Driver must fetch FW hard reset capability at every FW boot stage:
preboot, CPU boot, CPU application.
If hard reset is triggered, driver will take into consideration
only the last capability received.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
In case the clock gating was enabled in preboot we need to disable it
at the H/W initialization stage before touching the MME/TPC registers.
Otherwise, the ASIC can get stuck. If the security is enabled in
the firmware level, the CGM is always disabled and the driver can't
enable it.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
hw_queues_mirror was renamed to cs_mirror, so revise accordingly a
comment that refers to this list.
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
We don't need to set EB on signal packets from collective slave
queues as it degrades performance. Because the slaves are the network
queues, the engine barrier doesn't actually guarantee that the
packet has been sent.
Signed-off-by: Alon Mizrahi <amizrahi@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
FW hard reset capability indication is now moved to preboot stage.
Driver will check if HW is dirty only after it validated preboot
is up. If HW is dirty, driver will perform a hard reset according
to the FW capability.
In addition, FW defines a new message which driver need to send in
order to initiate a hard reset.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
As we only fetch the CPU_PLL frequency in gaudi, we don't need a
generic get_pll_frequency function which takes a pll index as input
Signed-off-by: Alon Mizrahi <amizrahi@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
When the F/W security is enabled, goya needs to fetch the PSOC pll
frequency through a dedicated interface
Signed-off-by: Alon Mizrahi <amizrahi@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
We want the fixes in here, and this resolves a merge issue with
drivers/misc/habanalabs/common/memory.c.
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Add a new CB IOCTL opcode that enables a user to query about a CB and
get its usage count.
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Modify the CS counter of a CB to be atomic, so no locking is required
when it is being modified or read.
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
hl_cs_sanity_checks() extracts the CS type bits of the CS flags, by
masking out the non-type bits.
To save the need for updating the function whenever new bits for
non-type flags are added, add an explicit mask for the CS type bits.
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Some messages should be changed to debug mode as we want to keep
minimal prints during normal operation of the device.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
If huge range is not valid, driver uses the host range also for
huge page allocations, but driver never frees its allocation.
This introduces a memory leak every time a user closes its context.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Currently, if the f/w is in preboot/u-boot they don't perform the new
reset mechanism. Therefore, the driver needs to reset the device.
To prevent reset of PCI_IF, the driver needs to first configure the
reset units.
If the security is enabled, the driver can't configure the reset units.
In that situation, don't reset the card.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
These defines are 64-bit defines so they need ull suffix.
Signed-off-by: Alon Mizrahi <amizrahi@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
add support for user to request a timestamp upon
cs completion.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
We want to indicate to the user that a certain command submission
is finished long time ago and it is no longer in database.
This means no further information regarding this cs can be obtained.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
We have the ECC type field from the firmware but the driver didn't
print it, so we need to add that field to the ECC print message.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Make a function in gaudi.c to be static
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: kernel test robot <lkp@intel.com>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Once firmware security is enabled, driver must fetch pll frequencies
through the firmware message interface instead of reading the registers
directly.
Signed-off-by: Alon Mizrahi <amizrahi@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
We introduce a new wrapper which allows us to mmu map any size
to any host va_range available. In addition we remove duplicated
code from various places in driver and using this new wrapper
instead.
This wrapper supports mapping only contiguous physical
memory blocks and will be used for mappings that are done to the
driver ASID.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
We have several types of command submissions and the user wants to know
which type of command submission has not finished in time when that
event occurs. This is very helpful for debug.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
As part of the security effort in which FW will be handling
sensitive HW registers, hard reset flow will be done by FW
and will be triggered by driver.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
F/W message returns 64bit value but up until now we casted it to
a 32bit variable, instead of receiving 64bit in the first place.
Signed-off-by: Alon Mizrahi <amizrahi@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
After the MMU-code refactoring, the existing MMU debugfs operations
are no longer working so we need to fix them.
In addition, remove the duplicate code that was in the debugfs code
and use the already existing MMU-code.
Signed-off-by: Moti Haimovski <mhaimovski@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Multiple locks are usually a source of problems, which in the MMU
case can be avoided since it is relatively rare that both MMU
tables are updated at the same time.
Therefore, use a single shared lock instead of two separate ones.
Signed-off-by: Moti Haimovski <mhaimovski@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Add support for reserving va block with alignment different than
page size. This is a pre-requisite for allocations needed in future
ASICs
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Add log prints for security and eFuse boot error bits
Signed-off-by: Guy Nisan <gnisan@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
During hard-reset, the driver rejects further IOCTL calls and prints
an error message. That error message should be printed with the correct
device instead of using only the control device.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Once FW security is enabled there is no access to HBM ecc registers,
need to read values from FW using a dedicated interface.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Driver must fetch FW hard reset capability during boot time,
in order to skip the hard reset flow if necessary.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Whether an ASIC has MMU towards its DRAM is an ASIC property, so
move it to the asic fixed properties structure.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Instead of using a dedicated va range for each internal pool,
we introduce a new way for reserving a va block from an existing
va range. This is a more generic way of reserving va blocks for
future use.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
We want to handle the scenario in which the driver was not able
to kill all user processes due to many memory mappings.
We need to retry again after some period while releasing the cores.
The devices will be unusable and "in-reset" status during that time.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Future command submission types might be submitted to HW not via the
QMAN queues path. However, it would be still required to have the TDR
mechanism for these CS, and thus the patch renames the TDR fields and
replaces the hw_queues_ prefix with cs_.
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Use an array of va_ranges instead of keeping each va_range separately,
we do this for better readability and in order to support access to
a specific range in a much elegant manner.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Driver must verify if HW is dirty before trying to fetch preboot
information. Hence, we move this validation to a prior stage of
the boot sequence.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Due to using dma_mmap_coherent() to perform mmap of dma memory, we
had to clear the vm_pgoff field before calling that function.
However, that broke the userspace (profiler tool) as they relied
on searching the /proc/self/maps for these values to correctly
"disassemble" the topology recipe.
To re-enable that functionality, the driver can simply restore the
value of vm_pgoff before returning to userspace but after calling
dma_mmap_coherent().
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
The new state indicates that device should be reset in order
to re-gain funcionality.
This unique state can occur if reset_on_lockup is disabled
and an actual lockup has occurred.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
One of the first steps of a hard reset flow is to close all open user
contexts. This user process teradown might take some time due to long
cleanup in our driver or some other reason even before our cleanup flow.
Hence fix the relevant print and comment to be more accurate.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Since the very large grace period is over and this functionality
prevents us to implement the new reset sequence and apply security
settings, we need to remove the code toggling the PCIE_EN bit in the
straps register.
Remove it for good.
Signed-off-by: Igor Grinberg <igrinberg@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
We print twice the firmware status regarding security, once in
common code and once in asic code. Remove the print in asic code
and leave the common code print.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Current CS jobs are no longer needed after their completion.
However, jobs of future workload might be in use even after they are
completed. To allow that, the patch adds a refcount to the job object,
and decouples its completion handling from its deallocation.
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
We need to have the MAX CS be much larger than the size of the
different queues. In GAUDI we have around 8 groups of queues, and each
group has 1K queue size. To prevent head-of-the-line blocking, we need
to make sure there is sufficient number of available CS allocations
even if one or more of those queues are full.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
failure in reading pre-boot verion is not handled correctly,
upon failure we need to reset the device in order to be able
to reinstall the driver.
Signed-off-by: farah kassabri <fkassabri@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Several header files are repeatedly included in many files.
Move these files to habanalabs.h which is included by all.
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
As in standard wait cs, we must release a signal fence once
a collective wait cs was dropped and not submitted.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
There are no internal queues if H/W queues are being used.
In this case we can skip the redundant traversal over the queues array,
looking for internal queues.
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Slightly refactor the cs_do_release() function, to reduce nesting level
and to ease the handling of future CS types.
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Refactor the CS IOCTL handling by gathering common code into
sub-functions, in order to ease future additions of new CS types.
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Once FW security is enabled there is no access to PLL registers,
need to read values from FW using a dedicated interface.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
This commit refactors the MMU code to support PCI MMU page tables
residing on host and DCORE MMU residing on the device DRAM at the
same time.
This is needed for future devices as on GAUDI and GOYA we have
a single MMU where its page tables always reside on DRAM.
Signed-off-by: Moti Haimovski <mhaimovski@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
In cases of multi-tenants, administrators may want to prevent data
leakage between users running on the same device one after another.
To do that the driver can scrub the internal memory (both SRAM and
DRAM) after a user finish to use the memory.
Because in GAUDI the driver allows only one application to use the
device at a time, it can scrub the memory when user app close FD.
In future devices where we have MMU on the DRAM, we can scrub the DRAM
memory with a finer granularity (page granularity) when the user
allocates the memory.
This feature is not supported in Goya.
To allow users that want to debug their applications, we add a kernel
module parameter to load the driver with this feature disabled.
Signed-off-by: farah kassabri <fkassabri@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Skip relevant HW configurations once FW security is enabled
because these configurations are being performed by FW.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Add support for fetching security indication from FW.
This indication is needed in order to skip unnecessary
initializations done by FW.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Fix cs counters structure in uapi to be one flat structure instead
of two instances of the same other structure.
use atomic read/increment for context counters so we could use
one structure for both aggregated and context counters.
Signed-off-by: farah kassabri <fkassabri@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Today driver is able to load a whole FW binary into a specific
location on ASIC. We add support for loading sections from the
same FW binary into different loactions.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Although we get a valid cs type from the callee, in case new values
will be added in the future, it is best to check the expected values
in that function.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
In GAUDI we don't have an MMU towards the HBM device memory. Therefore,
the user access that memory directly through physical address (via the
different engines) without the need to go through the driver to
allocate/free memory on the HBM.
For system monitoring purposes, the driver will keep track of the HBM
usage. This can be done as long as the user accurately reports the
allocations and releases of HBM memory, through the existing MEMORY
IOCTL uapi.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Implement sync stream collective for GAUDI. Need to allocate additional
resources for that and add ctx_fini() to clean up those resources.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
DMA5 QMAN is designated to be used for reduction process, hence it will
be no longer configured as external queue.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Define new API for collective wait support and modify sync stream
common flow. In addition add kernel CB allocation support for
internal queues.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
In the future there will be situations where queues can accept either
kernel allocated CBs or user allocated CBs, depending on different
states.
Therefore, instead of using a boolean variable of kernel/user allocated
CB, we need to use a bitmask to indicate that, which will allow to
combine the two options.
Add a flag to the uapi so the user will be able to indicate whether
the CB was allocated by kernel or by user. Of course the driver
validates that.
Signed-off-by: Tal Cohen <talcohen@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Initialize the QMANs that are responsible to submit doorbells to the NIC
engines. Add support for stopping and disabling them, and reset them as
part of the hard-reset procedure of GAUDI. This will allow the user to
submit work to the NICs.
Add support for receiving events on QMAN errors from the firmware.
However, the nic_ports_mask is still initialized to 0. That means this code
won't initialize the QMANs just yet. That will be in a later patch.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Configure the security properties of the NIC IP. This is to prevent the
user process from doing something with the NIC that he shouldn't do. e.g.
crash the server, steal data, etc.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Add new structures and messages that the driver use to interact with the
firmware to receive information and events (errors) about GAUDI's NIC.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Add auto-generated header files that describe the NIC QMANs registers
used by the driver.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
We already check if queue index is smaller than max queues a few lines
above this check so no need to check this again.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Support advanced monitor functionality to monitor more than a
single SOB. In addition expand all CB generation functions
with buffer offset in order to put in them multiple packets that are
generated by different functions.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
In case we are running without MMU enabled (debug mode), no need to
initialize the VM module in the driver.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
No need to print when the driver starts to initialize the H/W. Drivers
should be silent when everything is OK.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
The driver now loads the firmware in two stages. For debugging purposes
we need to support situations where only the first stage firmware is
loaded.
Therefore, use a bitmask to determine which F/W is loaded
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
F/W can be loaded but device CPU queues disabled. In that case, HWMON
should be disabled. This is only relevant when debugging
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Currently mmu_prepare is located at context switch.
Since we support a single context, no reason to reconfigure
the MMU registers every context switch.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
In case we will have multiple contexts/processes, we can't just
increment aggregated counters. We need to make them atomic as they can
be incremented by multiple processes
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Driver never puts its device and control_device objects, hence
a memory leak is introduced every driver removal.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
If huge range is not valid, driver uses the host range also for
huge page allocations, but driver never frees its allocation.
This introduces a memory leak every time a user closes its context.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
There is missing statement and missing "break;" in the ECC handling
code in gaudi.c
This will cause a wrong behavior upon certain ECC interrupts.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
This interrupt cause is not relevant because of how the user use the
QMAN arbitration mechanism. We must mask it as the log explodes with it.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
We must relocate the coresight mmu configuration to the coresight
flow to make it work in case the first submission is to configure
the profiler.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
All throughout the driver, normal kernel pointers are
stored as 'u64' struct members, which is kind of silly
and requires casting through a uintptr_t to void* every
time they are used.
There is one line that missed the intermediate uintptr_t
case, which leads to a compiler warning:
drivers/misc/habanalabs/common/command_buffer.c: In function 'hl_cb_mmap':
drivers/misc/habanalabs/common/command_buffer.c:512:44: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
512 | rc = hdev->asic_funcs->cb_mmap(hdev, vma, (void *) cb->kernel_address,
Rather than adding one more cast, just fix the type and
remove all the other casts.
Fixes: 0db575350c ("habanalabs: make use of dma_mmap_coherent")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
LDMA registers are configured with a fixed value.
We add new define set which gives the configuration
a proper meaning.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
The device should be idle after a context is closed. If not, print a
notice.
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
During debugging of error we sometimes need to know whether the error
happened when a user context was open. Add debug prints when opening and
closing user contexts.
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Some engines use resources that belong to the kernel context (e.g. MMU
mappings). In case the halt-engines doesn't work properly due to H/W
restriction, we need to make sure the kernel context lives on until after
the hw_fini. The hw_fini resets the ASIC after that no engine is alive and
we can safely close the kernel context.
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
We don't try to allocate huge pages here so remove the huge word.
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Our firmware use some scratchpad registers in the device for different
roles. Update the file to the latest version of the firmware code.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Future F/W versions will have enhanced security measures and the driver
won't be able to do certain configurations that it always did and those
configurations will be done by the firmware.
We use the firmware's preboot version to determine whether security
measures are enabled or not. Because we need this very early in our code,
the read of the preboot version is moved to the earliest possible place,
right after the device's PCI initialization.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
This is a workaround for H/W bug H3-2116, where if there are more than 16
outstanding completions in the DMA transpose engine, there can be a
deadlock in the engine.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Add new packet to fetch PLL information from firmware. This will be needed
in the future when the driver won't be able to access the PLL registers
directly
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
There are cases in which the device should access the host memory of a
CB through the device MMU, and thus this memory should be mapped.
The patch adds a flag to the CB IOCTL, in which a user can ask the
driver to perform the mapping when creating a CB.
The mapping is allowed only if a dedicated VA range was allocated for
the specific ASIC.
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Future changes require using a context while handling a command buffer,
and thus need to save the context in the command buffer object.
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Now that the driver no longer uses dma_buf, we can remove the select of
DMA_SHARED_BUFFER from kconfig.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
The user sometimes wants to check if a CS has completed to clean resources.
In that case, the user doesn't want to sleep but just to check if the CS
has finished and continue with his code.
Add a new definition to the API of the wait on CS. The new definition says
that if the timeout is 0, the driver won't sleep at all but return
immediately after checking if the CS has finished.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
The firmware running in the boot stage takes more time to execute due to
increased security mechanisms. Therefore, we need to increase the timeout
we wait for the boot fit to finish loading.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
This commit modify the existing debugfs code to support future devices that
have a 6 HOPs MMU implementation instead of 5 HOPs implementation.
Signed-off-by: Moti Haimovski <mhaimovski@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
This commit adds the number of HOPs supported by the device to the
device MMU properties.
Signed-off-by: Moti Haimovski <mhaimovski@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
As preparation to MMU v2, rework MMU to be device oriented
instantiated according to the device in hand.
Signed-off-by: Moti Haimovski <mhaimovski@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
In the future we will have MMU v2 code, so we need to prepare the
driver for it. The first step is to rename the current MMU file to
mmu_v1.c.
Signed-off-by: Moti Haimovski <mhaimovski@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Change the acquiring of a device virtual address for mapping by using the
smallest possible alignment, rather than the biggest, depending on the
page size used by the user for allocating the memory. This will lower the
virtual space memory consumption.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
For consistency with GAUDI code, add check of the relevant flag in the
device structure before resetting the GOYA device in case of firmware
event.
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
For future ASICs, we increase this field by one nibble. This field was not
used by the current ASICs so this change doesn't break anything.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Because the device CPU compiler aligns structures to 8 bytes,
struct cpucp_info has an alignment issue as some parts
in the structure are not aligned to 8 bytes.
It is preferred that we explicitly insert placeholders inside
the structure to avoid confusion
in order to validate this scenario, we printed both pointers:
__u8 cpucp_version[VERSION_MAX_LEN]; (0xffff899c67ed4cbc)
__le64 dram_size; (0xffff899c67ed4d40)
we see difference of 132 bytes although the first array
is only 128 bytes long, Meaning compiler added a 4 byte padding.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
There were a couple of comments where the name ArmCP was still used. Rename
it to CPU-CP.
In addition, rename ArmCP or ARM in log messages to "device CPU".
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
There is a case where the user reaches the maximum number of CS in-flight.
In that case, the driver rejects the new CS of the user with EAGAIN. Count
that event so the user can query the driver later to see if it happened.
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Add dma_mmap_coherent() for goya and gaudi to match their use of
dma_alloc_coherent(), see the Link tag for why.
Link: https://lore.kernel.org/lkml/20200609091727.GA23814@lst.de/
Cc: Christoph Hellwig <hch@lst.de>
Cc: Zhang Li <li.zhang@bitmain.com>
Cc: Ding Z Nan <oshack@hotmail.com>
Signed-off-by: Hillf Danton <hdanton@sina.com>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
The driver use vm_pgoff to hold the CB idr handle. Before we actually call
the mapping function, we need to clear the handle so there won't be any
garbage left in vm_pgoff.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Arrange the hl_mmap code to be more structured and expandable for the
future. Add better defines that describe our usage of the vm_pgoff.
Note that I shamelessly took the code and defines from the amdkfd driver
(my previous driver).
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
When shifting a boolean variable by more than 31 bits and putting the
result into a u64 variable, we need to cast the boolean into unsigned 64
bits to prevent possible overflow.
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
ArmCP mandates that the device CPU is always an ARM processor, which might
be wrong in the future.
Most of this change is an internal renaming of variables, functions and
defines but there are two entries in sysfs which have armcp in their
names. Add identical cpucp entries but don't remove yet the armcp entries.
Those will be deprecated next year. Add the documentation about it in sysfs
documentation.
Signed-off-by: Moti Haimovski <mhaimovski@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Add driver implementation for reading the total energy consumption
from the device ARM FW.
Signed-off-by: farah kassabri <fkassabri@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Include linux/bitfield.h only in habanalabs.h, instead of in each and
every file that needs it, as habanalabs.h is already included by all.
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
change busy engines bitmask to 64 bits in order to represent
more engines, needed for future ASIC support.
Signed-off-by: farah kassabri <fkassabri@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Eliminate following warning:
warning: Shifting signed 32-bit value by 31 bits is undefined behavior
Reported-by: kernel test robot <lkp@intel.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
The driver waits for the TPC vector pipe to be empty before checking if the
TPC kernel has finished executing, but the code doesn't validate that the
pipe was indeed empty, it just wait for it without checking the return
value.
Reported-by: kernel test robot <lkp@intel.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
new_dma_pkt->ctl is assigned a value and then is reassigned a new value
without the first value ever being used.
Reported-by: kernel test robot <lkp@intel.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Use the standard FIELD_PREP() macro instead of << operator to perform
bitmask operations. This ensures type check safety and eliminate compiler
warnings.
Reported-by: kernel test robot <lkp@intel.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Use the standard macros to define bitmasks.
Reported-by: kernel test robot <lkp@intel.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
If both parts of if-else are goto statements, we can remove the else and
put the else goto statement after the if statement.
Reported-by: kernel test robot <lkp@intel.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
%u is used for unsigned so we need to cast the int variable to u32 to avoid
compiler warning.
Reported-by: kernel test robot <lkp@intel.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Although the possible values for CB's ID are only 32 bits, there are a few
places in the code where this field is shifted and passed into a function
which expects 64 bits.
Reported-by: kernel test robot <lkp@intel.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
If there is a failure during the testing of a queue,
to ease up debugging - print the queue id.
Signed-off-by: Dotan Barak <dbarak@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Allow user application to write to this register in order
to be able to configure the quiet period of the QMAN between grants.
Signed-off-by: farah kassabri <fkassabri@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
driver will now get notified upon any PCI error occurred and
will respond according to the severity of the error.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Although the driver defines the first user-available sync manager object
and monitor in habanalabs.h, we would like to also expose this information
via the INFO IOCTL so the runtime can get this information dynamically.
This is because in future ASICs we won't need to define it statically.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Update firmware header with new API for getting pcie info
such as tx/rx throughput and replay counter.
These counters are needed by customers for monitor and maintenance
of multiple devices.
Add new opcodes to the INFO ioctl to retrieve these counters.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
habanalabs driver uses dma-fence mechanism for synchronization.
dma-fence mechanism was designed solely for GPUs, hence we purpose
a simpler mechanism based on completions to replace current
dma-fence objects.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Future ASIC names are longer than 15 chars so increase the variable length
to 32 chars.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
All initiator coordinates received upon an 'MMU page fault RAZWI
event' should be the routers coordinates, the only exception is the
DMA initiators for which the reported coordinates correspond to
their actual location.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
This commit fixes a potential debugfs issue that may occur when
reading the clock gating mask into the user buffer since the
user buffer size was not taken into consideration.
Signed-off-by: Moti Haimovski <mhaimovski@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
During inbound iATU configuration we can get errors while
configuring PCI registers, there is a certain scenario in which these
errors are not reflected and driver is loaded with wrong configuration.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
vmalloc can return different return code than NULL and a valid
pointer. We must validate it in order to dereference a non valid
pointer.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
We must validate FW size in order not to corrupt memory in case
a malicious FW file will be present in system.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
The null check on a failed workqueue create is currently null checking
hdev->cq_wq rather than the pointer hdev->cq_wq[i] and so the test
will never be true on a failed workqueue create. Fix this by checking
hdev->cq_wq[i].
Addresses-Coverity: ("Dereference before null check")
Fixes: 5574cb2194 ("habanalabs: Assign each CQ with its own work queue")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
In Gaudi, the default max power setting is different between PCI and PMC
cards. Therefore, the driver need to set the default after knowing what is
the card type.
The current code has a bug where it limits the maximum power of the PMC
card to 200W after a reset occurs.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Allocation size can go up to 64bit but truncated to 32bit,
we should make sure it is not truncated and validate no address
overflow.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Once clock gating is set we enable clock gating according to mask,
we should also disable clock gating according to relevant bits.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
User input must be validated before using it to
access internal structures.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
The condition was reversed. It should have been less than instead of
greater than. The result is that we never enter the loop.
Fixes: fcc6a4e606 ("habanalabs: Extract ECC information from FW")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
This has to be a long instead of a u32 because we write a long value.
On 64 bit systems, this will cause memory corruption.
Fixes: c216477363 ("habanalabs: add debugfs support")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
During command buffer parsing, driver extracts packet id
from user buffer. Driver must validate this packet id, since it is
being used in order to extract information from internal structures.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
In case the driver fails to configure the PCI controller iATU, it needs to
unmap the PCI bars before exiting so if the driver is removed, the bars
won't be left mapped.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Gcc report warning as follows:
drivers/misc/habanalabs/common/command_submission.c:373:6: warning:
variable 'ctx_asid' set but not used [-Wunused-but-set-variable]
373 | int ctx_asid, rc;
| ^~~~~~~~
This variable is not used in function cs_timedout(), this commit
remove it to fix the warning.
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Link: https://lore.kernel.org/r/20200729155902.33976-1-weiyongjun1@huawei.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There's no need to try to be cute with the include file locations in the
Makefile, so just specify exactly where the files are.
Bonus is this fixes the problem of building with O= as well as trying to
just build the subdirectory alone.
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Omer Shpigelman <oshpigelman@habana.ai>
Cc: Tomer Tayar <ttayar@habana.ai>
Cc: Moti Haimovski <mhaimovski@habana.ai>
Cc: Ofir Bitton <obitton@habana.ai>
Cc: Ben Segal <bpsegal20@gmail.com>
Cc: Christine Gharzuzi <cgharzuzi@habana.ai>
Cc: Pawel Piskorski <ppiskorski@habana.ai>
Link: https://lore.kernel.org/r/20200728171851.55842-1-gregkh@linuxfoundation.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Add a missing free of the cs_pending array in the error flow of context
initialization.
Fixes: c16d45f42b ("habanalabs: Use pending CS amount per ASIC")
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
gaudi_mmu_invalidate_cache() doesn't use the flags parameter, and thus
it can be set to 0 when the function is called in the gaudi only files.
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Device is now enabled before the hw_init() because part of the
initialization requires communication with the device firmware to get
information that is required for the initialization itself
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Create a device MMU-mapped internal command buffer pool, in order to allow
the driver to allocate CBs for the signal/wait operations
that are fetched by the queues when they are configured with the user's
address space ID.
We must pre-map this internal pool due to performance issues.
This pool is needed for future ASIC support and it is currently unused in
GOYA and GAUDI.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Update the boot interface file from the latest version from firmware.
Defines for secure boot were added.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Omer Shpigelman <oshpigelman@habana.ai>
For internal needs of our CI we need to move all the common code into a
common folder instead of putting them in the root folder of the driver.
Same applies to the common header files under include/
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Omer Shpigelman <oshpigelman@habana.ai>
In GAUDI we use QMAN0 DMA for clearing the MMU memory region
at initialization. if this operation fails it places the DMA in an error
state and then when trying to initialize QMAN0 we fail and erroneously
assume its the QMAN that failed.
This commit adds a check and clear of such DMA errors at initialization so
we will have a better understanding of what went wrong.
Signed-off-by: Moti Haimovski <mhaimovski@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
In order for the user to be aware of wrong inputs, we must return
error in case the amount of jobs per cs exceeds the corresponding
queue size.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
We identified a possible race during job completion when working
with a single multi-threaded work queue. In order to overcome this
race we suggest using a single threaded work queue per completion
queue, hence we guarantee jobs completion in order.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Currently the driver halts the device CPU in the halt engines function,
which halts all the engines of the ASIC. The problem is that if later on we
stop the reset process (due to inability to clean memory mappings in time),
the CPU will remain in halt mode. This creates many issues, such as
thermal/power control and FLR handling.
Therefore, move the halting of the device CPU to the very end of the reset
process, just before writing to the registers to initiate the reset. In
addition, the driver now needs to send a message to the device F/W to
disable it from sending interrupts to the host machine because during halt
engines function the driver disables the MSI/MSI-X interrupts.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Remove an old hash that is not in use anymore.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Instead of using the free slots amount on the compute CQ to determine
whether we can submit work to queues, use the queues pi/ci.
This is needed in future ASICs where we don't have CQ per queue.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Currently the amount of maximum queues is statically configured.
Using a static value is causing redundunt cycles when traversing
all queues and consumes more memory than actually needed.
In this patch we configure each asic with the exact number of
queues needed.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Soft-reset isn't supported in GAUDI. Remove the code that performs it and
print error in case the user wants to do it via sysfs.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Divide iATU initialization into inbound/outbound methods.
We must separate it in order to enable different match mode
per PCIe region.
In addition, added support for PCI address match mode.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
ECC (Error Correcting Code) interrupts are going to be handled
by the FW. Hence, we define an interface in which the driver can
obtain the relevant ECC information.
This information is needed for monitoring and can also lead
to a hard reset if ECC error is not correctable.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Add command submission statistics structure which can be obtained
through the info ioctl. Each drop counter describes the reason for
which the command submission was dropped.
This information is needed for the user to be aware of the specific
reason for which the submitted work was dropped. The user can then
utilize the driver more efficiently.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Extract detection of the cpu boot status to a function
to allow code reuse
Signed-off-by: Christine Gharzuzi <cgharzuzi@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
rephrase some error/warning/notice messages to make them more accessible to
ordinary users.
There is no need to print context ASID as the driver currently doesn't
support multiple contexts.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
After recent concurrent cs amount increase, we must also
increase queues depth since much more concurrent work can be done.
All external queue depths were increased to 4096 as gaudi's
internal queue depths were also increased to 1024.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Rephrase F/W error message to make it more understandable to ordinary
users.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
The profiler needs to know the PLL values for correctly showing the
profiling data. Because our firmware can use different PLL configurations,
we need to read the PLL values from the ASIC to pass them to the profiler.
Signed-off-by: Adam Aharon <aaharon@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Once there is a 64-bit field in a structure, GCC compiler for ARM aligns
the structure to 8 bytes. In order to avoid confusion when these
structures are being passed between CPUs from different architectures, we
explicitly align the structure to 8 bytes.
Reviewed-by: Omer Shpigelman <oshpigelman@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Currently sync stream is limited only for external queues. We want to
remove this constraint by adding a new queue property dedicated for sync
stream. In addition we move the initialization and reset methods to the
common code since we can re-use them with slight changes.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Training schemes requires much more concurrent command submissions than
inference does. In addition, training command submissions can be completed
in a non serialized manner. Hence, we add support in which each ASIC will
be able to configure the amount of concurrent pending command submissions,
rather than use a predefined amount. This change will enhance performance
by allowing the user to add more concurrent work without waiting for the
previous work to be completed.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
We no longer need to initialize the rate limiters in GAUDI A1.
Reviewed-by: Omer Shpigelman <oshpigelman@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Queue index is received from the user. Therefore, we must validate it
before using it to access the queue props array.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
We see that sometimes the CPU in GOYA and GAUDI is occupied by the
power/thermal loop and can't answer requests from the driver fast enough.
Therefore, to avoid false notifications on timeouts, increase the timeout
to 4 seconds on each message sent to the device CPU.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
For debugging purposes, we need to allow the root user better control of
the clock gating feature of the DMA and compute engines. Therefore, change
the clock gating debugfs interface to be bitmask instead of true/false.
Each bit represents a different engine, according to gaudi_engine_id enum.
See debugfs documentation for more details.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Omer Shpigelman <oshpigelman@habana.ai>
WREG_BULK is a special packet that has a variable length. Therefore, we
can't parse it when validating CBs that go to the PCI DMA queue. In case
the user needs to use it, it can put multiple WREG32 packets instead.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Omer Shpigelman <oshpigelman@habana.ai>
gaudi_pb_set_block()'s argument 'base' was incorrectly named 'block' in
its function header.
Fixes the following W=1 kernel build warning(s):
drivers/misc/habanalabs/gaudi/gaudi_security.c:454: warning: Function parameter or member 'base' not described in 'gaudi_pb_set_block'
drivers/misc/habanalabs/gaudi/gaudi_security.c:454: warning: Excess function parameter 'block' description in 'gaudi_pb_set_block'
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Link: https://lore.kernel.org/r/20200701085853.164358-11-lee.jones@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
W=1 kernel builds report a lack of description of gaudi_set_asic_funcs()'s
'hdev' argument. In reality it is documented, but the formatting
was not as expected '@.*:'. Instead, there was a misplaced asterisk
which was confusing the kerneldoc validator.
Squashes the following W=1 warning:
drivers/misc/habanalabs/gaudi/gaudi.c:6746: warning: Function parameter or member 'hdev' not described in 'gaudi_set_asic_funcs'
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Link: https://lore.kernel.org/r/20200701085853.164358-10-lee.jones@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
No attempt to check the return value of RREG32() has been made
since the call was introduced a year ago.
Fixes W=1 kernel build warning:
drivers/misc/habanalabs/goya/goya_coresight.c: In function ‘goya_debug_coresight’:
drivers/misc/habanalabs/goya/goya_coresight.c:643:6: warning: variable ‘val’ set but not used [-Wunused-but-set-variable]
643 | u32 val;
| ^~~
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Link: https://lore.kernel.org/r/20200701085853.164358-9-lee.jones@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
'dma_mask' is not passed directly into hl_pci_set_dma_mask() as
an argument. Instead, it is pulled from struct hl_device *hdev.
Fixed the following W=1 warning:
drivers/misc/habanalabs/pci.c:328: warning: Excess function parameter 'dma_mask' description in 'hl_pci_set_dma_mask
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Link: https://lore.kernel.org/r/20200701085853.164358-8-lee.jones@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Seeing as 'addr' is unsigned, it would be impossible for the assigned
value to be anything other than zero or positive.
Squashes the following W=1 warnings:
drivers/misc/habanalabs/goya/goya.c: In function ‘goya_debugfs_read32’:
drivers/misc/habanalabs/goya/goya.c:3945:19: warning: comparison of unsigned expression >= 0 is always true [-Wtype-limits]
3945 | } else if ((addr >= DRAM_PHYS_BASE) &&
| ^~
drivers/misc/habanalabs/goya/goya.c: In function ‘goya_debugfs_write32’:
drivers/misc/habanalabs/goya/goya.c:4002:19: warning: comparison of unsigned expression >= 0 is always true [-Wtype-limits]
4002 | } else if ((addr >= DRAM_PHYS_BASE) &&
| ^~
drivers/misc/habanalabs/goya/goya.c: In function ‘goya_debugfs_read64’:
drivers/misc/habanalabs/goya/goya.c:4047:19: warning: comparison of unsigned expression >= 0 is always true [-Wtype-limits]
4047 | } else if ((addr >= DRAM_PHYS_BASE) &&
| ^~
drivers/misc/habanalabs/goya/goya.c: In function ‘goya_debugfs_write64’:
drivers/misc/habanalabs/goya/goya.c:4091:19: warning: comparison of unsigned expression >= 0 is always true [-Wtype-limits]
4091 | } else if ((addr >= DRAM_PHYS_BASE) &&
| ^~
drivers/misc/habanalabs/pci.c:328: warning: Excess function parameter 'dma_mask' description in 'hl_pci_set_dma_mask'
drivers/misc/habanalabs/goya/goya_coresight.c: In function ‘goya_debug_coresight’:
drivers/misc/habanalabs/goya/goya_coresight.c:643:6: warning: variable ‘val’ set but not used [-Wunused-but-set-variable]
643 | u32 val;
| ^~~
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Link: https://lore.kernel.org/r/20200701085853.164358-7-lee.jones@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
W=1 kernel builds report a lack of descriptions for various
function arguments. In reality they are documented, but the
formatting was not as expected '@.*:'. Instead, '-'s were
used as separators.
While we're here, the headers for functions various functions
were written in kerneldoc format, but lack the kerneldoc
identifier '/**'. Let's promote them so they can gain access
to the checker.
This change fixes the following W=1 warnings:
drivers/misc/habanalabs/irq.c:24: warning: Function parameter or member 'eq_work' not described in 'hl_eqe_work'
drivers/misc/habanalabs/irq.c:24: warning: Function parameter or member 'hdev' not described in 'hl_eqe_work'
drivers/misc/habanalabs/irq.c:24: warning: Function parameter or member 'eq_entry' not described in 'hl_eqe_work'
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Link: https://lore.kernel.org/r/20200701085853.164358-6-lee.jones@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
hl_pci_bars_map() has a miss-typed argument name in the function
description. hl_pci_elbi_write() was missing documented arguments.
The headers for functions hl_pci_bars_unmap(), hl_pci_elbi_write()
and hl_pci_reset_link_through_bridge() were written in kerneldoc
format, but lack the kerneldoc identifier '/**'. Let's promote
them so they can gain access to the checker.
These changes fix the following W=1 kernel build warnings:
drivers/misc/habanalabs/pci.c:27: warning: Function parameter or member 'name' not described in 'hl_pci_bars_map'
drivers/misc/habanalabs/pci.c:27: warning: Excess function parameter 'bar_name' description in 'hl_pci_bars_map'
drivers/misc/habanalabs/pci.c:147: warning: Function parameter or member 'addr' not described in 'hl_pci_iatu_write'
drivers/misc/habanalabs/pci.c:147: warning: Function parameter or member 'data' not described in 'hl_pci_iatu_write'
drivers/misc/habanalabs/pci.c:324: warning: Excess function parameter 'dma_mask' description in 'hl_pci_set_dma_mask'
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Link: https://lore.kernel.org/r/20200701085853.164358-5-lee.jones@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Looks as though documentation for these function arguments have
been missing since the driver's inception last year.
Fixes the following W=1 kernel build warnings:
drivers/misc/habanalabs/firmware_if.c:26: warning: Function parameter or member 'fw_name' not described in 'hl_fw_load_fw_to_device'
drivers/misc/habanalabs/firmware_if.c:26: warning: Function parameter or member 'dst' not described in 'hl_fw_load_fw_to_device'
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Link: https://lore.kernel.org/r/20200701085853.164358-4-lee.jones@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In kerneldoc format, data structures have to start with 'struct'
else the kerneldoc tooling/parsers/validators get confused.
Squashes the following W=1 warning:
drivers/misc/habanalabs/irq.c:19: warning: cannot understand function prototype: 'struct hl_eqe_work '
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Link: https://lore.kernel.org/r/20200626130525.389469-10-lee.jones@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In GAUDI the current timer value for the hardware to check if it is
in IDLE state is too low. As a result, there are occasions where the H/W
wrongly reports it is not IDLE. The driver checks that before submitting
work on behalf of the driver during initialization, so a false report might
cause the driver to fail during device initialization.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
The fence release flow is different if the CS was never submitted. In that
case, we don't have an hw_sob object attached that we need to "put". While
if the CS was aborted, we do need to "put" the hw_sob.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
The current timeout is too low for some of the workloads and we see false
errors as a result.
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
The function name conflicts with a static inline function in
arch/m68k/include/asm/mcfmmu.h
Reported-by: kernel test robot <lkp@intel.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
The PS flow for MMU cache invalidation caused timeouts in stress tests.
Use PS + PI flow so no timeouts should happen whatsoever.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
In Gaudi, the user can't execute scalar load_and_exe on external queue
because it can be a security hole. The driver doesn't parse the commands
being loaded and it can be msg_prot, which the user isn't allowed to use.
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
MMU cache invalidation timeout indicates that the device is unstable and
therefore unusable.
Hence in such case do hard reset and return an error to the user if was
called from ioctl.
In addition, change the print to error level and rephrase its text.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
When the MMU is heavily used by the engines, unmapping might take a lot of
time due to a full MMU cache invalidation done as part of the unmap flow.
Hence we might not be able to kill all open processes before going to hard
reset the device, as it involves unmapping of all user memory.
In case of a failure in killing all open processes, we should stop the
hard reset flow as it might lead to a kernel crash - one thread (killing
of a process) is updating MMU structures that other thread (hard reset) is
freeing.
Stopping a hard reset flow leaves the device as nonoperational and the
user can then initiate a hard reset via sysfs to reinitialize the device.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
GAUDI does not support soft-reset as it leaves the NIC ports in an awkward
state, where their QMANs were reset but the NIC itself is still working.
In addition, there is not much sense in doing soft-reset when training is
done on multiple GAUDIs.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
A new sequence is introduced to invalidate the MMU cache in order to avoid
timeouts.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
It's the default.
Also so much for "we're not going to tell the graphics people how to
review their code", dma_fence is a pretty core piece of gpu driver
infrastructure. And it's very much uapi relevant, including piles of
corresponding userspace protocols and libraries for how to pass these
around.
Would be great if habanalabs would not use this (from a quick look
it's not needed at all), since open source the userspace and playing
by the usual rules isn't on the table. If that's not possible (because
it's actually using the uapi part of dma_fence to interact with gpu
drivers) then we have exactly what everyone promised we'd want to
avoid.
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
The patch_cb_size is not updated for Wreg32 in its validate function, so
updated in goya_validate_cb.
Signed-off-by: Rachel Stahl <rstahl@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Instead of writing similar event handling code for each ASIC, move the code
to the common firmware file. This code will be used for GAUDI and all
future ASICs.
In addition, add two new fields to the auto-generated events file: valid
and description. This will save the need to manually write the events
description in the source code and simplify the code.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Enable the GAUDI ASIC code in the pci probe callback of the driver so the
driver will handle GAUDI ASICs.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Add the GAUDI code to initialize the ASIC's profiler. The profile receives
its initialization values from the user, same as in Goya, but the code to
initialize is in the driver because the configuration space of the
device is not directly exposed to the user.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Add the code to initialize the security module of GAUDI. Similar to Goya,
we have two dedicated mechanisms for security: Range Registers and
Protection bits. Those mechanisms protect sensitive memory and
configuration areas inside the device.
In addition, in Gaudi we moved to a 3-level security scheme, where the F/W
runs with the highest security level (Privileged), the driver runs with a
less secured level (Secured) and the user is neither privileged nor
secured. The security module in the driver configures the Secured parts so
the user won't be able to access them. The Privileged parts are configured
by the F/W.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
The hwmgr module is responsible for messages sent to GAUDI F/W that are
not common to all habanalabs ASICs.
In GAUDI, we provide the user a simplified mode of controlling the ASIC
clock frequency. Instead of three different clocks, we present a single
clock property that the user can configure via sysfs.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Add the ASIC-dependent code for GAUDI. Supply (almost) all of the function
callbacks that the driver's common code need to initialize, finalize and
submit workloads to the GAUDI ASIC.
It also contains the code to initialize the F/W of the GAUDI ASIC and to
receive events from the F/W.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Add the relevant GAUDI ASIC registers header files. These files are
generated automatically from a tool maintained by the VLSI engineers.
There are more files which are not upstreamed because only very few defines
from those files are used in the driver. For those files, we copied the
relevant defines into gaudi_regs.h and gaudi_masks.h, to reduce the size of
this patch.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
For Gaudi the driver gets two new additional properties from the F/W:
1. The card's type - PCI or PMC
2. The card's location in the Gaudi's box (relevant only for PMC).
The card's location is also passed to the user in the HW IP info structure
as it needs this property for establishing communication between Gaudis.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
In Gaudi there is a feature of clock gating certain engines.
Therefore, add this property to the device structure.
In addition, due to a limitation of this feature, the driver needs to
dynamically enable or disable this feature during run-time. Therefore, add
ASIC interface functions to enable/disable this function from the common
code.
Moreover, this feature must be turned off when the user wishes to debug the
ASIC by reading/writing registers and/or memory through the driver's
debugfs. Therefore, add an option to enable/disable clock gating via the
debugfs interface.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
For Gaudi, the driver doesn't change the PM profile automatically due to
device-controlled PM capabilities. Therefore, set the PM profile to auto
only for Goya so the driver's code to automatically change the profile
won't run on Gaudi.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Gaudi requires longer waiting during reset due to closing of network ports.
Add this explanation to the relevant comment in the code and add a
dedicated define for this reset timeout period, instead of multiplying
another define.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Coresight is not supported on simulator, therefore add a boolean for
checking that (currently used by un-upstreamed code).
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Add the following two operations to the CS IOCTL:
Signal:
The signal operation is basically a command submission, that is created by
the driver upon user request. It will be implemented using a dedicated PQE
that will increment a specific SOB. There will be a new flag:
HL_CS_FLAGS_SIGNAL. When the user set this flag in the CS IOCTL structure,
the driver will execute a dedicated code path that will prepare this
special PQE and submit it. The user only needs to provide a queue index on
which to put the signal.
Wait:
The wait operation is also a command submission that is created by the
driver upon user request. It will be implemented using a dedicated PQE that
will contain packets of "ARM a monitor" + FENCE packet. There will be a new
flag: HL_CS_FLAGS_WAIT. When the user set this flag in the CS structure,
the driver will execute a dedicated code path that will prepare this
special PQE and submit it.
The user needs to provide the following parameters:
1. queue ID
2. an array of signal_seq numbers and the number of signals to wait on
(the length of signal_seq_arr).
The IOCTL will return the CS sequence number of the wait it put on the
queue ID.
Currently, the code supports signal_seq_nr==1. But this API definition will
allow us to put a single PQE that waits on multiple signals.
To correctly configure the monitor and fence, the driver will need to
retrieve the specified signal CS object that contains the relevant SOB and
its expected value. In case the signal CS has already been completed, there
is no point of adding a wait operation. In this case, the driver will
return to the user *without* putting anything on the PQ. The return code
should reflect to the user that the signal was completed, as we won't
return a CS sequence number for this wait.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Define a structure representing the h/w sync object (SOB).
a SOB can contain up to 2^15 values. Each signal CS will increment the SOB
by 1, so after some time we will reach the maximum number the SOB can
represent. When that happens, the driver needs to move to a different SOB
for the signal operation.
A SOB can be in 1 of 4 states:
1. Working state with value < 2^15
2. We reached a value of 2^15, but the signal operations weren't completed
yet OR there are pending waits on this signal. For the next submission, the
driver will move to another SOB.
3. ALL the signal operations on the SOB have finished AND there are no more
pending waits on the SOB AND we reached a value of 2^15 (This basically
means the refcnt of the SOB is 0 - see explanation below). When that
happens, the driver can clear the SOB by simply doing WREG32 0 to it and
set the refcnt back to 1.
4. The SOB is cleared and can be used next time by the driver when it needs
to reuse an SOB.
Per SOB, the driver will maintain a single refcnt, that will be initialized
to 1. When a signal or wait operation on this SOB is submitted to the PQ,
the refcnt will be incremented. When a signal or wait operation on this SOB
completes, the refcnt will be decremented. After the submission of the
signal operation that increments the SOB to a value of 2^15, the refcnt is
also decremented.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
This feature requires handling h/w resources which are a bit different from
one ASIC to the other. Therefore, we need to define a set of interfaces the
ASIC code provides to the common code to signal, wait, reset sync object
and to reset and init a queue.
As this feature is not supported in Goya, provide an empty implementation
of those functions.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
This is a pre-requisite to upstreaming GAUDI support.
Signal/wait operations are done by the user to perform sync between two
Primary Queues (PQs). The sync is done using the sync manager and it is
usually resolved inside the device, but sometimes it can be resolved in the
host, i.e. the user should be able to wait in the host until a signal has
been completed.
The mechanism to define signal and wait operations is done by the driver
because it needs atomicity and serialization, which is already done in the
driver when submitting work to the different queues.
To implement this feature, the driver "takes" a couple of h/w resources,
and this is reflected by the defines added to the uapi file.
The signal/wait operations are done via the existing CS IOCTL, and they use
the same data structure. There is a difference in the meaning of some of
the parameters, and for that we added unions to make the code more
readable.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
PCI drivers should use this define to declare their PCI ID table.
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Make all the CB handles printed in the same way and not some as decimal and
some as hex numbers.
Signed-off-by: Dotan Barak <dbarak@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Update the mapping to the latest one used by the Firmware. No impact on the
driver in this update.
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Set the STMTCSR.COMPEN bit to enable leading-zero trace data
compression functionality for the extended stimulus ports.
Signed-off-by: Adam Aharon <aaharon@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Load CPU device boot loader during driver boot time in order to avoid flash
write for every boot loader update.
To preserve backward-compatibility, skip the device boot load if the device
doesn't request it.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
The user must leave space for 2xMSG_PROT in the external CB, so adjust the
define of max size accordingly. The driver, however, can still create a CB
with the maximum size of 2MB. Therefore, we need to add a check
specifically for the user requested size.
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Align the protection bits configuration of all TPC cores to be as of TPC
core 0.
Fixes: a513f9a7ec ("habanalabs: make tpc registers secured")
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Allow user access to TPC LFSR register, as it might be accessed by TPC
kernels.
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Add a new opcode to the INFO IOCTL that retrieves the device time
alongside the host time, to allow a user application that want to measure
device time together with host time (such as a profiler) to synchronize
these times.
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
set function to be static as it is not called from outside its file.
Signed-off-by: kbuild test robot <lkp@intel.com>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
When we have DMA QMAN with multiple streams, we need to know whether the
command buffer contains at least one DMA packet in order to configure the
barriers correctly when adding the 2xMSG_PROT at the end of the JOB. If
there is no DMA packet, then there is no need to put engine barrier. This
is relevant only for GAUDI as GOYA doesn't have streams so the engine can't
be busy by another stream.
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Retrieve from the firmware the DMA mask value we need to set according to
the device's PCI controller configuration. This is needed when working on
POWER9 machines, as the device's PCI controller is configured in a
different way in those machines.
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Add comments for the various errors and states of the firmware during boot.
Add a mapping of a new register that will tell the driver whether the
firmware executed the request from the driver or if it has encountered an
error.
Add a new enum for the possible values of this register.
Reviewed-by: Omer Shpigelman <oshpigelman@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
When doing training, the DL framework (e.g. tensorflow) performs hundreds
of thousands of memory allocations and mappings. In case the driver needs
to perform hard-reset during training, the driver kills the application and
unmaps all those memory allocations. Unfortunately, because of that large
amount of mappings, the driver isn't able to do that in the current timeout
(5 seconds). Therefore, increase the timeout significantly to 30 seconds
to avoid situation where the driver resets the device with active mappings,
which sometime can cause a kernel bug.
BTW, it doesn't mean we will spend all the 30 seconds because the reset
thread checks every one second if the unmap operation is done.
Reviewed-by: Omer Shpigelman <oshpigelman@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
When the system administrator asks the driver to soft or hard reset the
device through sysfs, the driver should display a warning in the kernel log
to explain why it suddenly resets the device.
Reviewed-by: Omer Shpigelman <oshpigelman@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Move the code of device CPU initialization from being ASIC-Dependent to
common code. In addition, add support for the new error reporting feature
of the firmware boot code.
Reviewed-by: Omer Shpigelman <oshpigelman@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
We want to remove the following restrictions/assumptions in our driver:
1. The H/W queue index is also the completion queue index.
2. The H/W queue index is also the IRQ number of the completion queue.
3. All queues of the same type have consecutive indexes.
Therefore we add the support for H/W queues of the same type with
nonconsecutive indexes and completion queue index and IRQ number different
than the H/W queue index.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Stop-on-error mode in DMA is useful as it stops the transaction
immediately upon error e.g. page fault.
But it may cause the next command submission to fail as is leaves the DMA
in unstable state.
Therefore we remove the stop-on-error configuration from the DMA.
Stop-on-err is still available for debug.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Upon reset of the ASIC, the driver would have waited for the CPU to come
out of reset before finishing the reset process. This was done for the
purpose of making the CPU available to answer FLR requests. However, when a
VM shuts down, the driver isn't removed so a reset never happens.
Therefore, remove this waiting period as we don't need it.
Reviewed-by: Omer Shpigelman <oshpigelman@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
When moving from manual to automatic power management mode in GOYA, the
driver didn't correctly place the device in LOW power mode. As a result, if
an application was run immediately after the move, it would have run with
low frequencies.
Reviewed-by: Omer Shpigelman <oshpigelman@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
If a GAUDI device is present in the system, display an error message that
it is not supported by the current kernel.
Reviewed-by: Omer Shpigelman <oshpigelman@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Add print upon clock slow down due to power consumption or overheating.
In addition, add print when back to optimal clock.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Sparse reports a warning at goya_hw_queues_unlock()
warning: context imbalance in goya_hw_queues_unlock() - unexpected unlock
The root cause is a missing annotation at goya_hw_queues_unlock()
Add the missing __releases(&goya->hw_queues_lock) annotation
Signed-off-by: Jules Irenge <jbi.octave@gmail.com>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Sparse reports a warning at goya_hw_queues_lock()
warning: context imbalance in goya_hw_queues_lock() - wrong count at exit
The root cause is a missing annotation at goya_hw_queues_lock()
Add the missing __acquires(&goya->hw_queues_lock) annotation
Signed-off-by: Jules Irenge <jbi.octave@gmail.com>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
The "parse_cnt" variable is incremented while validating the CS chunks,
but it is actually not being used.
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Add support for hwmon_in_highest, hwmon_temp_highest and hwmon_curr_highest
attributes. These attributes retrieve the historical maximum voltage,
temperature and current that were sampled, respectively.
Signed-off-by: Christine Gharzuzi <cgharzuzi@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
The hl read and write routines implement the hwmon_ops read and write
interface routines respectively.
These routines are expected to return a completion status when called,
which was not the case until this commit.
This commit modifies these routines to return 0 upon success and a
negative error value upon failure.
Signed-off-by: Moti Haimovski <mhaimovski@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
This commit adds support for offsetting the temperatures reading
by a specified value as defined in
https://www.kernel.org/doc/Documentation/hwmon/sysfs-interface
using the standard sysfs defined for hwmon.
This is required by system administrators to inject errors to test
their monitoring applications in data centers.
Signed-off-by: Moti Haimovski <mhaimovski@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
The compute engines can perform millions of transactions per second. If
there is a bug in the S/W stack, we could get a lot of interrupts and spam
the kernel log. Therefore, ratelimit these prints
Reviewed-by: Omer Shpigelman <oshpigelman@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Allow debug user to write/read 64-bit data through debugfs.
This will expedite the dump process of the (large) internal
memories of the device done during debug.
Signed-off-by: Moti Haimovski <mhaimovski@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
DRAM_PHYS_BASE is already taken into account in MMU_PAGE_TABLES_ADDR.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
There is an extra ; after the end of a function, which needs to be removed
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
CS with no chunks for execute phase is invalid, so its
context_switch/restore phase should not be run.
Hence, move the check of the execute chunks number to the beginning of
hl_cs_ioctl().
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
As HL_MAX_JOBS_PER_CS is 512, it is possible that more than 255 CS jobs
will be submitted for a certain queue. Hence, modify the
"jobs_in_queue_cnt" parameter of the "hl_cs" structure to be u16 instead
of u8.
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Host memory may be allocated with huge pages.
A different virtual range may be used for mapping in this case.
Add Huge PCI MMU (HPMMU) properties to support it.
This patch is a prerequisite for future ASICs support and has no effect on
Goya ASIC as currently a single virtual host range is used for all page
sizes.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
When no patched command buffer (CB) is created, use the user CB size as
the job size.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
During device memory memset, the driver allocates and use a CB (command
buffer). To reuse existing code, it keeps a pointer to the CB in two
variables, user_cb and patched_cb. Therefore, there is no need to "put"
both the user_cb and patched_cb, as it will cause an underflow of the
refcnt of the CB.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
During hard reset we must not write to the device.
Hence avoid halting CoreSight during user context close if it is done
during hard reset.
In addition, we must not re-enable clock gating afterwards as it was
deliberately disabled in the beginning of the hard reset flow.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
The driver must halt the engines before doing hard-reset, otherwise the
device can go into undefined state. There is a place where the driver
didn't do that and this patch fixes it.
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Fixes gcc '-Wunused-but-set-variable' warning:
drivers/misc/habanalabs/goya/goya.c: In function goya_pldm_init_cpu:
drivers/misc/habanalabs/goya/goya.c:2195:6: warning: variable val set but not used [-Wunused-but-set-variable]
drivers/misc/habanalabs/goya/goya.c: In function goya_hw_init:
drivers/misc/habanalabs/goya/goya.c:2505:6: warning: variable val set but not used [-Wunused-but-set-variable]
Fixes: 9494a8dd8d ("habanalabs: add h/w queues module")
Signed-off-by: Chen Wandun <chenwandun@huawei.com>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
In case a user submits a CS, and the submission fails, and the user doesn't
check the return value and instead use the error return value as a valid
sequence number of a CS and ask to wait on it, the driver will print an
error and return an error code for that wait.
The real problem happens if now the user ignores the error of the wait, and
try to wait again and again. This can lead to a flood of error messages
from the driver and even soft lockup event.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Prevent accesses to the device (register read/write) from debugfs entries
during reset as that can cause the device to get stuck.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
During hard-reset, there can be multiple events received from the H/W. For
each event, the driver opens a worker thread to handle it. For some of the
events, the driver will read/write registers in the code that handles the
event.
In case of hard-reset, we must prevent reads/writes to the registers during
the reset operation because the device might get stuck if that happens.
Therefore, flush the EQ workers before resetting the device (in hard-reset
only). Additional events won't arrive as we synced and disabled the
interrupts.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
In the hl_device_reset we ask about the hard_reset argument when we want to
differentiate between soft and hard reset, except for three places where
we use "from_hard_reset_thread". Replace one of those locations with the
hard_reset argument as it is guaranteed that if we reached to that
line in the code during hard_reset, it is from a kernel thread.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Expose both soft and hard reset counts via INFO IOCTL.
This will allow system management applications to easily check
if the device has undergone reset.
Signed-off-by: Moti Haimovski <mhaimovski@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Instead of doing if inside if, just write them with && operator.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Omer Shpigelman <oshpigelman@habana.ai>
Make the code more concise and maintainable by using defines for the F/W
files.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Omer Shpigelman <oshpigelman@habana.ai>
Successful device initialization is mentioned in kernel log with the
message "Successfully added device to habanalabs driver". There is no point
of spamming the log with additional messages about successful queue
testing, which are implied by the above mentioned message.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Omer Shpigelman <oshpigelman@habana.ai>
Now that the VA block free list is not updated on context close in order
to optimize this flow, no need in the sanity checks of the list contents
as these will fail for sure.
In addition, remove the "context closing with VA in use" print during hard
reset as this situation is a side effect of the failure that caused the
hard reset.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reduce context close time by performing MMU cache invalidation once at the
end of the unmap loop rather in each iteration, in order to avoid hard
reset with open contexts.
Reset with open contexts can potentially lead to a kernel crash as the
generic pool of the MMU hops is destroyed while it is not empty because
some unmap operations are not done.
The commit affect mainly when running on simulator.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reduce context close time by skipping the VA block free list update in
order to avoid hard reset with open contexts.
Reset with open contexts can potentially lead to a kernel crash as the
generic pool of the MMU hops is destroyed while it is not empty because
some unmap operations are not done.
The commit affect mainly when running on simulator.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reduce context close time by skipping hash table lookup if possible in
order to avoid hard reset with open contexts.
Reset with open contexts can potentially lead to a kernel crash as the
generic pool of the MMU hops is destroyed while it is not empty because
some unmap operations are not done.
This commit affect mainly when running on simulator.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
During hard reset we should not access the device except of necessary
reset operations because the device might be stuck or unresponsive.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Split the properties used for MMU mappings to DRAM and PCI (host) types.
This is a prerequisite for future ASICs support.
Note that in Goya ASIC, the PMMU and DMMU are the same (except of page
sizes) as only one MMU mechanism is used for both of the mapping types.
Hence this patch should not have any effect on current behavior.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Some cosmetics around the MMU code to make it more self-explanatory.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Add the ability to invalidate the necessary MMU cache only.
This ability is a prerequisite for future ASICs support.
Note that in Goya ASIC, a single cache is used for both host/DRAM
mappings and hence this patch should not have any effect on current
behavior.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Some of the functions in the memory module code were too long and/or
contained multiple operations that are not always done together. Re-factor
the code by dividing those functions to smaller functions which are more
readable and maintainable.
Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
The two defines that control the maximum size of a command buffer and the
maximum number of JOBS per CS need to be exported to the user as they are
part of the API towards user-space.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Omer Shpigelman <oshpigelman@habana.ai>
If the queues are full and we return -EAGAIN to the user, there is no need
to print an error, as that case isn't an error and the user is expected to
re-submit the work.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Omer Shpigelman <oshpigelman@habana.ai>
In training, there is a need for a large amount of patching to the recipe.
This results in many command buffers contains a lot of DMA packets. The
number of command buffers per CS is larger than the current maximum of 64,
which is an arbitrary number that is enough for inference, but it has no
real affect on the code and/or resources of the host machine.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Omer Shpigelman <oshpigelman@habana.ai>
ETR should always be non-secured as it is used by the users to record
profiling/trace data.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Omer Shpigelman <oshpigelman@habana.ai>
We have a single ETR block in the SOC, so use explicit register
name defines for initializing this block. This makes it more readable and
maintainable.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Omer Shpigelman <oshpigelman@habana.ai>
Move the read of the F/W boot versions before exiting on possible failures
of the F/W boot. This will help debug boot failures as we will be able to
know the F/W boot version.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Omer Shpigelman <oshpigelman@habana.ai>
To enable userspace processes, e.g. management utilities, to display the
card name to the user, add the card name property to the HW_IP
structure that is copied to the user in the INFO IOCTL.
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Oded Gabbay