| Age | Commit message (Collapse) | Author |
|---|
|
Extend SCMI protocols accounting mechanism to address possible module
usage and add the support to possibly define new protocols as loadable
modules.
Keep the standard protocols built into the SCMI core.
Link: https://lore.kernel.org/r/20210316124903.35011-38-cristian.marussi@arm.com
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
|
|
Remove all the events registration code used to ease the transition to the
new interface based on protocol handles.
Link: https://lore.kernel.org/r/20210316124903.35011-35-cristian.marussi@arm.com
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
|
|
Now that all the SCMI driver users have been migrated to the new interface
remove the legacy interface and all the transient code.
Link: https://lore.kernel.org/r/20210316124903.35011-14-cristian.marussi@arm.com
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
|
|
Convert internals of protocol implementation to use protocol handles and
expose a new protocol operations interface for SCMI driver using the new
get/put common operations, while keeping the old handle->perf_ops still
around to ease transition.
Remove handle->perf_priv now unused.
Link: https://lore.kernel.org/r/20210316124903.35011-12-cristian.marussi@arm.com
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
|
|
Convert all refactored events registration routines to use protocol
handles.
In order to maintain bisectability and to allow protocols and drivers
to be later ported to the new protocol handle interface one by one,
introduce here also some transient code that will be removed later
in order to ease such transition.
Link: https://lore.kernel.org/r/20210316124903.35011-8-cristian.marussi@arm.com
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
|
|
Add a new refactored protocol events registration helper and invoke it
from the centralized initialization process triggered by get_ops() and
friends.
Also add a `get_num_sources` as a new optional callback amongst protocol
events operations. Finally remove events registration call-sites from
within the legacy protocol init routines.
Link: https://lore.kernel.org/r/20210316124903.35011-7-cristian.marussi@arm.com
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
|
|
Extend common protocol registration routines and provide some new generic
protocols get/put helpers that can track protocols usage and automatically
perform the proper initialization and de-initialization on demand when
required.
Convert all standard protocols to use this new registration scheme while
keeping them all still using the usual initialization logic bound to SCMI
devices probing.
Link: https://lore.kernel.org/r/20210316124903.35011-2-cristian.marussi@arm.com
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
|
|
Add a new interface to the existing perf_ops and export the information
about the power values scale.
This would be used by the cpufreq driver and Energy Model framework to
set the performance domains scale: milli-Watts or abstract scale.
Suggested-by: Morten Rasmussen <morten.rasmussen@arm.com>
Reviewed-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
Acked-by: Sudeep Holla <sudeep.holla@arm.com>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
|
|
Few commands provide the list of description partially and require
to be called consecutively until all the descriptors are fetched
completely. In such cases, we don't release the buffers and reuse
them for consecutive transmits.
However, currently we don't reset the Rx size which will be set as
per the response for the last transmit. This may result in incorrect
response size being interpretted as the firmware may repond with size
greater than the one set but we read only upto the size set by previous
response.
Let us reset the receive buffer size to max possible in such cases as
we don't know the exact size of the response.
Link: https://lore.kernel.org/r/20201012141746.32575-1-sudeep.holla@arm.com
Fixes: b6f20ff8bd94 ("firmware: arm_scmi: add common infrastructure and support for base protocol")
Reported-by: Etienne Carriere <etienne.carriere@linaro.org>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
|
|
In preparation to enable building SCMI as a single module, let us move
the SCMI protocol registration call into the driver. This enables us
to also add unregistration of the SCMI protocols.
The main reason for this is to keep it simple instead of maintaining
it as separate modules and dealing with all possible initcall races
and deferred probe handling. We can move it as separate modules if
needed in future.
Link: https://lore.kernel.org/r/20200907195046.56615-4-sudeep.holla@arm.com
Tested-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
|
|
These are never modified, so make them const to allow the compiler to
put them in read-only memory.
Link: https://lore.kernel.org/r/20200906230452.33410-4-rikard.falkeborn@gmail.com
Signed-off-by: Rikard Falkeborn <rikard.falkeborn@gmail.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
|
|
Event reports are used to convey information describing events to the
registered user-callbacks: they are necessarily derived from the underlying
raw SCMI events' messages but they are not meant to expose or directly
mirror any of those messages data layout, which belong to the protocol
layer.
Using fixed size types for report fields, mirroring messages structure,
is at odd with this: get rid of them using more generic, equivalent,
typing.
Substitute scmi_event_header fixed size fields with generic types too and
shuffle around fields definitions to minimize implicit padding while
adapting involved functions.
Link: https://lore.kernel.org/r/20200710133919.39792-3-cristian.marussi@arm.com
Signed-off-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
|
|
Make SCMI perf protocol register with the notification core.
Link: https://lore.kernel.org/r/20200701155348.52864-7-cristian.marussi@arm.com
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Cristian Marussi <cristian.marussi@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
|
|
Add a new fast_switch_possible interface to the existing perf_ops to
export the information of whether or not fast_switch is possible for a
given device.
This can be used by the cpufreq driver and framework to choose proper
mechanism for frequency change.
Link: https://lore.kernel.org/r/20200617094332.8391-1-nicola.mazzucato@arm.com
Suggested-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Nicola Mazzucato <nicola.mazzucato@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>
|
|
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 76497732932f ("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>
|
|
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 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: 823839571d76 ("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>
|
|
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>
|
|
The SCMI protocol can be used to get power estimates from firmware
corresponding to each performance state of a device. Although these power
costs are already managed by the SCMI firmware driver, they are not
exposed to any external subsystem yet.
Fix this by adding a new get_power() interface to the exisiting perf_ops
defined for the SCMI protocol.
Signed-off-by: Quentin Perret <quentin.perret@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
|
|
Replace all the memcpy() for copying name strings from the firmware with
strlcpy() to make sure we are bounded by the source buffer size and we
also always have NULL-terminated strings.
This is needed to avoid out of bounds accesses if the firmware returns
a non-terminated string.
Reported-by: Olof Johansson <olof@lixom.net>
Acked-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
|
|
Firmware can provide zero as values for sustained performance level and
corresponding sustained frequency in kHz in order to hide the actual
frequencies and provide only abstract values. It may endup with divide
by zero scenario resulting in kernel panic.
Let's set the multiplication factor to one if either one or both of them
(sustained_perf_level and sustained_freq) are set to zero.
Fixes: a9e3fbfaa0ff ("firmware: arm_scmi: add initial support for performance protocol")
Reported-by: Ionela Voinescu <ionela.voinescu@arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Signed-off-by: Olof Johansson <olof@lixom.net>
|
|
The "pi->dom_info" buffer is allocated in init() and it can't be NULL
here. These tests are sort of weird as well because if "pi->dom_info"
was NULL but "domain" was non-zero then it would lead to an Oops.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
|
|
Just after the initial patches were queued, Jonathan Cameron mentioned
that scmi_one_xfer_{get_put} were not very clear and suggested to use
scmi_xfer_{alloc,free}. While I agree to some extent, the reason not to
have alloc/free as these are preallocated buffers and these functions
just returns a reference to free slot in that preallocated array.
However it was agreed to drop "_one" as it's implicit that we are always
dealing with one slot anyways.
This patch updates the name accordingly dropping "_one" in both {get,put}
functions. Also scmi_one_xfer_init is renamed as scmi_xfer_get_init to
reflect the fact that it gets the free slots and then initialise it.
Reported-by: Jonathan Cameron <jonathan.cameron@huawei.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
|
|
Most of the scmi code follows the suggestion from Greg KH on a totally
different thread[0] to have the subsystem name first, followed by the
noun and finally the verb with couple of these exceptions.
This patch fixes them so that all the functions names are aligned to
that practice.
[0] https://www.spinics.net/lists/arm-kernel/msg583673.html
Acked-by: Rafael J. Wysocki <rafael@kernel.org>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
|
|
In order to implement fast CPU DVFS switching, we need to perform all
DVFS operations atomically. Since SCMI transfer already provide option
to choose between pooling vs interrupt driven(default), we can opt for
polling based transfers for set,get performance domain operations.
This patch adds option to choose between polling vs interrupt driven
SCMI transfers for set,get performance level operations.
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
|
|
The performance protocol is intended for the performance management of
group(s) of device(s) that run in the same performance domain. It
includes even the CPUs. A performance domain is defined by a set of
devices that always have to run at the same performance level.
For example, a set of CPUs that share a voltage domain, and have a
common frequency control, is said to be in the same performance domain.
The commands in this protocol provide functionality to describe the
protocol version, describe various attribute flags, set and get the
performance level of a domain. It also supports discovery of the list
of performance levels supported by a performance domain, and the
properties of each performance level.
This patch adds basic support for the performance protocol.
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
|
