressor needs to know whether it's a partial
or full decompression since only full decompression
can be decompressed in-place.

On kirin980 platform, sequential read is finally
increased to 812MiB/s after decompression inplace
is enabled.

Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Gao Xiang <gaoxiang25@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: erofs: switch to new decompression backend
Gao Xiang [Mon, 24 Jun 2019 07:22:57 +0000 (15:22 +0800)]
staging: erofs: switch to new decompression backend

This patch integrates new decompression framework to
erofs decompression path, and remove the old
decompression implementation as well.

On kirin980 platform, sequential read is slightly
improved to 778MiB/s after the new decompression
backend is used.

Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Gao Xiang <gaoxiang25@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: erofs: introduce LZ4 decompression inplace
Gao Xiang [Mon, 24 Jun 2019 07:22:56 +0000 (15:22 +0800)]
staging: erofs: introduce LZ4 decompression inplace

compressed data will be usually loaded into last pages of
the extent (the last page for 4k) for in-place decompression
(more specifically, in-place IO), as ilustration below,

         start of compressed logical extent
           |                          end of this logical extent
           |                           |
     ______v___________________________v________
... |  page 6  |  page 7  |  page 8  |  page 9  | ...
    |__________|__________|__________|__________|
           .                         ^ .        ^
           .                         |compressed|
           .                         |   data   |
           .                           .        .
           |<          dstsize        >|<margin>|
                                       oend     iend
           op                        ip

Therefore, it's possible to do decompression inplace (thus no
memcpy at all) if the margin is sufficient and safe enough [1],
and it can be implemented only for fixed-size output compression
compared with fixed-size input compression.

No memcpy for most of in-place IO (about 99% of enwik9) after
decompression inplace is implemented and sequential read will
be improved of course (see the following patches for test results).

[1] https://github.com/lz4/lz4/commit/b17f578a919b7e6b078cede2d52be29dd48c8e8c
    https://github.com/lz4/lz4/commit/5997e139f53169fa3a1c1b4418d2452a90b01602

Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Gao Xiang <gaoxiang25@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: erofs: introduce generic decompression backend
Gao Xiang [Mon, 24 Jun 2019 07:22:55 +0000 (15:22 +0800)]
staging: erofs: introduce generic decompression backend

This patch adds a new generic decompression framework
in order to replace the old LZ4-specific decompression code.

Even though LZ4 is still the only supported algorithm, yet
it is more cleaner and easy to integrate new algorithm than
the old almost hard-coded decompression backend.

Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Gao Xiang <gaoxiang25@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: erofs: move stagingpage operations to compress.h
Gao Xiang [Mon, 24 Jun 2019 07:22:54 +0000 (15:22 +0800)]
staging: erofs: move stagingpage operations to compress.h

stagingpages are behaved as bounce pages for temporary use.
Move to compress.h since the upcoming decompressor will
allocate stagingpages as well.

Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Gao Xiang <gaoxiang25@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: erofs: move per-CPU buffers implementation to utils.c
Gao Xiang [Mon, 24 Jun 2019 07:22:53 +0000 (15:22 +0800)]
staging: erofs: move per-CPU buffers implementation to utils.c

This patch moves per-CPU buffers to utils.c in order for
the upcoming generic decompression framework to use it.

Note that I tried to use generic per-CPU buffer or
per-CPU page approaches to clean up further, but obvious
performanace regression (about 2% for sequential read) was
observed.

Therefore let's leave it as it is instead, just move
to utils.c and I'll try to dig into the root cause later.

Signed-off-by: Gao Xiang <gaoxiang25@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: erofs: add compacted compression indexes support
Gao Xiang [Mon, 24 Jun 2019 07:22:52 +0000 (15:22 +0800)]
staging: erofs: add compacted compression indexes support

This patch aims at compacted compression indexes:
 1) cleanup z_erofs_map_blocks_iter and move into zmap.c;
 2) add compacted 4/2B decoding support.

On kirin980 platform, sequential read is increased about
6% (725MiB/s -> 770MiB/s) on enwik9 dataset if compacted 2B
feature is enabled.

Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Gao Xiang <gaoxiang25@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: erofs: add compacted ondisk compression indexes
Gao Xiang [Mon, 24 Jun 2019 07:22:51 +0000 (15:22 +0800)]
staging: erofs: add compacted ondisk compression indexes

This patch introduces new compacted compression indexes.

In contract to legacy compression indexes that
   each 4k logical cluster has an 8-byte index,
compacted ondisk compression indexes will have
   amortized 2 bytes for each 4k logical cluster (compacted 2B)
   amortized 4 bytes for each 4k logical cluster (compacted 4B)

In detail, several continuous clusters will be encoded in
a compacted pack with cluster types, offsets, and one blkaddr
at the end of the pack to leave 4-byte margin for better
decoding performance, as illustrated below:
   _____________________________________________
  |___@_____ encoded bits __________|_ blkaddr _|
  0       .                                     amortized * vcnt
  .          .
  .             .                   amortized * vcnt - 4
  .                .
  .___________________.
  |_type_|_clusterofs_|

Note that compacted 2 / 4B should be aligned with 32 / 8 bytes
in order to avoid each pack crossing page boundary.

Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Gao Xiang <gaoxiang25@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: comedi: use dma_mmap_coherent for DMA-able buffer mmap
Ian Abbott [Tue, 25 Jun 2019 11:26:59 +0000 (12:26 +0100)]
staging: comedi: use dma_mmap_coherent for DMA-able buffer mmap

Comedi's acquisition buffer allocation code can allocate the buffer from
normal kernel memory or from DMA coherent memory depending on the
`dma_async_dir` value in the comedi subdevice.  (A value of `DMA_NONE`
causes the buffer to be allocated from normal kernel memory.  Other
values cause the buffer to be allocated from DMA coherent memory.)   The
buffer currently consists of a bunch of page-sized blocks that are
vmap'ed into a contiguous range of virtual addresses. The pages are also
mmap'able into user address space.  For DMA'able buffers, these
page-sized blocks are allocated by `dma_alloc_coherent()`.

For DMA-able buffers, the DMA API is currently abused in various ways,
the most serious abuse being the calling of `virt_to_page()` on the
blocks allocated by `dma_alloc_coherent()` and passing those pages to
`vmap()` (for mapping to the kernels vmalloc address space) and via
`page_to_pfn()` to `remap_pfn_range()` (for mmap'ing to user space).  it
also uses the `__GFP_COMP` flag when allocating the blocks, and marks
the allocated pages as reserved (which is unnecessary for DMA coherent
allocations).

The code can be changed to get rid of the vmap'ed address altogether if
necessary, since there are only a few places in the comedi code that use
the vmap'ed address directly and we also keep a list of the kernel
addresses for the individual pages prior to the vmap operation. This
would add some run-time overhead to buffer accesses.  The real killer is
the mmap operation.

For mmap, the address range specified in the VMA needs to be mmap'ed to
the individually allocated page-sized blocks.  That is not a problem
when the pages are allocated from normal kernel memory as the individual
pages can be remapped by `remap_pfn_range()`, but it is a problem when
the page-sized blocks are allocated by `dma_alloc_coherent()` because
the DMA API currently has no support for splitting a VMA across multiple
blocks of DMA coherent memory (or rather, no support for mapping part of
a VMA range to a single block of DMA coherent memory).

In order to comply with the DMA API and allow the buffer to be mmap'ed,
the buffer needs to be allocated as a single block by a single call to
`dma_alloc_coherent()`, freed by a single call to `dma_free_coherent()`,
and mmap'ed to user space by a single call to `dma_mmap_coherent()`.
This patch changes the buffer allocation, freeing, and mmap'ing code to
do that, with the unfortunate consequence that buffer allocation is more
likely to fail.  It also no longer uses the `__GFP_COMP` flag when
allocating DMA coherent memory, no longer marks the
allocated pages of DMA coherent memory as reserved, and no longer vmap's
the DMA coherent memory pages (since they are contiguous anyway).

Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoRevert "staging: ks7010: Fix build error"
Greg Kroah-Hartman [Mon, 24 Jun 2019 08:45:34 +0000 (16:45 +0800)]
Revert "staging: ks7010: Fix build error"

This reverts commit 3e5bc68fa596874500e8c718605aa44d5e42a34c as it
causes build errors in linux-next.

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoMerge 5.2-rc6 into staging-next
Greg Kroah-Hartman [Sun, 23 Jun 2019 11:29:03 +0000 (13:29 +0200)]
Merge 5.2-rc6 into staging-next

We want the fixes and this resolves a merge issue as well.

Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agoLinux 5.2-rc6
Linus Torvalds [Sat, 22 Jun 2019 23:01:36 +0000 (16:01 -0700)]
Linux 5.2-rc6

5 years agoMerge tag 'iommu-fix-v5.2-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/joro...
Linus Torvalds [Sat, 22 Jun 2019 21:08:47 +0000 (14:08 -0700)]
Merge tag 'iommu-fix-v5.2-rc5' of git://git./linux/kernel/git/joro/iommu

Pull iommu fix from Joerg Roedel:
 "Revert a commit from the previous pile of fixes which causes new
  lockdep splats. It is better to revert it for now and work on a better
  and more well tested fix"

* tag 'iommu-fix-v5.2-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu:
  Revert "iommu/vt-d: Fix lock inversion between iommu->lock and device_domain_lock"

5 years agoRevert "iommu/vt-d: Fix lock inversion between iommu->lock and device_domain_lock"
Peter Xu [Fri, 21 Jun 2019 02:32:05 +0000 (10:32 +0800)]
Revert "iommu/vt-d: Fix lock inversion between iommu->lock and device_domain_lock"

This reverts commit 7560cc3ca7d9d11555f80c830544e463fcdb28b8.

With 5.2.0-rc5 I can easily trigger this with lockdep and iommu=pt:

    ======================================================
    WARNING: possible circular locking dependency detected
    5.2.0-rc5 #78 Not tainted
    ------------------------------------------------------
    swapper/0/1 is trying to acquire lock:
    00000000ea2b3beb (&(&iommu->lock)->rlock){+.+.}, at: domain_context_mapping_one+0xa5/0x4e0
    but task is already holding lock:
    00000000a681907b (device_domain_lock){....}, at: domain_context_mapping_one+0x8d/0x4e0
    which lock already depends on the new lock.
    the existing dependency chain (in reverse order) is:
    -> #1 (device_domain_lock){....}:
           _raw_spin_lock_irqsave+0x3c/0x50
           dmar_insert_one_dev_info+0xbb/0x510
           domain_add_dev_info+0x50/0x90
           dev_prepare_static_identity_mapping+0x30/0x68
           intel_iommu_init+0xddd/0x1422
           pci_iommu_init+0x16/0x3f
           do_one_initcall+0x5d/0x2b4
           kernel_init_freeable+0x218/0x2c1
           kernel_init+0xa/0x100
           ret_from_fork+0x3a/0x50
    -> #0 (&(&iommu->lock)->rlock){+.+.}:
           lock_acquire+0x9e/0x170
           _raw_spin_lock+0x25/0x30
           domain_context_mapping_one+0xa5/0x4e0
           pci_for_each_dma_alias+0x30/0x140
           dmar_insert_one_dev_info+0x3b2/0x510
           domain_add_dev_info+0x50/0x90
           dev_prepare_static_identity_mapping+0x30/0x68
           intel_iommu_init+0xddd/0x1422
           pci_iommu_init+0x16/0x3f
           do_one_initcall+0x5d/0x2b4
           kernel_init_freeable+0x218/0x2c1
           kernel_init+0xa/0x100
           ret_from_fork+0x3a/0x50

    other info that might help us debug this:
     Possible unsafe locking scenario:
           CPU0                    CPU1
           ----                    ----
      lock(device_domain_lock);
                                   lock(&(&iommu->lock)->rlock);
                                   lock(device_domain_lock);
      lock(&(&iommu->lock)->rlock);

     *** DEADLOCK ***
    2 locks held by swapper/0/1:
     #0: 00000000033eb13d (dmar_global_lock){++++}, at: intel_iommu_init+0x1e0/0x1422
     #1: 00000000a681907b (device_domain_lock){....}, at: domain_context_mapping_one+0x8d/0x4e0

    stack backtrace:
    CPU: 2 PID: 1 Comm: swapper/0 Not tainted 5.2.0-rc5 #78
    Hardware name: LENOVO 20KGS35G01/20KGS35G01, BIOS N23ET50W (1.25 ) 06/25/2018
    Call Trace:
     dump_stack+0x85/0xc0
     print_circular_bug.cold.57+0x15c/0x195
     __lock_acquire+0x152a/0x1710
     lock_acquire+0x9e/0x170
     ? domain_context_mapping_one+0xa5/0x4e0
     _raw_spin_lock+0x25/0x30
     ? domain_context_mapping_one+0xa5/0x4e0
     domain_context_mapping_one+0xa5/0x4e0
     ? domain_context_mapping_one+0x4e0/0x4e0
     pci_for_each_dma_alias+0x30/0x140
     dmar_insert_one_dev_info+0x3b2/0x510
     domain_add_dev_info+0x50/0x90
     dev_prepare_static_identity_mapping+0x30/0x68
     intel_iommu_init+0xddd/0x1422
     ? printk+0x58/0x6f
     ? lockdep_hardirqs_on+0xf0/0x180
     ? do_early_param+0x8e/0x8e
     ? e820__memblock_setup+0x63/0x63
     pci_iommu_init+0x16/0x3f
     do_one_initcall+0x5d/0x2b4
     ? do_early_param+0x8e/0x8e
     ? rcu_read_lock_sched_held+0x55/0x60
     ? do_early_param+0x8e/0x8e
     kernel_init_freeable+0x218/0x2c1
     ? rest_init+0x230/0x230
     kernel_init+0xa/0x100
     ret_from_fork+0x3a/0x50

domain_context_mapping_one() is taking device_domain_lock first then
iommu lock, while dmar_insert_one_dev_info() is doing the reverse.

That should be introduced by commit:

7560cc3ca7d9 ("iommu/vt-d: Fix lock inversion between iommu->lock and
              device_domain_lock", 2019-05-27)

So far I still cannot figure out how the previous deadlock was
triggered (I cannot find iommu lock taken before calling of
iommu_flush_dev_iotlb()), however I'm pretty sure that that change
should be incomplete at least because it does not fix all the places
so we're still taking the locks in different orders, while reverting
that commit is very clean to me so far that we should always take
device_domain_lock first then the iommu lock.

We can continue to try to find the real culprit mentioned in
7560cc3ca7d9, but for now I think we should revert it to fix current
breakage.

CC: Joerg Roedel <joro@8bytes.org>
CC: Lu Baolu <baolu.lu@linux.intel.com>
CC: dave.jiang@intel.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Tested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
5 years agoMerge tag 'pci-v5.2-fixes-1' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci
Linus Torvalds [Sat, 22 Jun 2019 16:42:29 +0000 (09:42 -0700)]
Merge tag 'pci-v5.2-fixes-1' of git://git./linux/kernel/git/helgaas/pci

Pull PCI fix from Bjorn Helgaas:
 "If an IOMMU is present, ignore the P2PDMA whitelist we added for v5.2
  because we don't yet know how to support P2PDMA in that case (Logan
  Gunthorpe)"

* tag 'pci-v5.2-fixes-1' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci:
  PCI/P2PDMA: Ignore root complex whitelist when an IOMMU is present

5 years agoMerge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Linus Torvalds [Sat, 22 Jun 2019 16:39:03 +0000 (09:39 -0700)]
Merge tag 'scsi-fixes' of git://git./linux/kernel/git/jejb/scsi

Pull SCSI fixes from James Bottomley:
 "Three driver fixes (and one version number update): a suspend hang in
  ufs, a qla hard lock on module removal and a qedi panic during
  discovery"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: qla2xxx: Fix hardlockup in abort command during driver remove
  scsi: ufs: Avoid runtime suspend possibly being blocked forever
  scsi: qedi: update driver version to 8.37.0.20
  scsi: qedi: Check targetname while finding boot target information

5 years agoMerge tag 'powerpc-5.2-5' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc...
Linus Torvalds [Sat, 22 Jun 2019 16:09:42 +0000 (09:09 -0700)]
Merge tag 'powerpc-5.2-5' of git://git./linux/kernel/git/powerpc/linux

Pull powerpc fixes from Michael Ellerman:
 "This is a frustratingly large batch at rc5. Some of these were sent
  earlier but were missed by me due to being distracted by other things,
  and some took a while to track down due to needing manual bisection on
  old hardware. But still we clearly need to improve our testing of KVM,
  and of 32-bit, so that we catch these earlier.

  Summary: seven fixes, all for bugs introduced this cycle.

   - The commit to add KASAN support broke booting on 32-bit SMP
     machines, due to a refactoring that moved some setup out of the
     secondary CPU path.

   - A fix for another 32-bit SMP bug introduced by the fast syscall
     entry implementation for 32-bit BOOKE. And a build fix for the same
     commit.

   - Our change to allow the DAWR to be force enabled on Power9
     introduced a bug in KVM, where we clobber r3 leading to a host
     crash.

   - The same commit also exposed a previously unreachable bug in the
     nested KVM handling of DAWR, which could lead to an oops in a
     nested host.

   - One of the DMA reworks broke the b43legacy WiFi driver on some
     people's powermacs, fix it by enabling a 30-bit ZONE_DMA on 32-bit.

   - A fix for TLB flushing in KVM introduced a new bug, as it neglected
     to also flush the ERAT, this could lead to memory corruption in the
     guest.

  Thanks to: Aaro Koskinen, Christoph Hellwig, Christophe Leroy, Larry
  Finger, Michael Neuling, Suraj Jitindar Singh"

* tag 'powerpc-5.2-5' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  KVM: PPC: Book3S HV: Invalidate ERAT when flushing guest TLB entries
  powerpc: enable a 30-bit ZONE_DMA for 32-bit pmac
  KVM: PPC: Book3S HV: Only write DAWR[X] when handling h_set_dawr in real mode
  KVM: PPC: Book3S HV: Fix r3 corruption in h_set_dabr()
  powerpc/32: fix build failure on book3e with KVM
  powerpc/booke: fix fast syscall entry on SMP
  powerpc/32s: fix initial setup of segment registers on secondary CPU

5 years agoBluetooth: Fix regression with minimum encryption key size alignment
Marcel Holtmann [Sat, 22 Jun 2019 13:47:01 +0000 (15:47 +0200)]
Bluetooth: Fix regression with minimum encryption key size alignment

When trying to align the minimum encryption key size requirement for
Bluetooth connections, it turns out doing this in a central location in
the HCI connection handling code is not possible.

Original Bluetooth version up to 2.0 used a security model where the
L2CAP service would enforce authentication and encryption.  Starting
with Bluetooth 2.1 and Secure Simple Pairing that model has changed into
that the connection initiator is responsible for providing an encrypted
ACL link before any L2CAP communication can happen.

Now connecting Bluetooth 2.1 or later devices with Bluetooth 2.0 and
before devices are causing a regression.  The encryption key size check
needs to be moved out of the HCI connection handling into the L2CAP
channel setup.

To achieve this, the current check inside hci_conn_security() has been
moved into l2cap_check_enc_key_size() helper function and then called
from four decisions point inside L2CAP to cover all combinations of
Secure Simple Pairing enabled devices and device using legacy pairing
and legacy service security model.

Fixes: d5bb334a8e17 ("Bluetooth: Align minimum encryption key size for LE and BR/EDR connections")
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=203643
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
Cc: stable@vger.kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Sat, 22 Jun 2019 05:23:35 +0000 (22:23 -0700)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) Fix leak of unqueued fragments in ipv6 nf_defrag, from Guillaume
    Nault.

 2) Don't access the DDM interface unless the transceiver implements it
    in bnx2x, from Mauro S. M. Rodrigues.

 3) Don't double fetch 'len' from userspace in sock_getsockopt(), from
    JingYi Hou.

 4) Sign extension overflow in lio_core, from Colin Ian King.

 5) Various netem bug fixes wrt. corrupted packets from Jakub Kicinski.

 6) Fix epollout hang in hvsock, from Sunil Muthuswamy.

 7) Fix regression in default fib6_type, from David Ahern.

 8) Handle memory limits in tcp_fragment more appropriately, from Eric
    Dumazet.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (24 commits)
  tcp: refine memory limit test in tcp_fragment()
  inet: clear num_timeout reqsk_alloc()
  net: mvpp2: debugfs: Add pmap to fs dump
  ipv6: Default fib6_type to RTN_UNICAST when not set
  net: hns3: Fix inconsistent indenting
  net/af_iucv: always register net_device notifier
  net/af_iucv: build proper skbs for HiperTransport
  net/af_iucv: remove GFP_DMA restriction for HiperTransport
  net: dsa: mv88e6xxx: fix shift of FID bits in mv88e6185_g1_vtu_loadpurge()
  hvsock: fix epollout hang from race condition
  net/udp_gso: Allow TX timestamp with UDP GSO
  net: netem: fix use after free and double free with packet corruption
  net: netem: fix backlog accounting for corrupted GSO frames
  net: lio_core: fix potential sign-extension overflow on large shift
  tipc: pass tunnel dev as NULL to udp_tunnel(6)_xmit_skb
  ip6_tunnel: allow not to count pkts on tstats by passing dev as NULL
  ip_tunnel: allow not to count pkts on tstats by setting skb's dev to NULL
  tun: wake up waitqueues after IFF_UP is set
  net: remove duplicate fetch in sock_getsockopt
  tipc: fix issues with early FAILOVER_MSG from peer
  ...

5 years agotcp: refine memory limit test in tcp_fragment()
Eric Dumazet [Fri, 21 Jun 2019 13:09:55 +0000 (06:09 -0700)]
tcp: refine memory limit test in tcp_fragment()

tcp_fragment() might be called for skbs in the write queue.

Memory limits might have been exceeded because tcp_sendmsg() only
checks limits at full skb (64KB) boundaries.

Therefore, we need to make sure tcp_fragment() wont punish applications
that might have setup very low SO_SNDBUF values.

Fixes: f070ef2ac667 ("tcp: tcp_fragment() should apply sane memory limits")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Christoph Paasch <cpaasch@apple.com>
Tested-by: Christoph Paasch <cpaasch@apple.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma
Linus Torvalds [Fri, 21 Jun 2019 21:47:09 +0000 (14:47 -0700)]
Merge tag 'for-linus' of git://git./linux/kernel/git/rdma/rdma

Pull rdma fixes from Doug Ledford:
 "This is probably our last -rc pull request. We don't have anything
  else outstanding at the moment anyway, and with the summer months on
  us and people taking trips, I expect the next weeks leading up to the
  merge window to be pretty calm and sedate.

  This has two simple, no brainer fixes for the EFA driver.

  Then it has ten not quite so simple fixes for the hfi1 driver. The
  problem with them is that they aren't simply one liner typo fixes.
  They're still fixes, but they're more complex issues like livelock
  under heavy load where the answer was to change work queue usage and
  spinlock usage to resolve the problem, or issues with orphaned
  requests during certain types of failures like link down which
  required some more complex work to fix too. They all look like
  legitimate fixes to me, they just aren't small like I wish they were.

  Summary:

   - 2 minor EFA fixes

   - 10 hfi1 fixes related to scaling issues"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
  RDMA/efa: Handle mmap insertions overflow
  RDMA/efa: Fix success return value in case of error
  IB/hfi1: Handle port down properly in pio
  IB/hfi1: Handle wakeup of orphaned QPs for pio
  IB/hfi1: Wakeup QPs orphaned on wait list after flush
  IB/hfi1: Use aborts to trigger RC throttling
  IB/hfi1: Create inline to get extended headers
  IB/hfi1: Silence txreq allocation warnings
  IB/hfi1: Avoid hardlockup with flushlist_lock
  IB/hfi1: Correct tid qp rcd to match verbs context
  IB/hfi1: Close PSM sdma_progress sleep window
  IB/hfi1: Validate fault injection opcode user input

5 years agoMerge tag 'nfs-for-5.2-3' of git://git.linux-nfs.org/projects/anna/linux-nfs
Linus Torvalds [Fri, 21 Jun 2019 20:45:41 +0000 (13:45 -0700)]
Merge tag 'nfs-for-5.2-3' of git://git.linux-nfs.org/projects/anna/linux-nfs

Pull more NFS client fixes from Anna Schumaker:
 "These are mostly refcounting issues that people have found recently.
  The revert fixes a suspend recovery performance issue.

   - SUNRPC: Fix a credential refcount leak

   - Revert "SUNRPC: Declare RPC timers as TIMER_DEFERRABLE"

   - SUNRPC: Fix xps refcount imbalance on the error path

   - NFS4: Only set creation opendata if O_CREAT"

* tag 'nfs-for-5.2-3' of git://git.linux-nfs.org/projects/anna/linux-nfs:
  SUNRPC: Fix a credential refcount leak
  Revert "SUNRPC: Declare RPC timers as TIMER_DEFERRABLE"
  net :sunrpc :clnt :Fix xps refcount imbalance on the error path
  NFS4: Only set creation opendata if O_CREAT

5 years agox86/vdso: Prevent segfaults due to hoisted vclock reads
Andy Lutomirski [Fri, 21 Jun 2019 15:43:04 +0000 (08:43 -0700)]
x86/vdso: Prevent segfaults due to hoisted vclock reads

GCC 5.5.0 sometimes cleverly hoists reads of the pvclock and/or hvclock
pages before the vclock mode checks.  This creates a path through
vclock_gettime() in which no vclock is enabled at all (due to disabled
TSC on old CPUs, for example) but the pvclock or hvclock page
nevertheless read.  This will segfault on bare metal.

This fixes commit 459e3a21535a ("gcc-9: properly declare the
{pv,hv}clock_page storage") in the sense that, before that commit, GCC
didn't seem to generate the offending code.  There was nothing wrong
with that commit per se, and -stable maintainers should backport this to
all supported kernels regardless of whether the offending commit was
present, since the same crash could just as easily be triggered by the
phase of the moon.

On GCC 9.1.1, this doesn't seem to affect the generated code at all, so
I'm not too concerned about performance regressions from this fix.

Cc: stable@vger.kernel.org
Cc: x86@kernel.org
Cc: Borislav Petkov <bp@alien8.de>
Reported-by: Duncan Roe <duncan_roe@optusnet.com.au>
Signed-off-by: Andy Lutomirski <luto@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agoSUNRPC: Fix a credential refcount leak
Trond Myklebust [Thu, 20 Jun 2019 14:47:40 +0000 (10:47 -0400)]
SUNRPC: Fix a credential refcount leak

All callers of __rpc_clone_client() pass in a value for args->cred,
meaning that the credential gets assigned and referenced in
the call to rpc_new_client().

Reported-by: Ido Schimmel <idosch@idosch.org>
Fixes: 79caa5fad47c ("SUNRPC: Cache cred of process creating the rpc_client")
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Tested-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
5 years agoRevert "SUNRPC: Declare RPC timers as TIMER_DEFERRABLE"
Anna Schumaker [Tue, 18 Jun 2019 18:57:33 +0000 (14:57 -0400)]
Revert "SUNRPC: Declare RPC timers as TIMER_DEFERRABLE"

Jon Hunter reports:
  "I have been noticing intermittent failures with a system suspend test on
   some of our machines that have a NFS mounted root file-system. Bisecting
   this issue points to your commit 431235818bc3 ("SUNRPC: Declare RPC
   timers as TIMER_DEFERRABLE") and reverting this on top of v5.2-rc3 does
   appear to resolve the problem.

   The cause of the suspend failure appears to be a long delay observed
   sometimes when resuming from suspend, and this is causing our test to
   timeout."

This reverts commit 431235818bc3a919ca7487500c67c3144feece80.

Reported-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
5 years agonet :sunrpc :clnt :Fix xps refcount imbalance on the error path
Lin Yi [Mon, 10 Jun 2019 02:16:56 +0000 (10:16 +0800)]
net :sunrpc :clnt :Fix xps refcount imbalance on the error path

rpc_clnt_add_xprt take a reference to struct rpc_xprt_switch, but forget
to release it before return, may lead to a memory leak.

Signed-off-by: Lin Yi <teroincn@163.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
5 years agoNFS4: Only set creation opendata if O_CREAT
Benjamin Coddington [Fri, 7 Jun 2019 10:37:30 +0000 (06:37 -0400)]
NFS4: Only set creation opendata if O_CREAT

We can end up in nfs4_opendata_alloc during task exit, in which case
current->fs has already been cleaned up.  This leads to a crash in
current_umask().

Fix this by only setting creation opendata if we are actually doing an open
with O_CREAT.  We can drop the check for NULL nfs4_open_createattrs, since
O_CREAT will never be set for the recovery path.

Suggested-by: Trond Myklebust <trondmy@hammerspace.com>
Signed-off-by: Benjamin Coddington <bcodding@redhat.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>