- Mar 11, 2022
-
-
Greg Kroah-Hartman authored
Merge tag 'misc-habanalabs-next-2022-02-28' of https://git.kernel.org/pub/scm/linux/kernel/git/ogabbay/linux into char-misc-next Oded writes: This tag contains habanalabs driver changes for v5.18: - Add new feature of recording time-stamp when a completion queue counter reaches a target value as determined by the userspace application. This is used by the graph compiler to accurately measure the time it takes for certain workloads to execute, which helps to fine-tune future compilations. - Add two new attributes to sysfs that expose the VRM and f/w OS version - Add a delay to the reset path that allows the driver to receive and handle additional events from the f/w before doing the reset. This can help when debugging why a reset event was received from the f/w. - Re-factor some of the sysfs code in the driver. Mainly, move functions from hwmgr.c to more relevant files and totally remove hwmgr.c file. - Fix multiple bugs such as races, use-after-free, ignoring error codes, etc. - As usual, multiple minor changes and small fixes. * tag 'misc-habanalabs-next-2022-02-28' of https://git.kernel.org/pub/scm/linux/kernel/git/ogabbay/linux: (50 commits) habanalabs: remove deprecated firmware states habanalabs: add an option to delay a device reset habanalabs: Add check for pci_enable_device habanalabs: Fix reset upon device release bug habanalabs: make sure device mem alloc is page aligned habanalabs/gaudi: add missing handling of NIC related events habanalabs/gaudi: handle axi errors from NIC engines habanalabs: allow user to set allocation page size habanalabs: use kernel-doc for memory ioctl documentation habanalabs: avoid using an uninitialized variable habanalabs: set max power on device init per ASIC habanalabs: use proper max_power variable for device utilization habanalabs: enable stop-on-error debugfs setting per ASIC habanalabs: change function to static habanalabs: add missing include of vmalloc.h habanalabs: fix use-after-free bug habanalabs: rephrase error messages in PCI initialization habanalabs: fix spelling mistake habanalabs: Timestamps buffers registration habanalabs: fix race when waiting on encaps signal ...
-
Greg Kroah-Hartman authored
Merge tag 'fpga-for-5.18-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/mdf/linux-fpga into char-misc-next Moritz writes: FPGA Manager changes for 5.18-rc1 Christophe's patch is part of a larger effort of removing the use of the deprecated pci-dma-compat API. All patches have been reviewed on the mailing list, and have been in the last few linux-next releases (as part of our for-next branch) without issues. Signed-off-by:
Moritz Fischer <mdf@kernel.org> * tag 'fpga-for-5.18-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/mdf/linux-fpga: fpga: dfl: pci: Remove usage of the deprecated "pci-dma-compat.h" API
-
- Mar 02, 2022
-
-
Ulf Hansson authored
The struct device *dev, is no longer needed at various functions, let's therefore drop it to fix the build errors/warnings. Fixes: 7570fb41 ("mmc: rtsx: Let MMC core handle runtime PM") Cc: Kai-Heng Feng <kai.heng.feng@canonical.com> Reported-by:
kernel test robot <lkp@intel.com> Signed-off-by:
Ulf Hansson <ulf.hansson@linaro.org> Link: https://lore.kernel.org/r/20220301115300.64332-1-ulf.hansson@linaro.org Signed-off-by:
Greg Kroah-Hartman <gregkh@linuxfoundation.org>
-
- Mar 01, 2022
-
-
Greg Kroah-Hartman authored
This reverts commit d0cac243. THe MFD maintainer has some remaining objections that need to be resolved before it can be accepted. Link: https://lore.kernel.org/r/20220228193928.3ec6ee98@canb.auug.org.au Reported-by:
Lee Jones <lee.jones@linaro.org> Reported-by:
Stephen Rothwell <sfr@canb.auug.org.au> Cc: Robert Marko <robert.marko@sartura.hr> Signed-off-by:
-
- Feb 28, 2022
-
-
Ofir Bitton authored
During driver and F/W handshake, driver waits for F/W to reach certain states in order to progress with the boot flow. Some of the states were deprecated a long time ago and were never present on official firmwares. Therefore, let's remove them from the handshake process. Signed-off-by:
Ofir Bitton <obitton@habana.ai> Reviewed-by:
Oded Gabbay <ogabbay@kernel.org> Signed-off-by:
-
Tomer Tayar authored
Several H/W events can be sent adjacently, even due to a single error. If a hard-reset is triggered as part of handling one of these events, the following events won't be handled. The debug info from these missed events is important, sometimes even more important than the one that was handled. To allow handling these close events, add an option to delay a device reset and use it when resetting due to H/W events. Signed-off-by:
Tomer Tayar <ttayar@habana.ai> Reviewed-by:
-
Jiasheng Jiang authored
As the potential failure of the pci_enable_device(), it should be better to check the return value and return error if fails. Fixes: 70b2f993 ("habanalabs: create common folder") Signed-off-by:
Jiasheng Jiang <jiasheng@iscas.ac.cn> Reviewed-by:
-
farah kassabri authored
In case user application was interrupted while some cs still in-flight or in the middle of completion handling in driver, the last refcount of the kernel private data for the user process will not be put in the fd close flow, but in the cs completion workqueue context. This means that the device reset-upon-device-release will be called from that context. During the reset flow, the driver flushes all the cs workqueue to ensure that any scheduled work has run to completion, and since we are running from the completion context we will have deadlock. Therefore, we need to skip flushing the workqueue in those cases. It is safe to do it because the user won't be able to release the device unless the workqueues are already empty. Signed-off-by:
farah kassabri <fkassabri@habana.ai> Reviewed-by:
-
Ohad Sharabi authored
Working with MMU that supports multiple page sizes requires that mapping of a page of a certain size will be aligned to the same size (e.g. the physical address of 32MB page shall be aligned to 32MB). To achieve this the gen_poll allocation is now using the "align" variant to comply with the alignment requirements. Signed-off-by:
Ohad Sharabi <osharabi@habana.ai> Reviewed-by:
-
Oded Gabbay authored
There are a few events that can arrive from the f/w and without proper handling can cause errors to appear in the kernel log without reason. Add the relevant handling that was missing. Signed-off-by: -
Oded Gabbay authored
Various AXI errors can occur in the NIC engines and are reported to the driver by the f/w. Add code to print the errors and ack them to the f/w. Signed-off-by: -
Ohad Sharabi authored
In future ASICs the MMU will be able to work with multiple page sizes, thus a new flag is added to allow the user to set the requested page size. This flag is added since the whole DRAM is allocated for the user and the user also should be familiar with the memory usage use case. As such, the user may choose to "over allocate" memory in favor of performance (for instance- large page allocations covers more memory in less TLB entries). For example: say available page sizes are of 1MB and 32MB. If user wants to allocate 40MB the user can either set page size to 1MB and allocate the exact amount of memory (but will result in 40 TLB entries) or the user can use 32MB pages, "waste" 8MB of physical memory but occupy only 2 TLB entries. Note that this feature will be available only to ASIC that supports multiple DRAM page sizes. Signed-off-by:
-
Ohad Sharabi authored
Re-format the comments for the memory ioctl structure to be in kernel-doc style. Signed-off-by:
-
Tomer Tayar authored
Fix the following compilation warning in hl_cb_ioctl() @ command_buffer.c: warning: ‘device_va’ may be used uninitialized in this function Signed-off-by:
-
Tomer Tayar authored
For current devices there is a need to send the max power value to F/W during device init, for example because there might be several card types. In future devices, this info will be programmed in the device's EEPROM and will be read by F/W, and hence the driver should not send it. Modify the sending of the relevant message to be done only for ASIC types that need it. Signed-off-by:
-
Tomer Tayar authored
The max_power variable which is used for calculating the device utilization is the ASIC specific property which is set during init. However, the max value can be modified via sysfs, and thus the updated value in the device structure should be used instead. Signed-off-by:
-
Tomer Tayar authored
On Goya and Gaudi, the stop-on-error configuration can be set via debugfs. However, in future devices, this configuration will always be enabled. Modify the debugfs node to be allowed only for ASICs that support this dynamic configuration. Signed-off-by:
-
Oded Gabbay authored
handle_registration_node() is called directly from the irq handler in irq.c, so it can be static. Signed-off-by: -
Oded Gabbay authored
Use of vfree(), vmalloc_user(), vmalloc() and remap_vmalloc_range() requires this include in some architectures. Signed-off-by: -
Oded Gabbay authored
When the code iterates over the free list of physical pages nodes, it deletes the physical page node which is used as the iterator. Therefore, we need to use the safe version of the iteration to prevent use-after-free. Signed-off-by: -
Oded Gabbay authored
The iATU is an internal h/w machine inside Habana's PCI controller. Mentioning it by name doesn't say anything to the user. It is better to say the PCI controller initialization was not done successfully. Signed-off-by: -
Oded Gabbay authored
The name of the property is hints_range_reservation Signed-off-by: -
farah kassabri authored
Timestamp registration API allows the user to register a timestamp record event which will make the driver set timestamp when CQ counter reaches the target value and write it to a specific location specified by the user. This is a non blocking API, unlike the wait_for_interrupt which is a blocking one. Signed-off-by:
-
Dani Liberman authored
Scenario: 1. CS which is part of encaps signal has been completed and now executing kref_put to its encaps signal handle. The refcount of the handle decremented to 0, and called the encaps signal handle release function - hl_encaps_handle_do_release. 2. At this point the user starts waiting on the signal, and finds the encaps signal handle in the handlers list and increment the habdle refcount to 1. 3. Immediately after, hl_encaps_handle_do_release removed the handle from the list and free its memory. 4. Wait function using the handle although it has been freed. This scenario caused the slab area which was previously allocated for the handle to be poison overwritten which triggered kernel bug the next time the OS needed to allocate this slab. Fixed by getting the refcount of the handle only in case it is not zero. Signed-off-by:
Dani Liberman <dliberman@habana.ai> Reviewed-by:
-
Dan Carpenter authored
Smatch warns that: drivers/misc/habanalabs/common/command_buffer.c:471 hl_cb_ioctl() error: uninitialized symbol 'device_va'. Which is true, but harmless. Anyway, it's easy to silence this by adding a error check. Signed-off-by:Dan Carpenter <dan.carpenter@oracle.com> Reviewed-by:
-
Oded Gabbay authored
We print detailed messages inside the internal ioctl functions. No need to print a generic message at the end, it doesn't add any information. Signed-off-by: -
Tomer Tayar authored
The heartbeat thread is active during soft-reset, and it tries to send messages to CPU-CP core. Within the soft-reset, in the time window in which the device is marked as disabled, any CPU-CP command is "silently" skipped and a success value it returned. However, in addition to the return value, the heartbeat function also checks the F/W result, but because no command is sent in this time window, the result variable won't hold the expected value and we will have a false heartbeat failure. To avoid it, modify the "silent" skip to be done only in hard-reset. The CPU-CP should be able to handle messages during soft-reset. In addition to the heartbeat problem, this should also solve other issues in other flows that send messages during soft-reset and use the F/W result as it w/o being aware to the reset. Signed-off-by:
-
Oded Gabbay authored
There is a race in the user interrupts code, where between checking the target value and adding the new pend to the list, there is a chance the interrupt happened. In that case, no one will complete the node, and we will get a timeout on it. Signed-off-by: -
Oded Gabbay authored
When timeout is 0, we need to return the busy status in case the target value wasn't reached upon entry to the ioctl. Also return the correct timestamp. Signed-off-by: -
Oded Gabbay authored
This is not something we can do a workaround. It is clearly an error and we should notify the user that it is an error. Signed-off-by: -
Oded Gabbay authored
Currently we only expose to the user the ID of the first available user interrupt. To make user interrupts allocation truly dynamic, we need to also expose the number of user interrupts. Signed-off-by: -
Oded Gabbay authored
Copy the latest versions of the f/w specs files. Signed-off-by: -
Tomer Tayar authored
Add a missing error check in the sysfs show function for max_power. Signed-off-by:
-
Dani Liberman authored
In case of soft reset failure, hard reset should be initiated, but reset flags were not set to enable it, which caused another soft reset followed by another failure. Updated reset flags to enable hard reset flow in case of soft reset failure. Signed-off-by:
-
Tomer Tayar authored
Add a missing error check in the sysfs show functions for clk_max_freq_mhz and clk_cur_freq_mhz_show. Signed-off-by:
-
Tomer Tayar authored
If reading PLL info from F/W fails, the PLL info is not set in the "result" variable, and hence shouldn't be copied to the caller's array. Signed-off-by:
-
Oded Gabbay authored
Freeing phys_pg_pack includes calling to scrubbing functions of the device's memory, taking locks and possibly even calling reset. This is not something that should be done while holding a device-wide spinlock. Therefore, save the relevant objects on a local linked-list and after releasing the spinlock, traverse that list and free the phys_pg_pack objects. Signed-off-by: -
Ohad Sharabi authored
In order to support several device MMU blocks with different architectures (e.g. different HOP table size) we need to move to per-MMU properties rather than keeping those properties as ASIC properties. Refactoring the code to use "per-MMU proprties" is a major effort. To start making the transition towards this goal but still support taking the properties from ASIC properties (for code that currently uses them) this patch copies some of the properties to the "per-MMU" properties and later, when implementing the per-MMU properties, we would be able to delete the MMU props from the ASIC props. Signed-off-by:
-
Oded Gabbay authored
In future ASICs, there is no kernel TDR for new workloads that are submitted directly from user-space to the device. Therefore, the driver can NEVER know that a workload has timed-out. So, when the user asks us to wait for interrupt on the workload's completion, and the wait has timed-out, it doesn't mean the workload has timed-out. It only means the wait has timed-out, which is NOT an error from driver's perspective. Signed-off-by: -
Rajaravi Krishna Katta authored
Adds new sysfs entry to display firmware os version /sys/class/habanalabs/hl<n>/fw_os_ver Signed-off-by:
Rajaravi Krishna Katta <rkatta@habana.ai> Reviewed-by:
-
