summaryrefslogtreecommitdiff
path: root/drivers/accel
AgeCommit message (Collapse)Author
2023-10-27accel/qaic: Enable 1 MSI fallback modeCarl Vanderlip
Several virtualization use-cases either don't support 32 MultiMSIs (Xen/VMware) or have significant drawbacks to their use (KVM's vIOMMU, which is required to support 32 MSI, needs to allocate an alternate system memory space for each device using vIOMMU (e.g. 8GB VM mem and 2 cards => 8 + 2 * 8 = 24GB host memory required)). Support these cases by enabling a 1 MSI fallback mode. Whenever all 32 MSIs requested are not available, a second request for a single MSI is made. Its success is the initiator of single MSI mode. This mode causes all interrupts generated by the device to be directed to the 0th MSI (firmware >=v1.10 will do this as a response to the PCIe MSI capability configuration). Likewise, all interrupt handlers for the device are registered to the 0th MSI. Since the DBC interrupt handler checks if the DBC is in use or if there is any pending changes, the 'spurious' interrupts are disregarded. If there is work to be done, the standard threaded IRQ handler is dispatched. On every interrupt, the MHI handler wakes up its threaded interrupt handler, and attempts to wake any waiters for MHI state events. Performance is within +-0.6% for test cases that typify real world use. Larger differences ([-4,+132]%, avg +47%) exist for very simple tasks (e.g. addition) compiled for single NSPs. It is assumed that the small work and many interrupts typically cause contention (e.g. 16 NSPs vs 4 CPUs), as evidenced by the standard deviation between runs also decreasing (r=-0.48 between delta(Performace_test) and delta(StdDev_test/Avg_test)) Signed-off-by: Carl Vanderlip <quic_carlv@quicinc.com> Reviewed-by: Pranjal Ramajor Asha Kanojiya <quic_pkanojiy@quicinc.com> Reviewed-by: Jeffrey Hugo <quic_jhugo@quicinc.com> Signed-off-by: Jeffrey Hugo <quic_jhugo@quicinc.com> Reviewed-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20231016170036.5409-1-quic_jhugo@quicinc.com
2023-10-26accel/ivpu/37xx: Fix missing VPUIP interruptsKarol Wachowski
Move sequence of masking and unmasking global interrupts from buttress interrupt handler to generic one that handles both VPUIP and BTRS interrupts. Unmasking global interrupts will re-trigger MSI for any pending interrupts. Lack of this sequence will cause the driver to miss any VPUIP interrupt that comes after reading VPU_37XX_HOST_SS_ICB_STATUS_0 and before clearing all active interrupt sources. Fixes: 35b137630f08 ("accel/ivpu: Introduce a new DRM driver for Intel VPU") Cc: stable@vger.kernel.org Signed-off-by: Karol Wachowski <karol.wachowski@linux.intel.com> Reviewed-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com> Signed-off-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20231024161952.759914-1-stanislaw.gruszka@linux.intel.com
2023-10-23BackMerge tag 'v6.6-rc7' into drm-nextDave Airlie
This is needed to add the msm pr which is based on a higher base. Signed-off-by: Dave Airlie <airlied@redhat.com>
2023-10-23accel/ivpu: Delete the TODO fileDeepak R Varma
The work items listed in the TODO file of this driver file are either completed or dropped. The file is no more significant according to the maintainers. Hence removing it from the sources. Suggested-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com> Signed-off-by: Deepak R Varma <drv@mailo.com> Reviewed-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com> Signed-off-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/ZTAucrOT69/tQK2o@runicha.com
2023-10-23accel/ivpu/37xx: Remove support for FPGA and simicsStanislaw Gruszka
We do not run 37xx VPU on non-silicon platforms any longer. Remove deprecated code to make it cleaner. Reviewed-by: Jeffrey Hugo <quic_jhugo@quicinc.com> Signed-off-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20231020104501.697763-7-stanislaw.gruszka@linux.intel.com
2023-10-23accel/ivpu: Do not initialize parameters on power upStanislaw Gruszka
Initialize HW specific parameters only once. We do not have to do this on every power_up (performed during initialization and on resume). Move corresponding code to ->info_init() Reviewed-by: Jeffrey Hugo <quic_jhugo@quicinc.com> Signed-off-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20231020104501.697763-6-stanislaw.gruszka@linux.intel.com
2023-10-23accel/ivpu: Print IPC type string instead of numberKrystian Pradzynski
Introduce ivpu_jsm_msg_type_to_str() helper to print type of IPC message. This will make reading logs and debugging IPC issues easier. Co-developed-by: Maciej Falkowski <maciej.falkowski@intel.com> Signed-off-by: Maciej Falkowski <maciej.falkowski@intel.com> Signed-off-by: Krystian Pradzynski <krystian.pradzynski@linux.intel.com> Reviewed-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com> Reviewed-by: Jeffrey Hugo <quic_jhugo@quicinc.com> Signed-off-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20231020104501.697763-5-stanislaw.gruszka@linux.intel.com
2023-10-23accel/ivpu: Read clock rate only if device is upKarol Wachowski
Do not unnecessarily wake up device to read clock rate. Return 0 as clk_rate if device is suspended. Signed-off-by: Karol Wachowski <karol.wachowski@linux.intel.com> Reviewed-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com> Reviewed-by: Jeffrey Hugo <quic_jhugo@quicinc.com> Signed-off-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20231020104501.697763-4-stanislaw.gruszka@linux.intel.com
2023-10-23accel/ivpu: Fix verbose version of REG_POLL macrosKrystian Pradzynski
Remove two out of four _POLL macros. For two remaining _POLL macros add message about polling register start and finish. Additionally avoid inconsequence when using REGV_WR/RD macros in MMU code - passing raw register offset instead of register name. Signed-off-by: Krystian Pradzynski <krystian.pradzynski@linux.intel.com> Reviewed-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com> Reviewed-by: Jeffrey Hugo <quic_jhugo@quicinc.com> Signed-off-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20231020104501.697763-3-stanislaw.gruszka@linux.intel.com
2023-10-23accel/ivpu: Use ratelimited warn and err in IPC/JSMKrystian Pradzynski
Quite often during test corner cases IPC, JSM functions can flood dmesg with warn or err messages. With that lost dmesg history. Change warn, err to ratelimited versions in IPC, JSM to suppress dmesg spam occurrence during fail test scenarios. Signed-off-by: Krystian Pradzynski <krystian.pradzynski@linux.intel.com> Reviewed-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com> Reviewed-by: Jeffrey Hugo <quic_jhugo@quicinc.com> Signed-off-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20231020104501.697763-2-stanislaw.gruszka@linux.intel.com
2023-10-19accel/ivpu: Extend address range for MMU mmapWludzik, Jozef
Allow to use whole address range in MMU context mmap which is up to 48 bits. Return invalid argument from MMU context mmap in case address is not aligned to MMU page size, address is below MMU page size or address is greater then 47 bits. This fixes problem disallowing to run large models on VPU4 Signed-off-by: Wludzik, Jozef <jozef.wludzik@intel.com> Reviewed-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com> Signed-off-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20231018110113.547208-1-stanislaw.gruszka@linux.intel.com
2023-10-19Revert "accel/ivpu: Use cached buffers for FW loading"Stanislaw Gruszka
This reverts commit 645d694559cab36fe6a57c717efcfa27d9321396. The commit cause issues with memory access from the device side. Switch back to write-combined memory mappings until the issues will be properly addressed. Add extra wmb() needed when boot_params->save_restore_ret_address() is modified. Reviewed-by: Karol Wachowski <karol.wachowski@linux.intel.com> Signed-off-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20231017121353.532466-1-stanislaw.gruszka@linux.intel.com
2023-10-19accel/ivpu: Don't enter d0i3 during FLRJacek Lawrynowicz
Avoid HW bug on some platforms where we enter D0i3 state and CPU is in low power states (C8 or above). Fixes: 852be13f3bd3 ("accel/ivpu: Add PM support") Cc: stable@vger.kernel.org Signed-off-by: Jacek Lawrynowicz <jacek.lawrynowicz@linux.intel.com> Reviewed-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com> Signed-off-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20231003064213.1527327-1-stanislaw.gruszka@linux.intel.com
2023-10-17Merge tag 'drm-habanalabs-next-2023-10-10' of ↵Dave Airlie
https://git.kernel.org/pub/scm/linux/kernel/git/ogabbay/linux into drm-next This tag contains habanalabs driver changes for v6.7. The notable changes are: - uAPI changes: - Expose tsc clock sampling to better sync clock information in profiler. - Enhance engine error reporting in the info ioctl. - Block access to the eventfd operations through the control device. - Disable the option of the user to register multiple times with the same offset for timestamp dump by the driver. If a user wants to use the same offset in the timestamp buffer for different interrupt, it needs to first de-register the offset. - When exporting dma-buf (for p2p), force the user to specify size/offset in multiples of PAGE_SIZE. This is instead of the driver doing the rounding to PAGE_SIZE, which has caused the driver to map more memory than was intended by the user. - New features and improvements: - Complete the move of the driver to the accel subsystem by removing the custom habanalabs class and major and registering to accel subsystem. - Move the firmware interface files to include/linux/habanalabs. This is a pre-requisite for upstreaming the NIC drivers of Gaudi (as they need to include those files). - Perform device hard-reset upon PCIe AXI drain event to prevent the failure from cascading to different IP blocks in the SoC. In secured environments, this is done automatically by the firmware. - Print device name when it is removed for better debuggability. - Add support for trace of dma map sgtable operations. - Optimize handling of user interrupts by splitting the interrupts to two lists. One list for fast handling and second list for handling with timestamp recording, which is slower. - Prevent double device hard-reset due to 2 adjacent H/W events. - Set device status 'malfunction' while in rmmod. - Firmware related fixes: - Extend preboot timeout because preboot loading might take longer than expected in certain cases. - Add a protection mechanism for the Event Queue. In case it is full, the firmware will be able to notify about it through a dedicated interrupt. - Perform device hard-reset in case scrubbing of memory has failed. - Bug fixes and code cleanups: - Small fixes of dma-buf handling in Gaudi2, such as handling an offset != 0, using the correct exported size, creation of sg table. - Fix spmu mask creation. - Fix bug in wait for cs completion for decoder workloads. - Cleanup Greco name from documentation. - Fix bug in recording timestamp during cs completion interrupt handling. - Fix CoreSight ETF configuration and flush logic. - Fix small bug in hpriv_list handling (the list that contains the private data per process that opens our device). Signed-off-by: Dave Airlie <airlied@redhat.com> # -----BEGIN PGP SIGNATURE----- # # iQEzBAABCgAdFiEE7TEboABC71LctBLFZR1NuKta54AFAmUlHoQACgkQZR1NuKta # 54DsXQf8CW+W4iWJf5UDTj/E/giu9rVRrsUsU0hhCcXbecIxRsLObYXtulENu5/u # VuEAo/tAvo0LUKi8pdIv6ernDKaxZ1+fimlfXMCzllAA/ts3yp1NgunprsIsx3tv # YgcJ2GNR8UlVZ1qYuZl+4dOTyD0yfRMROUXBe7wqKnUXOEepOiLBxq6W15tZiJnx # L+V0yGkNk6pAoADIXLW9EgEXiN/bJZCXGPWp06i/Nz7cHIHJGoV59wAqftqllCtk # 8ZMkLByjlQKPhc5AgWBtKE8EGVip3sm7b/Q2Gq0ZXdZiebyVJ+AjuuDOdtq1UCIw # Rcp2576E7rByIBu3RAFlrioWhuR5Zw== # =2ien # -----END PGP SIGNATURE----- # gpg: Signature made Tue 10 Oct 2023 19:51:00 AEST # gpg: using RSA key ED311BA00042EF52DCB412C5651D4DB8AB5AE780 # gpg: Can't check signature: No public key From: Oded Gabbay <ogabbay@kernel.org> Link: https://patchwork.freedesktop.org/patch/msgid/ZSUfiX4J7v4Wn0cU@ogabbay-vm-u22.habana-labs.com
2023-10-12accel/ivpu: Add ivpu_bo_vaddr() and ivpu_bo_size()Jacek Lawrynowicz
Use: - ivpu_bo_vaddr(bo) instead of bo->kvaddr - ivpu_bo_size(bo) instead of bo->base.size This is a preparation for switch to a drm_gem_shmem_object as a base for ivpu_bo, where: - bo->kvaddr becomes bo->base.vaddr - bo->base.size becomes bo->base.base.size Using ivpu_bo_vaddr() and ivpu_bo_size() increases the readability of the code. Signed-off-by: Jacek Lawrynowicz <jacek.lawrynowicz@linux.intel.com> Reviewed-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com> Reviewed-by: Jeffrey Hugo <quic_jhugo@quicinc.com> Signed-off-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20230901094957.168898-9-stanislaw.gruszka@linux.intel.com
2023-10-11Merge drm/drm-next into drm-misc-nextThomas Zimmermann
Updating drm-misc-next to the state of Linux v6.6-rc2. Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
2023-10-09accel/habanalabs/gaudi2: fix spmu mask creationOded Gabbay
event_types_num received from the user can be 0. In that case, the event_mask should be 0. In addition, to create a correct mask we need to match the number of event types to the bit location such that bit 0 represents a single event type, bit 1 represents 2 types and so on. Signed-off-by: Oded Gabbay <ogabbay@kernel.org> Reviewed-by: Ofir Bitton <obitton@habana.ai>
2023-10-09accel/habanalabs/gaudi2: perform hard-reset upon PCIe AXI drain eventTomer Tayar
Non-completed transactions from PCIe towards the device are handled by the AXI drain mechanism. This handling is in the PCIe level, but the transactions are still there in the device consuming some queues entries, and therefore the device must be reset. Modify to perform hard-reset upon PCIe AXI drain events. Signed-off-by: Tomer Tayar <ttayar@habana.ai> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs: fix bug in decoder wait for cs completionfarah kassabri
The decoder interrupts are handled in the interrupt context same as all user interrupts. In such case, the wait list should be protected by spin_lock_irqsave in order to avoid deadlock that might happen with the user submission flow. Signed-off-by: farah kassabri <fkassabri@habana.ai> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs: remove wrong doc for init_phys_pg_pack_from_userptrDafna Hirschfeld
The function does not pin the pages so remove that from the inline doc. Signed-off-by: Dafna Hirschfeld <dhirschfeld@habana.ai> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs: add missing debugfs function stubsArnd Bergmann
Two function stubs were removed in an earlier commit but are now needed again: drivers/accel/habanalabs/common/device.c: In function 'hl_device_init': drivers/accel/habanalabs/common/device.c:2231:14: error: implicit declaration of function 'hl_debugfs_device_init'; did you mean 'drm_debugfs_dev_init'? [-Werror=implicit-function-declaration] 2231 | rc = hl_debugfs_device_init(hdev); drivers/accel/habanalabs/common/device.c:2367:9: error: implicit declaration of function 'hl_debugfs_device_fini'; did you mean 'hl_debugfs_remove_file'? [-Werror=implicit-function-declaration] 2367 | hl_debugfs_device_fini(hdev); | ^~~~~~~~~~~~~~~~~~~~~~ Signed-off-by: Arnd Bergmann <arnd@arndb.de> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs: minor cosmetic update to habanalabs.hOded Gabbay
- Update copyright years - Align fields in struct hl_userptr - Fix comments Signed-off-by: Oded Gabbay <ogabbay@kernel.org> Reviewed-by: Ofir Bitton <obitton@habana.ai>
2023-10-09accel/habanalabs/gaudi: remove define used for simulatorOded Gabbay
We don't support simulator in upstream. Signed-off-by: Oded Gabbay <ogabbay@kernel.org> Reviewed-by: Ofir Bitton <obitton@habana.ai>
2023-10-09accel/habanalabs: remove leftover codeOded Gabbay
This code was added as part of a bigger feature which was never upstreamed, so remove this code. Signed-off-by: Oded Gabbay <ogabbay@kernel.org> Reviewed-by: Ofir Bitton <obitton@habana.ai>
2023-10-09accel/habanalabs: print device name when it is removedOded Gabbay
Notifies the user which device was removed. It is important in a server with multiple devices. Signed-off-by: Oded Gabbay <ogabbay@kernel.org> Reviewed-by: Ofir Bitton <obitton@habana.ai>
2023-10-09accel/habanalabs: remove unused fieldOded Gabbay
flags in struct wait_interrupt_data is not used anywhere so remove it. Signed-off-by: Oded Gabbay <ogabbay@kernel.org> Reviewed-by: Ofir Bitton <obitton@habana.ai>
2023-10-09accel/habanalabs/gaudi: remove unused structure definitionOded Gabbay
struct gaudi_nic_status is not used anywhere in the code. Signed-off-by: Oded Gabbay <ogabbay@kernel.org> Reviewed-by: Ofir Bitton <obitton@habana.ai>
2023-10-09accel/habanalabs: trace dma map sgtableOhad Sharabi
Traces the DMA [un]map_sgtable using the new traces we added. Signed-off-by: Ohad Sharabi <osharabi@habana.ai> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs: remove unused asic functionsOded Gabbay
asic_dma_{un}map_single() asic-specific functions are no longer called from the common code, so delete these functions. In addition, delete the gaudi2 implementation as they are also not called. Signed-off-by: Oded Gabbay <ogabbay@kernel.org> Reviewed-by: Ofir Bitton <obitton@habana.ai>
2023-10-09accel/habanalabs: update boot status printAriel Suller
FW shutdown preparation status was added to spec. Signed-off-by: Ariel Suller <asuller@habana.ai> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs: extend preboot timeout when preboot might take longerDafna Hirschfeld
There are cases such when FW runs MBIST, that preboot is expected to take longer than the usual. In such cases the firmware reports status SECURITY_READY/IN_PREBOOT and we extend the timeout waiting for it. This is currently implemented for Gaudi2 only. Signed-off-by: Dafna Hirschfeld <dhirschfeld@habana.ai> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs: add debug prints to dump content of SG table for dma-bufTomer Tayar
Add debug prints to dump the content of the SG table which is prepared when the dma-buf map op is called. Signed-off-by: Tomer Tayar <ttayar@habana.ai> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs: add missing offset handling for dma-bufTomer Tayar
On devices with virtual device memory (Gaudi2 onwards), user can provide an offset within an allocated device memory from which he wants to export a dma-buf object. The offset value is verified by driver, but it is not taken into consideration when the importer driver maps the dma-buf and the SG table it prepared. Add the missing offset handling. Signed-off-by: Tomer Tayar <ttayar@habana.ai> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs: set hl_dmabuf_priv.device_address only when neededTomer Tayar
The device_address member of 'struct hl_dmabuf_priv' is used only when virtual device memory is not supported and dma-buf is exported from address. Set the value of this field only when it is relevant, and add "phys" to its name so it would be clearer that it can't be a device virtual address. Signed-off-by: Tomer Tayar <ttayar@habana.ai> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs: fix SG table creation for dma-buf mappingTomer Tayar
In some cases the calculated number of required entries for the dma-buf SG table is wrong. For example, if the page size is larger than both the dma max segment size of the importer device and from the exported side, or if the exported size is part of a phys_pg_pack that is composed of several pages. In these cases, redundant entries will be added to the SG table. Modify the method that the number of entries is calculated, and the way they are prepared. Signed-off-by: Tomer Tayar <ttayar@habana.ai> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs: split user interrupts pending listfarah kassabri
Currently driver maintain one list for both pending user interrupts which seeks to wait till CQ reaches it's target value and also the ones that seeks to get timestamp records when the CQ reaches it's target value. This causes delay in handling the waiters which gets higher priority than the timestamp records. In order to solve this, let's split the list into two, one for each case and each one is protected by it's own spinlock. Waiters will be handled within the interrupt context first, then the timestamp records will be set. Freeing the timestamp related memory will be handled in a workqueue. Signed-off-by: farah kassabri <fkassabri@habana.ai> Reviewed-by: Tomer Tayar <ttayar@habana.ai> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs: optimize timestamp registration handlerfarah kassabri
Currently we use dynamic allocation inside the irq handler in order to allocate free node to be used for the free jobs. This operation is expensive, especially when we deal with large burst of events records that get released at the same time. The alternative is to have pre allocated pool of free nodes and just fetch nodes from this pool at irq handling time instead of allocating them. In case the pool becomes full, then the driver will fallback to dynamic allocations. As part of the optimization also update the unregister flow upon re-using a timestamp record, by making the operation much simpler and quicker. We already have the record in the registration flow and now we just seek to re-use with different interrupt. Therefore, no need to look for buffer according to the user handle. Signed-off-by: farah kassabri <fkassabri@habana.ai> Reviewed-by: Tomer Tayar <ttayar@habana.ai> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs: fix bug in timestamp interrupt handlingfarah kassabri
There is a potential race between user thread seeking to re-use a timestamp record with new interrupt id, while this record is still in the middle of interrupt handling and it is about to be freed. Imagine the driver set the record in_use to 0 and only then fill the free_node information. This might lead to unpleasant scenario where the new registration thread detects the record as free to use, and change the cq buff address. That will cause the free_node to get the wrong buffer address to put refcount to. Signed-off-by: farah kassabri <fkassabri@habana.ai> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs: tiny refactor of hl_map_dmabuf()Tomer Tayar
alloc_sgt_from_device_pages() includes relatively many parameters, and in a subsequent change another offset parameter is going to be added. Using structure fields directly when calling this function, and in hl_map_dmabuf() it is done twice, makes it a little bit difficult to understand the meaning of the parameters. To make it clearer, assign the required values into local variables with explicit names, and use the variables when calling the function. Signed-off-by: Tomer Tayar <ttayar@habana.ai> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs: export dma-buf only if size/offset multiples of PAGE_SIZETomer Tayar
It is currently allowed for a user to export dma-buf with size and offset that are not multiples of PAGE_SIZE. The exported memory is mapped for the importer device, and there it will be rounded to PAGE_SIZE, leading to actually exporting more than the user intended to. To make the user be aware of it, accept only size and offset which are multiple of PAGE_SIZE. Signed-off-by: Tomer Tayar <ttayar@habana.ai> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs: use exported size from dma_buf and not from phys_pg_packTomer Tayar
The 'exported_size' member in 'struct hl_vm_phys_pg_pack' is used to keep the exported dma-buf size, to be later used when the buffer is mapped. However it is possible that the same phys_pg_pack will be exported more than once, and independently of when the mapping takes place. Remove this member from the phys_pg_pack structure, and simply use the size in the dma-buf object as the exported size when mapping. Signed-off-by: Tomer Tayar <ttayar@habana.ai> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs: always pass exported size to alloc_sgt_from_device_pages()Tomer Tayar
For Gaudi1 the exported dma-buf is always composed of a single page, and therefore the exported size is equal to this page's size. When calling alloc_sgt_from_device_pages(), we pass 0 as the exported size and internally calculate it as "number of pages * page size". This makes alloc_sgt_from_device_pages() less clear, because the exported size parameter is not understood as a restriction on the pages' size. Modify to always pass the exported size explicitly. Signed-off-by: Tomer Tayar <ttayar@habana.ai> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs: prevent sending heartbeat before events are enabledfarah kassabri
After the heartbeat mechanism is now expanded to be used also for EQ health check, we shouldn't send heartbeat messages to FW before driver allow events to be received from FW. Because if the driver will send two heartbeats before it enables events to be received from FW, then the EQ health check will fail and reset the device. Signed-off-by: farah kassabri <fkassabri@habana.ai> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs/gaudi2: add eq health check using irqfarah kassabri
This is the second patch for applying the eq health check mechanism which will add support for the interrupt flow for gaudi2 asic. More info about the interrupt mechanism: set a dedicated msix for the eq error interrupt, and add interrupt handler for it. when FW detects some issue with EQ like EQ_FULL, it'll raise that interrupt and driver should reset the device. Driver will inform the FW which msix index to use through the already existing handshake mechanism which will send msix info message to fw. Signed-off-by: farah kassabri <fkassabri@habana.ai> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs/gaudi2: handle eq health heartbeat checkfarah kassabri
Add mechanism for fw eq health check. this will be done using two flows: using the heartbeat mechanism and raising a dedicated interrupt to indicate an eq failure like EQ full. This patch will add implementation for the eq heartbeat for gaudi2 asic. More info about the heartbeat mechanism: Expand the heartbeat mechanism to monitor a new event that will be sent from FW upon receiving heartbeat message. that way driver can know that the eq is working or not. Signed-off-by: farah kassabri <fkassabri@habana.ai> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs/gaudi2: print power-mode changesMoti Haimovski
Print to kernel log any device power mode changes events reported by the FW. Signed-off-by: Moti Haimovski <mhaimovski@habana.ai> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs: add tsc clock sampling to clock sync infoHen Alon
Add tsc clock to clock sync info, to enable using this clock for sampling and sync it with device time. Signed-off-by: Hen Alon <halon@habana.ai> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs: fix inline doc typosDafna Hirschfeld
Fix two typos Signed-off-by: Dafna Hirschfeld <dhirschfeld@habana.ai> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs: disable events ioctls on control deviceDafna Hirschfeld
Because it is not used and also, for graceful reset to work those ioctls should run on the compute device. Signed-off-by: Dafna Hirschfeld <dhirschfeld@habana.ai> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
2023-10-09accel/habanalabs: move cpucp interface to linux/habanalabsDavid Meriin
The CPUCP interface is moved to a shared folder outside of accel as a pre-requisite to upstream the NIC drivers that will also include this file. Signed-off-by: David Meriin <dmeriin@habana.ai> Reviewed-by: Oded Gabbay <ogabbay@kernel.org> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
generated by cgit (git 2.34.1) at 2025-09-23 18:52:45 +0000