For purpose of adding testing HW we would like
to get rid of pci dependency in generic mei code.
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fw status retrieval has pci specific implementation
so we push it back to the hw layer
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
For purpose of adding testing HW we would like
to get rid of pci dependency in generic mei code
This patch provides only straight forward changes
FW status and prob quirks need to be handled separately
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Setting of hw_ops and device has should be in
mei_device_init.
We add reference to the parent device and remove
pci dependent cfg
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fix new style warning: Missing a blank line after declarations
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
extract supported hbm features and commands from the hbm version
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We cannot handle user interrupt in context of hw initialization
so we only wait for time out which is reasonably short
Also we don't need to check error from wait, only flag value.
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We introduced unified FW status function in patch
mei: add per device configuration (lkml.org/lkml/2014/5/12/607)
This change made hw_ops functions unused and obsolete
therefore we remove these functions from source code.
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Move quirk FW type detector to cfg structure
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Add mei_cfg structure that holds per device configuration
data and hooks, as the first step we add firmware
status register offsets
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
According the spec the host should read H_CSR again
after asserting reset H_RST to ensure that reset was
read by the firmware
Cc: stable@vger.kernel.org
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
It my take time till ME_RDY will be cleared after the reset,
so we cannot check the bit before we got the interrupt
Cc: stable@vger.kernel.org
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We cleared H_RST for H_CSR on spurious interrupt generated when ME_RDY
while cleared and not while ME_RDY is set. The spurious interrupt
is not delivered on all platforms in this case the
driver may fail to initialize.
Cc: stable@vger.kernel.org
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Fetch FW status registers, as they are important in
in understanding of FW reset reasons
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
For power gating entry we write hbm pg entry request command and
then we set pg register
For power gating exit we clear pg register and wait for exit request
hbm command.
Exit power gating request might also be initiated by the firmware
w/o explicit driver request
The power gating state is tracked by pg_state member of me_hw
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Since the runtime pm and the internal power gating
cannot be in complete sync in regards to I/O
operations, we need to expose the device
hardware internal power gating state to mei layer
2. We add pg_state handler that translate the hw
internal pg state to mei layer
2. We add power gating event variable to keep
power track of power gating transitions
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Reviewed-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Enable power gating isolation only if hw and fw support it.
This is indicated by ME_PGIC_HRA bit in ME_CSR_HA register
and on HBM protocol version.
The information is exported to MEI layer through
new pg_is_enabled hw op.
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Remove includes of me-hw.h and me-hw-regs.h headers from
the mei generic code.
The mei layer should not depend on hw specific headers
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There is a race in ME hardware between data copy for host and interrupt
delivery. An interrupt can be delivered prior to whole data copied for the
host to read but rather then going trough the reset we just merely need to
wait for the next interrupt.
The bug is visible in read/write stress with multiple connections per client
This is a regression caused as a side effect of the commit:
commit 544f946014
mei: do not run reset flow from the interrupt thread
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Cc: stable <stable@vger.kernel.org> # 3.14
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
1. Propagate ENOTTY to user space if the client is not present
in the system
2. Use ETIME consistently on timeouts
3. Return EIO on write failures
4. Return ENODEV on recoverable device failures such as resets
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Since txe use doorbell and not circular buffer
we have to cheat in write slot counting, txe always consume all the
slots upon write. In order for it to work we need to track
slots using mei_hbuf_empty_slots() instead of tracking it in mei layer
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
A client has to acquire host buffer
before writing, we add lock like wrapper
to replace the code snippet
if (dev->hbuf_is_ready)
dev->hbuf_is_ready = false;
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Use more standard message writing for
oob data.
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In some rare case mei hw reset may take long time to settle.
Instead of blocking resume flow we span another driver reset flow in
separate work context
This allows as to shorten hw reset timeout to something more acceptable
by DPM_WATCHDOG_TIMEOUT
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
1. MEI_DEV_RESETTING device state spans only hardware reset flow
while starting dev state is saved into a local variable for further
reference, this let us to reduce big if statements in case we
are trying to avoid nested resets
2. During initializations if the reset ended in MEI_DEV_DISABLED device
state we bail out with -ENODEV
3. Remove redundant interrupts_enabled parameter as this
can be deduced from the starting dev_state
4. mei_reset propagates error code to the caller
5. Add mei_restart function to wrap the pci resume
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This fixes a potential deadlock in case of a firmware
initiated reset
mei_reset has a dialog with the interrupt thread hence
it has to be run from an another work item
Most of the mei_resets were called from mei_hbm_dispatch
which is called in interrupt thread context so this
function underwent major revamp. The error code is
propagated to the interrupt thread and if needed
the reset is scheduled from there.
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Both H_IS and H_IE needs to be set to receive H_RDY
interrupt
1. Assert H_IS to clear the interrupts during hw reset
and use mei_me_reg_write instead of mei_hcsr_set as the later
strips down the H_IS
2. fix interrupt disablement embarrassing typo
hcsr |= ~H_IE -> hcsr &= ~H_IE;
this will remove the unwanted interrupt on power down
3. remove useless debug print outs
Cc: Shuah Khan <shuah.kh@samsung.com>
Cc: Konstantin Khlebnikov <khlebnikov@openvz.org>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
1. MEI_INTEROP_TIMEOUT is in seconds not in jiffies
so we use mei_secs_to_jiffies macro
While cold boot is fast this is relevant in resume
2. wait_event_interruptible_timeout can return with
-ERESTARTSYS so do not override it with -ETIMEDOUT
3.Adjust error message
Tested-by: Shuah Khan <shuah.kh@samsung.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
ME HW ready bit is down after hw reset was asserted or on error.
Only on error we need to enter the reset flow, additional reset
need to be prevented when reset was triggered during
initialization , power up/down or a reset is already in progress
Tested-by: Shuah Khan <shuah.kh@samsung.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The hw may have multiple steps for resetting
so we need to check if it has really succeeded.
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This enum leaks out to userspace via error messages, so fix the spelling.
Signed-off-by: Bill Nottingham <notting@redhat.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
1. Rename init_clients_state to hbm_state and use
MEI_HBM_ prefix for HBM states
2. Remove recvd_msg and use hbm state for synchronizing
hbm protocol has successful start.
We can wake up the hbm event from start response handler
and remove the hack from the interrupt thread
3. mei_hbm_start_wait function encapsulate start completion
waiting
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
mei_timer and mei_host_client_init belongs to mei framework
and are not ME hw specific.
AMTHIF and WD are available only for ME but are above the hardware layer
so move the initialization back from mei_me_dev_init to mei_device_init.
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
similar to read/write add also irq completion handler
that is called for the irq thread
rename missnamed mei_irq_complete_handler to
mei_cl_complete_handler as it operates on a single client
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This fixes failure during initialization on Lynx Point LP devices.
ME driver needs to release the device from the reset
only after the FW has completed its flow and indicated
it by delivering an interrupt to the host.
This is the correct behavior for all the ME devices yet the
the previous versions are less susceptive to the implementation
that ignored FW reset completion indication.
We add mei_me_hw_reset_release function which is called
after reset from the interrupt thread or directly
from mei_reset during power down.
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This callback wraps up hardware dependent details
of the hardware initialization.
This callback also contains host ready setting
so we can remove host_set_ready callback
In ME we switch to waiting on event so
we can streamline the initialization flow.
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
1. Move the mei_data2slots to mei_dev.h as it will be used
by the all supported HW.
2. Change return value from u8 to u32 to catch possible overflows
3. Eliminate computing the slots number twice in the same function
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
we rename the mei_host_buffer_is_empty to keep naming
convention of hbuf and also make the query more generic
to be correct also for other under laying hardware
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
interrupt handler are platform specifics so we move
them to hw-mei.c. For sake of that we need to export
write, read, and complete handlers from the interrupt.c
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We add struct mei_hw_ops to virtualize access to hw specific
configurations. This allows us to separate the compilation
of the ME interface from the ME hardware specifics
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This is initial step of move the ME hw specifics
out of mei_device structure into mei_me_hw
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Move csr reading into me hardware functional calls.
Since we gave up on registers caching we remove some of the unnecessary
queries in mei_hw_init ane mei_reset functions.
We add mei_hw_config function to wrap up host buffer depth configuration.
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Now when mei_hcsr_set is local to hw-me.c
we can benefit form the fact that it wraps
H_IS removal from the host csr.
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>