- 12 Jul, 2022 40 commits
-
-
Oded Gabbay authored
This fixes a sparse warning of "cast truncates bits from constant value" Signed-off-by:Oded Gabbay <ogabbay@kernel.org>
-
Oded Gabbay authored
packets are defined as LE so we need to convert before assigning values to them. Signed-off-by: -
Oded Gabbay authored
function name in comment didn't match actual function name. Signed-off-by: -
Oded Gabbay authored
The values in this enum are not used by h/w but are a contract between userspace and the kernel driver so they must be defined in the uapi file. Signed-off-by: -
Dafna Hirschfeld authored
Set a default value for memory scrubbing Signed-off-by:
Dafna Hirschfeld <dhirschfeld@habana.ai> Reviewed-by:
-
Dafna Hirschfeld authored
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 authored
In the callback scrub_device_mem, use 'memory_scrub_val' from debugfs for the scrubbing value. Signed-off-by:
-
Dafna Hirschfeld authored
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 authored
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:
-
Ofir Bitton authored
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:
-
Yuri Nudelman authored
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:
-
Koby Elbaz authored
kernel test robot: "warning: variable 'index' is used uninitialized whenever 'if' condition is false" Reported-by:
kernel test robot <lkp@intel.com> Signed-off-by:
Koby Elbaz <kelbaz@habana.ai> Reviewed-by:
-
Dafna Hirschfeld authored
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:
-
Oded Gabbay authored
function name should not be preceded with @ Signed-off-by: -
Oded Gabbay authored
kvcalloc is same as kvmalloc_array with GFP_ZERO. Signed-off-by: -
Oded Gabbay authored
Use %p instead of %llx for printing pointers. Signed-off-by: -
Oded Gabbay authored
fence pointer can be NULL in this path, as shown by an earlier check. Signed-off-by: -
ran shalit authored
Multiple SRAM SERR events are treated as critical events, and host should be notified about it. Thus, adding is_critical indication as part of SRAM ECC failure packet. Signed-off-by:
ran shalit <rshalit@habana.ai> Reviewed-by:
-
Tal Cohen authored
When a device error occurs, user process would like to get some indication on the error by reading some device HW info. If the device is unavailable, user process can't perform any HW device reading. Signed-off-by:
Tal Cohen <talcohen@habana.ai> Reviewed-by:
-
Oded Gabbay authored
This asic callback function is not called anymore from the common code. The asic-specific function itself is called but from within the asic-specific code. Signed-off-by: -
Yuri Nudelman authored
Device descriptor is accessed before an assignment Signed-off-by:
-
Ofir Bitton authored
When validating NIC queues, queue offset calculation must be performed only for NIC queues. Signed-off-by:
-
farah kassabri authored
Driver performs no validity check for the user cq counter offset used in both wait_for_interrupt and register_for_timestamp APIs. Signed-off-by:
farah kassabri <fkassabri@habana.ai> Reviewed-by:
-
Koby Elbaz authored
Once FW raised an event following a MME2 QMAN error, the driver should have gone to the corresponding status registers, trying to gather more info on the error, yet it was accidentally accessing MME1 QMAN address space. Generally, we have x4 MMEs, while 0 & 2 are marked MASTER, and 1 & 3 are marked SLAVE. The former can be addressed, yet addressing the latter is considered an access violation, and will result in a hung system, which is what unintentionally happened above. Note that this cannot happen in a secured system, since these registers are protected with range registers. Signed-off-by:
-
Dani Liberman authored
When sending a packet to FW right after it made reset, we will get packet timeout. Since it is expected behavior, we don't need to print an error in such case. Hence, when driver is in hard reset it will avoid from printing error messages about packet timeout. Signed-off-by:
Dani Liberman <dliberman@habana.ai> Reviewed-by:
-
Tal Cohen authored
The Driver needs to inform the User process whenever one of its CS is timed out. The Driver shall recognize the CS timeout and shall send an eventfd notification, towards user space, whenever a timeout is expired on a CS. Signed-off-by:
-
Tal Cohen authored
Device reset event, indicates that the device shall be reset - after a short delay. In such case, the driver sends a notification towards the User process. This allows the User process to be able to take several debug actions for system diagnostic purposes. Signed-off-by:
-
Tal Cohen authored
In order to prepare the driver code for device reset event notification, change the event handler function flow to call device reset from one code block. In addition, the commit fixes an issue that reset was performed w/o checking the 'hard_reset_on_fw_event' state and w/o setting the HL_DRV_RESET_DELAY flag. Signed-off-by:
-
Tal Cohen authored
The info ioctl retrieves information on the last undefined opcode occurred. Signed-off-by:
-
Tal Cohen authored
when an undefined opcode error occurres, the driver collects the relevant information from the Qman and stores it inside the hdev data structure. An event fd indication is sent towards the user space. Note: another commit shall be followed which will add support to read the error info by an ioctl. Signed-off-by:
-
Tomer Tayar authored
hl_get_compute_ctx() is used to get the pointer to the compute context from the hpriv object. The function is called in code paths that are not necessarily initiated by user, so it is possible that a context release process will happen in parallel. This can lead to a race condition in which hl_get_compute_ctx() retrieves the context pointer, and just before it increments the context refcount, the context object is released and a freed memory is accessed. To avoid this race, add a mutex to protect the context pointer in hpriv. With this lock, hl_get_compute_ctx() will be able to detect if the context has been released or is about to be released. struct hl_ctx_mgr has a mutex for contexts IDR with a similar "ctx_lock" name, so rename it to just "lock" to avoid a confusion with the new lock. Signed-off-by:
Tomer Tayar <ttayar@habana.ai> Reviewed-by:
-
Yuri Nudelman authored
Often, the user is not interested in the completion timestamp of all command submissions. A common situation is, for example, when the user submits a burst of, possibly, several thousands of commands, then request the completion timestamp of only couple of specific key commands from all the burst. The problem is that currently, the outcome of the early commands may be lost, due to a large amount of later commands, that the user does not really care about. This patch creates a separate store with the outcomes of commands the user has mark explicitly as interested in. This store does not mix the marked commands with the unmarked ones, hence the data there will survive for much longer. Signed-off-by:
-
Oded Gabbay authored
Due to code changes in the past few years, the original comment of how parser->user_cb_size is checked was not correct anymore. Fix it to reflect current code and add more explanation as the code is more complex now. Reported-by:
Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by:
-
Tal Cohen authored
positive flags naming will make more clear code while adding more 'error info' structures Signed-off-by:
-
Tal Cohen authored
raising the tpc assert event in an internal function will make the code cleaner as we are going to be adding more events Signed-off-by:
-
Dan Rapaport authored
The compiler is padding the members of the struct to be aligned to 64-bit. The content of the padded bytes is and not zeroed explicitly, hence might copy undefined data. We add a padding member to the struct to get a zeroed 64-bit align struct. Signed-off-by:
Dan Rapaport <drapaport@habana.ai> Reviewed-by:
-
Dafna Hirschfeld authored
Arrays of struct attribute are expected to be NULL terminated. This is required by API methods such as device_add_groups. This fixes a crash when loading the driver for Goya device. Signed-off-by:
-
Jiapeng Chong authored
Fix the following W=1 kernel warnings: drivers/misc/habanalabs/common/mmu/mmu_v1.c:425: warning: expecting prototype for hl_mmu_fini(). Prototype was for hl_mmu_v1_fini() instead. drivers/misc/habanalabs/common/mmu/mmu_v1.c:449: warning: expecting prototype for hl_mmu_ctx_init(). Prototype was for hl_mmu_v1_ctx_init() instead. Reported-by:
Abaci Robot <abaci@linux.alibaba.com> Signed-off-by:
Jiapeng Chong <jiapeng.chong@linux.alibaba.com> Reviewed-by:
-
Jiapeng Chong authored
Fix the following W=1 kernel warnings: drivers/misc/habanalabs/common/pci/pci.c:454: warning: expecting prototype for hl_fw_fini(). Prototype was for hl_pci_fini() instead. Reported-by:
-
Dan Carpenter authored
If hl_mmu_prefetch_cache_range() fails then this code calls mutex_unlock(&ctx->mmu_lock) when it's no longer holding the mutex. Fixes: 9e495e24 ("habanalabs: do MMU prefetch as deferred work") Signed-off-by:
-
