When CONFIG_ARM_PSCI_FW is disabled but CONFIG_HAVE_ARM_SMCCC is enabled,
arm-scmi runs into a link failure:
arm-linux-gnueabi-ld: drivers/firmware/arm_scmi/smc.o: in function `smc_send_message':
smc.c:(.text+0x200): undefined reference to `arm_smccc_1_1_get_conduit'
Change from HAVE_ARM_SMCCC to ARM_PSCI_FW config dependency for now.
We rely on PSCI bindings anyways for the conduit and this should be
fine.
Link: https://lore.kernel.org/r/20200507144905.11397-1-sudeep.holla@arm.com
Fixes: 1dc6558062 ("firmware: arm_scmi: Add smc/hvc transport")
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Upon reception of an unexpected bogus delayed response, clear the channel
and bail-out safely.
Link: https://lore.kernel.org/r/20200420152315.21008-5-cristian.marussi@arm.com
Fixes: 4d09852b6f ("firmware: arm_scmi: Add support for notifications message processing")
Signed-off-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Clear channel properly when done processing a delayed response.
This will let the platform firmware know that the channel is now free to
use it for any new delayed response or notification.
Link: https://lore.kernel.org/r/20200420152315.21008-4-cristian.marussi@arm.com
Signed-off-by: Cristian Marussi <cristian.marussi@arm.com>
(sudeep.holla: Updated commit log to reflect that channel is now free for
platform to use)
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
When an unexpected response message is received we currently warn the user
and bail-out, ensure to also free the channel by invoking the transport
independent operation .clear_channel()
Link: https://lore.kernel.org/r/20200420152315.21008-3-cristian.marussi@arm.com
Signed-off-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
SCMI transport operation .clear_notification() is indeed a generic method
to clear the channel in a transport dependent way, as such it could be a
useful helper also in other contexts.
Rename such method as .clear_channel(), renaming accordingly also its
already existent call-sites.
No functional change.
Link: https://lore.kernel.org/r/20200420152315.21008-2-cristian.marussi@arm.com
Signed-off-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Add the mechanisms to distinguish notifications from delayed responses
and command responses. Also add support to properly fetch notification
messages upon reception. Notifications processing does not continue any
further after the fetch phase.
Link: https://lore.kernel.org/r/20200327143438.5382-5-cristian.marussi@arm.com
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
[Reworked/renamed scmi_handle_xfer_delayed_resp()]
Signed-off-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Add common transport-layer methods to:
- fetch a notification instead of a response
- clear a pending notification
Add also all the needed support in mailbox/shmem transports.
Link: https://lore.kernel.org/r/20200327143438.5382-4-cristian.marussi@arm.com
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Add commands' enumerations and messages definitions for all existing
notify-enable commands across all protocols.
Link: https://lore.kernel.org/r/20200327143438.5382-3-cristian.marussi@arm.com
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
With all the plumbing in place, let's just add the separate dedicated
receive buffers to handle notifications that can arrive asynchronously
from the platform firmware to OS.
Also add one check to see if the platform supports any receive channels
before allocating the receive buffers: since those buffers are optionally
supported though, the whole xfer initialization is also postponed to be
able to check for their existence in advance.
Link: https://lore.kernel.org/r/20200327143438.5382-2-cristian.marussi@arm.com
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
[Changed parameters in __scmi_xfer_info_init()]
Signed-off-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
The scmi protocol core driver checks for the channel availability
before evaluating the shmem property. If the individual protocols
don't have separate channel assigned to them, the channel alloted
for the BASE protocol is reused automatically.
Therefore there is no need to check for the shmem property in the
parent node if it is absent in the child protocol node.
Link: https://lore.kernel.org/r/20200327163654.13389-5-sudeep.holla@arm.com
Tested-by: Peng Fan <peng.fan@nxp.com>
Reviewed-by: Peng Fan <peng.fan@nxp.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Instead of declaring the channel availabilty unconditionally, let us
check for the presence of "shmem" property and return the channel
availablity accordingly.
Link: https://lore.kernel.org/r/20200327163654.13389-4-sudeep.holla@arm.com
Tested-by: Peng Fan <peng.fan@nxp.com>
Reviewed-by: Peng Fan <peng.fan@nxp.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
The scmi protocol core driver check for non NULL mark_txdone before
invoking the same. There is no need to provide a empty stub. SMC/HVC
calls are synchronous and the call return indicates the completion.
Link: https://lore.kernel.org/r/20200327163654.13389-3-sudeep.holla@arm.com
Tested-by: Peng Fan <peng.fan@nxp.com>
Reviewed-by: Peng Fan <peng.fan@nxp.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
In order to support multiple SMC/HVC transport channels with associated
shared memory, it is better to maintain the mutex per channel instead of
existing global one.
Move the smc_mutex into the scmi_smc structure and also rename it to
shmem_lock which is more appropriate for it's use.
Link: https://lore.kernel.org/r/20200327163654.13389-2-sudeep.holla@arm.com
Tested-by: Peng Fan <peng.fan@nxp.com>
Reviewed-by: Peng Fan <peng.fan@nxp.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Use the value of "arm,smc-id" property from the device tree as the first
argument for SMCCC call leaving all the other arguments as zero for now.
There is no Rx, only Tx because of smc/hvc not support Rx.
Link: https://lore.kernel.org/r/1583673879-20714-3-git-send-email-peng.fan@nxp.com
Signed-off-by: Peng Fan <peng.fan@nxp.com>
[sudeep.holla: reworded commit log/subject and fixed !HAVE_ARM_SMCCC build]
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:
struct foo {
int stuff;
struct boo array[];
};
By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertenly introduced[3] to the codebase from now on.
This issue was found with the help of Coccinelle.
[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 7649773293 ("cxgb3/l2t: Fix undefined behaviour")
Replace the zero-length member "opp" in scmi_msg_resp_perf_describe_levels
structure with flexible-array.
Link: https://lore.kernel.org/r/20200211231252.GA14830@embeddedor
Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:
struct foo {
int stuff;
struct boo array[];
};
By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertenly introduced[3] to the codebase from now on.
This issue was found with the help of Coccinelle.
[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 7649773293 ("cxgb3/l2t: Fix undefined behaviour")
Replace the zero-length member "msg_payload" in scmi_shared_mem
structure with flexible-array.
Link: https://lore.kernel.org/r/20200211231045.GA13956@embeddedor
Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
[ rebased the change as files are moved around ]
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
The SCMI specification is fairly independent of the transport protocol,
which can be a simple mailbox (already implemented) or anything else.
The current Linux implementation however is very much dependent on the
mailbox transport layer.
This patch makes the SCMI core code (driver.c) independent of the
mailbox transport layer and moves all mailbox related code to a new
file: mailbox.c and all struct shared_mem related code to a new file:
shmem.c.
We can now implement more transport protocols to transport SCMI
messages.
The transport protocols just need to provide struct scmi_transport_ops,
with its version of the callbacks to enable exchange of SCMI messages.
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Link: https://lore.kernel.org/r/8698a3cec199b8feab35c2339f02dc232bfd773b.1580448239.git.viresh.kumar@linaro.org
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
The SCMI transport layer communicates via mailboxes and shared memory with
firmware running on a microcontroller. It is platform specific how long it
takes to pass a SCMI message. The most sensitive requests are coming from
CPUFreq subsystem, which might be used by the scheduler.
Thus, there is a need to measure these delays and capture anomalies.
This change introduces trace events wrapped around transfer code.
According to Jim's suggestion a unique transfer_id is to distinguish
similar entries which might have the same message id, protocol id and
sequence. This is a case then there are some timeouts in transfers.
Suggested-by: Jim Quinlan <james.quinlan@broadcom.com>
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
The scmi bus now supports adding multiple devices per protocol,
and since scmi_protocol_init is called for each scmi device created,
we must avoid allocating protocol private data and initialising the
protocol itself if it is already initialised.
In order to achieve the same, we can simple replace the idr pointer
from protocol initialisation function to a dummy function.
Suggested-by: Cristian Marussi <cristian.marussi@arm.com>
Reviewed-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
In order to avoid querying the individual protocol versions multiple
time with more that one device created for each protocol, we can simple
store the copy in the protocol specific private data and use them whenever
required.
Reviewed-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
The scmi bus now has support to match the driver with devices not only
based on their protocol id but also based on their device name if one is
available. This was added to cater the need to support multiple devices
and drivers for the same protocol.
Let us add the name "genpd" to scmi_device_id table in the driver so
that in matches only with device with the same name and protocol id
SCMI_PROTOCOL_POWER.
Reviewed-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Platform drivers now have the option to have the platform core create
and remove any needed sysfs attribute files. Using the same, let's add
the scmi firmware and protocol version attributes as well as vendor and
sub-vendor identifiers to sysfs.
It helps to identify the firmware details from the sysfs entries similar
to ARM SCPI implementation.
Reviewed-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Now that scmi bus provides option to create named scmi device, let us
create the default devices with names. This will help to add names for
matching to respective drivers and eventually to add multiple devices
and drivers per protocol.
Reviewed-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Now that the scmi bus supports adding multiple devices per protocol,
and since scmi_create_protocol_device calls scmi_mbox_chan_setup,
we must avoid allocating and initialising the mbox channel if it is
already initialised.
Reviewed-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Currently only one scmi device is created for each protocol enumerated.
However, there is requirement to make use of some procotols by multiple
kernel subsystems/frameworks. One such example is SCMI PERFORMANCE
protocol which can be used by both cpufreq and devfreq drivers.
Similarly, SENSOR protocol may be used by hwmon and iio subsystems,
and POWER protocol may be used by genpd and regulator drivers.
In order to achieve that, let us extend the scmi bus to match based
not only protocol id but also the scmi device name if one is available.
Reviewed-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
If device_register() fails, both put_device() and kfree() are called,
ending with a double free of the scmi_dev.
Calling kfree() is needed only when a failure happens between the
allocation of the scmi_dev and its registration, so move it to there
and remove it from the error flow.
Fixes: 46edb8d132 ("firmware: arm_scmi: provide the mandatory device release callback")
Signed-off-by: Wen Yang <wenyang@linux.alibaba.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
The logic to ring the scmi performance fastchannel ignores the
value read from the doorbell register in case of !CONFIG_64BIT.
This bug also shows up as warning with '-Wunused-but-set-variable' gcc
flag:
drivers/firmware/arm_scmi/perf.c: In function scmi_perf_fc_ring_db:
drivers/firmware/arm_scmi/perf.c:323:7: warning: variable val set but
not used [-Wunused-but-set-variable]
Fix the same by aligning the logic with CONFIG_64BIT as used in the
macro SCMI_PERF_FC_RING_DB().
Fixes: 823839571d ("firmware: arm_scmi: Make use SCMI v2.0 fastchannel for performance protocol")
Reported-by: Hulk Robot <hulkci@huawei.com>
Reported-by: Zheng Yongjun <zhengyongjun3@huawei.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
SCMIv2.0 adds a new Reset Management Protocol to manage various reset
states a given device or domain can enter. Device(s) that can be
collectively reset through a common reset signal constitute a reset
domain for the firmware.
A reset domain can be reset autonomously or explicitly through assertion
and de-assertion of the signal. When autonomous reset is chosen, the
firmware is responsible for taking the necessary steps to reset the
domain and to subsequently bring it out of reset. When explicit reset is
chosen, the caller has to specifically assert and then de-assert the
reset signal by issuing two separate RESET commands.
Add the basic SCMI reset infrastructure that can be used by Linux
reset controller driver.
Reviewed-by: Peng Fan <peng.fan@nxp.com>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
SCMI v2.0 adds support for "FastChannel" which do not use a message
header as they are specialized for a single message.
Only PERFORMANCE_LIMITS_{SET,GET} and PERFORMANCE_LEVEL_{SET,GET}
commands are supported over fastchannels. As they are optional, they
need to be discovered by PERFORMANCE_DESCRIBE_FASTCHANNEL command.
Further {LIMIT,LEVEL}_SET commands can have optional doorbell support.
Add support for making use of these fastchannels.
Cc: Ionela Voinescu <Ionela.Voinescu@arm.com>
Cc: Chris Redpath <Chris.Redpath@arm.com>
Cc: Quentin Perret <Quentin.Perret@arm.com>
Reviewed-by: Peng Fan <peng.fan@nxp.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
SCMI v2.0 adds support for "FastChannel", a lightweight unidirectional
channel that is dedicated to a single SCMI message type for controlling
a specific platform resource. They do not use a message header as they
are specialized for a single message.
Only PERFORMANCE_LIMITS_{SET,GET} and PERFORMANCE_LEVEL_{SET,GET}
commands are supported over fastchannels. As they are optional, they
need to be discovered by PERFORMANCE_DESCRIBE_FASTCHANNEL command.
Further {LIMIT,LEVEL}_SET commands can have optional doorbell support.
Add support for discovery of these fastchannels.
Cc: Ionela Voinescu <Ionela.Voinescu@arm.com>
Cc: Chris Redpath <Chris.Redpath@arm.com>
Cc: Quentin Perret <Quentin.Perret@arm.com>
Reviewed-by: Peng Fan <peng.fan@nxp.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Instead of type-casting the {tx,rx}.buf all over the place while
accessing them to read/write __le{32,64} from/to the firmware, let's
use the existing {get,put}_unaligned_le{32,64} accessors to hide all
the type cast ugliness.
Suggested-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
CLOCK_PROTOCOL_ATTRIBUTES provides attributes to indicate the maximum
number of pending asynchronous clock rate changes supported by the
platform. If it's non-zero, then we should be able to use asynchronous
clock rate set for any clocks until the maximum limit is reached.
Tracking the current count of pending asynchronous clock set rate
requests, we can decide if the incoming/new request for clock set rate
can be handled asynchronously or not until the maximum limit is
reached.
Cc: linux-clk@vger.kernel.org
Reviewed-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
CLOCK_PROTOCOL_ATTRIBUTES provides attributes to indicate the maximum
number of pending asynchronous clock rate changes supported by the
platform. If it's non-zero, then we should be able to use asynchronous
clock rate set for any clocks until the maximum limit is reached.
In order to add that support, let's drop the config flag passed to
clk_ops->rate_set and handle the asynchronous requests dynamically.
Cc: Stephen Boyd <sboyd@kernel.org>
Cc: linux-clk@vger.kernel.org
Acked-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
SENSOR_DESCRIPTION_GET provides attributes to indicate if the sensor
supports asynchronous read. We can read that flag and use asynchronous
reads for any sensors with that attribute set.
Let's use the new scmi_do_xfer_with_response to support asynchronous
sensor reads.
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
SENSOR_DESCRIPTION_GET provides attributes to indicate if the sensor
supports asynchronous read. Ideally we should be able to read that flag
and use asynchronous reads for any sensors with that attribute set.
In order to add that support, let's drop the async flag passed to
sensor_ops->reading_get and dynamically switch between sync and async
flags based on the attributes as provided by the firmware.
Cc: linux-hwmon@vger.kernel.org
Acked-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Messages that are sent to platform, also known as commands and can be:
1. Synchronous commands that block the channel until the requested work
has been completed. The platform responds to these commands over the
same channel and hence can't be used to send another command until the
previous command has completed.
2. Asynchronous commands on the other hand, the platform schedules the
requested work to complete later in time and returns almost immediately
freeing the channel for new commands. The response indicates the success
or failure in the ability to schedule the requested work. When the work
has completed, the platform sends an additional delayed response message.
Using the same transmit buffer used for sending the asynchronous command
even for the delayed response corresponding to it simplifies handling of
the delayed response. It's the caller of asynchronous command that is
responsible for allocating the completion flag that scmi driver can
complete to indicate the arrival of delayed response.
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
In order to identify the message type when a response arrives, we need
a mechanism to unpack the message header similar to packing. Let's
add one.
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Currently we pre-allocate transmit buffers only and use the first free
slot in that pre-allocated buffer for transmitting any new message that
are generally originated from OS to the platform firmware.
Notifications or the delayed responses on the other hand are originated
from the platform firmware and consumes by the OS. It's better to have
separate and dedicated pre-allocated buffers to handle the notifications.
We can still use the transmit buffers for the delayed responses.
In addition, let's prepare existing scmi_xfer_{get,put} for acquiring
and releasing a slot to identify the right(tx/rx) buffers.
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
With scmi_mbox_chan_setup enabled to identify and setup both Tx and Rx,
let's consolidate setting up of both the channels under the function
scmi_mbox_txrx_setup.
Since some platforms may opt not to support notifications or delayed
response, they may not need support for Rx. Hence Rx is optional and
failure of setting one up is not considered fatal.
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
The transmit(Tx) channels are specified as the first entry and the
receive(Rx) channels are the second entry as per the device tree
bindings. Since we currently just support Tx, index 0 is hardcoded at
all required callsites.
In order to prepare for adding Rx support, let's remove those hardcoded
index and add boolean parameter to identify Tx/Rx channels when setting
them up.
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Re-shuffling few functions to keep definitions and their usages close.
This is also needed to avoid too many unnecessary forward declarations
while adding new features(delayed response and notifications).
Keeping this separate to avoid mixing up of these trivial change that
doesn't affect functionality into the ones that does.
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Sometimes platfom may take too long to respond to the command and OS
might timeout before platform transfer the ownership of the shared
memory region to the OS with the response.
Since the mailbox channel associated with the channel is freed and new
commands are dispatch on the same channel, OS needs to wait until it
gets back the ownership. If not, either OS may end up overwriting the
platform response for the last command(which is fine as OS timed out
that command) or platform might overwrite the payload for the next
command with the response for the old.
The latter is problematic as platform may end up interpretting the
response as the payload. In order to avoid such race, let's wait until
the OS gets back the ownership before we prepare the shared memory with
the payload for the next command.
Reported-by: Jim Quinlan <james.quinlan@broadcom.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
In preparation to adding support for other two types of messages that
SCMI specification mentions, let's replace the term 'command' with the
correct term 'message'.
As per the specification the messages are of 3 types:
commands(synchronous or asynchronous), delayed responses and notifications.
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
While adding new comments found couple of typos that are better fixed.
s/informfation/information/
s/statues/status/
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
scmi_xfer_get_init ensures both transmit and receive buffer lengths are
within the maximum limits. If receive buffer length is not supplied by
the caller, it's set to the maximum limit value. Receive buffer length
is never modified after that. So there's no need for the extra check
when receive transmit completion for a command essage.
Further, if the response header length is greater than the prescribed
receive buffer length, the response buffer is truncated to the latter.
Reported-by: Jim Quinlan <james.quinlan@broadcom.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Looks like more code developed during the draft versions of the
specification slipped through and they don't match the final
released version. This seem to have happened only with sensor
protocol.
Renaming few command and function names here to match exactly with
the released version of SCMI specification for ease of maintenance.
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Sudeep Holla