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: Nkernel test robot <lkp@intel.com>
Reviewed-by: NTomer Tayar <ttayar@habana.ai>
Signed-off-by: NOded 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: NOfir Bitton <obitton@habana.ai>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: NOded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NOded Gabbay <oded.gabbay@gmail.com>

User input must be validated before using it to
access internal structures.
Signed-off-by: NOfir Bitton <obitton@habana.ai>
Reviewed-by: NOded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: NOded 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: NHulk Robot <hulkci@huawei.com>
Signed-off-by: NWei Yongjun <weiyongjun1@huawei.com>
Link: https://lore.kernel.org/r/20200729155902.33976-1-weiyongjun1@huawei.comSigned-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

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: NOfir Bitton <obitton@habana.ai>
Reviewed-by: NOded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: NOded Gabbay <oded.gabbay@gmail.com>

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: NOded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: NOmer Shpigelman <oshpigelman@habana.ai>

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: NOfir Bitton <obitton@habana.ai>
Reviewed-by: NOded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: NOded 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: NOfir Bitton <obitton@habana.ai>
Reviewed-by: NOded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: NOded 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: NOfir Bitton <obitton@habana.ai>
Reviewed-by: NOded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: NOded 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: NOded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: NTomer Tayar <ttayar@habana.ai>

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: NOfir Bitton <obitton@habana.ai>
Reviewed-by: NOded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: NOded 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: NOfir Bitton <obitton@habana.ai>
Reviewed-by: NOded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: NOded 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: NOded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: NTomer Tayar <ttayar@habana.ai>

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: NOfir Bitton <obitton@habana.ai>
Reviewed-by: NOded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: NOded Gabbay <oded.gabbay@gmail.com>

Use the u64_to_user_ptr(x) kernel macro to correctly cast u64 to void*
Reported-by: Nkbuild test robot <lkp@intel.com>
Reviewed-by: NOmer Shpigelman <oshpigelman@habana.ai>
Signed-off-by: NOded Gabbay <oded.gabbay@gmail.com>
Link: https://lore.kernel.org/r/20200601065648.8775-2-oded.gabbay@gmail.comSigned-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

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: NDaniel Vetter <daniel.vetter@intel.com>
Reviewed-by: NOded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: NOded 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: NOmer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: NOded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: NOded 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: NOmer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: NOded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: NOded 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: NOmer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: NOded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: NOded 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: NTomer Tayar <ttayar@habana.ai>
Signed-off-by: NOded 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: NTomer Tayar <ttayar@habana.ai>
Reviewed-by: NOded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: NOded Gabbay <oded.gabbay@gmail.com>

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: NTomer Tayar <ttayar@habana.ai>
Reviewed-by: NOded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: NOded 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: NOmer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: NOded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: NOded 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: NOded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: NTomer Tayar <ttayar@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: NOded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: NOmer Shpigelman <oshpigelman@habana.ai>

This patch adds a support for a new H/W queue type.
This type of queue is for DMA and compute engines jobs, for which
completion notification are sent by H/W.
Command buffer for this queue can be created either through the CB
IOCTL and using the retrieved CB handle, or by preparing a buffer on the
host or device SRAM/DRAM, and using the device address to that buffer.
The patch includes the handling of the 2 options, as well as the
initialization of the H/W queue and its jobs scheduling.
Signed-off-by: NTomer Tayar <ttayar@habana.ai>
Reviewed-by: NOded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: NOded Gabbay <oded.gabbay@gmail.com>

Jobs on some queues must be provided with a handle to a driver command
buffer object, while for other queues, jobs must be provided with an
address to a command buffer.
Currently the distinction is done based on the queue type, which is less
flexible if the same queue type behaves differently on different
types of ASICs.
This patch adds a new queue property for this target, which is
configured per queue type per ASIC type.
Signed-off-by: NTomer Tayar <ttayar@habana.ai>
Reviewed-by: NOded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: NOded Gabbay <oded.gabbay@gmail.com>

We want to stop using the acronym KMD. Therefore, replace all locations
(except for register names we can't modify) where KMD is written to other
terms such as "Linux kernel driver" or "Host kernel driver", etc.
Signed-off-by: NOded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: NOmer Shpigelman <oshpigelman@habana.ai>

Users and sysadmins usually want to know what is the device utilization as
a level 0 indication if they are efficiently using the device.

Add a new opcode to the INFO IOCTL that will return the device utilization
over the last period of 100-1000ms. The return value is 0-100,
representing as percentage the total utilization rate.
Signed-off-by: NOded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: NOmer Shpigelman <oshpigelman@habana.ai>

When rejecting CS because of too many in-flight CS, print a debug message
about it as it useful to know when the user is debugging (it indicates a
back-pressure from the driver as the device is not fast enough to consume
the CS)
Signed-off-by: NOded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: NOmer Shpigelman <oshpigelman@habana.ai>

This patch fix a bug in the host memory polling macro. The bug is that the
memory being polled can be written by the device, which always writes it
in LE. However, if the host is running Linux in BE mode, we need to
convert the value that was written by the device before matching it to the
required value that the caller has given to the macro.
Signed-off-by: NBen Segal <bpsegal20@gmail.com>
Reviewed-by: NOded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: NOded Gabbay <oded.gabbay@gmail.com>

This patch changes two polling functions to macros, in order to make their
API the same as the standard readl_poll_timeout so we would be able to
define the "condition for exit" when calling these macros.

This will simplify the code as it will eliminate the need to check both
for timeout and for the (cond) in the calling function.
Signed-off-by: NOded Gabbay <oded.gabbay@gmail.com>

use_virt_addr member was used for telling whether to treat the
addresses in the CB as virtual during parsing. We disabled it only
when calling the parser from the driver memset device function,
and since this call had been removed, it should always be enabled.
Signed-off-by: NDalit Ben Zoor <dbenzoor@habana.ai>
Signed-off-by: NOded Gabbay <oded.gabbay@gmail.com>

This patch only does renaming of certain variables and structure members,
and their accompanied comments.

This is done to better reflect the actions these variables and members
represent.

There is no functional change in this patch.
Signed-off-by: NOded Gabbay <oded.gabbay@gmail.com>

This patch improves two error messages to help the user to
better understand what error occurred.
Signed-off-by: NOded Gabbay <oded.gabbay@gmail.com>

This patch adds accounting for active CS. Active means that the CS was
submitted to the H/W queues and was not completed yet.

This is necessary to support suspend operation. Because the device will be
reset upon suspend, we can only suspend after all active CS have been
completed. Hence, we need to perform accounting on their number.
Signed-off-by: NOded Gabbay <oded.gabbay@gmail.com>

This patch fix a bug in the driver, where if the TPC or MME remains in
non-IDLE even after all the command submissions are done (due to user bug
or malicious user), then future command submissions will fail in the
context-switch stage and the driver will remain in "stuck" mode.

The fix is to do a soft-reset of the device in case the context-switch
fails, because the device should be IDLE during context-switch. If it is
not IDLE, then something is wrong and we should reset the compute engines.
Signed-off-by: NOded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

At the start of some IOCTLs we check if the device is disabled or in reset.
If it is, we return -EBUSY and print a message to kernel log.

Because these IOCTLs can be called at very high frequency, use ratelimit
to avoid spamming the kernel log. Also use the same type of message -
dev_warn - in all the relevant IOCTLs.
Signed-off-by: NOded Gabbay <oded.gabbay@gmail.com>

This patch adds debugfs support to the driver. It allows the user-space to
display information that is contained in the internal structures of the
driver, such as:
- active command submissions
- active user virtual memory mappings
- number of allocated command buffers

It also enables the user to perform reads and writes through Goya's PCI
bars.
Reviewed-by: NMike Rapoport <rppt@linux.ibm.com>
Signed-off-by: NOded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

This patch adds the main flow for the user to submit work to the device.

Each work is described by a command submission object (CS). The CS contains
3 arrays of command buffers: One for execution, and two for context-switch
(store and restore).

For each CB, the user specifies on which queue to put that CB. In case of
an internal queue, the entry doesn't contain a pointer to the CB but the
address in the on-chip memory that the CB resides at.

The driver parses some of the CBs to enforce security restrictions.

The user receives a sequence number that represents the CS object. The user
can then query the driver regarding the status of the CS, using that
sequence number.

In case the CS doesn't finish before the timeout expires, the driver will
perform a soft-reset of the device.
Reviewed-by: NMike Rapoport <rppt@linux.ibm.com>
Signed-off-by: NOded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>