Before ME watchdog was exported through standard watchdog interface
it was closed and started together with the mei device.

The major issue is that closing ME watchdog disabled also MEI device,
to fix this the watchdog state machine has to be independent from MEI
state machine.
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

1. rename defines to more be descriptive
2. remove duplicated defines from interface.h
3. add common prefix MEI_
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

After watchdog was disabled the driver would stall
due to wrong calculation of credits reduction

The cat&paste bug was introduced in the commit
7bdf72d3
mei: introduce mei_data2slots wrapper
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

After watchdog was disabled the driver would stall
due to wrong calculation of credits reduction

The cat&paste bug was introduced in the commit
7bdf72d3
mei: introduce mei_data2slots wrapper
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

1. add MEI_DEV_ prefix for mei device state enums
2. rename mei_state to dev_state
3. add constant to string translation for debug purposes
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

1. straight up lines that doesn't cross 80 characters
2. don't break strings
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

change statements of types

	if (ok)
		do something
	else
		return err

into

	if (err)
		return err

	do something
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Introduce mei_data2slots wrapper for sake of
readability. This wrapper close up the open code
for computing slots from a message length:
rond up dwords count from payload and header byte size
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

1. Use unified _hbuf_ prefix for host/write buffer functions.
2. Cleanup the code w/o functional changes.
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

1. We record host (write) buffer depth during reset
 so we don't need to query HCSR register later on.
 The host buffer depth doesn't change after the reset
2. Use mei_hbuf_max_data function to compute payload size in bytes
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

check for overflow when retrieving empty write slots
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

we cannot use if(!mei_flow_ctrl_creds()) logic as
mei_flow_ctrl_creds also negative error codes
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

This bug caused severe connectivity issue in the LMS application
(LMS is described in  Documentation/misc-devices/mei/mei.txt)

The bug was introduced in patch:
commit 1ccb7b62
staging/mei: propagate error codes up in the write flow

The patch has reverted the return value logic of some fo function but
the conditional in _mei_irq_thread_read function was not swapped
making read always entering the error path
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The header exports API for application layer

1. move under include/linux and add to  the export list
2. update include path n the sources
3. update TODO
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

It's been cleaned up, and there's nothing else left to do, so move it
out of staging into drivers/misc/ where all can use it now.

Cc: Tomas Winkler <tomas.winkler@intel.com>
Cc: Oren Weil <oren.jer.weil@intel.com>
Cc: Alan Cox <alan@linux.intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Change mei_write_message's return type from bool to int
to enable propagation of the error code up to caller functions.
The function now returns -EIO when low level write fails and 0
on success.
A similar change is done in intermediate caller functions:
mei_send_flow_control, mei_connect, and mei_disconnect

This makes code more alike to typical Linux kernel error
handling.
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The below patch fixes some typos in drivers/staging/mei/* that I have found while
doing a little bit of reading.
Signed-off-by: NJustin P. Mattock <justinmattock@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

In mei_irq_thread_read_bus_message we reused mei_hdr
allocated on read buffer to write the stop message.
There is no bug associated with this but for code
clarity we use write buffer also for message header.
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Replace bitfield struct hbm_cmd with simple u8 as we
always access the value as whole. This allows us to remove
few ugly type casts
For possible further uses and documentation purposes we add
corresponding bitmask defines
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

~ENODEV is a different number than -ENODEV
Signed-off-by: NDevin J. Pohly <djpohly@gmail.com>
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

1. convert all read buffers to unsigned char and drop useless castings
2. simplify mei_read_slots implementation
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

1. Update Copyright to 2012
2. Also fix mei.h copyright format
   checkpaatch complained:
   WARNING: please, no spaces at the start of a line
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Correct spelling "reseting" to "resetting" in
drivers/staging/mei/interrupt.c
Signed-off-by: NMasanari Iida <standby24x7@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Correct spelling "reseting" to "resetting" in
drivers/staging/mei/interrupt.c
Signed-off-by: NMasanari Iida <standby24x7@gmail.com>
Signed-off-by: NJiri Kosina <jkosina@suse.cz>

add mei_watchdog_register/unregister wrappers for
cleaner  encapsulation
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

1. Remove redundant blanks
2. Straighten code lines that don't go over 80 characters
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

The irq handler processes queued mei clients connect requests, which were
not transferred to the device in ioctl context due to unavailability
of the write buffer.

The handler may transfer the connection request only if there is
no other ongoing  requests for the same mei id.
This condition was implemented inversely which depending on
the write buffer availability lead to seemingly random
failures during connection attempts.

Cc: Oren Weil <oren.jer.weil@intel.com>
Signed-off-by: NNatalia Ovsyanikov <natalia.ovsyanikov@intel.com>
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

1. No need to check if a list is empty before list_for_each_ looping as
this is already checked by loop stopping conditional.

The side effect is reduced indentation depth
from:
	if (!list_empty)
		list_for_each()
to:
	list_for_each()

2. drop cb_ prefix from cl_pos, cl_next variables used in list_for_each
loops. The code is more compact and readable
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

status was never writen
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

We can remove BUG_ON in mei_irq_thread_read_client_message()
as the testing for response buffer size overflow has already
graceful handling in place.
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

moving the final state, clearing of the client maps and
updating of mei state out from mei_host_client_properties function.
Acked-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NOren Weil <oren.jer.weil@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

the wd_work and mei_wd_timer() function was not just for watchdog.
mei managing the AMT watchdog going to be replace by the kernel watchdog
interface.

the mei_timer() will still manage the mei driver timeouts.
Signed-off-by: NOren Weil <oren.jer.weil@intel.com>
Acked-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

adding watchdog ping to send ping/heartbeat to the the AMT watchdog client.
in addition removing the heartbeat sending from the driver timers function.
Signed-off-by: NOren Weil <oren.jer.weil@intel.com>
Acked-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Adding kernel watchdog interface (/dev/watchdog) to the MEI Driver to support AMT Watchdog feature.
This patch and the following one will replace MEI Driver self management of the AMT watchdog
with the standard kernel watchdog interface.
Signed-off-by: NOren Weil <oren.jer.weil@intel.com>
Acked-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

enable msi when supported
also in that case we can drop the quick handler
from the threaded interrupt that protected us from
handling USB interrupts
Signed-off-by: NMaarten Lankhorst <m.b.lankhorst@gmail.com>
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Fix warning: reported in http://marc.info/?l=linux-kernel&m=130812960402606&w=2

drivers/staging/mei/interrupt.c: warning: 'buffer' may be used uninitialized in this function:  => 198

It is a real issue and wrong path in execution is taken when
list is empty or (cl && _mei_irq_thread_state_ok(cl, mei_hdr))
evaluates to false
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

1. remove list used for loop. There were only 2 loops used in non
   time critical places so we can safely unroll them
2. normalize functions names operating on io_list to mei_io_list_<op>
3. rename mei_fe_same_id to mei_cl_cmp_id  used for comparing list
   elements containing struct mei_cl
4. group together  io_list functions in the header file
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NOren Weil <oren.jer.weil@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NOren Weil <oren.jer.weil@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

remove write only/unsed variables
mei_dev.write_hang and mei_io_lis.tdevice_extension
Signed-off-by: NTomas Winkler <tomas.winkler@intel.com>
Signed-off-by: NOren Weil <oren.jer.weil@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>