greybus: loopback: remove module specific identifier from debugfs name

The datapoint we are using to return metrics across modules and cports
shouldn't have a module identifier in it i.e.

/sys/kernel/debug/gb_loopback/raw_latency_endo0
not
/sys/kernel/debug/gb_loopback/raw_latency_endo0:X

This patch removes the module_id used up to this point. Including module_id
actually ends up making life harder in user-space so dropping it.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Patrick Titiano <ptitiano@baylibre.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: loopback: drop redundant endo0 string from debugfs entry name

dev_name() will return the endo0 component of the string on it's own,
there's no need to include it in the snprintf() when construting the
debugfs name. This fixes 'endo0' appearing more than once in the debugfs
name - shamefully slipped through testing cb570c93783f
('greybus/loopback: use dev_name to populate sysfsname').

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: loopback: masked out threads should sleep

If a thread is masked out it should not consume CPU cycles during a test.
We set an arbitrary 100 millisecond sleep time for each masked out thread.
Reasonably blunt instrument to ensure threads with nothing to do don't end
up thrashing the acquisition/release of mutexes.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Patrick Titiano <ptitiano@baylibre.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: loopback: move sysfs entries to /sys/bus/greybus/devices/endo0

Currently we have sysfs entries that are created when the first incoming
connection is created as sub-nodes of the module associated with that
connection e.g. /sys/bus/grebus/devices/endo0:X where X is the module
identifier associated with the new connection. This is conceptually
incorrect since the sysfs entries we create actually aren't bound to a
module. Depending on the order connections are brought up we can also have
a situation where /sys/bus/greybus/devices/endo0:X has high-level control
sysfs data-points but /sys/bus/greybus/devices/endo0:Y does not. Rather
than needlessly replicate data-points across each endo0:X, endo0:Y, endo0:Z
sysfs directories it is more sensible to locate the entries in
/sys/bus/greybus/devices/endo0.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Patrick Titiano <ptitiano@baylibre.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: es1, es2: hook tracepoints to hardware send/recv operations

This patch hooks tracepoints for the handoff point to/from hardware. With
these tracepoints in place we can view the time between gb_message_send and
usb_submit_urb and similarly we can view the time between cport_in_callback
and gb_message_recv_response/gb_message_recv_request

- trace_gb_host_device_send
- trace_gb_host_device_recv

It provides standard tracepoints at

/sys/kernel/debug/tracing/events/greybus/gb_host_device_send
/sys/kernel/debug/tracing/events/greybus/gb_host_device_recv

Giving outputs like

gb_host_device_recv: greybus:2-1 if_id=0000 l=10
gb_host_device_send: greybus:2-1 if_id=0000 l=10

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: tracepoints: add tracepoints for host_device tx/rx

This patch adds new tracepoint declarations to greybus_trace.h to allow for
capture of greybus host device tx and rx events. These two tracepoints
allow an observer to see the point where the hardware interface driver
performs the relevant read or write to receive or write the data it's been
given from the higher layer greybus driver.

The following two new tracepoints are declared:
- trace_gb_host_device_send
- trace_gb_host_device_recv

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: loopback: drop dependency on internal timestamps

This patch drops tracking of internal latencies, it's possible to derive
these times via kernel tracepoints and some user-space scripting. Since
there's no other use of the internal timestamp than the loopback driver we
remove the connection.c, connection.h, es1.c, es2.c and loopback.c
inter-dependency in one go.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: es2: comment out unused functions

The Qualcomm kernel builds with -Werror so the existing es2.c driver
breaks the build due to unused static functions.  As we are still
hashing out exactly how to implement this logic at the moment, just
comment out the functions to make the build be clean, no logic changes
happen here at all.

Reported-by: Michael Scott <michael.scott@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: operation, core: hook tracepoints into message opertions

This patch hooks tracepoints for greybus messages

- trace_gb_message_send
- trace_gb_message_recv_request
- trace_gb_message_recv_response
- trace_gb_message_cancel_outgoing
- trace_gb_message_cancel_incoming

It provides standard tracepoints at

/sys/kernel/debug/tracing/events/greybus/gb_message_send
/sys/kernel/debug/tracing/events/greybus/gb_message_recv_response
/sys/kernel/debug/tracing/events/greybus/gb_message_recv_request
/sys/kernel/debug/tracing/events/greybus/gb_message_cancel_outgoing
/sys/kernel/debug/tracing/events/greybus/gb_message_cancel_incoming

Giving outputs like

gb_message_recv_request: greybus:1-1.1:0 op=0001 if_id=0000 hd_id=0000 l=2
gb_message_send: greybus:1-1.1:0 op=0001 if_id=0000 hd_id=0000 l=2

Similarly perf events can be viewed with standard perf tools e.g.

root@beaglebone:~# perf list 'greybus:*'
  greybus:gb_message_send                            [Tracepoint event]
  greybus:gb_message_recv_request                    [Tracepoint event]
  greybus:gb_message_recv_response                   [Tracepoint event]
  greybus:gb_message_cancel_outgoing                 [Tracepoint event]
  greybus:gb_message_cancel_incoming                 [Tracepoint event]

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: tracepoints: add tracepoint definitions

This patch adds greybus_trace.h with the following trace definitions

- trace_gb_message_send
- trace_gb_message_recv_request
- trace_gb_message_recv_response
- trace_gb_message_cancel_incoming
- trace_gb_message_cancel_outgoing

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: Documentation: Update sysfs-bus-greybus

The file names here weren't in sync with what we have today and the
updates give a better picture of the same.

Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: hd: remove connection-create/destroy callbacks

These host-driver callbacks were intended to allow host drivers to
prepare a cport, something which can now be handled by the cport
enable/disable callbacks instead.

The current create/destroy are somewhat confusingly named as they were
not supposed to create or destroy connections. They were however called
from the unrelated helper functions that do create and destroy SVC
connections.

Furthermore, no errors were returned should the create callback fail,
which should have caused the connection initialisation to fail.

Remove these unused callbacks for now, and let us use the cport
enable/disable callbacks that should be able handle all host cport
initialisation (possibly after also adding an interface to provide
information for endpoint-cport mapping).

Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: hd: add optional cport enable and disable callbacks

Add optional cport enable and disable callbacks to the greybus host
drivers, that can be used to initialise and allocate/release resources
associated with a cport during connection setup/teardown (e.g. software
queues and hardware state).

Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: connection: add protocol-version helper

Add protocol-version to improve readability.

Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: connection: add control-connected helper

Add control-connected helper to improve readability.

Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: connection: clean up control-disconnected helper

Rename helper to the more descriptive
gb_connection_control_disconnected().

Use u16 for cport number, remove redundant cport number from warning
message, and shorten a long line.

Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: connection: clean up init error paths

Clearly mark error-path labels as such and clean up control flow.

Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: connection: clean up svc-connection creation

Move SVC-connection creation to its own helper.

Note that the connection_create host-driver callback is really
unrelated to the SVC connection, and will be removed by a later patch.
It is is included for now as the connection_destroy callback is
currently made from the SVC-connection-destroy helper.

Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: es{1,2}: remove obselete define

The SVC Control request is obsolete and not used anymore. Remove the related
define.

Signed-off-by: Fabien Parent <fparent@baylibre.com>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: svc: fix endianness for svc message

Some fields in svc request were not being set with the correct
endianness, which will trigger the following sparse issues as example:

greybus/svc.c:116:22: warning: incorrect type in assignment (different base types)
greybus/svc.c:116:22:    expected unsigned short [unsigned] [assigned] [usertype] attr
greybus/svc.c:116:22:    got restricted __le16 [usertype] <noident>

Signed-off-by: Rui Miguel Silva <rui.silva@linaro.org>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: greybus_protocols: add missing version-request definition

Add the missing version-request definition that was falsely claimed to
be empty.

Update the generic version-request helper and SVC version-request
handler to use the request definition rather than rely on the response
happening to have the same layout, something which also improves
readability.

Remove a misplaced "Control Protocol" header while at it.

Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: svc: fix version response

The SVC-protocol driver currently accepts the version offered by the
SVC, but still responded with a hard-coded version.

Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: svc: fix ida memory leak

The device-id map was never deallocated on SVC-connection tear down.

Also make the map per-SVC-connection (there should still be only one)
rather than use a global pointer.

Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: es2: add some documentation about endpoints mapping

Signed-off-by: Alexandre Bailon <abailon@baylibre.com>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: es2: change (un)map methods to static

Endpoints pair will only be managed by es2 driver.
map_cport_to_ep() and unmap_cport() should be static.

Signed-off-by: Alexandre Bailon <abailon@baylibre.com>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: es2: rename misnamed variables and methods

Some methods and variables name were a lot confusing.
Replace it or add ep_pair in methods or varaibles name
to make sources less confusing.

Signed-off-by: Alexandre Bailon <abailon@baylibre.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: sdio: fix work-queue leak and use-after-free

A single global work-queue pointer was used for the per-connection
workqueue, something which would lead to memory leaks and all sorts of
bad things if there are ever more than one SDIO connection in a system.

Also add the missing error handling when allocating the queue.

Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: sdio: fix tear-down use-after-free

The mmc-driver private data must not be accessed after mmc_free_host()
has released it.

Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: vibrator: fix tear-down race

Do not release the minor number until after the device has been
deregistered.

Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: firmware: Send AP-ready operation

Module's Bootrom needs a way to know (currently), when to start sending
requests to the AP. The version request is sent before connection_init()
routine is called, and if the module sends the request right after
receiving version request, the connection->private field will be NULL.

Fix this TEMPORARILY by sending an AP_READY request.

Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: firmware: firmware image name has .tftf instead of .fw

That's how the bootrom-tool names it, and that's how the kernel should
expect it.

Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: firmware: Fix incorrect firmware file's name

All the id-fields are 32 bit long instead of 16 bits and so we will need
8 characters per field instead of four. Also the stage field is only one
byte long and so needs just two characters to represent it.

Reported-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: firmware: firmware's file name is 32 characters long

28 is the wrong value and should be 32 instead.

Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: svc: Implement DME peer get/set attributes helpers

These are required to get/set DME attributes of the modules. This is
implemented based on the greybus specifications.

Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: loopback: use dev_name to populate sysfsname

dev_name() will give a nice string representing the end0:X:Y:Z:W name
mitigating the need to pick apart the various nested data structures and
print out their various identifiers.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Suggested-by: Viresh Kumar <viresh.kumar@linaro.org>
Reviewed-by: Patrick Titiano <ptitiano@baylibre.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: loopback: add gb_loopback_operation_sync

In order to extract timestamps from gb_message instead of gb_connection we
will need access to the gb_operation structure. A first step to that is to
create our own gb_loopback_operation_sync which will call
gb_operation_request_send_sync() directly. Once loopback is using this
function internally it will be possible to convert to gb_message based
timestamps and drop gb_connection based timestamps in two seperate patches.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Patrick Titiano <ptitiano@baylibre.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: loopback: sort list of connections for masking purposes

In user-space we specify a list of connections as a bit-mask with the
assumption that a list such is indexed as indicated below.

end0:3:3:1:1 = 1
end0:3:3:2:3 = 2
end0:3:3:3:4 = 4

Current code assigns bitmask ids based on the order of discovery, however
user-space has no idea what the order of discovery is. This patch sorts the
linked list of connections associated with the loopback driver and assigns
a bit-id based on the sorted list - not the order of discovery. This
change therefore enforces the end-users idea that end0:3:3:1:1 above is
always denoted by bit 1 - even if from the AP's perspective it was the last
entry discovered.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Patrick Titiano <ptitiano@baylibre.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: loopback: add response len to loopback protocol

This patch adds a len field to the loopback protocol.
This field is validated in gb_loopback_transfer() and stuffed in
gb_loopback_request_recv().

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Patrick Titiano <ptitiano@baylibre.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: loopback: convert pr_info to dev_err

This patch fixes and invalid use of pr_info() in favour of dev_err();

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: loopback: ensure sysfs entries are cleaned up on exit

bdd4bba4 ('greybus/loopback: add module level sys/debug fs data points')
added a sysfs entry attached to gb_dev but missed the jump to out_sysfs_dev
This patchs fixes the missing jump to out_sysfs_dev.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reported-by: Viresh Kumar <viresh.kumar@linaro.org>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: loopback: ensure debugfs entires are cleaned up on exit

bdd4bba4 ('greybus/loopback: add module level sys/debug fs data points')
added a debugfs entry attached to gb_dev but omitted the cleanup on gb_init
error and gb_exit. This patchs fixes the missing debugfs_remove().

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reported-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: loopback: exit kfree after mutex release

init doesn't have a lock for kzalloc so exit shouldn't have lock with the
corresponding kfree.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: loopback: hold a coarse lock while init/exit run

This patch holds gb_dev.mutex for the duration of init and exit to reduce
complexity while ensuring that init and exit run atomically with respect
to slave threads @ gb_loopback_fn().

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: loopback: ensure count decrement happens before sysfs_remove_groups

This patches fixes a case where gb_dev.count is decremented too late in the
exit() routine.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reported-by: Viresh Kumar <viresh.kumar@linaro.org>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: connection: Add sysfs 'ap_cport_id' file for connections

Its a very useful piece of information, i.e. the cport id of the AP to
which the cport of the module is connected, and is required lots of
times. It isn't known in advance as it is allocated at runtime.

This patch creates another file 'ap_cport_id', only for the connection
directories, which will give the cport id of the AP.

Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: svc: destroy the route on module hot-unplug

We created two-way routes between the AP and module's interface on
hotplug, and forgot to remove them on hot-unplug. The same is also
required while handling errors in hotplug case.

Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: svc: No need of per-direction route-create requests

The route-create request creates bi-directional routes and there is no
need to make separate calls for setting up routes on both the
directions.

Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: Add flags to struct gb_protocol

This helps in removing special per-protocol code, with the help of
generic flags passed by protocol drivers.

Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: connection: Call connection_destroy() from gb_connection_svc_connection_destroy()

connection_create() is called right after svc is requested to create the
connection and so connection_destroy() must be called just before we
request the SVC to destroy the connection.

Over that, this fixes the inconsistency where connection_create() is
called for all connections except SVC connection, but
connection_destroy() is called always.

Acked-by: Alexandre Bailon <abailon@baylibre.com>
Fixes: 5313ca607afb ("Greybus driver: add a new callbacks to driver")
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: connection: destroy svc connection on connection exit or errors

While initializing a connection, the AP requests the SVC to create a
connection between a cport on AP and a cport on the Module.

The opposite of that is missing, when connection is destroyed or if
errors occur after creating the connection. Fix it.

Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: connection: call gb_svc_connection_create() from gb_connection_init()

There are two operations which very much work together:
- AP asks the SVC to create a connection between a cport of AP and a
  cport of module.
- AP tells the module that the connection is created.

Its better (logically) to do these two operations together and so call
gb_svc_connection_create() from gb_connection_init() instead. Also check
its return value properly.

Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: manifest: release cport descriptors to avoid 'excess descriptors' warning

If we fail to initialize a cport of a bundle, we abort the entire
bundle. But that leads to following (unnecessary) warnings as few of the
cport descriptors, belonging to the aborted bundle were never parsed:

"greybus: excess descriptors in interface manifest"

Fix that by releasing all cport descriptors for the aborted bundle.

Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: manifest: don't reject the interface on failing to initialize a cport

A 'bundle' represents a device in greybus. It may require multiple
cports for its functioning. If we fail to setup any cport of a bundle,
we better reject the complete bundle as the device may not be able to
function properly then.

But, failing to setup a cport of bundle X doesn't mean that the device
corresponding to bundle Y will not work properly. Bundles should be
treated as separate independent devices.

While parsing manifest for an interface, treat bundles as separate
entities and don't reject entire interface and its bundles on failing to
initialize a cport. But make sure the bundle which needs the cport, gets
destroyed properly.

We now release the bundle descriptor before parsing the cports, but that
shouldn't make any difference.

Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: loopback: make sure to list_del on connection_exit

gb_loopback_connection_exit does a kfree on a data structure associated
with a loopback connection but fails to do a corresponding list_del(). On
subsequent enumerations this can lead to a NULL pointer dereference. Each
list_add in gb_loopback_connection_init() must have a corresponding
list_del in gb_loopback_connection_exit(), this patch adds the relevant
list_del() and ensures that an appropriate mutex protecting gb_dev.list is
held while doing so.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reported-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: es2: update the bulk_ep_set value accepted by map_to_cpor_ep()

The endpoint set 0 is currently considered as invalid.
But 0 mean muxed cports on ep1 and ep2,
then it must not return -EINVAL.

Signed-off-by: Alexandre Bailon <abailon@baylibre.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: Greybus driver: add a new callbacks to driver

Add connection_create and connection_destroy callbacks.

ES2 can map a cport to a pair of endpoints.
Because ES2 have only a few pair of endpoints, ES2 need to have
access to some high level connection information such as protocol id
to effectively map the cports.
These callback will provide these information and help ES2 to map cports.

Signed-off-by: Alexandre Bailon <abailon@baylibre.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: spi: Allow spi-core to allocate bus numbers dynamically

And that fixes these warnings generated with BDB:

WARNING: at /home/viresh/ara/jetson-kernel-build/tegra/fs/sysfs/dir.c:530 sysfs_add_one+0xa4/0xb4()
sysfs: cannot create duplicate filename '/class/spi_master/spi0'
Modules linked in: gb_es2(O) gb_phy(O) greybus(O)
CPU: 2 PID: 111 Comm: kworker/u8:3 Tainted: G        W  O 3.10.40-gf32f9c5ca7e8 #2
Workqueue: events_unbound svc_process_hotplug [greybus]
[<c0016844>] (unwind_backtrace+0x0/0x13c) from [<c0012fc4>] (show_stack+0x18/0x1c)
[<c0012fc4>] (show_stack+0x18/0x1c) from [<c0067d4c>] (warn_slowpath_common+0x5c/0x74)
[<c0067d4c>] (warn_slowpath_common+0x5c/0x74) from [<c0067d9c>] (warn_slowpath_fmt+0x38/0x48)
[<c0067d9c>] (warn_slowpath_fmt+0x38/0x48) from [<c01bc218>] (sysfs_add_one+0xa4/0xb4)
[<c01bc218>] (sysfs_add_one+0xa4/0xb4) from [<c01bcc24>] (sysfs_do_create_link_sd+0xc0/0x20c)
[<c01bcc24>] (sysfs_do_create_link_sd+0xc0/0x20c) from [<c042e088>] (device_add+0x2e8/0x5f4)
[<c042e088>] (device_add+0x2e8/0x5f4) from [<c04ae008>] (spi_register_master+0x15c/0x654)
[<c04ae008>] (spi_register_master+0x15c/0x654) from [<bf015df4>] (gb_spi_connection_init+0x164/0x19c [gb_phy])
[<bf015df4>] (gb_spi_connection_init+0x164/0x19c [gb_phy]) from [<bf002604>] (gb_connection_bind_protocol+0x160/0x1b4 [greybus])
[<bf002604>] (gb_connection_bind_protocol+0x160/0x1b4 [greybus]) from [<bf002880>] (gb_connection_create_range+0x228/0x2fc [greybus])
[<bf002880>] (gb_connection_create_range+0x228/0x2fc [greybus]) from [<bf002994>] (gb_connection_create+0x40/0x48 [greybus])
[<bf002994>] (gb_connection_create+0x40/0x48 [greybus]) from [<bf000be0>] (gb_manifest_parse+0x61c/0x628 [greybus])
[<bf000be0>] (gb_manifest_parse+0x61c/0x628 [greybus]) from [<bf0019ac>] (gb_interface_init+0x130/0x170 [greybus])
[<bf0019ac>] (gb_interface_init+0x130/0x170 [greybus]) from [<bf003bf0>] (svc_process_hotplug+0x214/0x258 [greybus])
[<bf003bf0>] (svc_process_hotplug+0x214/0x258 [greybus]) from [<c0087ecc>] (process_one_work+0x13c/0x454)
[<c0087ecc>] (process_one_work+0x13c/0x454) from [<c0088c20>] (worker_thread+0x140/0x3dc)
[<c0088c20>] (worker_thread+0x140/0x3dc) from [<c008f20c>] (kthread+0xe0/0xe4)
[<c008f20c>] (kthread+0xe0/0xe4) from [<c000f098>] (ret_from_fork+0x14/0x20)

Reported-by: Mitchell Tasman <tasman@leaflabs.com>
Suggested-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: connection: protocol can be NULL in gb_connection_exit()

gb_connection_exit() is getting called from gb_connection_destroy() now,
which will get called from failure path of gb_connection_create_range()
(in a later commit). And at that point connection->protocol will be
NULL.

Don't print an error message if this happens in gb_connection_exit().

Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: connection: fail to bind if connection init fails

gb_connection_init() can fail and will return proper error code in that
case, but the caller is ignoring it currently.

Fix that by properly handling errors returned from gb_connection_init()
and propagating them to callers of gb_connection_bind_protocol().

Signed-off-by: Fabien Parent <fparent@baylibre.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: connection: call gb_connection_exit() from gb_connection_destroy()

Both the routines are always called together and in the same sequence.
Rather than duplicating this at different places, make
gb_connection_destroy() call gb_connection_exit().

This also makes it more sensible, as gb_connection_init() is never
called directly by the users and so its its counterpart shouldn't be
called directly as well.

Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: connection: Propagate error properly

We just got an error, propagate the exact return value instead of 0.

Reviewed-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: connection: no need to verify connection->protocol

connection->protocol will always be valid in gb_connection_init() as it
is called only from a single routine, after initializing the 'protocol'
field.

No need to check it again.

Reviewed-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: connection: convert connected dev_warn() to dev_err()

Failures from control-connected operations are fatal errors and
must be reported with dev_err() instead of dev_warn().

Fix it.

Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
[ johan: do not promote disconnected warnings, update summary ]
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: connection: staticize gb_connection_init()

Its not used by external users, mark it static. This required some
shuffling of the code.

Reviewed-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: bundle: kill unnecessary forward declaration of routine

Move the function to an earlier place, to kill the unnecessary forward
declaration.

Reviewed-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: svc: unexport few internal functions

There are no external users of these, and probably would never be. Make
them static.

Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: svc: Kill unnecessary wrapper functions

Not sure why they were created, but there is no need for them. Kill
them.

Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: svc: No need to return errors from [gb_]svc_connection_destroy()

These routines are responsible to destroy a connection that is going
away, the return value is of no use. At best, print an error message to
show that we got an error.

Make their return type void.

Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: greybus_protocols: Pack all request/response structure

These structures are exchanged between the AP and the module and must be
packed to avoid any unwanted holes.

Its all working currently because compiler doesn't add any pad bytes for
these structures, as their elements are already aligned to their size.

But these structures can change in future and we better mark them
packed.

Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: greybus_protocols: Pack structure as required by the module firmware

These structures are expected to be packed by the module firmware code,
but the kernel wasn't following it until now.

Its all working currently because compiler doesn't add any pad bytes for
these structures, as their elements are already aligned to their size.

But these structures can change in future and we better mark them
packed.

Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: greybus_protocols: svc: Add missing comment

All request/responses either have a structure representing them or a
comment saying the request/response payload doesn't exist.

The comment was missing for route create response message, add it.

Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: svc: reject invalid state requests

The request sequence for SVC protocol is fixed at least upto SVC_HELLO
request. The first request has to be Protocol Version, followed by
SVC_HELLO. Any other request can follow them, but these two.

Add another field in 'struct gb_svc' that keeps track of current state
of the protocol driver. It tracks only upto SVC_HELLO, as we don't need
to track later ones.

Also add a comment, about the order in which the requests are allowed
and why a race can't happen while accessing 'state'.

This removes the WARN_ON() in gb_svc_hello() as we track state
transition with 'state' field.

This also fixes a crash, when the hotplug request is received before
fully initializing the svc connection. The crash mostly happens while
accessing svc->connection->bundle, which is NULL, but can happen at
other places too, as svc connection isn't fully initialized.

Reported-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
[johan: add 0x-prefix to warning message ]
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: svc: Include system headers at the top

System headers should get included before greybus.h. Its followed
everywhere except svc.c. Fix it.

Reported-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: build: android: avoid building for targets that don't have kernel binary

For sdk related targets, the greybus build will error out due to missing
kernel dependency.  Let's avoid those targets by checking TARGET_NO_KERNEL.

Signed-off-by: Vishal Bhoj <vishal.bhoj@linaro.org>
Signed-off-by: Michael Scott <michael.scott@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: fix cport-id defines

The CPORT_ID_MAX define has been used by host drivers as a device limit,
but also for sanity checks when parsing manifests.

Now that it's only used for sanity checks we can increase it to the
specification maximum (4095) and get rid of the config-option that could
be used to override the previous limit (128).

Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>

greybus: es1: fix build-time cport constraint

The CPort count of es1 is now defined by CPORT_COUNT.

Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>

greybus: core: fix hd-creation error path

Make sure to return an errno when a host-device buffer-size check fails.

Fixes: 1f92f6404614 ("core: return error code when creating host device")
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>

greybus: es{1,2}: remove control endpoint field

There is no need to store the endpoint number of the control requests since
the default control endpoint is used and the USB standard defines for it a fixed
endpoint number of 0.

Remove every instance of the field control_endpoint and replace it with a
hardcoded 0 value.

Signed-off-by: Fabien Parent <fparent@baylibre.com>
Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: es2: get cport count from apb1 usb device

Use the control request REQUEST_CPORT_COUNT in order to get the number of
CPorts supported by the UniPro IP.

Signed-off-by: Fabien Parent <fparent@baylibre.com>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: es2: dynamically allocate array for cport <-> ep mapping

In order to be able to dynamically determine the number of CPorts supported
by the UniPro IP instead of hardcoding the value we need to dynamically
allocate the array that is doing the cport-ep mapping.

Signed-off-by: Fabien Parent <fparent@baylibre.com>
Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: add num_cports field to greybus hd

This commit is doing the preparation work in order to get the number of cports
supported from the UniPro IP instead of using a constant defined in a Kconfig
file.

Greybus host device is now holding the cport count, and all the code will now
use this value instead of the constant CPORT_ID_MAX when referring to an AP's
CPort ID.

Signed-off-by: Fabien Parent <fparent@baylibre.com>
[johan: es1 supports 256 cports, minor style changes ]
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: es2: rename misnamed CPORT_MAX into CPORT_COUNT

Rename the misnamed macro CPORT_MAX into CPORT_COUNT. CPORT_MAX could let
people think that the macro is holding the value of the last CPort ID
usable.

Signed-off-by: Fabien Parent <fparent@baylibre.com>
Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: svc: get hd directly from connection

There is no need to perform connection->bundle->intf->hd as the same can
be done with connection->hd.

Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: pwm: replace pr_err with dev_err

Replace pr_err with the more descriptive dev_err. Also include the error
code on failure to register the PWM chip.

Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>

greybus: interface: fix potential attribute-buffer overflow

Use scnprintf in the generic attribute helper, which does not currently
check for buffer overflow.

The attribute helper is used to print generic strings, which could
potentially overflow the buffer. Note that the only strings currently
exported are taken from greybus string descriptors and should therefore
be limited to 255 chars.

Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Reviewed-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

greybus: svc: fix hot-plug-state allocation flag

Use GFP_KERNEL for hot-plug state allocation in
gb_svc_intf_hotplug_recv, which is called from a request handler (i.e.
a work queue).

Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>

greybus: svc: fix device-id allocation flag

Use GFP_KERNEL for device-id allocation in svc_process_hotplug, which is
called from a work queue.

Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>

greybus: endo: fix endo-id allocation flag

Use GFP_KERNEL for endo ida allocation in gb_endo_register, which is not
called from atomic context.

Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>

greybus: sdio: fix command type defines

Broadcast command with response and without response where swapped
related to what is defined in greybus specification.

Make it coherent with the document.

Signed-off-by: Rui Miguel Silva <rui.silva@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: firmware: fix potential stack corruption

Use snprintf when generating the firmware name to avoid stack corruption
if the fixed-size buffer overflows.

Note that the current buffer size appears to expect 16-bit ids while
the they are actually 32-bit, something which could trigger the
corruption.

Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>

greybus: es1/es2: set transfer flag to send a zero-length packet

Greybus messages with a multiple size of 512B generate timeouts
(any other message size doesn't).
512B is exactly the packet size of a bulk out endpoint.
Hence USB device is expecting a short (< 512B)
or zero-length packet to finish the transfer,
which is never generated and causes the timeout.

Set the transfer flag to send a zero-length packet in this situation.

Signed-off-by: Alexandre Bailon <abailon@baylibre.com>
Reviewed-by: Patrick Titiano <ptitiano@baylibre.com>
Reviewed-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>

greybus: loopback: fix typo in comment

Alex previously post a patch to fix this typo. Somehow it fell through the
cracks in the meantime. Do it again 'stastic' is a word 'statistic' is not.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: loopback: add module level sys/debug fs data points

Code this far has used the first connection's sysfs entry to present
variables intended to control the entire test - across multiple
connections. This patch changes that so that the module level variables
only appear at the end0:x level in sysfs.

Example:

Total counts for errors over the entire set of connections will be here
/sys/bus/greybus/devices/endo0:x/error_dev

In contrast an error for each connection will be presented like this
/sys/bus/greybus/devices/endo0:x:y:z:w/error_con

x = <module-id>
y = <interface-id>
z = <bundle-id>
w = <cport-id>

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: loopback: remove checkpatch error causing macro

checkpatch.pl is choking on a later change to dev_stats_attrs, where
checkpatch expects to see the values encapsulated in curly brackets.
Encapsulating in curly brackets will cause a compiler error. To resolve the
dichotomy this patch drops the macros and adds the arrary declarations
directly.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: loopback: initialized ms_wait negate warning

ms_wait = 0; caught by a compiler warning.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: loopback: add bitmask of connections to include in test

Feature add which enables the ability to select a bit-mask of connections
to run when executing a loopback test set. This is a feature add to
facilitate testing on the firmware side minus the necessity to recompile
firmware to support unicast (v) multicast (v) bitmask.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: loopback: graph round-trip time for all threads

This patch adds the ability to time the delta between all threads like this

t1 = timestmap();
thread1:gb_operation_sync();
thread2:gb_operation_sync();
t2 = timestamp();

In order to enable that behaviour without forcing an undesirable
checkpointing scheme this patch introduces a kfifo for each thread to store
the raw timestamps and calculate a time difference.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: loopback: functionally decompose gb_loopback_calc_latency

The __gb_loopback_calc_latency will be useful in later patches. Provide it
here and use as intended. Later on we just want to use the timestamp
rollover detection, so split it out now.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: loopback: add gb_loopback_nsec_to_usec_latency

A helper function to convert from a nanosecond value to a latency value
expressed in mircoseconds. This will be used again in subsequent patches.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

greybus: loopback: support synchronized tests over multiple cports

The loopback code as currently implemented allows free running threads to
blast data to cports in isolation, however no overall stastics are gathered
for the aggregate throughput to a given module.

This patch moves derivation of stastics for all cports to one structure
so that 1 to n cports will report their overall stastics in one place.
Later patches update the sysfs/debugfs accessor functions to provide the
per-module and per-connection data separately.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>