staging: wlang-ng: avoid new typedef: hfa384x_pdr_level_compc_measurements_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_pdr_level_compc_measurements_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_pdr_vgdac_measurements_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_pdr_vgdac_measurements_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_pdr_refdac_measurements_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_pdr_refdac_measurements_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_pdr_nicid_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_pdr_nicid_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_pdr_cfisuprange_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_pdr_cfisuprange_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_pdr_mfisuprange_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_pdr_mfisuprange_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_pdr_nic_ramsize_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_pdr_nic_ramsize_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_pdr_mkk_measurements_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_pdr_mkk_measurements_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_pdr_nic_serial_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_pdr_nic_serial_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_pdr_pcb_tracenum_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_pdr_pcb_tracenum_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_pdr_pcb_partnum_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_pdr_pcb_partnum_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_usbin_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_usbin_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_usbout_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_usbout_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_usb_error_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_usb_error_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_usb_bufavail_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_usb_bufavail_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_usb_rmemresp_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_usb_rmemresp_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_usb_wmemresp_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_usb_wmemresp_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_usb_rridresp_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_usb_rridresp_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_usb_wridresp_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_usb_wridresp_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_usb_cmdresp_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_usb_cmdresp_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_usb_infofrm_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_usb_infofrm_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_usb_rxfrm_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_usb_rxfrm_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_usb_rmemreq_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_usb_rmemreq_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_usb_wmemreq_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_usb_wmemreq_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_usb_rridreq_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_usb_rridreq_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_usb_wridreq_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_usb_wridreq_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_usb_cmdreq_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_usb_cmdreq_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_usb_txfrm_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_usb_txfrm_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_InfFrame_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_InfFrame_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_infodata_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_infodata_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_KeyIDChanged_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_KeyIDChanged_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_PSUserCount_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_PSUserCount_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_AuthReq_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_AuthReq_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_AssocStatus_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_AssocStatus_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_LinkStatus_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_LinkStatus_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_HScanResult_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_HScanResult_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_HScanResultSub_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_HScanResultSub_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_ChInfoResult_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_ChInfoResult_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_ChInfoResultSub_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_ChInfoResultSub_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_ScanResult_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_ScanResult_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_ScanResultSub_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_ScanResultSub_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_CommTallies32_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_CommTallies32_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_CommTallies16_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_CommTallies16_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_rx_frame_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_rx_frame_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_tx_frame_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_tx_frame_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_dbmcommsquality_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_dbmcommsquality_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_commsquality_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_commsquality_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_downloadbuffer_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_downloadbuffer_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_WPAData_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_WPAData_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_authenticateStation_data_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_authenticateStation_data_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_JoinRequest_data_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_JoinRequest_data_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_HostScanRequest_data_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_HostScanRequest_data_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_caplevel_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_caplevel_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: Fix block comments style warnings in hfa384x.h

 This patch fixes the following checkpatch.pl warnings in hfa384x.h:
 WARNING: Block comments use * on subsequent lines
 WARNING: Block comments use a trailing */ on a separate line

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_compident_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_compident_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: wlang-ng: avoid new typedef: hfa384x_bytestr32_t

This patch fixes the following checkpatch.pl warning in hfa384x.h:
WARNING: do not add new typedefs

It applies for typedef hfa384x_bytestr32_t

Signed-off-by: Sergio Paracuellos <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: android: ion: fix checks found by checkpatch.pl

Alignment fixes

Signed-off-by: Wayne Porter <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: android: ion: fix warning found by checkpatch.pl

Fix checkpatch.pl warning:
WARNING: braces {} are not necessary for single statement blocks

Signed-off-by: Wayne Porter <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: media: replace unsigned with unsigned int

Fix checkpatch error "Prefer 'unsigned int' to bare use of 'unsigned'"
for readability.

Signed-off-by: Elizabeth Ferdman <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: sm750fb: replace roundedDiv with DIV_ROUND_CLOSEST

Replace local implementation of rounded division (roundedDiv macro) with
the in-kernel implementation (DIV_ROUND_CLOSEST macro) in ddk750_chip.c

Signed-off-by: Moshe Green <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: sm750fb: conform to block comment style

Fix 2 checkpatch errors:
-Block comments use * on subsequent lines,
-Block comments use a trailing */ on a separate line
to conform to block commenting style.

Signed-off-by: Elizabeth Ferdman <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: greybus: light: check the correct value of delay_on

When checking the value of delay_on to set the channel as active, it was
checked the pointer and not the value, as it should be.

Fixes: cc43368a3c ("greybus: lights: Control runtime pm suspend/resume on AP side")
Signed-off-by: Rui Miguel Silva <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: greybus: Add remaining audio files to maintained list

Some Greybus Audio Device Class Protocol driver files are missing
in the MAINTAINERS file even though they are maintained so add
them.

CC: Vaibhav Agarwal <[email protected]>
Signed-off-by: Mark Greer <[email protected]>
Reviewed-by: Viresh Kumar <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: greybus: light: check delay_{on|off} before use

Even though we trust leds core that the pointers should be valid, we are
safer to check delay_{on|off} before use.

Also, this avoid a smatch warning:
drivers/staging/greybus/light.c:484 gb_blink_set()
warn: variable dereferenced before check 'delay_on' (see line 476)

Signed-off-by: Rui Miguel Silva <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: greybus: light: fix attributes allocation

Fix allocation of attributes with the correct size, this also fix smatch
warning:

drivers/staging/greybus/light.c:293 channel_attr_groups_set()
warn: double check that we're allocating correct size: 8 vs 16

Signed-off-by: Rui Miguel Silva <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: greybus: sdio: fix cmd_flags check for none response

When checking for command flags field if response is not available we
really need to compare it with the right define and not bitwise AND it.

smatch warn:
        drivers/staging/greybus/sdio.c:481 gb_sdio_command()
        warn: bitwise AND condition is false here

Reported-by: Dan Carpenter <[email protected]>
Signed-off-by: Rui Miguel Silva <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

staging: add bcm2708 vchiq driver

Signed-off-by: popcornmix <[email protected]>
vchiq: create_pagelist copes with vmalloc memory

Signed-off-by: Daniel Stone <[email protected]>
vchiq: fix the shim message release

Signed-off-by: Daniel Stone <[email protected]>
vchiq: export additional symbols

Signed-off-by: Daniel Stone <[email protected]>
VCHIQ: Make service closure fully synchronous (drv)

This is one half of a two-part patch, the other half of which is to
the vchiq_lib user library. With these patches, calls to
vchiq_close_service and vchiq_remove_service won't return until any
associated callbacks have been delivered to the callback thread.

VCHIQ: Add per-service tracing

The new service option VCHIQ_SERVICE_OPTION_TRACE is a boolean that
toggles tracing for the specified service.

This commit also introduces vchi_service_set_option and the associated
option VCHI_SERVICE_OPTION_TRACE.

vchiq: Make the synchronous-CLOSE logic more tolerant

vchiq: Move logging control into debugfs

vchiq: Take care of a corner case tickled by VCSM

Closing a connection that isn't fully open requires care, since one
side does not know the other side's port number. Code was present to
handle the case where a CLOSE is sent immediately after an OPEN, i.e.
before the OPENACK has been received, but this was incorrectly being
used when an OPEN from a client using port 0 was rejected.

(In the observed failure, the host was attempting to use the VCSM
service, which isn't present in the 'cutdown' firmware. The failure
was intermittent because sometimes the keepalive service would
grab port 0.)

This case can be distinguished because the client's remoteport will
still be VCHIQ_PORT_FREE, and the srvstate will be OPENING. Either
condition is sufficient to differentiate it from the special case
described above.

vchiq: Avoid high load when blocked and unkillable

vchiq: Include SIGSTOP and SIGCONT in list of signals not-masked by vchiq to allow gdb to work

vchiq_arm: Complete support for SYNCHRONOUS mode

vchiq: Remove inline from suspend/resume

vchiq: Allocation does not need to be atomic

vchiq: Fix wrong condition check

The log level is checked from within the log call. Remove the check in the call.

Signed-off-by: Pranith Kumar <[email protected]>
BCM270x: Add vchiq device to platform file and Device Tree

Prepare to turn the vchiq module into a driver.

Signed-off-by: Noralf Trønnes <[email protected]>
bcm2708: vchiq: Add Device Tree support

Turn vchiq into a driver and stop hardcoding resources.
Use devm_* functions in probe path to simplify cleanup.
A global variable is used to hold the register address. This is done
to keep this patch as small as possible.
Also make available on ARCH_BCM2835.
Based on work by Lubomir Rintel.

Signed-off-by: Noralf Trønnes <[email protected]>
vchiq: Change logging level for inbound data

vchiq_arm: Two cacheing fixes

1) Make fragment size vary with cache line size
Without this patch, non-cache-line-aligned transfers may corrupt
(or be corrupted by) adjacent data structures.

Both ARM and VC need to be updated to enable this feature. This is
ensured by having the loader apply a new DT parameter -
cache-line-size. The existence of this parameter guarantees that the
kernel is capable, and the parameter will only be modified from the
safe default if the loader is capable.

2) Flush/invalidate vmalloc'd memory, and invalidate after reads

vchiq: fix NULL pointer dereference when closing driver

The following code run as root will cause a null pointer dereference oops:

        int fd = open("/dev/vc-cma", O_RDONLY);
        if (fd < 0)
                err(1, "open failed");
        (void)close(fd);

[ 1704.877721] Unable to handle kernel NULL pointer dereference at virtual address 00000000
[ 1704.877725] pgd = b899c000
[ 1704.877736] [00000000] *pgd=37fab831, *pte=00000000, *ppte=00000000
[ 1704.877748] Internal error: Oops: 817 [#1] PREEMPT SMP ARM
[ 1704.877765] Modules linked in: evdev i2c_bcm2708 uio_pdrv_genirq uio
[ 1704.877774] CPU: 2 PID: 3656 Comm: stress-ng-fstat Not tainted 3.19.1-12-generic-bcm2709 #12-Ubuntu
[ 1704.877777] Hardware name: BCM2709
[ 1704.877783] task: b8ab9b00 ti: b7e68000 task.ti: b7e68000
[ 1704.877798] PC is at __down_interruptible+0x50/0xec
[ 1704.877806] LR is at down_interruptible+0x5c/0x68
[ 1704.877813] pc : [<80630ee8>]    lr : [<800704b0>]    psr: 60080093
sp : b7e69e50  ip : b7e69e88  fp : b7e69e84
[ 1704.877817] r10: b88123c8  r9 : 00000010  r8 : 00000001
[ 1704.877822] r7 : b8ab9b00  r6 : 7fffffff  r5 : 80a1cc34  r4 : 80a1cc34
[ 1704.877826] r3 : b7e69e50  r2 : 00000000  r1 : 00000000  r0 : 80a1cc34
[ 1704.877833] Flags: nZCv  IRQs off  FIQs on  Mode SVC_32  ISA ARM  Segment user
[ 1704.877838] Control: 10c5387d  Table: 3899c06a  DAC: 00000015
[ 1704.877843] Process do-oops (pid: 3656, stack limit = 0xb7e68238)
[ 1704.877848] Stack: (0xb7e69e50 to 0xb7e6a000)
[ 1704.877856] 9e40:                                     80a1cc3c 00000000 00000010 b88123c8
[ 1704.877865] 9e60: b7e69e84 80a1cc34 fff9fee9 ffffffff b7e68000 00000009 b7e69ea4 b7e69e88
[ 1704.877874] 9e80: 800704b0 80630ea4 fff9fee9 60080013 80a1cc28 fff9fee9 b7e69edc b7e69ea8
[ 1704.877884] 9ea0: 8040f558 80070460 fff9fee9 ffffffff 00000000 00000000 00000009 80a1cb7c
[ 1704.877893] 9ec0: 00000000 80a1cb7c 00000000 00000010 b7e69ef4 b7e69ee0 803e1ba4 8040f514
[ 1704.877902] 9ee0: 00000e48 80a1cb7c b7e69f14 b7e69ef8 803e1c9c 803e1b74 b88123c0 b92acb18
[ 1704.877911] 9f00: b8812790 b8d815d8 b7e69f24 b7e69f18 803e2250 803e1bc8 b7e69f5c b7e69f28
[ 1704.877921] 9f20: 80167bac 803e222c 00000000 00000000 b7e69f54 b8ab9ffc 00000000 8098c794
[ 1704.877930] 9f40: b8ab9b00 8000efc4 b7e68000 00000000 b7e69f6c b7e69f60 80167d6c 80167b28
[ 1704.877939] 9f60: b7e69f8c b7e69f70 80047d38 80167d60 b7e68000 b7e68010 8000efc4 b7e69fb0
[ 1704.877949] 9f80: b7e69fac b7e69f90 80012820 80047c84 01155490 011549a8 00000001 00000006
[ 1704.877957] 9fa0: 00000000 b7e69fb0 8000ee5c 80012790 00000000 353d8c0f 7efc4308 00000000
[ 1704.877966] 9fc0: 01155490 011549a8 00000001 00000006 00000000 00000000 76cf3ba0 00000003
[ 1704.877975] 9fe0: 00000000 7efc42e4 0002272f 76e2ed66 60080030 00000003 00000000 00000000
[ 1704.877998] [<80630ee8>] (__down_interruptible) from [<800704b0>] (down_interruptible+0x5c/0x68)
[ 1704.878015] [<800704b0>] (down_interruptible) from [<8040f558>] (vchiu_queue_push+0x50/0xd8)
[ 1704.878032] [<8040f558>] (vchiu_queue_push) from [<803e1ba4>] (send_worker_msg+0x3c/0x54)
[ 1704.878045] [<803e1ba4>] (send_worker_msg) from [<803e1c9c>] (vc_cma_set_reserve+0xe0/0x1c4)
[ 1704.878057] [<803e1c9c>] (vc_cma_set_reserve) from [<803e2250>] (vc_cma_release+0x30/0x38)
[ 1704.878069] [<803e2250>] (vc_cma_release) from [<80167bac>] (__fput+0x90/0x1e0)
[ 1704.878082] [<80167bac>] (__fput) from [<80167d6c>] (____fput+0x18/0x1c)
[ 1704.878094] [<80167d6c>] (____fput) from [<80047d38>] (task_work_run+0xc0/0xf8)
[ 1704.878109] [<80047d38>] (task_work_run) from [<80012820>] (do_work_pending+0x9c/0xc4)
[ 1704.878123] [<80012820>] (do_work_pending) from [<8000ee5c>] (work_pending+0xc/0x20)
[ 1704.878133] Code: e50b1034 e3a01000 e50b2030 e580300c (e5823000)

..the fix is to ensure that we have actually initialized the queue before we attempt
to push any items onto it.  This occurs if we do an open() followed by a close() without
any activity in between.

Signed-off-by: Colin Ian King <[email protected]>
vchiq_arm: Sort out the vmalloc case

See: https://github.com/raspberrypi/linux/issues/1055

vchiq: hack: Add include depecated dma include file

[gregkh] added dependancy on CONFIG_BROKEN to make things sane for now.

Cc: Eric Anholt <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

IB/hfi1: Fix trace of atomic ack

The length is incorrect, causing the trace data to
be truncated.

Add the additional 8 bytes that should have been there.
Also trace out the atomic ack in hex to aid debugging.

Reviewed-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Mike Marciniszyn <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Update SMA ingress checks for response packets

Fix "unsupported method" error by skipping ingress pkey checks on
response SMA packets.

Reviewed-by: Dennis Dalessandro <[email protected]>
Reviewed-by: Mike Marciniszyn <[email protected]>
Signed-off-by: Jianxin Xiong <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Use EPROM platform configuration read

The driver will now try to read directly from the EPROM as its
first choice for the platform configuration file.

Reviewed-by: Easwar Hariharan <[email protected]>
Signed-off-by: Dean Luick <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Add ability to read platform config from the EPROM

Add a function to read the platform configuration file from
the EPROM.

Reviewed-by: Easwar Hariharan <[email protected]>
Signed-off-by: Dean Luick <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Restore EPROM read ability

Partially revert commit d07903174202 ("IB/hfi1: Remove
EPROM functionality from data device"), bringing back
the ability to read from the EPROM.

This code will be used for driver-only acccess to the EPROM, hence
change EPROM read to save to a buffer instead of copy touser. Also
allow any offset and remove missed includes and leftover declarations.

Reviewed-by: Easwar Hariharan <[email protected]>
Signed-off-by: Dean Luick <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Document new sysfs entries for hfi1 driver

This patch adds description for the sdma engine related sysfs entries
for the HFI1 driver.

Reviewed-by: Dennis Dalessandro <[email protected]>
Reviewed-by: Sebastian Sanchez <[email protected]>
Reviewed-by: Jianxin Xiong <[email protected]>
Signed-off-by: Tadeusz Struk <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Add new debugfs sdma_cpu_list file

Add a debugfs sdma_cpu_list file that can be used to examine the CPU to
sdma engine assignments for the whole device.

Reviewed-by: Dennis Dalessandro <[email protected]>
Reviewed-by: Sebastian Sanchez <[email protected]>
Reviewed-by: Jianxin Xiong <[email protected]>
Signed-off-by: Tadeusz Struk <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Add irq affinity notification handler

This patch adds an irq affinity notification handler.
When a user changes interrupt affinity settings for an sdma engine,
the driver needs to make changes to its internal sde structures and
also update the affinity_hint.

Reviewed-by: Dennis Dalessandro <[email protected]>
Reviewed-by: Sebastian Sanchez <[email protected]>
Reviewed-by: Jianxin Xiong <[email protected]>
Signed-off-by: Tadeusz Struk <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Add a new VL sysfs attribute for sdma engines

This patch adds a read-only "VL" attribute for the sysfs entry of each
sdma engine. It will allow the user to check VL to sdma engine mappings.

Reviewed-by: Dennis Dalessandro <[email protected]>
Reviewed-by: Sebastian Sanchez <[email protected]>
Reviewed-by: Jianxin Xiong <[email protected]>
Signed-off-by: Tadeusz Struk <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Add sysfs interface for affinity setup

Some users want more control over which cpu cores are being used by the
driver. For example, users might want to restrict the driver to some
specified subset of the cores so that they can appropriately partition
processes, irq handlers, and work threads.
To allow the user to fine tune system affinity settings new sysfs
attributes are introduced per sdma engine.  This patch adds a new
attribute type for sdma engine and a new cpu_list attribute.
When the user writes a cpu range to the cpu_list attribute the driver
will create an internal cpu->sdma map, which will be used later as a
look-up table to choose an optimal engine for a user requests.

Reviewed-by: Dean Luick <[email protected]>
Reviewed-by: Dennis Dalessandro <[email protected]>
Reviewed-by: Sebastian Sanchez <[email protected]>
Reviewed-by: Jianxin Xiong <[email protected]>
Signed-off-by: Tadeusz Struk <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Fix resource release in context allocation

Correct resource free in allocate_ctxt() function.
When context creation fails allocated resources are properly
released and pointer in receive context data table is set back
to NULL.

Reviewed-by: Dean Luick <[email protected]>
Reviewed-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Jakub Pawlak <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Remove unused variable from devdata

We no longer use an error tasklet. Remove it from the hfi1_devdata
structure.

Reviewed-by: Ira Weiny <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Cleanup tasklet refs in comments

The code no longer uses tasklets for the send engine. However it does
use a tasklet for sdma but the send routines use a workqueue now days.
Update the comments to reflect that. Make things more generic with
saying "send engine" because that is what is being referred to.

Reviewed-by: Ira Weiny <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Adjust hardware buffering parameter

It was determined that 0x880 is a better value for hardware buffering,
use it.

Reviewed-by: Dean Luick <[email protected]>
Reviewed-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Harish Chegondi <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Act on external device timeout

Add missing external device timeout notification.  Recognize
it as a failed LNI signal from the 8051 firmware.

Reviewed-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Dean Luick <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Fix defered ack race with qp destroy

There is a a bug in defered ack stuff that causes a race with the
destroy of a QP.

A packet causes a defered ack to be pended by putting the QP
into an rcd queue.

A return from the driver interrupt processing will process that rcd
queue of QPs and attempt to do a direct send of the ack.   At this
point no locks are held and the above QP could now be put in the reset
state in the qp destroy logic.   A refcount protects the QP while it
is in the rcd queue so it isn't going anywhere yet.

If the direct send fails to allocate a pio buffer,
hfi1_schedule_send() is called to trigger sending an ack from the
send engine. There is no state test in that code path.

The refcount is then dropped from the driver.c caller
potentially allowing the qp destroy to continue from its
refcount wait in parallel with the workqueue scheduling of the qp.

Cc: [email protected]
Reviewed-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Mike Marciniszyn <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Combine shift copy and byte copy for SGE reads

Prevent over-reading the SGE length by using byte
reads for non quad-word reads.

Reviewed-by: Dean Luick <[email protected]>
Signed-off-by: Sebastian Sanchez <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Do not read more than a SGE length

In certain cases, if the tail of an SGE is not
8-byte aligned, bytes beyond the end to an 8-byte
alignment can be read. Change the copy routine
to avoid the over-read. Instead, stop on the final
whole quad-word, then read the remaining bytes.

Reviewed-by: Dean Luick <[email protected]>
Signed-off-by: Sebastian Sanchez <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Extend i2c timeout

Allow a longer timeout for i2c due to clock stretching and
inaccurate jiffy timing when under a spin lock.  This timeout
is consistent with other i2c-algo-bit users.

Reviewed-by: Mike Marciniszyn <[email protected]>
Reviewed-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Dean Luick <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Increase default settings of max_cqes and max_qps

The ib_write_bw test allows using up to 16384 QPs. When a relatively
large number of QPs (within that range) is used, the test can fail
because the number of CQ entries needed exceeds the limit set by the
driver.

This patch increases the default setting of max_cqes from 0x2FFFF
(196607) to 0x2FFFFF(3145727), which is sufficient to cover the
maximum number needed by the ib_write_bw test (2097152). The default
setting of max_qps is also increased from 16384 to 32768 to allow
the test to run successfully with 16383 or 16384 QPs.

Reviewed-by: Mike Marciniszyn <[email protected]>
Signed-off-by: Jianxin Xiong <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Remove filtering of Set(PkeyTable) in HFI SMA

The FM should have full control to set the pkeys in the
driver pkey table. Remove filtering done by the driver.

Reviewed-by: Ira Weiny <[email protected]>
Signed-off-by: Sebastian Sanchez <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/qib: Remove qpt_mask global

There is no need to have a global qpt_mask as that does not support the
multiple chip model which qib has. Instead rely on the value which
exists already in the device data (dd).

Fixes: 898fa52b4ac3 "IB/qib: Remove qpn, qp tables and related variables from qib"
Reviewed-by: Mike Marciniszyn <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Consolidate pio control masks into single definition

This allows for adding additional pages of adaptive pio
opcode control including manufacturer specific ones.

Reviewed-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Mike Marciniszyn <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/rdmavt, IB/hfi1: Add lockdep asserts for lock debug

This patch adds lockdep asserts in key code paths for
insuring lock correctness.

Reviewed-by: Ira Weiny <[email protected]>
Reviewed-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Mike Marciniszyn <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/rdmavt: Add qp init function

Add an rvt_qp_init() to initialize specific
common fields as the qp is created or reset.

The routine is shared by the rvt_reset_qp() and
the rvt_create_qp().

The intent is that lock dep assertions will only
appear in the rvt_reset_qp().

Reviewed-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Mike Marciniszyn <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/rdmavt: Move reset calldown to reset path

The reset calldown is misplaced.

It should only be called in the code that actually
transitions the QP to reset.

Reviewed-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Mike Marciniszyn <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Move iowait_init() to priv allocate

The call is misplaced in the reset calldown function
and causes issues with lockdep assertions that are to
be added.

Fixes: Commit a2c2d608957c ("staging/rdma/hfi1: Remove create_qp functionality")
Reviewed-by: Ira Weiny <[email protected]>
Reviewed-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Mike Marciniszyn <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/rdmavt: Correct sparse annotation

The __must_hold() is sufficent to correct the sparse
context imbalance inside a function.

Per Documentation/sparse.txt:
__must_hold - The specified lock is held on function entry and exit.

Fixes: Commit c0a67f6ba356 ("IB/rdmavt: Annotate rvt_reset_qp()")
Reviewed-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Mike Marciniszyn <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Fix locking scheme for affinity settings

Existing locking scheme in affinity.c file using the
&node_affinity.lock spinlock is not very elegant.
We acquire the lock to get hfi1_affinity_node entry,
unlock, and then use the entry without the lock held.
With more functions being added, which access and
modify the entries, this can lead to race conditions.
This patch makes this locking scheme more consistent.
It changes the spinlock to mutex. Since all the code
is executed in a user process context there is no need
for a spinlock. This also allows to keep the lock
not only while we look up for the node affinity entry,
but over the whole section where the entry is being used.

Reviewed-by: Ira Weiny <[email protected]>
Reviewed-by: Sebastian Sanchez <[email protected]>
Reviewed-by: Jianxin Xiong <[email protected]>
Signed-off-by: Tadeusz Struk <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Fix user-space buffers mapping with IOMMU enabled

The dma_XXX API functions return bus addresses which are
physical addresses when IOMMU is disabled. Buffer
mapping to user-space is done via remap_pfn_range() with PFN
based on bus address instead of physical. This results in
wrong pages being mapped to user-space when IOMMU is enabled.

Reviewed-by: Mitko Haralanov <[email protected]>
Reviewed-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Tymoteusz Kielan <[email protected]>
Signed-off-by: Andrzej Kacprowski <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Fix the count of user packets submitted to an SDMA engine

Each user SDMA request coming into the driver may contain multiple packets.
Each user packet may use multiple SDMA descriptors to fill the send buffer.
The field seqsubmitted in struct user_sdma_request counts the number of
user packets submitted to an SDMA engine. Sometimes, the intermediate count
may not be updated properly. However, once all the packets' descriptors
are successfully submitted to the SDMA engine, the final count is updated
correctly. But, if only some of the packets are submitted to the engine due
to an error, the intermediate count doesn't reflect the partial number of
packets submitted to the SDMA engine. This can cause a hang later in the
code as the count of packets submitted to the SDMA engine doesn't match the
the count of packets processed by the SDMA engine.

Reviewed-by: Dean Luick <[email protected]>
Signed-off-by: Harish Chegondi <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/hfi1: Move serdes tune inside link start function

All calls to tune_serdes and start_link are paired.  Move
tune_serdes inside start_link.

Reviewed-by: Easwar Hariharan <[email protected]>
Signed-off-by: Dean Luick <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>

IB/qib,IB/hfi: Use core common header file

Use common header file structs, defines, and accessors
in the drivers.  The old declarations are removed.

The repositioning of the includes allows for the removal
of hfi1_message_header and replaces its use with ib_header.

Also corrected are two issues with set_armed_to_active():
- The "packet" parameter is now a pointer as it should have been
- The etype is validated to insure that the header is correct

Reviewed-by: Dennis Dalessandro <[email protected]>
Reviewed-by: Don Hiatt <[email protected]>
Signed-off-by: Mike Marciniszyn <[email protected]>
Signed-off-by: Dennis Dalessandro <[email protected]>
Signed-off-by: Doug Ledford <[email protected]>