| Age | Commit message (Collapse) | Author |
|---|
|
There is some left-over code from the gaudi2 bring-up that wasn't
removed so far.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
On Gaudi2 the f/w always configures the PCIe iATU and allows access to
scratchpad registers. Therefore, we can know if the f/w is secured
by reading a status bit from the f/w registers.
Signed-off-by: farah kassabri <fkassabri@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
A common function that is called from multiple places can't be
located in degugfs.c because that file is only compiled if
debugfs is enabled in the kernel config file.
This can lead to undefined symbol compilation error.
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
Add a missing lock in hl_device_resume() when it assigns a value to the
'in_reset' indication.
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
In hl_hw_block_mmap(), the vma's 'vm_private_data' and 'vm_ops' fields
are assigned before filling the content of the private data.
In between there is a call to the ASIC hw_block_mmap() function, and if
it fails, the vma close function will be called with a bad private data
value.
Fix the order of assignments to avoid this issue.
In hl_hw_block_mmap() the vma's 'vm_private_data and vm_ops are assigned
before setting the
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
map_block() sets the block id handle even if get_hw_block_id() fails,
and in this case it uses block id 0 which might be a valid id.
Modify it to set the handle only if get_hw_block_id() succeeds.
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
When a CS is submitted, the ioctl handler checks the CS
flags and performs a sanity check, according to its value.
As new CS flags are added, the sanity check needs to be updated
according to the new flags.
Signed-off-by: Tal Cohen <talcohen@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
Currently, to get engines status, user needed to read debugfs file
with root permissions.
This new uapi allows user apace apps retrieve status, so for example,
in case of failure, status can be retrieved immediately by the
application itself which runs without root permissions.
Signed-off-by: Dani Liberman <dliberman@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
We don't use KDMA concurrently in the driver. The only use is through
debugfs and we don't protect concurrent access through it.
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
The macro argument <val> is cast-ed to u32 in some of the places.
Because this arg can be some arithmetic computation (e.g. address +
offset) the cast should be on the whole expression.
Signed-off-by: Ohad Sharabi <osharabi@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
Cosmetic commit, no logical changes. It just fixes the spelling
mistakes.
Signed-off-by: Bharat Jauhari <bjauhari@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
In order to improve scalability and reduce host overhead, it is better
to increase the default TDR timeout of Gaudi1 from 30 seconds to
10 minutes.
This will allow the DL Framework (e.g. PyTorch, TensorFlow) to remove
the host sync they are using now and improve overall performance on
scaleout training.
Note that one can always set the timeout to a custom value via
a kernel module parameter given during driver load.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
Up until now the module iterator called void callback functions
and so caller activating callback that may fail suffered from 2 issues:
1. The need to "plant" return called in the private data. This is a
drawback since the iterator itself should not be aware of the private
data of the caller.
2. Due to 1 even in a failure the iterator would keep iterating instead
of break upon error.
To overcome this an optional rc field added to the iterator context.
Signed-off-by: Ohad Sharabi <osharabi@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
In order to be more explicit we should use the term compute_reset
for describing the reset in which only the compute engines gets
reset.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
Change is_idle functions so it would be more usable outside debugfs.
Do this by replacing seq_file parameter with regular string.
Signed-off-by: Dani Liberman <dliberman@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
Doing compute reset can be the traditional inference soft reset
that is supported only in Goya.
Or it can be the new reset upon device release, which is supported
in Gaudi2 and above.
Therefore, wherever suitable, use the terminology of compute reset
instead of soft reset.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
The user might want to know the device is in reset after device
release, which is not an erroneous event as a regular reset.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
reset_info.is_in_soft_reset should be updated both before in_reset
and inside the spin lock of the reset info structure.
The reasons are:
- When we are inside soft reset, it implies we are in reset. Therefore,
if someone checks if we are in soft reset, he can deduce we are
in reset, while the opposite is not correct and might be misleading.
- Both these flags are changed together so they must be changed
inside the reset info spinlock.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
In case security is enabled on the device, some debugfs nodes will
fail. Hence, we do not expose them.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
For gaudi2 we need to send a value to F/W as part of the
PCI_ACCESS packet.
As a preparation, modify hl_fw_send_pci_access_msg() to have a 'value'
field.
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
- use conventional internal macro variables (double underscore prefix)
- adjust address casting
- on register poll using ELBI use ELBI read rather than BAR read on
error condition
- remove unused macro
Signed-off-by: Ohad Sharabi <osharabi@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
Fix warning of
"warning: ‘old_base’ may be used uninitialized in this function"
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
Use bitmap_zalloc()/bitmap_free() instead of hand-writing them.
It is less verbose and it improves the semantic.
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
timestamp could be unset in both _hl_interrupt_wait_ioctl() and
_hl_interrupt_wait_ioctl_user_addr() so it is better to explicitly
initialize it to 0 when declaring it.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
tmp is declared in the scope of the function cs_do_release() and
inside a block inside that function.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
Max power API is not supported in secured devices. Hence, we should
skip setting it during boot.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
If we send a packet to the f/w, and that packet is unsupported, we
want to be able to identify this situation and possibly ignore this.
Therefore, if the f/w returned an error, we need to propagate it
to the callers in the result value, if those callers were interested
in it.
In addition, no point of printing the error code here because each
caller prints its own error with a specific message.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
We need this property for backward compatibility against the f/w.
Signed-off-by: Sagiv Ozeri <sozeri@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
User application should be able to get notification for any decoder
completion. Hence, we introduce a new interface in which a user
can wait for all current decoder pending interrupts.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
Current naming convention can be misleading. Hence renaming some
variables and defines in order to be more explicit.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
Currently we are not waiting for preboot ready after hard reset.
This leads to a race in which COMMs protocol begins but will get no
response from the f/w.
Signed-off-by: Ohad Sharabi <osharabi@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
Enable the Gaudi2 ASIC code in the pci probe callback of the driver so
the driver will handle Gaudi2 ASICs.
Add the PCI ID to the PCI table and add the ASIC enum value to all
relevant places.
Fixup the device parameters initialization for Gaudi2.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
Gaudi2 has new MMU units. A PMMU for device->host accesses, and HMMU
for HBM accesses.
The page tables of both MMUs are located in the host's memory (referred
to in the code as host-resident pgt).
Signed-off-by: Moti Haimovski <mhaimovski@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
In Gaudi2 we moved to a different wait for command submission
completion model. Instead of receiving interrupt only on external
queues, we use the device's sync manager to notify us when the
entire command submission finishes.
This enables us to remove the categorization of queues to external
and internal, and treat each queue equally, without the need to parse
and patch any command buffer.
This change also requires refactoring to the IRQ handling of
CS completions.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
As the ASICs become more complex and have many more registers, we need
a better way to configure the security properties.
As a reminder, we have two dedicated mechanisms for security:
Range Registers and Protection bits. Those mechanisms protect sensitive
memory and configuration areas inside the device.
The generic module handles the low-level part of the configuration,
because the configuration mechanism is identical in all ASICs. The
difference is the address ranges and register names.
Any ASIC that use this block should first block all the register
blocks in the ASIC. Then, it should open only the registers that
need to be accessed by the user (This is opposed to Goya and Gaudi,
where we blocked only what should not be accesses by the user).
The module contains several functions, to unblock single register,
multiple registers, entire blocks, ranges, ranges with mask.
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 a couple of device variables that are used for testing
purposes and they are set to fixed values.
Remove the variables that are not relevant anymore and document the
remaining variables.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
New asic properties were added for Gaudi2. We want to initialize
and use them, when relevant, also for Goya and Gaudi.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
Add the ASIC-specific code for Gaudi2. Supply (almost) all of the
function callbacks that the driver's common code need to initialize,
finalize and submit workloads to the Gaudi2 ASIC.
It also contains the code to initialize the F/W of the Gaudi2 ASIC
and to receive events from the F/W.
It contains new debugfs entry to dump razwi events. razwi is a case
where the device's engines create a transaction that reaches an
invalid destination.
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
Region structure is derived from region type, hence no need to pass
it as an argument.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
Because in future ASICs the driver will allow the user to set the
page size we need to make sure this data is propagated in all APIs.
In addition, since this is already an ASIC property we no longer need
ASIC function for it.
Signed-off-by: Ohad Sharabi <osharabi@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
free_device_memory() ends with if and else, each has a return statement,
followed by another return statement that can never be reached.
Restructure the function and remove this dead code.
Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
We dropped support for page sizes that are not power of 2.
Signed-off-by: Ohad Sharabi <osharabi@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
This is a pre-requisite patch for adding tracepoints to the DMA memory
operations (allocation/free) in the driver.
The main purpose is to be able to cross data with the map operations and
determine whether memory violation occurred, for example free DMA
allocation before unmapping it from device memory.
To achieve this the DMA alloc/free code flows were refactored so that a
single DMA tracepoint will catch many flows.
Signed-off-by: Ohad Sharabi <osharabi@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
Set a default value for memory scrubbing
Signed-off-by: Dafna Hirschfeld <dhirschfeld@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
In future ASICs, it would be possible to have a non-idle
device when context is released. We thus need to postpone the
scrubbing. Postpone it to hpriv release if reset is not executed
or to device late init if reset is executed.
Signed-off-by: Dafna Hirschfeld <dhirschfeld@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
We use scrub_device_mem only to scrub the entire SRAM and entire
DRAM. Therefore there is no need to send addr and size
args to the callback.
Signed-off-by: Dafna Hirschfeld <dhirschfeld@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
There is no need to do memory scrub when unmapping anymore as it is
an overhead as long as we have a single user at any given time.
Remove that code and change return value of free_phys_pg_pack to void
Signed-off-by: Dafna Hirschfeld <dhirschfeld@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
For easier debug, it is desirable to have a simple way
to know whether the device is secured or not, hence we dump this
indication during boot.
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 a rare race condition in CB completion mechanism, that can
occur under a very high pressure of command submissions.
The preconditions for this to happen are:
1. There should be enough command submissions for the pre-allocated
patched CB pool to run out of commands. At this stage we start
allocating new patched CBs as they arrive.
2. CB size has to be exactly (128*n + 104)B for some n, i.e. 24B below
a cache line end.
The flow:
1. Two command buffers being completed on different streams, at the
same time. Denote those CB1 and CB2.
2. Each command buffer is injected with two messages, 16B each - one
for a HBW update of the completion queue, another to raise
interrupt.
3. Assume CB1 updated the completion queue and raise the interrupt.
4. Assume CB2 updated the completion queue but did not raise the
interrupt yet.
5. The host receives the interrupt. It goes over the completion queue
and sees two completions - CB1 and CB2. Release them both.
6. CB2 performs the last command. The problem is that the last command
is split between 2 cache lines. So to read the last 8B of the last
command, it has to access the host again. Problem is - CB2 is
already released. This causes a DMAR error.
The solution to this problem is simply to make sure the last two
commands in the CB are always in the same cache line, using NOP padding.
Signed-off-by: Yuri Nudelman <ynudelman@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
|
move the field memory_scrub_val from struct hl_dbg_device_entry
to struct hl_device. This is because we want to use
this field also if debugfs is off.
Signed-off-by: Dafna Hirschfeld <dhirschfeld@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
|
