habanalabs: set command buffer host VA dynamically
PATCH 1/7 habanalabs: set command buffer host VA dynamically
From: Dafna Hirschfeld <[email protected]>
Set the addresses for userspace command buffer dynamically instead of hard-coded. There is no reason for it to be hard-coded.
--- .../misc/habanalabs/common/command_buffer.c | 18 ++++++++++++++---- drivers/misc/habanalabs/common/habanalabs.h | 10 +++------- drivers/misc/habanalabs/common/memory.c | 2 +- drivers/misc/habanalabs/gaudi2/gaudi2.c | 3 --- drivers/misc/habanalabs/gaudi2/gaudi2P.h | 3 --- 5 files changed, 18 insertions(+), 18 deletions(-)
drivers/misc/habanalabs/common/command_buffer.c +14 -4
@@ -12,6 +12,8 @@#include <linux/slab.h> #include <linux/uaccess.h> +#define CB_VA_POOL_SIZE (4UL * SZ_1G) +static int cb_map_mem (struct hl_ctx *ctx,struct hl_cb *cb) {struct hl_device *hdev = ctx->hdev;
@@ -25,7 +27,7 @@ static int cb_map_mem(struct hl_ctx *ctx, struct hl_cb *cb)if (!hdev->supports_cb_mapping) {dev_err_ratelimited (hdev->dev, -"Cannot map CB because no VA range is allocated for CB mapping\n" ); +"Mapping a CB to the device's MMU is not supported\n" );return -EINVAL ; }
@@ -566,16 +568,23 @@ int hl_cb_va_pool_init(struct hl_ctx *ctx)return -ENOMEM ; } - rc =gen_pool_add (ctx->cb_va_pool, prop->cb_va_start_addr, - prop->cb_va_end_addr - prop->cb_va_start_addr,-1 ); + ctx->cb_va_pool_base =hl_reserve_va_block (hdev, ctx,HL_VA_RANGE_TYPE_HOST , +CB_VA_POOL_SIZE ,HL_MMU_VA_ALIGNMENT_NOT_NEEDED ); +if (!ctx->cb_va_pool_base) { + rc = -ENOMEM ; + goto err_pool_destroy; + } + rc =gen_pool_add (ctx->cb_va_pool, ctx->cb_va_pool_base,CB_VA_POOL_SIZE ,-1 );if (rc) {dev_err (hdev->dev,"Failed to add memory to VA gen pool for CB mapping\n" ); - goto err_pool_destroy; + goto err_unreserve_va_block; }return 0 ; +err_unreserve_va_block: +hl_unreserve_va_block (hdev, ctx, ctx->cb_va_pool_base,CB_VA_POOL_SIZE ); err_pool_destroy:gen_pool_destroy (ctx->cb_va_pool);
@@ -590,4 +599,5 @@ void hl_cb_va_pool_fini(struct hl_ctx *ctx)return ;gen_pool_destroy (ctx->cb_va_pool); +hl_unreserve_va_block (hdev, ctx, ctx->cb_va_pool_base,CB_VA_POOL_SIZE ); }
drivers/misc/habanalabs/common/habanalabs.h +3 -7
@@ -567,10 +567,6 @@ struct hl_hints_range {* @tpc_binning_mask: which TPCs are binned. 0 means usable and 1 means binned. * @dram_enabled_mask: which DRAMs are enabled. * @dram_binning_mask: which DRAMs are binned. 0 means usable, 1 means binned. -* @cb_va_start_addr: virtual start address of command buffers which are mapped -* to the device's MMU. -* @cb_va_end_addr: virtual end address of command buffers which are mapped to -* the device's MMU. * @dram_hints_align_mask: dram va hint addresses alignment mask which is used * for hints validity check. * @cfg_base_address: config space base address.
@@ -713,8 +709,6 @@ struct asic_fixed_properties {u64 tpc_binning_mask;u64 dram_enabled_mask;u64 dram_binning_mask; -u64 cb_va_start_addr; -u64 cb_va_end_addr;u64 dram_hints_align_mask;u64 cfg_base_address;u64 mmu_cache_mng_addr;
@@ -1803,6 +1797,7 @@ struct hl_cs_outcome_store {* @cb_va_pool: device VA pool for command buffers which are mapped to the * device's MMU. * @sig_mgr: encaps signals handle manager. +* @cb_va_pool_base: the base address for the device VA pool * @cs_sequence: sequence number for CS. Value is assigned to a CS and passed * to user so user could inquire about CS. It is used as * index to cs_pending array.
@@ -1838,6 +1833,7 @@ struct hl_ctx {struct hl_cs_counters_atomic cs_counters;struct gen_pool *cb_va_pool;struct hl_encaps_signals_mgr sig_mgr; +u64 cb_va_pool_base;u64 cs_sequence;u64 *dram_default_hops;spinlock_t cs_lock;
@@ -3600,7 +3596,7 @@ void hl_hw_block_mem_init(struct hl_ctx *ctx);void hl_hw_block_mem_fini (struct hl_ctx *ctx);u64 hl_reserve_va_block (struct hl_device *hdev,struct hl_ctx *ctx, -enum hl_va_range_type type,u32 size,u32 alignment); +enum hl_va_range_type type,u64 size,u32 alignment);int hl_unreserve_va_block (struct hl_device *hdev,struct hl_ctx *ctx,u64 start_addr,u64 size); int hl_pin_host_memory(struct hl_device *hdev, u64 addr, u64 size,
drivers/misc/habanalabs/common/memory.c +1 -1
@@ -755,7 +755,7 @@ static u64 get_va_block(struct hl_device *hdev,* - Return the start address of the virtual block. */ u64 hl_reserve_va_block (struct hl_device *hdev,struct hl_ctx *ctx, -enum hl_va_range_type type,u32 size,u32 alignment) +enum hl_va_range_type type,u64 size,u32 alignment) { return get_va_block(hdev, ctx->va_range[type], size, 0, max(alignment, ctx->va_range[type]->page_size),
drivers/misc/habanalabs/gaudi2/gaudi2.c +0 -3
@@ -2022,9 +2022,6 @@ static int gaudi2_set_fixed_properties(struct hl_device *hdev) prop->server_type =HL_SERVER_TYPE_UNKNOWN ; - prop->cb_va_start_addr =VA_HOST_SPACE_USER_MAPPED_CB_START ; - prop->cb_va_end_addr =VA_HOST_SPACE_USER_MAPPED_CB_END ; - prop->max_dec =NUMBER_OF_DEC ; prop->clk_pll_index =HL_GAUDI2_MME_PLL ;
drivers/misc/habanalabs/gaudi2/gaudi2P.h +0 -3
@@ -139,9 +139,6 @@#define VA_HOST_SPACE_HPAGE_START 0xFFF0800000000000ull#define VA_HOST_SPACE_HPAGE_END 0xFFF1000000000000ull/* 140TB */ -#define VA_HOST_SPACE_USER_MAPPED_CB_START 0xFFF1000000000000ull -#define VA_HOST_SPACE_USER_MAPPED_CB_END 0xFFF1000100000000ull/* 4GB */ -/* 140TB */ #define VA_HOST_SPACE_PAGE_SIZE (VA_HOST_SPACE_PAGE_END - VA_HOST_SPACE_PAGE_START)
Signature · 1 line
2.25.1
PATCH 2/7 habanalabs: perform context switch flow only if needed
From: Ofir Bitton <[email protected]>
Except Goya, none of our ASICs require context switch flow, hence we enable this flow only where it is needed.
--- drivers/misc/habanalabs/common/command_submission.c | 10 ++++++---- drivers/misc/habanalabs/common/habanalabs.h | 2 ++ drivers/misc/habanalabs/goya/goya.c | 1 + 3 files changed, 9 insertions(+), 4 deletions(-)
drivers/misc/habanalabs/common/command_submission.c +6 -4
@@ -1590,13 +1590,14 @@ static int hl_cs_ctx_switch(struct hl_fpriv *hpriv, union hl_cs_args *args,struct hl_device *hdev = hpriv->hdev;struct hl_ctx *ctx = hpriv->ctx;bool need_soft_reset = false; -int rc =0 , do_ctx_switch; +int rc =0 , do_ctx_switch =0 ;void __user *chunks;u32 num_chunks, tmp;u16 sob_count;int ret; - do_ctx_switch =atomic_cmpxchg (&ctx->thread_ctx_switch_token,1 ,0 ); +if (hdev->supports_ctx_switch) + do_ctx_switch =atomic_cmpxchg (&ctx->thread_ctx_switch_token,1 ,0 );if (do_ctx_switch || (args->in.cs_flags &HL_CS_FLAGS_FORCE_RESTORE )) {mutex_lock (&hpriv->restore_phase_mutex);
@@ -1667,9 +1668,10 @@ static int hl_cs_ctx_switch(struct hl_fpriv *hpriv, union hl_cs_args *args, } } - ctx->thread_ctx_switch_wait_token =1 ; +if (hdev->supports_ctx_switch) + ctx->thread_ctx_switch_wait_token =1 ; - }else if (!ctx->thread_ctx_switch_wait_token) { + }else if (hdev->supports_ctx_switch && !ctx->thread_ctx_switch_wait_token) { rc =hl_poll_timeout_memory (hdev, &ctx->thread_ctx_switch_wait_token, tmp, (tmp ==1 ),100 ,jiffies_to_usecs (hdev->timeout_jiffies), false);
drivers/misc/habanalabs/common/habanalabs.h +2 -0
@@ -3199,6 +3199,7 @@ struct hl_reset_info {* Used only for testing. * @heartbeat: Controls if we want to enable the heartbeat mechanism vs. the f/w, which verifies * that the f/w is always alive. Used only for testing. +* @supports_ctx_switch: true if a ctx switch is required upon first submission. */ struct hl_device {struct pci_dev *pdev;
@@ -3335,6 +3336,7 @@ struct hl_device {u8 compute_ctx_in_release;u8 supports_mmu_prefetch;u8 reset_upon_device_release; +u8 supports_ctx_switch;/* Parameters for bring-up */ u64 nic_ports_mask;
drivers/misc/habanalabs/goya/goya.c +1 -0
@@ -1025,6 +1025,7 @@ static int goya_sw_init(struct hl_device *hdev) hdev->asic_prop.supports_compute_reset = true; hdev->asic_prop.allow_inference_soft_reset = true; hdev->supports_wait_for_multi_cs = false; + hdev->supports_ctx_switch = true; hdev->asic_funcs->set_pci_memory_regions (hdev);
Signature · 1 line
2.25.1
PATCH 4/7 habanalabs/gaudi2: log critical events with no rate limit
In reply to Oded GabbayFrom: farah kassabri <[email protected]>
When we have a storm of errors of HBM ECC SERR we can reach a situation where driver start hard reset flow without logging the error cause that caused the hard reset due to logs rate limiting.
--- drivers/misc/habanalabs/gaudi2/gaudi2.c | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-)
drivers/misc/habanalabs/gaudi2/gaudi2.c +11 -4
@@ -8200,10 +8200,17 @@ static bool gaudi2_handle_hbm_mc_sei_err(struct hl_device *hdev, u16 event_type,return true; } -dev_err_ratelimited (hdev->dev, -"System Error Interrupt - HBM(%u) MC(%u) MC_CH(%u) MC_PC(%u). Critical(%u). Error cause: %s\n" , - hbm_id, mc_id, sei_data->hdr.mc_channel, sei_data->hdr.mc_pseudo_channel, - sei_data->hdr.is_critical, hbm_mc_sei_cause[cause_idx]); +if (sei_data->hdr.is_critical) +dev_err (hdev->dev, +"System Critical Error Interrupt - HBM(%u) MC(%u) MC_CH(%u) MC_PC(%u). Error cause: %s\n" , + hbm_id, mc_id, sei_data->hdr.mc_channel, sei_data->hdr.mc_pseudo_channel, + hbm_mc_sei_cause[cause_idx]); + +else +dev_err_ratelimited (hdev->dev, +"System Non-Critical Error Interrupt - HBM(%u) MC(%u) MC_CH(%u) MC_PC(%u). Error cause: %s\n" , + hbm_id, mc_id, sei_data->hdr.mc_channel, sei_data->hdr.mc_pseudo_channel, + hbm_mc_sei_cause[cause_idx]);/* Print error-specific info */ switch (cause_idx) {
Signature · 1 line
2.25.1
PATCH 3/7 habanalabs: ignore EEPROM errors during boot
In reply to Oded GabbayFrom: Ofir Bitton <[email protected]>
EEPROM errors reported by firmware are basically warnings and should not fail the boot process.
--- drivers/misc/habanalabs/common/firmware_if.c | 9 +++++++++ drivers/misc/habanalabs/include/common/hl_boot_if.h | 5 +++++ 2 files changed, 14 insertions(+)
drivers/misc/habanalabs/common/firmware_if.c +9 -0
@@ -573,6 +573,15 @@ static bool fw_report_boot_dev0(struct hl_device *hdev, u32 err_val,dev_dbg (hdev->dev,"Device status0 %#x\n" , sts_val);/* All warnings should go here in order not to reach the unknown error validation */ +if (err_val &CPU_BOOT_ERR0_EEPROM_FAIL ) { +dev_warn (hdev->dev, +"Device boot warning - EEPROM failure detected, default settings applied\n" ); +/* This is a warning so we don't want it to disable the +* device +*/ + err_val &= ~CPU_BOOT_ERR0_EEPROM_FAIL ; + } +if (err_val &CPU_BOOT_ERR0_DRAM_SKIPPED ) {dev_warn (hdev->dev,"Device boot warning - Skipped DRAM initialization\n" );
drivers/misc/habanalabs/include/common/hl_boot_if.h +5 -0
@@ -34,6 +34,7 @@ enum cpu_boot_err {CPU_BOOT_ERR_BINNING_FAIL =19 ,CPU_BOOT_ERR_TPM_FAIL =20 ,CPU_BOOT_ERR_TMP_THRESH_INIT_FAIL =21 , +CPU_BOOT_ERR_EEPROM_FAIL =22 ,CPU_BOOT_ERR_ENABLED =31 ,CPU_BOOT_ERR_SCND_EN =63 ,CPU_BOOT_ERR_LAST =64 /* we have 2 registers of 32 bits */
@@ -115,6 +116,9 @@ enum cpu_boot_err {* CPU_BOOT_ERR0_TMP_THRESH_INIT_FAIL Failed to set threshold for tmperature * sensor. * +* CPU_BOOT_ERR_EEPROM_FAIL Failed reading EEPROM data. Defaults +* are used. +* * CPU_BOOT_ERR0_ENABLED Error registers enabled. * This is a main indication that the * running FW populates the error
@@ -139,6 +143,7 @@ enum cpu_boot_err {#define CPU_BOOT_ERR0_BINNING_FAIL (1 << CPU_BOOT_ERR_BINNING_FAIL)#define CPU_BOOT_ERR0_TPM_FAIL (1 << CPU_BOOT_ERR_TPM_FAIL)#define CPU_BOOT_ERR0_TMP_THRESH_INIT_FAIL (1 << CPU_BOOT_ERR_TMP_THRESH_INIT_FAIL) +#define CPU_BOOT_ERR0_EEPROM_FAIL (1 << CPU_BOOT_ERR_EEPROM_FAIL)#define CPU_BOOT_ERR0_ENABLED (1 << CPU_BOOT_ERR_ENABLED)#define CPU_BOOT_ERR1_ENABLED (1 << CPU_BOOT_ERR_ENABLED)
Signature · 1 line
2.25.1
PATCH 5/7 habanalabs: allow control device open during reset
In reply to Oded GabbayFrom: Ofir Bitton <[email protected]>
Monitoring apps would like to query device state at any time so we should allow it also during reset because it doesn't involve accessing the h/w.
--- drivers/misc/habanalabs/common/device.c | 22 +++++++++++++++++++ drivers/misc/habanalabs/common/habanalabs.h | 2 ++ .../misc/habanalabs/common/habanalabs_drv.c | 4 ++-- 3 files changed, 26 insertions(+), 2 deletions(-)
drivers/misc/habanalabs/common/device.c +22 -0
@@ -352,6 +352,28 @@ bool hl_device_operational(struct hl_device *hdev, } } +bool hl_ctrl_device_operational (struct hl_device *hdev, +enum hl_device_status *status) +{ +enum hl_device_status current_status; + + current_status =hl_device_status (hdev); +if (status) + *status = current_status; + +switch (current_status) { +case HL_DEVICE_STATUS_MALFUNCTION : +return false; +case HL_DEVICE_STATUS_IN_RESET : +case HL_DEVICE_STATUS_IN_RESET_AFTER_DEVICE_RELEASE : +case HL_DEVICE_STATUS_NEEDS_RESET : +case HL_DEVICE_STATUS_OPERATIONAL : +case HL_DEVICE_STATUS_IN_DEVICE_CREATION : +default : +return true; + } +} +static void hpriv_release (struct kref *ref) {u64 idle_mask[HL_BUSY_ENGINES_MASK_EXT_SIZE ] = {0 };
drivers/misc/habanalabs/common/habanalabs.h +2 -0
@@ -3486,6 +3486,8 @@ int hl_device_open(struct inode *inode, struct file *filp);int hl_device_open_ctrl (struct inode *inode,struct file *filp);bool hl_device_operational (struct hl_device *hdev,enum hl_device_status *status); +bool hl_ctrl_device_operational (struct hl_device *hdev, +enum hl_device_status *status);enum hl_device_status hl_device_status (struct hl_device *hdev);int hl_device_set_debug_mode (struct hl_device *hdev,struct hl_ctx *ctx,bool enable);int hl_hw_queues_create (struct hl_device *hdev);
drivers/misc/habanalabs/common/habanalabs_drv.c +2 -2
@@ -268,9 +268,9 @@ int hl_device_open_ctrl(struct inode *inode, struct file *filp)mutex_lock (&hdev->fpriv_ctrl_list_lock); -if (!hl_device_operational (hdev,NULL )) { +if (!hl_ctrl_device_operational (hdev,NULL )) {dev_err_ratelimited (hdev->dev_ctrl, -"Can't open %s because it is disabled or in reset\n" , +"Can't open %s because it is disabled\n" ,dev_name (hdev->dev_ctrl)); rc = -EPERM ; goto out_err;
Signature · 1 line
2.25.1
PATCH 6/7 habanalabs/gaudi2: dump detailed information upon RAZWI
In reply to Oded GabbayFrom: Ofir Bitton <[email protected]>
In order to improve debuggability, we add all available information when a RAZWI event occur.
--- drivers/misc/habanalabs/gaudi2/gaudi2.c | 206 ++++++++++++++++++------ 1 file changed, 155 insertions(+), 51 deletions(-)
drivers/misc/habanalabs/gaudi2/gaudi2.c +155 -51
@@ -1531,17 +1531,57 @@ static const u32 rtr_coordinates_to_rtr_id[NUM_OF_RTR_PER_DCORE * NUM_OF_DCORES]RTR_ID_X_Y (17 ,11 ) }; +enum rtr_id { +DCORE0_RTR0 , +DCORE0_RTR1 , +DCORE0_RTR2 , +DCORE0_RTR3 , +DCORE0_RTR4 , +DCORE0_RTR5 , +DCORE0_RTR6 , +DCORE0_RTR7 , +DCORE1_RTR0 , +DCORE1_RTR1 , +DCORE1_RTR2 , +DCORE1_RTR3 , +DCORE1_RTR4 , +DCORE1_RTR5 , +DCORE1_RTR6 , +DCORE1_RTR7 , +DCORE2_RTR0 , +DCORE2_RTR1 , +DCORE2_RTR2 , +DCORE2_RTR3 , +DCORE2_RTR4 , +DCORE2_RTR5 , +DCORE2_RTR6 , +DCORE2_RTR7 , +DCORE3_RTR0 , +DCORE3_RTR1 , +DCORE3_RTR2 , +DCORE3_RTR3 , +DCORE3_RTR4 , +DCORE3_RTR5 , +DCORE3_RTR6 , +DCORE3_RTR7 , +}; +static const u32 gaudi2_tpc_initiator_rtr_id[NUM_OF_TPC_PER_DCORE *NUM_OF_DCORES +1 ] = { -1 ,1 ,2 ,2 ,3 ,3 ,14 ,14 ,13 ,13 ,12 ,12 ,19 ,19 ,18 ,18 ,17 , -17 ,28 ,28 ,29 ,29 ,30 ,30 ,0 +DCORE0_RTR1 ,DCORE0_RTR1 ,DCORE0_RTR2 ,DCORE0_RTR2 ,DCORE0_RTR3 ,DCORE0_RTR3 , +DCORE1_RTR6 ,DCORE1_RTR6 ,DCORE1_RTR5 ,DCORE1_RTR5 ,DCORE1_RTR4 ,DCORE1_RTR4 , +DCORE2_RTR3 ,DCORE2_RTR3 ,DCORE2_RTR2 ,DCORE2_RTR2 ,DCORE2_RTR1 ,DCORE2_RTR1 , +DCORE3_RTR4 ,DCORE3_RTR4 ,DCORE3_RTR5 ,DCORE3_RTR5 ,DCORE3_RTR6 ,DCORE3_RTR6 , +DCORE0_RTR0 };static const u32 gaudi2_dec_initiator_rtr_id[NUMBER_OF_DEC ] = { -0 ,0 ,15 ,15 ,16 ,16 ,31 ,31 ,0 ,0 +DCORE0_RTR0 ,DCORE0_RTR0 ,DCORE1_RTR7 ,DCORE1_RTR7 ,DCORE2_RTR0 ,DCORE2_RTR0 , +DCORE3_RTR7 ,DCORE3_RTR7 ,DCORE0_RTR0 ,DCORE0_RTR0 };static const u32 gaudi2_nic_initiator_rtr_id[NIC_NUMBER_OF_MACROS ] = { -15 ,15 ,15 ,15 ,15 ,16 ,16 ,16 ,16 ,31 ,31 ,31 +DCORE1_RTR7 ,DCORE1_RTR7 ,DCORE1_RTR7 ,DCORE1_RTR7 ,DCORE1_RTR7 ,DCORE2_RTR0 , +DCORE2_RTR0 ,DCORE2_RTR0 ,DCORE2_RTR0 ,DCORE3_RTR7 ,DCORE3_RTR7 ,DCORE3_RTR7 };struct sft_info {
@@ -1554,11 +1594,11 @@ static const struct sft_info gaudi2_edma_initiator_sft_id[NUM_OF_EDMA_PER_DCORE };static const u32 gaudi2_pdma_initiator_rtr_id[NUM_OF_PDMA ] = { -0 ,0 +DCORE0_RTR0 ,DCORE0_RTR0 };static const u32 gaudi2_rot_initiator_rtr_id[NUM_OF_ROT ] = { -16 ,31 +DCORE2_RTR0 ,DCORE3_RTR7 };struct mme_initiators_rtr_id {
@@ -7062,10 +7102,6 @@ static void gaudi2_razwi_rr_hbw_shared_printf_info(struct hl_device *hdev, razwi_lo =le32_to_cpu (razwi_info->hbw.rr_aw_razwi_lo_reg); razwi_xy =le32_to_cpu (razwi_info->hbw.rr_aw_razwi_id_reg); } - -dev_err_ratelimited (hdev->dev, -"%s-RAZWI SHARED RR HBW WR error, captured address HI 0x%x LO 0x%x, Initiator coordinates 0x%x\n" , - name, razwi_hi, razwi_lo, razwi_xy); }else {if (read_razwi_regs) { razwi_hi =RREG32 (rtr_mstr_if_base_addr +RR_SHRD_HBW_AR_RAZWI_HI );
@@ -7076,11 +7112,11 @@ static void gaudi2_razwi_rr_hbw_shared_printf_info(struct hl_device *hdev, razwi_lo =le32_to_cpu (razwi_info->hbw.rr_ar_razwi_lo_reg); razwi_xy =le32_to_cpu (razwi_info->hbw.rr_ar_razwi_id_reg); } - -dev_err_ratelimited (hdev->dev, -"%s-RAZWI SHARED RR HBW AR error, captured address HI 0x%x LO 0x%x, Initiator coordinates 0x%x\n" , - name, razwi_hi, razwi_lo, razwi_xy); } + +dev_err_ratelimited (hdev->dev, +"%s-RAZWI SHARED RR HBW %s error, address %#llx, Initiator coordinates 0x%x\n" , + name, is_write ?"WR" :"RD" , (u64 )razwi_hi <<32 | razwi_lo, razwi_xy); } static void gaudi2_razwi_rr_lbw_shared_printf_info(struct hl_device *hdev,
@@ -7338,7 +7374,79 @@ static void gaudi2_check_if_razwi_happened(struct hl_device *hdev)gaudi2_ack_module_razwi_event_handler (hdev,RAZWI_ROT , mod_idx,0 ,NULL ); } -static void gaudi2_razwi_unmapped_addr_hbw_printf_info (struct hl_device *hdev, +static const char *gaudi2_get_initiators_name (u32 rtr_id) +{ +switch (rtr_id) { +case DCORE0_RTR0 : +return "DEC0/1/8/9, TPC24, PDMA0/1, PMMU, PCIE_IF, EDMA0/2, HMMU0/2/4/6, CPU" ; +case DCORE0_RTR1 : +return "TPC0/1" ; +case DCORE0_RTR2 : +return "TPC2/3" ; +case DCORE0_RTR3 : +return "TPC4/5" ; +case DCORE0_RTR4 : +return "MME0_SBTE0/1" ; +case DCORE0_RTR5 : +return "MME0_WAP0/SBTE2" ; +case DCORE0_RTR6 : +return "MME0_CTRL_WR/SBTE3" ; +case DCORE0_RTR7 : +return "MME0_WAP1/CTRL_RD/SBTE4" ; +case DCORE1_RTR0 : +return "MME1_WAP1/CTRL_RD/SBTE4" ; +case DCORE1_RTR1 : +return "MME1_CTRL_WR/SBTE3" ; +case DCORE1_RTR2 : +return "MME1_WAP0/SBTE2" ; +case DCORE1_RTR3 : +return "MME1_SBTE0/1" ; +case DCORE1_RTR4 : +return "TPC10/11" ; +case DCORE1_RTR5 : +return "TPC8/9" ; +case DCORE1_RTR6 : +return "TPC6/7" ; +case DCORE1_RTR7 : +return "DEC2/3, NIC0/1/2/3/4, ARC_FARM, KDMA, EDMA1/3, HMMU1/3/5/7" ; +case DCORE2_RTR0 : +return "DEC4/5, NIC5/6/7/8, EDMA4/6, HMMU8/10/12/14, ROT0" ; +case DCORE2_RTR1 : +return "TPC16/17" ; +case DCORE2_RTR2 : +return "TPC14/15" ; +case DCORE2_RTR3 : +return "TPC12/13" ; +case DCORE2_RTR4 : +return "MME2_SBTE0/1" ; +case DCORE2_RTR5 : +return "MME2_WAP0/SBTE2" ; +case DCORE2_RTR6 : +return "MME2_CTRL_WR/SBTE3" ; +case DCORE2_RTR7 : +return "MME2_WAP1/CTRL_RD/SBTE4" ; +case DCORE3_RTR0 : +return "MME3_WAP1/CTRL_RD/SBTE4" ; +case DCORE3_RTR1 : +return "MME3_CTRL_WR/SBTE3" ; +case DCORE3_RTR2 : +return "MME3_WAP0/SBTE2" ; +case DCORE3_RTR3 : +return "MME3_SBTE0/1" ; +case DCORE3_RTR4 : +return "TPC18/19" ; +case DCORE3_RTR5 : +return "TPC20/21" ; +case DCORE3_RTR6 : +return "TPC22/23" ; +case DCORE3_RTR7 : +return "DEC6/7, NIC9/10/11, EDMA5/7, HMMU9/11/13/15, ROT1, PSOC" ; +default : +return "N/A" ; + } +} + +static void gaudi2_razwi_unmapped_addr_hbw_printf_info (struct hl_device *hdev,u32 rtr_id,u64 rtr_ctrl_base_addr,bool is_write) {u32 razwi_hi, razwi_lo;
@@ -7347,50 +7455,47 @@ static void gaudi2_razwi_unmapped_addr_hbw_printf_info(struct hl_device *hdev, razwi_hi =RREG32 (rtr_ctrl_base_addr +DEC_RAZWI_HBW_AW_ADDR_HI ); razwi_lo =RREG32 (rtr_ctrl_base_addr +DEC_RAZWI_HBW_AW_ADDR_LO ); -dev_err_ratelimited (hdev->dev, -"RAZWI PSOC unmapped HBW WR error, ctr_base 0x%llx, captured address HI 0x%x, LO 0x%x\n" , - rtr_ctrl_base_addr, razwi_hi, razwi_lo); -/* Clear set indication */ WREG32 (rtr_ctrl_base_addr +DEC_RAZWI_HBW_AW_SET ,0x1 ); }else { razwi_hi =RREG32 (rtr_ctrl_base_addr +DEC_RAZWI_HBW_AR_ADDR_HI ); - razwi_lo =RREG32 (rtr_ctrl_base_addr +DEC_RAZWI_HBW_AR_ADDR_LO ); -dev_err_ratelimited (hdev->dev, -"RAZWI PSOC unmapped HBW AR error, ctr_base 0x%llx, captured address HI 0x%x, LO 0x%x\n" , - rtr_ctrl_base_addr, razwi_hi, razwi_lo); -/* Clear set indication */ WREG32 (rtr_ctrl_base_addr +DEC_RAZWI_HBW_AR_SET ,0x1 ); } + +dev_err_ratelimited (hdev->dev, +"RAZWI PSOC unmapped HBW %s error, rtr id %u, address %#llx\n" , + is_write ?"WR" :"RD" , rtr_id, (u64 )razwi_hi <<32 | razwi_lo); + +dev_err_ratelimited (hdev->dev, +"Initiators: %s\n" ,gaudi2_get_initiators_name (rtr_id)); } -static void gaudi2_razwi_unmapped_addr_lbw_printf_info (struct hl_device *hdev, -u64 rtr_ctrl_base_addr,bool is_write) +static void gaudi2_razwi_unmapped_addr_lbw_printf_info (struct hl_device *hdev,u32 rtr_id, +u64 rtr_ctrl_base_addr,bool is_write) {u32 razwi_addr;if (is_write) { razwi_addr =RREG32 (rtr_ctrl_base_addr +DEC_RAZWI_LBW_AW_ADDR ); -dev_err_ratelimited (hdev->dev, -"RAZWI PSOC unmapped LBW WR error, ctr_base 0x%llx, captured address 0x%x\n" , - rtr_ctrl_base_addr, razwi_addr); -/* Clear set indication */ WREG32 (rtr_ctrl_base_addr +DEC_RAZWI_LBW_AW_SET ,0x1 ); }else { razwi_addr =RREG32 (rtr_ctrl_base_addr +DEC_RAZWI_LBW_AR_ADDR ); -dev_err_ratelimited (hdev->dev, -"RAZWI PSOC unmapped LBW AR error, ctr_base 0x%llx, captured address 0x%x\n" , - rtr_ctrl_base_addr, razwi_addr); -/* Clear set indication */ WREG32 (rtr_ctrl_base_addr +DEC_RAZWI_LBW_AR_SET ,0x1 ); } + +dev_err_ratelimited (hdev->dev, +"RAZWI PSOC unmapped LBW %s error, rtr id %u, address %#x\n" , + is_write ?"WR" :"RD" , rtr_id, razwi_addr); + +dev_err_ratelimited (hdev->dev, +"Initiators: %s\n" ,gaudi2_get_initiators_name (rtr_id)); }/* PSOC RAZWI interrupt occurs only when trying to access a bad address */
@@ -7408,21 +7513,16 @@ static void gaudi2_ack_psoc_razwi_event_handler(struct hl_device *hdev) } razwi_mask_info =RREG32 (mmPSOC_GLOBAL_CONF_RAZWI_MASK_INFO); - - xy = (razwi_mask_info &PSOC_GLOBAL_CONF_RAZWI_MASK_INFO_AXUSER_L_MASK ) - >>PSOC_GLOBAL_CONF_RAZWI_MASK_INFO_AXUSER_L_SHIFT ; + xy =FIELD_GET (PSOC_GLOBAL_CONF_RAZWI_MASK_INFO_AXUSER_L_MASK , razwi_mask_info);dev_err_ratelimited (hdev->dev, -"PSOC RAZWI interrupt: Mask %d, WAS_AR %d, WAS_AW %d, AXUSER_L 0x%x AXUSER_H 0x%x\n" , - (razwi_mask_info &PSOC_GLOBAL_CONF_RAZWI_MASK_INFO_MASK_MASK ) - >>PSOC_GLOBAL_CONF_RAZWI_MASK_INFO_MASK_SHIFT , - (razwi_mask_info &PSOC_GLOBAL_CONF_RAZWI_MASK_INFO_WAS_AR_MASK ) - >>PSOC_GLOBAL_CONF_RAZWI_MASK_INFO_WAS_AR_SHIFT , - (razwi_mask_info &PSOC_GLOBAL_CONF_RAZWI_MASK_INFO_WAS_AW_MASK ) - >>PSOC_GLOBAL_CONF_RAZWI_MASK_INFO_WAS_AW_SHIFT , xy, - (razwi_mask_info & -PSOC_GLOBAL_CONF_RAZWI_MASK_INFO_AXUSER_H_MASK ) - >>PSOC_GLOBAL_CONF_RAZWI_MASK_INFO_AXUSER_H_SHIFT ); +"PSOC RAZWI interrupt: Mask %d, AR %d, AW %d, AXUSER_L 0x%x AXUSER_H 0x%x\n" , +FIELD_GET (PSOC_GLOBAL_CONF_RAZWI_MASK_INFO_MASK_MASK , razwi_mask_info), +FIELD_GET (PSOC_GLOBAL_CONF_RAZWI_MASK_INFO_WAS_AR_MASK , razwi_mask_info), +FIELD_GET (PSOC_GLOBAL_CONF_RAZWI_MASK_INFO_WAS_AW_MASK , razwi_mask_info), + xy, +FIELD_GET (PSOC_GLOBAL_CONF_RAZWI_MASK_INFO_AXUSER_H_MASK , razwi_mask_info)); +if (xy ==0 ) {dev_err_ratelimited (hdev->dev,"PSOC RAZWI interrupt: received event from 0 rtr coordinates\n" );
@@ -7452,16 +7552,20 @@ static void gaudi2_ack_psoc_razwi_event_handler(struct hl_device *hdev) lbw_ar_set =RREG32 (rtr_ctrl_base_addr +DEC_RAZWI_LBW_AR_SET );if (hbw_aw_set) -gaudi2_razwi_unmapped_addr_hbw_printf_info (hdev, rtr_ctrl_base_addr, true); +gaudi2_razwi_unmapped_addr_hbw_printf_info (hdev, rtr_id, + rtr_ctrl_base_addr, true);if (hbw_ar_set) -gaudi2_razwi_unmapped_addr_hbw_printf_info (hdev, rtr_ctrl_base_addr, false); +gaudi2_razwi_unmapped_addr_hbw_printf_info (hdev, rtr_id, + rtr_ctrl_base_addr, false);if (lbw_aw_set) -gaudi2_razwi_unmapped_addr_lbw_printf_info (hdev, rtr_ctrl_base_addr, true); +gaudi2_razwi_unmapped_addr_lbw_printf_info (hdev, rtr_id, + rtr_ctrl_base_addr, true);if (lbw_ar_set) -gaudi2_razwi_unmapped_addr_lbw_printf_info (hdev, rtr_ctrl_base_addr, false); +gaudi2_razwi_unmapped_addr_lbw_printf_info (hdev, rtr_id, + rtr_ctrl_base_addr, false); clear:/* Clear Interrupts only on pldm or if f/w doesn't handle interrupts */
Signature · 1 line
2.25.1
PATCH 7/7 habanalabs: send device active message to f/w
In reply to Oded GabbayFrom: farah kassabri <[email protected]>
As part of the RAS that is done by the f/w, we should send a message to the f/w when a user either acquires or releases the device.
--- drivers/misc/habanalabs/common/device.c | 2 ++ drivers/misc/habanalabs/common/firmware_if.c | 15 +++++++++++++++ drivers/misc/habanalabs/common/habanalabs.h | 3 +++ drivers/misc/habanalabs/common/habanalabs_drv.c | 2 ++ drivers/misc/habanalabs/gaudi/gaudi.c | 6 ++++++ drivers/misc/habanalabs/gaudi2/gaudi2.c | 12 ++++++++++++ drivers/misc/habanalabs/gaudi2/gaudi2P.h | 1 + drivers/misc/habanalabs/goya/goya.c | 6 ++++++ drivers/misc/habanalabs/include/common/cpucp_if.h | 11 +++++++++++ 9 files changed, 58 insertions(+)
drivers/misc/habanalabs/common/device.c +2 -0
@@ -492,6 +492,8 @@ static int hl_device_release(struct inode *inode, struct file *filp) hdev->last_open_session_duration_jif = jiffies - hdev->last_successful_open_jif; + hdev->asic_funcs->send_device_activity (hdev, false); +return 0 ; }
drivers/misc/habanalabs/common/firmware_if.c +15 -0
@@ -454,6 +454,21 @@ void hl_fw_cpu_accessible_dma_pool_free(struct hl_device *hdev, size_t size, size); } +int hl_fw_send_device_activity (struct hl_device *hdev,bool open) +{ +struct cpucp_packet pkt; +int rc; + +memset (&pkt,0 ,sizeof (pkt)); + pkt.ctl =cpu_to_le32 (CPUCP_PACKET_ACTIVE_STATUS_SET <<CPUCP_PKT_CTL_OPCODE_SHIFT ); + pkt.value =cpu_to_le64 (open); + rc = hdev->asic_funcs->send_cpu_message (hdev, (u32 *) &pkt,sizeof (pkt),0 ,NULL ); +if (rc) +dev_err (hdev->dev,"failed to send device activity msg(%u)\n" , open); + +return rc; +} +int hl_fw_send_heartbeat (struct hl_device *hdev) {struct cpucp_packet hb_pkt;
drivers/misc/habanalabs/common/habanalabs.h +3 -0
@@ -1528,6 +1528,7 @@ struct engines_data {* @access_dev_mem: access device memory * @set_dram_bar_base: set the base of the DRAM BAR * @set_engine_cores: set a config command to enigne cores +* @send_device_activity: indication to FW about device availability */ struct hl_asic_funcs {int (*early_init)(struct hl_device *hdev);
@@ -1664,6 +1665,7 @@ struct hl_asic_funcs {u64 (*set_dram_bar_base)(struct hl_device *hdev,u64 addr);int (*set_engine_cores)(struct hl_device *hdev,u32 *core_ids,u32 num_cores,u32 core_command); +int (*send_device_activity)(struct hl_device *hdev,bool open); };
@@ -3717,6 +3719,7 @@ int hl_fw_dram_replaced_row_get(struct hl_device *hdev, struct cpucp_hbm_row_info *info);int hl_fw_dram_pending_row_get (struct hl_device *hdev,u32 *pend_rows_num);int hl_fw_cpucp_engine_core_asid_set (struct hl_device *hdev,u32 asid); +int hl_fw_send_device_activity (struct hl_device *hdev,bool open);int hl_pci_bars_map (struct hl_device *hdev,const char *const name[3 ],bool is_wc[3 ]);int hl_pci_elbi_read (struct hl_device *hdev,u64 addr,u32 *data);
drivers/misc/habanalabs/common/habanalabs_drv.c +2 -0
@@ -204,6 +204,8 @@ int hl_device_open(struct inode *inode, struct file *filp) goto out_err; } + rc = hdev->asic_funcs->send_device_activity (hdev, true); +list_add (&hpriv->dev_node, &hdev->fpriv_list);mutex_unlock (&hdev->fpriv_list_lock);
drivers/misc/habanalabs/gaudi/gaudi.c +6 -0
@@ -9132,6 +9132,11 @@ static void gaudi_add_device_attr(struct hl_device *hdev, struct attribute_group dev_vrm_attr_grp->attrs = gaudi_vrm_dev_attrs; } +static int gaudi_send_device_activity (struct hl_device *hdev,bool open) +{ +return 0 ; +} +static const struct hl_asic_funcs gaudi_funcs = { .early_init = gaudi_early_init, .early_fini = gaudi_early_fini,
@@ -9224,6 +9229,7 @@ static const struct hl_asic_funcs gaudi_funcs = { .mmu_get_real_page_size = hl_mmu_get_real_page_size, .access_dev_mem = hl_access_dev_mem, .set_dram_bar_base = gaudi_set_hbm_bar_base, + .send_device_activity = gaudi_send_device_activity, };/**
drivers/misc/habanalabs/gaudi2/gaudi2.c +12 -0
@@ -10031,6 +10031,17 @@ static int gaudi2_get_monitor_dump(struct hl_device *hdev, void *data)return -EOPNOTSUPP ; } +int gaudi2_send_device_activity (struct hl_device *hdev,bool open) +{ +struct gaudi2_device *gaudi2 = hdev->asic_specific; + +if (!(gaudi2->hw_cap_initialized &HW_CAP_CPU_Q ) || hdev->fw_major_version <37 ) +return 0 ; + +/* TODO: add check for FW version using minor ver once it's known */ +return hl_fw_send_device_activity (hdev, open); +} +static const struct hl_asic_funcs gaudi2_funcs = { .early_init = gaudi2_early_init, .early_fini = gaudi2_early_fini,
@@ -10127,6 +10138,7 @@ static const struct hl_asic_funcs gaudi2_funcs = { .access_dev_mem = hl_access_dev_mem, .set_dram_bar_base = gaudi2_set_hbm_bar_base, .set_engine_cores = gaudi2_set_engine_cores, + .send_device_activity = gaudi2_send_device_activity, };void gaudi2_set_asic_funcs (struct hl_device *hdev)
drivers/misc/habanalabs/gaudi2/gaudi2P.h +1 -0
@@ -553,5 +553,6 @@ void gaudi2_pb_print_security_errors(struct hl_device *hdev, u32 block_addr, u32 u32 offended_addr);int gaudi2_init_security (struct hl_device *hdev);void gaudi2_ack_protection_bits_errors (struct hl_device *hdev); +int gaudi2_send_device_activity (struct hl_device *hdev,bool open);#endif /* GAUDI2P_H_ */
drivers/misc/habanalabs/goya/goya.c +6 -0
@@ -5420,6 +5420,11 @@ static int goya_scrub_device_dram(struct hl_device *hdev, u64 val)return -EOPNOTSUPP ; } +static int goya_send_device_activity (struct hl_device *hdev,bool open) +{ +return 0 ; +} +static const struct hl_asic_funcs goya_funcs = { .early_init = goya_early_init, .early_fini = goya_early_fini,
@@ -5512,6 +5517,7 @@ static const struct hl_asic_funcs goya_funcs = { .mmu_get_real_page_size = hl_mmu_get_real_page_size, .access_dev_mem = hl_access_dev_mem, .set_dram_bar_base = goya_set_ddr_bar_base, + .send_device_activity = goya_send_device_activity, };/*
drivers/misc/habanalabs/include/common/cpucp_if.h +11 -0
@@ -636,6 +636,10 @@ enum pq_init_status {* passes the max size it allows the CpuCP to write to the structure, to prevent * data corruption in case of mismatched driver/FW versions. * Relevant only to Gaudi. +* +* CPUCP_PACKET_ACTIVE_STATUS_SET - +* LKD sends FW indication whether device is free or in use, this indication is reported +* also to the BMC. */ enum cpucp_packet_id {
@@ -691,6 +695,13 @@ enum cpucp_packet_id {CPUCP_PACKET_RESERVED4 ,/* not used */ CPUCP_PACKET_RESERVED5 ,/* not used */ CPUCP_PACKET_MONITOR_DUMP_GET ,/* debugfs */ +CPUCP_PACKET_RESERVED6 ,/* not used */ +CPUCP_PACKET_RESERVED7 ,/* not used */ +CPUCP_PACKET_RESERVED8 ,/* not used */ +CPUCP_PACKET_RESERVED9 ,/* not used */ +CPUCP_PACKET_RESERVED10 ,/* not used */ +CPUCP_PACKET_ACTIVE_STATUS_SET ,/* internal */ +CPUCP_PACKET_ID_MAX /* must be last */ };#define CPUCP_PACKET_FENCE_VAL 0xFE8CE7A5
Signature · 1 line
2.25.1