Expose through debugfs device telemetry data for QAT GEN4 devices.
This allows to gather metrics about the performance and the utilization
of a device. In particular, statistics on (1) the utilization of the
PCIe channel, (2) address translation, when SVA is enabled and (3) the
internal engines for crypto and data compression.
If telemetry is supported by the firmware, the driver allocates a DMA
region and a circular buffer. When telemetry is enabled, through the
`control` attribute in debugfs, the driver sends to the firmware, via
the admin interface, the `TL_START` command. This triggers the device to
periodically gather telemetry data from hardware registers and write it
into the DMA memory region. The device writes into the shared region
every second.
The driver, every 500ms, snapshots the DMA shared region into the
circular buffer. This is then used to compute basic metric
(min/max/average) on each counter, every time the `device_data` attribute
is queried.
Telemetry counters are exposed through debugfs in the folder
/sys/kernel/debug/qat_<device>_<BDF>/telemetry.
For details, refer to debugfs-driver-qat_telemetry in Documentation/ABI.
This patch is based on earlier work done by Wojciech Ziemba.
Signed-off-by: Lucas Segarra Fernandez <lucas.segarra.fernandez@intel.com>
Reviewed-by: Giovanni Cabiddu <giovanni.cabiddu@intel.com>
Reviewed-by: Damian Muszynski <damian.muszynski@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The Compress and Verify (CnV) feature check and ensures data integrity
in the compression operation. The implementation of CnV keeps a record
of the CnV errors that have occurred since the driver was loaded.
Expose CnV error stats by providing the "cnv_errors" file under
debugfs. This includes the number of errors detected up to now and
the type of the last error. The error count is provided on a per
Acceleration Engine basis and it is reset every time the driver is loaded.
Signed-off-by: Lucas Segarra Fernandez <lucas.segarra.fernandez@intel.com>
Reviewed-by: Giovanni Cabiddu <giovanni.cabiddu@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
QAT devices implement a mechanism that allows them to go autonomously
to a low power state depending on the load.
Expose power management info by providing the "pm_status" file under
debugfs. This includes PM state, PM event log, PM event counters, PM HW
CSRs, per-resource type constrain counters and per-domain power gating
status specific to the QAT device.
This information is retrieved from (1) the FW by means of
ICP_QAT_FW_PM_INFO command, (2) CSRs and (3) counters collected by the
device driver.
In addition, add logic to keep track and report power management event
interrupts and acks/nacks sent to FW to allow/prevent state transitions.
Signed-off-by: Lucas Segarra Fernandez <lucas.segarra.fernandez@intel.com>
Reviewed-by: Giovanni Cabiddu <giovanni.cabiddu@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Under some circumstances, firmware in the QAT devices could become
unresponsive. The Heartbeat feature provides a mechanism to detect
unresponsive devices.
The QAT FW periodically writes to memory a set of counters that allow
to detect the liveness of a device. This patch adds logic to enable
the reporting of those counters, analyze them and report if a device
is alive or not.
In particular this adds
(1) heartbeat enabling, reading and detection logic
(2) reporting of heartbeat status and configuration via debugfs
(3) documentation for the newly created sysfs entries
(4) configuration of FW settings related to heartbeat, e.g. tick period
(5) logic to convert time in ms (provided by the user) to clock ticks
This patch introduces a new folder in debugfs called heartbeat with the
following attributes:
- status
- queries_sent
- queries_failed
- config
All attributes except config are reading only. In particular:
- `status` file returns 0 when device is operational and -1 otherwise.
- `queries_sent` returns the total number of heartbeat queries sent.
- `queries_failed` returns the total number of heartbeat queries failed.
- `config` allows to adjust the frequency at which the firmware writes
counters to memory. This period is given in milliseconds and it is
fixed for GEN4 devices.
Signed-off-by: Damian Muszynski <damian.muszynski@intel.com>
Reviewed-by: Giovanni Cabiddu <giovanni.cabiddu@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Expose FW counters statistics by providing the "fw_counters" file
under debugfs. Currently the statistics include the number of
requests sent to the FW and the number of responses received
from the FW for each Acceleration Engine, for all the QAT product
line.
This patch is based on earlier work done by Marco Chiappero.
Co-developed-by: Adam Guerin <adam.guerin@intel.com>
Signed-off-by: Adam Guerin <adam.guerin@intel.com>
Signed-off-by: Lucas Segarra Fernandez <lucas.segarra.fernandez@intel.com>
Reviewed-by: Giovanni Cabiddu <giovanni.cabiddu@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Move initialization of debugfs entries to a separate file.
This simplifies the exclusion of the debugfs logic in the QAT driver
when the kernel is built with CONFIG_DEBUG_FS=n.
In addition, it will allow to consolidate the addition of debugfs
entries to a single location in the code.
This implementation adds infrastructure to create (and remove) debugfs
entries at two different stages. The first, done when a device is probed,
allows to keep debugfs entries persistent between a transition in device
state (up to down or vice versa). The second, done after the initialization
phase, allows to have debugfs entries that are accessible only when
the device is up.
In addition, move the creation of debugfs entries for configuration
to the newly created function adf_dbgfs_init() and replace symbolic
permissions with octal permissions when creating the debugfs files.
This is to resolve the following warning reported by checkpatch:
WARNING: Symbolic permissions 'S_IRUSR' are not preferred. Consider using octal permissions '0400'.
Signed-off-by: Damian Muszynski <damian.muszynski@intel.com>
Reviewed-by: Giovanni Cabiddu <giovanni.cabiddu@intel.com>
Signed-off-by: Giovanni Cabiddu <giovanni.cabiddu@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
