Alex Elder [Tue, 12 May 2015 02:16:34 +0000 (21:16 -0500)]
greybus: loopback: fix the type attribute check
In gb_loopback_check_attr(), the value of gb->type is checked for
validity. The only valid values are 0, 1, and 2. But the check
allows the value 3. Fix that.
Signed-off-by: Alex Elder <elder@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
In commit 1efb53a220b7 ("ASoC: simple-card: Remove support for setting
differing DAI formats"), the .fmt field was removed from struct
asoc_simple_dai. Fix this build breakage by not trying to set it.
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com> Acked-by: John Stultz <john.stultz@linaro.org>
Introduce INSTALL_MOD_PATH to allow for building and installing of the
greybus modules from a different location.
This lets you build the greybus modules on a PC and then install the
modules to an SD card in the appropriate location relative to the SD such
as /media/sdcard/lib/modules/version and subsequent running of depmod in
the same location. If INSTALL_MOD_PATH isn't specified the default
behaviour of installing and depmoding to /lib/modules/version is
maintained.
So here's the current greybus audio driver, which I
wanted to send out for more serious review and
consideration for merging.
I've tried to integrate much of the feedback from the
last round and most of the hotplug issues that I've found
have been resolved. I've tested this via gbsim, and the
Android ARA HAL layer seems to work with it.
Mark has also successfully played audio with this driver,
adding a few hacks to get the codec's i2c connection to
probe.
Current issues:
* Hotplug problem - When gbsim is killed, or the module
removed, the greybus driver gets stuck since the android
mediaserver process is holding the audio device open.
Killing the mediaserver allows things to clean up and
allows greybus to accept new gbsim connections. I have
a workaround patch to the soc-core.c logic which converts
the snd_card_free() call to snd_card_free_when_closed()
which allows the greybus connection cleanup to finish.
Remaining todos:
* Probably need to break apart the mgmt_setup function
to integrate better with the constraint logic. I took
a really basic stab at this, but more is probably
needed.
* Figure out how to properly find and tie in the
codec's I2C bus-id to the driver.
This code requires that the kernel support the following
config options, which I've enabled in a separate kernel
patch:
CONFIG_SND_SIMPLE_CARD
CONFIG_SND_SOC_SPDIF
CONFIG_SND_SOC_RT5645
I really can't calim to be the sole author of this, since
many many fixes and tweaks that have been folded in have
come from Mark Greer. His analsysis and debugging is really
what has made this dummy-framework driver evolve into an
actual audio driver. So much credit and thanks to Mark!
Signed-off-by: John Stultz <john.stultz@linaro.org> Signed-off-by: Mark A. Greer <mgreer@animalcreek.com> Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
Alex Elder [Thu, 7 May 2015 18:03:52 +0000 (13:03 -0500)]
greybus: eliminate extra response flag definitions
All protocols use the same value to distinguish between request and
response message types. This is a requirement.
Use GB_MESSAGE_TYPE_RESPONSE rather than GB_OPERATION_TYPE_RESPONSE
for the name of the flag used to distiguish between request and
response messages.
Get rid of the redundant response flag definitions that are
associated with specific protocols.
Describe the symbolic values as "operation types" rather than
"message types" where they are defined. The message type for a
request is the same as the operation type; the message type for a
response is the operation type OR'd with GB_MESSAGE_TYPE_RESPONSE.
Signed-off-by: Alex Elder <elder@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
Alex Elder [Thu, 7 May 2015 18:00:21 +0000 (13:00 -0500)]
greybus: battery: use feature tag rather than kernel version
Conditionally define a new symbol DRIVER_OWNS_PSY_STRUCT, which is
set in "kernel_ver.h" based on on the kernel version. Use it to
distinguish code used for kernels that differ in whether a power
supply structure is owned by the driver, or by the power supply
core.
Signed-off-by: Alex Elder <elder@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
Alex Elder [Thu, 7 May 2015 18:00:20 +0000 (13:00 -0500)]
greybus: battery: free struct on error in caller
When a battery connection is initialized, a gb_battery structure for
it is allocated in gb_battery_connection_init(). Currently that
function ends by calling init_and_register(); in the event an error
occurs, init_and_register() is responsible for freeing the allocated
gb_battery structure.
Make the code a bit better balanced by having the function that
allocates the structure be responsible for freeing it in case of
error.
Signed-off-by: Alex Elder <elder@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
greybus: battery: update for 4.1 power supply api changes
The 4.1-rc1 kernel changed the power supply apis such that the
structures are now owned by the power supply core, and not the
individual drivers. This broke the greybus battery driver, so update it
to support both the old and the new apis.
The API changes were such that I can't "hide" them in kernel_ver.h, but
rather the driver itself needs to have ugly #ifdefs in it. I tried to
keep it to a minimum, making a sub-function for initializing the power
supply device that is implemented differently for different kernel
versions.
When this is submitted upstream, or if we ever move our AP development
to 4.1 or greater, the support for older kernels can be removed.
Alex Elder [Fri, 17 Apr 2015 19:41:47 +0000 (14:41 -0500)]
greybus: bundle: use kstrdup() for state file
The kernfs code guarantees we'll get a NUL-terminated buffer.
Use kstrdup() rather than kzalloc() + memcpy() in state_store()
making it slightly clearer what we're doing. This has the added
benefit of guaranteeing that the stored string has no NUL character
inside it.
Signed-off-by: Alex Elder <elder@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
greybus: manifest: Warn if descriptor size > expected size
A descriptor passed to AP can be bigger than what AP expects, if
manifest's minor version is higher with same major number as the AP. As
it can have some extra data in descriptor.
But, if AP and manifest versions are identical, or if the AP's minor
version is greater than the manifest version, we should at least warn
(if not fail).
Doing this would require some changes to record the manifest version
somewhere reachable by identify_descriptor().
For now, just warn if descriptor is bigger than expected.
There can be three Endo types: mini, medium and large. And that's what
Endo 'type' should refer to.
But we have named the 16 bit number that uniquely represents a valid
endo, as its type. 'id' seems to be a more suitable name to that instead
of 'type'. Lets rename it.
A bundle has a state file, that is managed by the endo userspace
process. This file can be written to and any process that is polling on
the file will be woken up and can read the new value. It's a "cheap"
IPC for programs that are not allowed to do anything other than
read/write to kernel sysfs files.
Johan Hovold [Thu, 16 Apr 2015 07:53:59 +0000 (09:53 +0200)]
greybus: loopback: fix 64-bit divisions
The code uses 64-bit divisions, which should be avoided, and also
prevents the module from loading on 32-bit systems:
gb_loopback: Unknown symbol __aeabi_uldivmod (err 0)
Fix by using the kernel's 64-bit by 32-bit division implementation
do_div.
Compile tested only. I did not look very closely at the code itself.
Perhaps this could be worked around in some other way, but this silences
the linker warning and allows the module to be loaded.
Reviewed-by: Alex Elder <elder@linaro.org> Signed-off-by: Johan Hovold <johan@hovoldconsulting.com> Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
greybus: Documentation/sysfs-bus-greybus: update kernel version and date
The kernel is now on the 4.XX numbering scheme, and it's going to be a
while before we merge this code, so pick a date sometime in the future
to be safe.
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com> Reviewed-by: Alex Elder <elder@linaro.org>
This hooks up the endo, and modules, into the device tree. All modules
for a specific endo are created when the host device is initialized.
When an interface is registered, the correct module for it is found and
that module is used for the sysfs tree. When the interface is removed,
the reference on the module is dropped.
When the host device goes away, the whole endo and modules are removed
at once.
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com> Reviewed-by: Alex Elder <elder@linaro.org>
This adds the attributes power_control and present to a module. It also
removes the unneeded module_id attribute, as that comes from the name of
the module itself.
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com> Reviewed-by: Alex Elder <elder@linaro.org>
Johan Hovold [Tue, 7 Apr 2015 09:27:21 +0000 (11:27 +0200)]
greybus: drop host-driver buffer headroom
Drop the host-driver buffer headroom that was used to transfer the cport
id on ES1 and ES2.
Rather than transferring additional bytes on the wire and having to deal
with buffer-alignment issues (e.g. requiring the headroom to be a
multiple of 8 bytes) simply drop the headroom functionality.
Host drivers are expected set up their transfer descriptors separately
from the data buffers and any intermediate drivers (e.g. for Greybus
over USB) can (ab)use the operation message pad bytes for now.
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com> Reviewed-by: Alex Elder <elder@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
Johan Hovold [Tue, 7 Apr 2015 09:27:19 +0000 (11:27 +0200)]
greybus: es1: fix transfer-buffer alignment
Fix transfer-buffer alignment of outgoing transfers which are currently
byte aligned.
Some USB host drivers cannot handle byte-aligned buffers and will
allocate temporary buffers, which the data is copied to or from on every
transfer. This affects for example musb (e.g. Beaglebone Black) and
ehci-tegra (e.g. Jetson).
Instead of transferring pad bytes on the wire, let's (ab)use the pad
bytes of the operation message header to transfer the cport id. This
gives us properly aligned buffers and more efficient transfers in both
directions.
By using both pad bytes, we can also remove the arbitrary limitation of
256 cports.
Note that the protocol between the host driver and the UniPro bridge is
not necessarily Greybus. As long as the firmware clears the pad bytes
before forwarding the data, and the host driver does the same before
passing received data up the stack, this should be considered "legal"
use.
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com> Reviewed-by: Alex Elder <elder@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
Make sure to allocate the message transfer-buffer separately from the
containing message structure to avoid data corruption on systems without
DMA-coherent caches.
The message structure contains state that is updated while the buffer
may be used for DMA, something which could lead to data corruption due
to cache-line sharing on some architectures.
Use the (renamed) message cache for the message structure itself and
allocate the buffer separately.
If the additional allocation is a concern, the message structures
could eventually be allocated as part of the operation structure.
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com> Reviewed-by: Alex Elder <elder@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
Remove unused and unnecessary buffer-alignment define that host driver
were supposed to use.
We can handle unaligned incoming buffers just fine by accessing the
operation-message header via a copy in the receive path, rather than
requiring host drivers to make sure the alignment is correct.
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com> Reviewed-by: Alex Elder <elder@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
Johan Hovold [Tue, 7 Apr 2015 09:27:13 +0000 (11:27 +0200)]
greybus: operation: fix unaligned memory accesses in receive path
The buffer received from our current host driver is 1-byte aligned and
will therefore cause unaligned memory accesses if simply cast to an
operation-message header.
Fix this by making a properly aligned copy of the header in
gb_connection_recv_response before accessing its fields.
Note that this does not affect protocol drivers as the whole buffer is
copied when creating the corresponding request or response before being
forwarded.
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com> Reviewed-by: Alex Elder <elder@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
greybus: manifest: Use interface descriptor instead of module descriptor to get information
A module can have more than one interfaces and we get hotplug events or
manifests for interfaces, not modules. Details like version, vendor,
product id, etc. can be different for different interfaces within the
same module and so shall be fetched from interface descriptor instead of
module descriptor.
So what we have been doing for module descriptors until now must be done
for interface descriptors. There can only be one interface descriptor in
the manifest. Module descriptor isn't used anymore and probably most of
its fields can be removed now.
greybus: bundle: Create bundles using bundle descriptors
Currently we are creating bundles based on interface descriptors. An interface
can have one or more bundles associated with it and so a bundle must be created
based on a bundle descriptor.
greybus: interface: Fetch interface id instead of module id during setup
There can be more than one interface on a module and we need to know the
interface for which the event has occurred.
But at the same time we may not need the module id at all. During initial phase
when AP is probed, the AP will receive the unique Endo id which shall be enough
to draw relationships between interface and module ids.
Code for that isn't available today and so lets create another routine to get
module id (which needs to be fixed separately), which will simply return
interface id passed to it.
Now that we have interface id, update rest of the code to use it.
Devices registered with the device-core needs to be freed by calling
device_unregister(). For module we are calling just put_device() and for
bundle, connection and interface we are calling device_del().
All of these are incomplete and so none of them get freed, i.e. the
.release() routine is never called for their devices.
Module being a special case that it needs to maintain a refcount or a
list of interfaces to trace its usage count. I have chosen refcount.
And so once the refcount is zero, we can Unregister the device and
module will get free as well.
Because of this bug in freeing devices, their sysfs directories were not
getting removed properly and after a manifest is parsed with the help of
gbsim, removing modules was creating problems. The sysfs directory
'greybus' wasn't getting removed. And inserting the modules again
resulted in warnings and insmod failure.
WARNING: CPU: 3 PID: 4277 at
/build/buildd/linux-3.13.0/fs/sysfs/dir.c:486
sysfs_warn_dup+0x86/0xa0()
Mark Greer [Tue, 31 Mar 2015 23:49:56 +0000 (16:49 -0700)]
greybus: Initial I2S definitions
These are definitions from Mark that I've consolidated into
one header file. I'd like to get these merged at some point
soon, so the audio driver and gbsim work can avoid having
out-of-tree dependencies.
Signed-off-by: Mark A. Greer <mgreer@animalcreek.com> Reviewed-by: Alex Elder <elder@linaro.org> Signed-off-by: John Stultz <john.stultz@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
greybus: Documentation/sysfs: add a proposed sysfs tree for greybus
This adds a proposed sysfs layout for greybus to Documentation to make
it easier for people to discuss / test things. It includes a module, an
interface, a bundle, and a gpbridge binding to that bundle.
This was discussed on the projectara software mailing list.
greybus: kernel_ver.h: add sysfs_create_groups() and sysfs_remove_groups()
These functions showed up in 3.12 or so, and we are stuck on 3.10 for
various reasons, so provide backports in kernel_ver.h so that we can
rely on these functions.
greybus: loopback: use the attribute groups, not group
We should use the attribute groups, not group, for the device, so
add and remove it. No one should ever be updating a sysfs group for a
device, as that can be pretty dangerous if you don't duplicate _all_
existing attribute for that device, and I don't think we were doing that
here.
Alexandre Bailon [Tue, 31 Mar 2015 07:51:59 +0000 (09:51 +0200)]
greybus: Add loopback protocol
Add a simple Greybus protocol in order to stress USB and Greybus.
This protocol currently support 2 requests: ping and transfer.
ping request is useful to measure latency.
Kernel send a ping request and firmware should respond with a ping.
The transfer request request is useful to stress Greybus and USB.
Kernel can send data from 0 to 4k and the firmware must send back the data to kernel.
This behaviour of gb-loopback module is controlled via sysfs.
Curently, connection sysfs folder is updated with new entries:
- type: Type of loopback message to send
* 0 => Don't send message
* 1 => Send ping message continuously (message without payload)
* 2 => Send transer message continuously (message with payload)
- size: Size of transfer message payload: 0-4096 bytes
- ms_wait: Time to wait between two messages: 0-1024 ms
Module also export some statistics about connection:
- latency: Time to send and receive one message
- frequency: Number of packet sent per second on this cport
- throughput: Quantity of data sent and received on this cport
- error
All this statistics are cleared everytime type, size or ms_wait entries are updated.
Alex Elder [Fri, 27 Mar 2015 20:20:49 +0000 (15:20 -0500)]
greybus: es2: test apb1_log_task safely
When usb_log_enable() is called, the global apb1_log_task is used to
hold the result of kthread_run(). It is possible for kthread_run()
to return an error pointer, so tests of apb_log_task against NULL
are insufficient to determine its validity.
Note that kthread_run() never returns NULL so we don't have to check
for that. But apb1_log_task is initially NULL, so that global must
be both non-null and not an error in order to be considered valid.
Signed-off-by: Alex Elder <elder@linaro.org> Signed-off-by: Greg Kroah-Hartman <greg@kroah.com>
Alex Elder [Fri, 27 Mar 2015 20:20:49 +0000 (15:20 -0500)]
greybus: es1: test apb1_log_task safely
When usb_log_enable() is called, the global apb1_log_task is used to
hold the result of kthread_run(). It is possible for kthread_run()
to return an error pointer, so tests of apb_log_task against NULL
are insufficient to determine its validity.
Note that kthread_run() never returns NULL so we don't have to check
for that. But apb1_log_task is initially NULL, so that global must
be both non-null and not an error in order to be considered valid.
Signed-off-by: Alex Elder <elder@linaro.org> Signed-off-by: Greg Kroah-Hartman <greg@kroah.com>
Johan Hovold [Fri, 27 Mar 2015 11:45:49 +0000 (12:45 +0100)]
greybus: operation: refactor response handling
Send response to incoming requests from the operation request handler
rather than in every protocol request_recv callback.
This simplifies request_recv error handling and allows for further code
reuse.
Note that if we ever get protocols that need to hold off sending
responses we could implement this by letting them return a special
value (after acquiring the necessary operation references) to suppress
the response from being sent by greybus core.
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com> Reviewed-by: Alex Elder <elder@linaro.org> Signed-off-by: Greg Kroah-Hartman <greg@kroah.com>
Johan Hovold [Fri, 27 Mar 2015 11:45:47 +0000 (12:45 +0100)]
greybus: hid: fix missing input verification of report events
Add minimal verification of incoming report size, before using it to
determine what buffer and size to pass on to HID core.
Add comment about protocol needing to be revisited. If we are going to
be parsing the report data received, then those fields have to be
defined in the Greybus specification at least.
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com> Reviewed-by: Alex Elder <elder@linaro.org> Signed-off-by: Greg Kroah-Hartman <greg@kroah.com>
Fix the payload size of incoming requests, which should not include the
operation message-header size.
When creating requests we pass the sizes of request and response
payloads and greybus core allocates buffers and adds the required
headers. Specifically, the payload sizes do not include the
message-header size.
This is currently not the case for incoming requests however, something
which prevents protocol drivers from implementing appropriate input
verification and could lead to random data being treated as a valid
message in case of a short request.
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com> Reviewed-by: Alex Elder <elder@linaro.org> Signed-off-by: Greg Kroah-Hartman <greg@kroah.com>
Johan Hovold [Fri, 27 Mar 2015 11:41:17 +0000 (12:41 +0100)]
greybus: operation: fix null-deref on operation destroy
Incoming operations are created without a response message. If a
protocol driver fails to send a response, or if the operation were to be
cancelled before it has been fully processed, we get a null-pointer
dereference when the operation is released.
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com> Reviewed-by: Alex Elder <elder@linaro.org> Signed-off-by: Greg Kroah-Hartman <greg@kroah.com>
Johan Hovold [Fri, 27 Mar 2015 11:41:16 +0000 (12:41 +0100)]
greybus: operation: fix null-deref on operation cancel
Incoming operations are created without a response message. If an
operation were to be cancelled before it has been fully processed (e.g.
on connection destroy), we would get a null-pointer dereference in
gb_operation_cancel.
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com> Reviewed-by: Alex Elder <elder@linaro.org> Signed-off-by: Greg Kroah-Hartman <greg@kroah.com>
Johan Hovold [Fri, 27 Mar 2015 11:41:13 +0000 (12:41 +0100)]
greybus: operation: fix use-after-free when sending responses
Fix use-after-free when sending responses due to reference imbalance.
Make sure to take a reference to the operation when sending responses.
This reference is dropped in greybus_data_sent when the message has been
sent, while the initial reference is dropped in gb_operation_work after
processing the corresponding request.
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com> Reviewed-by: Alex Elder <elder@linaro.org> Signed-off-by: Greg Kroah-Hartman <greg@kroah.com>
Johan Hovold [Fri, 27 Mar 2015 11:41:12 +0000 (12:41 +0100)]
greybus: operation: fix callback handling and documentation
Fix up obsolete comments referring to null callback pointers for
synchronous operations, and make sure a callback is always provided when
sending a request.
Also document that the callback is responsible for dropping the initial
(and not necessarily final) reference to the operation.
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com> Reviewed-by: Alex Elder <elder@linaro.org> Signed-off-by: Greg Kroah-Hartman <greg@kroah.com>
Johan Hovold [Fri, 27 Mar 2015 11:41:10 +0000 (12:41 +0100)]
greybus: operation: fix missing symbol exports
Add missing EXPORT_SYMBOL_GPL for gb_operation_response_alloc,
gb_operation_result, gb_operation_get, gb_operation_request_send and
gb_operation_cancel, which are all supposed to be accessible from
protocol handlers.
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com> Reviewed-by: Alex Elder <elder@linaro.org> Signed-off-by: Greg Kroah-Hartman <greg@kroah.com>
Viresh Kumar [Fri, 27 Mar 2015 11:02:56 +0000 (16:32 +0530)]
greybus: kernel_ver.h: include <linux/kernel.h> to fix warning
And this is the warning I was getting on kernel version > 3.14
CC [M] greybus/connection.o
In file included from
include/asm-generic/gpio.h:4:0,
from arch/arm/include/asm/gpio.h:9,
from include/linux/gpio.h:48,
from greybus/kernel_ver.h:59,
from greybus/connection.c:12:
include/linux/kernel.h:35:0: warning: "U16_MAX" redefined
kernel_ver.h is taking care of defining U16_MAX only if is not defined earlier,
but it is often included as the first .h file. <linux/kernel.h> might be
included later, which always defines it, unconditionally. And so this warning.
projects / karo-tx-linux.git / log