The firmware is having a hard time with 4k buffers and memory
allocation, so decrease the size on the host side to 2k.  Also move away
from using PAGE_SIZE to denote 4k as that's not the case on all
architectures, and someone, someday, might get a rude surprise.
Reviewed-by: NAlex Elder <elder@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

The unregister_chrdev_region() does twice here.
The chrdev region was unregistered
inside tty_unregister_driver().
Signed-off-by: NPhong Tran <tranmanphong@gmail.com>
Reviewed-by: NAlex Elder <elder@linaro.org>
Reviewed-by: NBryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

There is a sparse warning. The endo id map is also
used in endo.c. Should define in endo.h
Signed-off-by: NPhong Tran <tranmanphong@gmail.com>
Reviewed-by: NAlex Elder <elder@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

Normally, its a good practice to free resources in the reverse order in
which they are allocated, so that all the dependencies can be sorted out
properly.

This is true while creating/destroying devices as well. For example
consider this scenario (I faced a crash with control protocol due to
this). For a new module, we will first create a bundle+connection for
the control cport and then create other bundles/connections after
parsing manifest.

And while destroying interface on module hot unplug, we are removing the
devices in the order they are added. And so the bundle/connection for
the control cport are destroyed first. But, control cport was still
required while destroying other bundles/connections.

To solve this problem, lets destroy the resources in the reverse order
in which they are added.
Signed-off-by: NViresh Kumar <viresh.kumar@linaro.org>
Reviewed-by: NAlex Elder <elder@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

Provide a little more information in two pr_err() calls.
Also enclose a reported condition in parentheses, to match
the style used everywhere else in the file.
Signed-off-by: NAlex Elder <elder@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

Rename greybus_deregister() to be greybus_deregister_driver(), so
its name mirrors the greybus_register_driver() function it matches.
Define greybus_deregister() to be a trivial macro.
Signed-off-by: NAlex Elder <elder@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

In gb_uart_request_recv(), the receive data size is in little-endian
format.  Do the proper byte swapping of that value before using it.
Found by "make check".
Signed-off-by: NAlex Elder <elder@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

Give gb_connection_hd_find() static scope; it's never used
outside "connection.c".
Signed-off-by: NAlex Elder <elder@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

The SVC protocol driver should have been defined as a basic
Greybus protocol driver, not a GP Bridge protocol driver.
Signed-off-by: NAlex Elder <elder@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

Prefix module-id with endo-id to uniquely identify it for the entire
kernel.
Reviewed-by: NAlex Elder <elder@linaro.org>
Signed-off-by: NViresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

Currently we name the endo device as "endo". And it shows up with the
same name in sysfs directory: /sys/bus/greybus/devices/.

But each device in kernel should be represented by a unique id in
kernel, and "endo" isn't unique.

Lets generate unique ids for endo devices. The ida mechanism for
allocating ids may be overkill but it works.
Reviewed-by: NAlex Elder <elder@linaro.org>
Signed-off-by: NViresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

This isn't unique just for the bundle but the complete interface. Its
wrong to call it bundle_cport_id. Lets name it intf_cport_id to make
things clear.
Signed-off-by: NViresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

'buffer' isn't used in this function, remove it.
Reviewed-by: NJohan Hovold <johan@hovoldconsulting.com>
Reviewed-by: NAlex Elder <elder@linaro.org>
Signed-off-by: NViresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

Use defines for the data format command.
Tidy up naming of gb_tty variables.
Signed-off-by: NBryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

gb_uart_request_recv job in life is to process unsolicited greybus
mesages from the UART.
Hook the incoming UART data and pass to the TTY layer.
Line-state changes still TBD.
Signed-off-by: NBryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

Make kmalloc for the send buffer a one time alloc based on the MTU for
a given greybus link.

The write_room for an gb_operation_sync then will be the size of the
buffer we use for a single operation.
Signed-off-by: NBryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

gb_operation_sync returns 0 on success but the calling function
expects the number of bytes written on success or a negative errno
Signed-off-by: NBryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

For each new UART connection we need to do a tty_port_init else
we'll crash when trying to access the tty mutex later on.
Base the TTY major/minor numbers on non-zero values.
Supply an empty operations structure for the newly regitered port.
Signed-off-by: NBryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

Arbitrary number 255 is both not aligned and probably too big.
Move the UART count down to 16 which is still large but, more realistic.
8 may be too few for future testing setups, 16 should accomodate any.
Signed-off-by: NBryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

Update tabs and naming of structures to match the naming used in the greybus
specification more closely.
Signed-off-by: NBryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

gbsim depends on the structures and defines in greybus_protocols.h
generally in order to simulate firmware. Move UART defines into this
header to facilitate.
Signed-off-by: NBryan O'Donoghue <bryan.odonoghue@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

If alloc minor is error, gb_tty should free.
Signed-off-by: NPhong Tran <tranmanphong@gmail.com>
Reviewed-by: NViresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

Because of the missing '\n', this is how the output of reading endo
sysfs files looks:

root# cat /sys/bus/greybus/devices/endo/id
0x4755root#

Fix it by including \n to the end of the printed string.
Signed-off-by: NViresh Kumar <viresh.kumar@linaro.org>
Reviewed-by: NAlex Elder <elder@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

Add Greybus GPIO IRQ-type defines rather than rely on the current
Linux implementation.
Signed-off-by: NJohan Hovold <johan@hovoldconsulting.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

Remove unused debounce define from protocol header.
Signed-off-by: NJohan Hovold <johan@hovoldconsulting.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

With the Endo "id" attribute in place, there's no need to encode
the ID of an Endo in its sysfs path.  So get rid of it.
Signed-off-by: NAlex Elder <elder@linaro.org>
Reviewed-by: NViresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

This commit:
    7e761e2 endo: rework some attributes
added a new "endo_id" attribute, located under a new "Endo"
directory in sysfs.  The resulting path looks like:
    Documentation/sysfs/endo-TYPE/Endo/endo_id
There's no need to have a separate "Endo" subdirectory to contain
Endo-specific attributes.

That commit also added "svc_" to some other paths related to the
SVC, like:
    Documentation/sysfs/endo-TYPE/SVC/svc_version
The additional "svc_" is redundant.

This patch retouches those paths a bit, mainly to remove some
redundancy.  It also makes the pathname components all lower case.
As a result, the above two paths now look like:
    Documentation/sysfs/endo-TYPE/id
    Documentation/sysfs/endo-TYPE/svc/version

All other Endo sysfs files are updated similarly.
Signed-off-by: NAlex Elder <elder@linaro.org>
Reviewed-by: NViresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

Impose a few editorial conventions on the Greybus-related sysfs
files under "Documentation".
    - Capitalize "Endo" (except in path names)
    - Capitalize "ID" (except in path names)
    - Use "..." to indicate unspecified path components (because
      ".." means something else).
    - Add the "0x" prior to the "XXXX" representing the Endo ID.
Signed-off-by: NAlex Elder <elder@linaro.org>
Reviewed-by: NViresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

Update the files documenting Greybus-related sysfs files under
Documentation/ to reflect the addition of the two recently-added
Endo attributes.
Signed-off-by: NAlex Elder <elder@linaro.org>
Reviewed-by: NViresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

This is not a kernel module. It should not use
the module license macro.
Signed-off-by: NPhong Tran <tranmanphong@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

This macro should be DATA not MGMT.
Signed-off-by: NPhong Tran <tranmanphong@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

I neglected to update the "#ifndef/#define" when I renamed
"greybus_protocols.h".  Fix that.
Signed-off-by: NAlex Elder <elder@linaro.org>
Reviewed-by: NViresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

Update the remaining operation types now that the ack operation is gone
to avoid leaving a hole in the type definitions.
Signed-off-by: NJohan Hovold <johan@hovoldconsulting.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

Change to gb_gpbridge_protocol_driver for
making the consitent with other drivers.
Signed-off-by: NPhong Tran <tranmanphong@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

data of hotplug request should exchange to native
CPU format before assignment.
Signed-off-by: NPhong Tran <tranmanphong@gmail.com>
Reviewed-by: NViresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

The current interrupt implementation uses the simple irq-flow handler,
which means that the interrupt subsystem makes no irq-chip callbacks
when handling an interrupt. Specifically, no end-of-interrupt message is
sent when the threaded handler has run. This means that we may currently
re-enable an interrupt before it has been serviced (i.e. the irq-event
operation may complete before the threaded handler has run).

The simple flow handler also silently drops a second interrupt arriving
while a handler is running. This means that we may lose a second edge
interrupt with the current firmware.

Switch to a new one-shot interrupt protocol, where the primary handler
(firmware) always masks and acks an interrupt before sending an event to
the AP. The AP is responsible for unmasking the interrupt when it has
been handled. By having the firmware ack an edge interrupt before
sending the event, a second edge interrupt will no longer get lost.

This one-shot protocol can be implemented in the kernel by using the
level irq-flow handler, one-shot interrupts with threaded handlers and
bus-lock synchronisation for slow buses. Note that the same flow handler
is used for both edge and level interrupts.
Signed-off-by: NJohan Hovold <johan@hovoldconsulting.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

The irq-chip callbacks are made in atomic context where we must not do
any synchronous greybus operations.

Fix the current gpio-interrupt implementation by using the bus-lock
functionality provided by the irq subsystem.
Signed-off-by: NJohan Hovold <johan@hovoldconsulting.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

Rename irq mask and unmask functions to match the callback names.
Signed-off-by: NJohan Hovold <johan@hovoldconsulting.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>

Use irq_find_mapping directly rather than go through the legacy gpio
interface.
Signed-off-by: NJohan Hovold <johan@hovoldconsulting.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@google.com>