Merge tag 'misc-habanalabs-next-2020-07-24' of git://people.freedesktop.org/~gabbayo/linux into char-misc-next

Oded writes:

This tag contains the following changes for kernel 5.9-rc1:

- Remove rate limiters from GAUDI configuration (no longer needed).
- Set maximum amount of in-flight CS per ASIC type and increase
  the maximum amount for GAUDI.
- Refactor signal/wait command submissions code
- Calculate trace frequency from PLLs to show accurate profiling data
- Rephrase error messages to make them more clear to the common user
- Add statistics of dropped CS (counter per possible reason for drop)
- Get ECC information from firmware
- Remove support for partial SoC reset in Gaudi
- Halt device CPU only when reset is certain to happen. Sometimes we abort
  the reset procedure and in that case we can't leave device CPU in halt
  mode.
- set each CQ to its own work queue to prevent a race between completions
  on different CQs.
- Use queue pi/ci in order to determine queue occupancy. This is done to
  make the code reusable between current and future ASICs.
- Add more validations for user inputs.
- Refactor PCIe controller configuration to make the code reusable between
  current and future ASICs.
- Update firmware interface headers to latest version
- Move all common code to a dedicated common sub-folder

* tag 'misc-habanalabs-next-2020-07-24' of git://people.freedesktop.org/~gabbayo/linux: (28 commits)
  habanalabs: Fix memory leak in error flow of context initialization
  habanalabs: use no flags on MMU cache invalidation
  habanalabs: enable device before hw_init()
  habanalabs: create internal CB pool
  habanalabs: update hl_boot_if.h from firmware
  habanalabs: create common folder
  habanalabs: check for DMA errors when clearing memory
  habanalabs: verify queue can contain all cs jobs
  habanalabs: Assign each CQ with its own work queue
  habanalabs: halt device CPU only upon certain reset
  habanalabs: remove unused hash
  habanalabs: use queue pi/ci in order to determine queue occupancy
  habanalabs: configure maximum queues per asic
  habanalabs: remove soft-reset support from GAUDI
  habanalabs: PCIe iATU refactoring
  habanalabs: Extract ECC information from FW
  habanalabs: Add dropped cs statistics info struct
  habanalabs: extract cpu boot status lookup
  habanalabs: rephrase error messages
  habanalabs: Increase queues depth
  ...

1 files changed, 24 insertions, 3 deletions

diff --git a/include/uapi/misc/habanalabs.h b/include/uapi/misc/habanalabs.h
index f6267a8d7416..d5c4f983b7a8 100644
--- a/include/uapi/misc/habanalabs.h
+++ b/include/uapi/misc/habanalabs.h
@@ -263,6 +263,7 @@ enum hl_device_status {
  *                         time the driver was loaded.
  * HL_INFO_TIME_SYNC     - Retrieve the device's time alongside the host's time
  *                         for synchronization.
+ * HL_INFO_CS_COUNTERS   - Retrieve command submission counters
  */
 #define HL_INFO_HW_IP_INFO		0
 #define HL_INFO_HW_EVENTS		1
@@ -274,6 +275,7 @@ enum hl_device_status {
 #define HL_INFO_CLK_RATE		8
 #define HL_INFO_RESET_COUNT		9
 #define HL_INFO_TIME_SYNC		10
+#define HL_INFO_CS_COUNTERS		11
 
 #define HL_INFO_VERSION_MAX_LEN	128
 #define HL_INFO_CARD_NAME_MAX_LEN	16
@@ -338,6 +340,25 @@ struct hl_info_time_sync {
 	__u64 host_time;
 };
 
+/**
+ * struct hl_info_cs_counters - command submission counters
+ * @out_of_mem_drop_cnt: dropped due to memory allocation issue
+ * @parsing_drop_cnt: dropped due to error in packet parsing
+ * @queue_full_drop_cnt: dropped due to queue full
+ * @device_in_reset_drop_cnt: dropped due to device in reset
+ */
+struct hl_cs_counters {
+	__u64 out_of_mem_drop_cnt;
+	__u64 parsing_drop_cnt;
+	__u64 queue_full_drop_cnt;
+	__u64 device_in_reset_drop_cnt;
+};
+
+struct hl_info_cs_counters {
+	struct hl_cs_counters cs_counters;
+	struct hl_cs_counters ctx_cs_counters;
+};
+
 struct hl_info_args {
 	/* Location of relevant struct in userspace */
 	__u64 return_pointer;
@@ -530,13 +551,13 @@ union hl_wait_cs_args {
 	struct hl_wait_cs_out out;
 };
 
-/* Opcode to alloc device memory */
+/* Opcode to allocate device memory */
 #define HL_MEM_OP_ALLOC			0
 /* Opcode to free previously allocated device memory */
 #define HL_MEM_OP_FREE			1
-/* Opcode to map host memory */
+/* Opcode to map host and device memory */
 #define HL_MEM_OP_MAP			2
-/* Opcode to unmap previously mapped host memory */
+/* Opcode to unmap previously mapped host and device memory */
 #define HL_MEM_OP_UNMAP			3
 
 /* Memory flags */