Mirrors/u-boot
2
0
Fork 0
mirror of https://github.com/AsahiLinux/u-boot synced 2024-12-26 21:13:48 +00:00
Code Issues Projects Releases Packages Wiki Activity
u-boot/arch/mips/mach-octeon/include/mach/cvmx-pow.h
Heinrich Schuchardt 185f812c41 doc: replace @return by Return: 
Sphinx expects Return: and not @return to indicate a return value.

find . -name '*.c' -exec \
sed -i 's/^\(\s\)\*\(\s*\)@return\(\s\)/\1*\2Return:\3/' {} \;

find . -name '*.h' -exec \
sed -i 's/^\(\s\)\*\(\s*\)@return\(\s\)/\1*\2Return:\3/' {} \;

Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
2022-01-19 18:11:34 +01:00

2991 lines
89 KiB
C
Raw Blame History

/* SPDX-License-Identifier: GPL-2.0 */
/*
* Copyright (C) 2020 Marvell International Ltd.
*
* Interface to the hardware Scheduling unit.
*
* New, starting with SDK 1.7.0, cvmx-pow supports a number of
* extended consistency checks. The define
* CVMX_ENABLE_POW_CHECKS controls the runtime insertion of POW
* internal state checks to find common programming errors. If
* CVMX_ENABLE_POW_CHECKS is not defined, checks are by default
* enabled. For example, cvmx-pow will check for the following
* program errors or POW state inconsistency.
* - Requesting a POW operation with an active tag switch in
* progress.
* - Waiting for a tag switch to complete for an excessively
* long period. This is normally a sign of an error in locking
* causing deadlock.
* - Illegal tag switches from NULL_NULL.
* - Illegal tag switches from NULL.
* - Illegal deschedule request.
* - WQE pointer not matching the one attached to the core by
* the POW.
*/
#ifndef __CVMX_POW_H__
#define __CVMX_POW_H__
#include "cvmx-wqe.h"
#include "cvmx-pow-defs.h"
#include "cvmx-sso-defs.h"
#include "cvmx-address.h"
#include "cvmx-coremask.h"
/* Default to having all POW constancy checks turned on */
#ifndef CVMX_ENABLE_POW_CHECKS
#define CVMX_ENABLE_POW_CHECKS 1
#endif
/*
* Special type for CN78XX style SSO groups (0..255),
* for distinction from legacy-style groups (0..15)
*/
typedef union {
u8 xgrp;
/* Fields that map XGRP for backwards compatibility */
struct __attribute__((__packed__)) {
u8 group : 5;
u8 qus : 3;
};
} cvmx_xgrp_t;
/*
* Softwsare-only structure to convey a return value
* containing multiple information fields about an work queue entry
*/
typedef struct {
u32 tag;
u16 index;
u8 grp; /* Legacy group # (0..15) */
u8 tag_type;
} cvmx_pow_tag_info_t;
/**
* Wait flag values for pow functions.
*/
typedef enum {
CVMX_POW_WAIT = 1,
CVMX_POW_NO_WAIT = 0,
} cvmx_pow_wait_t;
/**
* POW tag operations. These are used in the data stored to the POW.
*/
typedef enum {
CVMX_POW_TAG_OP_SWTAG = 0L,
CVMX_POW_TAG_OP_SWTAG_FULL = 1L,
CVMX_POW_TAG_OP_SWTAG_DESCH = 2L,
CVMX_POW_TAG_OP_DESCH = 3L,
CVMX_POW_TAG_OP_ADDWQ = 4L,
CVMX_POW_TAG_OP_UPDATE_WQP_GRP = 5L,
CVMX_POW_TAG_OP_SET_NSCHED = 6L,
CVMX_POW_TAG_OP_CLR_NSCHED = 7L,
CVMX_POW_TAG_OP_NOP = 15L
} cvmx_pow_tag_op_t;
/**
* This structure defines the store data on a store to POW
*/
typedef union {
u64 u64;
struct {
u64 no_sched : 1;
u64 unused : 2;
u64 index : 13;
cvmx_pow_tag_op_t op : 4;
u64 unused2 : 2;
u64 qos : 3;
u64 grp : 4;
cvmx_pow_tag_type_t type : 3;
u64 tag : 32;
} s_cn38xx;
struct {
u64 no_sched : 1;
cvmx_pow_tag_op_t op : 4;
u64 unused1 : 4;
u64 index : 11;
u64 unused2 : 1;
u64 grp : 6;
u64 unused3 : 3;
cvmx_pow_tag_type_t type : 2;
u64 tag : 32;
} s_cn68xx_clr;
struct {
u64 no_sched : 1;
cvmx_pow_tag_op_t op : 4;
u64 unused1 : 12;
u64 qos : 3;
u64 unused2 : 1;
u64 grp : 6;
u64 unused3 : 3;
cvmx_pow_tag_type_t type : 2;
u64 tag : 32;
} s_cn68xx_add;
struct {
u64 no_sched : 1;
cvmx_pow_tag_op_t op : 4;
u64 unused1 : 16;
u64 grp : 6;
u64 unused3 : 3;
cvmx_pow_tag_type_t type : 2;
u64 tag : 32;
} s_cn68xx_other;
struct {
u64 rsvd_62_63 : 2;
u64 grp : 10;
cvmx_pow_tag_type_t type : 2;
u64 no_sched : 1;
u64 rsvd_48 : 1;
cvmx_pow_tag_op_t op : 4;
u64 rsvd_42_43 : 2;
u64 wqp : 42;
} s_cn78xx_other;
} cvmx_pow_tag_req_t;
union cvmx_pow_tag_req_addr {
u64 u64;
struct {
u64 mem_region : 2;
u64 reserved_49_61 : 13;
u64 is_io : 1;
u64 did : 8;
u64 addr : 40;
} s;
struct {
u64 mem_region : 2;
u64 reserved_49_61 : 13;
u64 is_io : 1;
u64 did : 8;
u64 node : 4;
u64 tag : 32;
u64 reserved_0_3 : 4;
} s_cn78xx;
};
/**
* This structure describes the address to load stuff from POW
*/
typedef union {
u64 u64;
/**
* Address for new work request loads (did<2:0> == 0)
*/
struct {
u64 mem_region : 2;
u64 reserved_49_61 : 13;
u64 is_io : 1;
u64 did : 8;
u64 reserved_4_39 : 36;
u64 wait : 1;
u64 reserved_0_2 : 3;
} swork;
struct {
u64 mem_region : 2;
u64 reserved_49_61 : 13;
u64 is_io : 1;
u64 did : 8;
u64 node : 4;
u64 reserved_32_35 : 4;
u64 indexed : 1;
u64 grouped : 1;
u64 rtngrp : 1;
u64 reserved_16_28 : 13;
u64 index : 12;
u64 wait : 1;
u64 reserved_0_2 : 3;
} swork_78xx;
/**
* Address for loads to get POW internal status
*/
struct {
u64 mem_region : 2;
u64 reserved_49_61 : 13;
u64 is_io : 1;
u64 did : 8;
u64 reserved_10_39 : 30;
u64 coreid : 4;
u64 get_rev : 1;
u64 get_cur : 1;
u64 get_wqp : 1;
u64 reserved_0_2 : 3;
} sstatus;
/**
* Address for loads to get 68XX SS0 internal status
*/
struct {
u64 mem_region : 2;
u64 reserved_49_61 : 13;
u64 is_io : 1;
u64 did : 8;
u64 reserved_14_39 : 26;
u64 coreid : 5;
u64 reserved_6_8 : 3;
u64 opcode : 3;
u64 reserved_0_2 : 3;
} sstatus_cn68xx;
/**
* Address for memory loads to get POW internal state
*/
struct {
u64 mem_region : 2;
u64 reserved_49_61 : 13;
u64 is_io : 1;
u64 did : 8;
u64 reserved_16_39 : 24;
u64 index : 11;
u64 get_des : 1;
u64 get_wqp : 1;
u64 reserved_0_2 : 3;
} smemload;
/**
* Address for memory loads to get SSO internal state
*/
struct {
u64 mem_region : 2;
u64 reserved_49_61 : 13;
u64 is_io : 1;
u64 did : 8;
u64 reserved_20_39 : 20;
u64 index : 11;
u64 reserved_6_8 : 3;
u64 opcode : 3;
u64 reserved_0_2 : 3;
} smemload_cn68xx;
/**
* Address for index/pointer loads
*/
struct {
u64 mem_region : 2;
u64 reserved_49_61 : 13;
u64 is_io : 1;
u64 did : 8;
u64 reserved_9_39 : 31;
u64 qosgrp : 4;
u64 get_des_get_tail : 1;
u64 get_rmt : 1;
u64 reserved_0_2 : 3;
} sindexload;
/**
* Address for a Index/Pointer loads to get SSO internal state
*/
struct {
u64 mem_region : 2;
u64 reserved_49_61 : 13;
u64 is_io : 1;
u64 did : 8;
u64 reserved_15_39 : 25;
u64 qos_grp : 6;
u64 reserved_6_8 : 3;
u64 opcode : 3;
u64 reserved_0_2 : 3;
} sindexload_cn68xx;
/**
* Address for NULL_RD request (did<2:0> == 4)
* when this is read, HW attempts to change the state to NULL if it is NULL_NULL
* (the hardware cannot switch from NULL_NULL to NULL if a POW entry is not available -
* software may need to recover by finishing another piece of work before a POW
* entry can ever become available.)
*/
struct {
u64 mem_region : 2;
u64 reserved_49_61 : 13;
u64 is_io : 1;
u64 did : 8;
u64 reserved_0_39 : 40;
} snull_rd;
} cvmx_pow_load_addr_t;
/**
* This structure defines the response to a load/SENDSINGLE to POW (except CSR reads)
*/
typedef union {
u64 u64;
/**
* Response to new work request loads
*/
struct {
u64 no_work : 1;
u64 pend_switch : 1;
u64 tt : 2;
u64 reserved_58_59 : 2;
u64 grp : 10;
u64 reserved_42_47 : 6;
u64 addr : 42;
} s_work;
/**
* Result for a POW Status Load (when get_cur==0 and get_wqp==0)
*/
struct {
u64 reserved_62_63 : 2;
u64 pend_switch : 1;
u64 pend_switch_full : 1;
u64 pend_switch_null : 1;
u64 pend_desched : 1;
u64 pend_desched_switch : 1;
u64 pend_nosched : 1;
u64 pend_new_work : 1;
u64 pend_new_work_wait : 1;
u64 pend_null_rd : 1;
u64 pend_nosched_clr : 1;
u64 reserved_51 : 1;
u64 pend_index : 11;
u64 pend_grp : 4;
u64 reserved_34_35 : 2;
u64 pend_type : 2;
u64 pend_tag : 32;
} s_sstatus0;
/**
* Result for a SSO Status Load (when opcode is SL_PENDTAG)
*/
struct {
u64 pend_switch : 1;
u64 pend_get_work : 1;
u64 pend_get_work_wait : 1;
u64 pend_nosched : 1;
u64 pend_nosched_clr : 1;
u64 pend_desched : 1;
u64 pend_alloc_we : 1;
u64 reserved_48_56 : 9;
u64 pend_index : 11;
u64 reserved_34_36 : 3;
u64 pend_type : 2;
u64 pend_tag : 32;
} s_sstatus0_cn68xx;
/**
* Result for a POW Status Load (when get_cur==0 and get_wqp==1)
*/
struct {
u64 reserved_62_63 : 2;
u64 pend_switch : 1;
u64 pend_switch_full : 1;
u64 pend_switch_null : 1;
u64 pend_desched : 1;
u64 pend_desched_switch : 1;
u64 pend_nosched : 1;
u64 pend_new_work : 1;
u64 pend_new_work_wait : 1;
u64 pend_null_rd : 1;
u64 pend_nosched_clr : 1;
u64 reserved_51 : 1;
u64 pend_index : 11;
u64 pend_grp : 4;
u64 pend_wqp : 36;
} s_sstatus1;
/**
* Result for a SSO Status Load (when opcode is SL_PENDWQP)
*/
struct {
u64 pend_switch : 1;
u64 pend_get_work : 1;
u64 pend_get_work_wait : 1;
u64 pend_nosched : 1;
u64 pend_nosched_clr : 1;
u64 pend_desched : 1;
u64 pend_alloc_we : 1;
u64 reserved_51_56 : 6;
u64 pend_index : 11;
u64 reserved_38_39 : 2;
u64 pend_wqp : 38;
} s_sstatus1_cn68xx;
struct {
u64 pend_switch : 1;
u64 pend_get_work : 1;
u64 pend_get_work_wait : 1;
u64 pend_nosched : 1;
u64 pend_nosched_clr : 1;
u64 pend_desched : 1;
u64 pend_alloc_we : 1;
u64 reserved_56 : 1;
u64 prep_index : 12;
u64 reserved_42_43 : 2;
u64 pend_tag : 42;
} s_sso_ppx_pendwqp_cn78xx;
/**
* Result for a POW Status Load (when get_cur==1, get_wqp==0, and get_rev==0)
*/
struct {
u64 reserved_62_63 : 2;
u64 link_index : 11;
u64 index : 11;
u64 grp : 4;
u64 head : 1;
u64 tail : 1;
u64 tag_type : 2;
u64 tag : 32;
} s_sstatus2;
/**
* Result for a SSO Status Load (when opcode is SL_TAG)
*/
struct {
u64 reserved_57_63 : 7;
u64 index : 11;
u64 reserved_45 : 1;
u64 grp : 6;
u64 head : 1;
u64 tail : 1;
u64 reserved_34_36 : 3;
u64 tag_type : 2;
u64 tag : 32;
} s_sstatus2_cn68xx;
struct {
u64 tailc : 1;
u64 reserved_60_62 : 3;
u64 index : 12;
u64 reserved_46_47 : 2;
u64 grp : 10;
u64 head : 1;
u64 tail : 1;
u64 tt : 2;
u64 tag : 32;
} s_sso_ppx_tag_cn78xx;
/**
* Result for a POW Status Load (when get_cur==1, get_wqp==0, and get_rev==1)
*/
struct {
u64 reserved_62_63 : 2;
u64 revlink_index : 11;
u64 index : 11;
u64 grp : 4;
u64 head : 1;
u64 tail : 1;
u64 tag_type : 2;
u64 tag : 32;
} s_sstatus3;
/**
* Result for a SSO Status Load (when opcode is SL_WQP)
*/
struct {
u64 reserved_58_63 : 6;
u64 index : 11;
u64 reserved_46 : 1;
u64 grp : 6;
u64 reserved_38_39 : 2;
u64 wqp : 38;
} s_sstatus3_cn68xx;
struct {
u64 reserved_58_63 : 6;
u64 grp : 10;
u64 reserved_42_47 : 6;
u64 tag : 42;
} s_sso_ppx_wqp_cn78xx;
/**
* Result for a POW Status Load (when get_cur==1, get_wqp==1, and get_rev==0)
*/
struct {
u64 reserved_62_63 : 2;
u64 link_index : 11;
u64 index : 11;
u64 grp : 4;
u64 wqp : 36;
} s_sstatus4;
/**
* Result for a SSO Status Load (when opcode is SL_LINKS)
*/
struct {
u64 reserved_46_63 : 18;
u64 index : 11;
u64 reserved_34 : 1;
u64 grp : 6;
u64 head : 1;
u64 tail : 1;
u64 reserved_24_25 : 2;
u64 revlink_index : 11;
u64 reserved_11_12 : 2;
u64 link_index : 11;
} s_sstatus4_cn68xx;
struct {
u64 tailc : 1;
u64 reserved_60_62 : 3;
u64 index : 12;
u64 reserved_38_47 : 10;
u64 grp : 10;
u64 head : 1;
u64 tail : 1;
u64 reserved_25 : 1;
u64 revlink_index : 12;
u64 link_index_vld : 1;
u64 link_index : 12;
} s_sso_ppx_links_cn78xx;
/**
* Result for a POW Status Load (when get_cur==1, get_wqp==1, and get_rev==1)
*/
struct {
u64 reserved_62_63 : 2;
u64 revlink_index : 11;
u64 index : 11;
u64 grp : 4;
u64 wqp : 36;
} s_sstatus5;
/**
* Result For POW Memory Load (get_des == 0 and get_wqp == 0)
*/
struct {
u64 reserved_51_63 : 13;
u64 next_index : 11;
u64 grp : 4;
u64 reserved_35 : 1;
u64 tail : 1;
u64 tag_type : 2;
u64 tag : 32;
} s_smemload0;
/**
* Result For SSO Memory Load (opcode is ML_TAG)
*/
struct {
u64 reserved_38_63 : 26;
u64 tail : 1;
u64 reserved_34_36 : 3;
u64 tag_type : 2;
u64 tag : 32;
} s_smemload0_cn68xx;
struct {
u64 reserved_39_63 : 25;
u64 tail : 1;
u64 reserved_34_36 : 3;
u64 tag_type : 2;
u64 tag : 32;
} s_sso_iaq_ppx_tag_cn78xx;
/**
* Result For POW Memory Load (get_des == 0 and get_wqp == 1)
*/
struct {
u64 reserved_51_63 : 13;
u64 next_index : 11;
u64 grp : 4;
u64 wqp : 36;
} s_smemload1;
/**
* Result For SSO Memory Load (opcode is ML_WQPGRP)
*/
struct {
u64 reserved_48_63 : 16;
u64 nosched : 1;
u64 reserved_46 : 1;
u64 grp : 6;
u64 reserved_38_39 : 2;
u64 wqp : 38;
} s_smemload1_cn68xx;
/**
* Entry structures for the CN7XXX chips.
*/
struct {
u64 reserved_39_63 : 25;
u64 tailc : 1;
u64 tail : 1;
u64 reserved_34_36 : 3;
u64 tt : 2;
u64 tag : 32;
} s_sso_ientx_tag_cn78xx;
struct {
u64 reserved_62_63 : 2;
u64 head : 1;
u64 nosched : 1;
u64 reserved_56_59 : 4;
u64 grp : 8;
u64 reserved_42_47 : 6;
u64 wqp : 42;
} s_sso_ientx_wqpgrp_cn73xx;
struct {
u64 reserved_62_63 : 2;
u64 head : 1;
u64 nosched : 1;
u64 reserved_58_59 : 2;
u64 grp : 10;
u64 reserved_42_47 : 6;
u64 wqp : 42;
} s_sso_ientx_wqpgrp_cn78xx;
struct {
u64 reserved_38_63 : 26;
u64 pend_switch : 1;
u64 reserved_34_36 : 3;
u64 pend_tt : 2;
u64 pend_tag : 32;
} s_sso_ientx_pendtag_cn78xx;
struct {
u64 reserved_26_63 : 38;
u64 prev_index : 10;
u64 reserved_11_15 : 5;
u64 next_index_vld : 1;
u64 next_index : 10;
} s_sso_ientx_links_cn73xx;
struct {
u64 reserved_28_63 : 36;
u64 prev_index : 12;
u64 reserved_13_15 : 3;
u64 next_index_vld : 1;
u64 next_index : 12;
} s_sso_ientx_links_cn78xx;
/**
* Result For POW Memory Load (get_des == 1)
*/
struct {
u64 reserved_51_63 : 13;
u64 fwd_index : 11;
u64 grp : 4;
u64 nosched : 1;
u64 pend_switch : 1;
u64 pend_type : 2;
u64 pend_tag : 32;
} s_smemload2;
/**
* Result For SSO Memory Load (opcode is ML_PENTAG)
*/
struct {
u64 reserved_38_63 : 26;
u64 pend_switch : 1;
u64 reserved_34_36 : 3;
u64 pend_type : 2;
u64 pend_tag : 32;
} s_smemload2_cn68xx;
struct {
u64 pend_switch : 1;
u64 pend_get_work : 1;
u64 pend_get_work_wait : 1;
u64 pend_nosched : 1;
u64 pend_nosched_clr : 1;
u64 pend_desched : 1;
u64 pend_alloc_we : 1;
u64 reserved_34_56 : 23;
u64 pend_tt : 2;
u64 pend_tag : 32;
} s_sso_ppx_pendtag_cn78xx;
/**
* Result For SSO Memory Load (opcode is ML_LINKS)
*/
struct {
u64 reserved_24_63 : 40;
u64 fwd_index : 11;
u64 reserved_11_12 : 2;
u64 next_index : 11;
} s_smemload3_cn68xx;
/**
* Result For POW Index/Pointer Load (get_rmt == 0/get_des_get_tail == 0)
*/
struct {
u64 reserved_52_63 : 12;
u64 free_val : 1;
u64 free_one : 1;
u64 reserved_49 : 1;
u64 free_head : 11;
u64 reserved_37 : 1;
u64 free_tail : 11;
u64 loc_val : 1;
u64 loc_one : 1;
u64 reserved_23 : 1;
u64 loc_head : 11;
u64 reserved_11 : 1;
u64 loc_tail : 11;
} sindexload0;
/**
* Result for SSO Index/Pointer Load(opcode ==
* IPL_IQ/IPL_DESCHED/IPL_NOSCHED)
*/
struct {
u64 reserved_28_63 : 36;
u64 queue_val : 1;
u64 queue_one : 1;
u64 reserved_24_25 : 2;
u64 queue_head : 11;
u64 reserved_11_12 : 2;
u64 queue_tail : 11;
} sindexload0_cn68xx;
/**
* Result For POW Index/Pointer Load (get_rmt == 0/get_des_get_tail == 1)
*/
struct {
u64 reserved_52_63 : 12;
u64 nosched_val : 1;
u64 nosched_one : 1;
u64 reserved_49 : 1;
u64 nosched_head : 11;
u64 reserved_37 : 1;
u64 nosched_tail : 11;
u64 des_val : 1;
u64 des_one : 1;
u64 reserved_23 : 1;
u64 des_head : 11;
u64 reserved_11 : 1;
u64 des_tail : 11;
} sindexload1;
/**
* Result for SSO Index/Pointer Load(opcode == IPL_FREE0/IPL_FREE1/IPL_FREE2)
*/
struct {
u64 reserved_60_63 : 4;
u64 qnum_head : 2;
u64 qnum_tail : 2;
u64 reserved_28_55 : 28;
u64 queue_val : 1;
u64 queue_one : 1;
u64 reserved_24_25 : 2;
u64 queue_head : 11;
u64 reserved_11_12 : 2;
u64 queue_tail : 11;
} sindexload1_cn68xx;
/**
* Result For POW Index/Pointer Load (get_rmt == 1/get_des_get_tail == 0)
*/
struct {
u64 reserved_39_63 : 25;
u64 rmt_is_head : 1;
u64 rmt_val : 1;
u64 rmt_one : 1;
u64 rmt_head : 36;
} sindexload2;
/**
* Result For POW Index/Pointer Load (get_rmt == 1/get_des_get_tail == 1)
*/
struct {
u64 reserved_39_63 : 25;
u64 rmt_is_head : 1;
u64 rmt_val : 1;
u64 rmt_one : 1;
u64 rmt_tail : 36;
} sindexload3;
/**
* Response to NULL_RD request loads
*/
struct {
u64 unused : 62;
u64 state : 2;
} s_null_rd;
} cvmx_pow_tag_load_resp_t;
typedef union {
u64 u64;
struct {
u64 reserved_57_63 : 7;
u64 index : 11;
u64 reserved_45 : 1;
u64 grp : 6;
u64 head : 1;
u64 tail : 1;
u64 reserved_34_36 : 3;
u64 tag_type : 2;
u64 tag : 32;
} s;
} cvmx_pow_sl_tag_resp_t;
/**
* This structure describes the address used for stores to the POW.
* The store address is meaningful on stores to the POW. The hardware assumes that an aligned
* 64-bit store was used for all these stores.
* Note the assumption that the work queue entry is aligned on an 8-byte
* boundary (since the low-order 3 address bits must be zero).
* Note that not all fields are used by all operations.
*
* NOTE: The following is the behavior of the pending switch bit at the PP
* for POW stores (i.e. when did<7:3> == 0xc)
* - did<2:0> == 0 => pending switch bit is set
* - did<2:0> == 1 => no affect on the pending switch bit
* - did<2:0> == 3 => pending switch bit is cleared
* - did<2:0> == 7 => no affect on the pending switch bit
* - did<2:0> == others => must not be used
* - No other loads/stores have an affect on the pending switch bit
* - The switch bus from POW can clear the pending switch bit
*
* NOTE: did<2:0> == 2 is used by the HW for a special single-cycle ADDWQ command
* that only contains the pointer). SW must never use did<2:0> == 2.
*/
typedef union {
u64 u64;
struct {
u64 mem_reg : 2;
u64 reserved_49_61 : 13;
u64 is_io : 1;
u64 did : 8;
u64 addr : 40;
} stag;
} cvmx_pow_tag_store_addr_t; /* FIXME- this type is unused */
/**
* Decode of the store data when an IOBDMA SENDSINGLE is sent to POW
*/
typedef union {
u64 u64;
struct {
u64 scraddr : 8;
u64 len : 8;
u64 did : 8;
u64 unused : 36;
u64 wait : 1;
u64 unused2 : 3;
} s;
struct {
u64 scraddr : 8;
u64 len : 8;
u64 did : 8;
u64 node : 4;
u64 unused1 : 4;
u64 indexed : 1;
u64 grouped : 1;
u64 rtngrp : 1;
u64 unused2 : 13;
u64 index_grp_mask : 12;
u64 wait : 1;
u64 unused3 : 3;
} s_cn78xx;
} cvmx_pow_iobdma_store_t;
/* CSR typedefs have been moved to cvmx-pow-defs.h */
/*enum for group priority parameters which needs modification*/
enum cvmx_sso_group_modify_mask {
CVMX_SSO_MODIFY_GROUP_PRIORITY = 0x01,
CVMX_SSO_MODIFY_GROUP_WEIGHT = 0x02,
CVMX_SSO_MODIFY_GROUP_AFFINITY = 0x04
};
/**
* @INTERNAL
* Return the number of SSO groups for a given SoC model
*/
static inline unsigned int cvmx_sso_num_xgrp(void)
{
if (OCTEON_IS_MODEL(OCTEON_CN78XX))
return 256;
if (OCTEON_IS_MODEL(OCTEON_CNF75XX))
return 64;
if (OCTEON_IS_MODEL(OCTEON_CN73XX))
return 64;
printf("ERROR: %s: Unknown model\n", __func__);
return 0;
}
/**
* @INTERNAL
* Return the number of POW groups on current model.
* In case of CN78XX/CN73XX this is the number of equivalent
* "legacy groups" on the chip when it is used in backward
* compatible mode.
*/
static inline unsigned int cvmx_pow_num_groups(void)
{
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE))
return cvmx_sso_num_xgrp() >> 3;
else if (octeon_has_feature(OCTEON_FEATURE_CN68XX_WQE))
return 64;
else
return 16;
}
/**
* @INTERNAL
* Return the number of mask-set registers.
*/
static inline unsigned int cvmx_sso_num_maskset(void)
{
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE))
return 2;
else
return 1;
}
/**
* Get the POW tag for this core. This returns the current
* tag type, tag, group, and POW entry index associated with
* this core. Index is only valid if the tag type isn't NULL_NULL.
* If a tag switch is pending this routine returns the tag before
* the tag switch, not after.
*
* Return: Current tag
*/
static inline cvmx_pow_tag_info_t cvmx_pow_get_current_tag(void)
{
cvmx_pow_load_addr_t load_addr;
cvmx_pow_tag_info_t result;
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
cvmx_sso_sl_ppx_tag_t sl_ppx_tag;
cvmx_xgrp_t xgrp;
int node, core;
CVMX_SYNCS;
node = cvmx_get_node_num();
core = cvmx_get_local_core_num();
sl_ppx_tag.u64 = csr_rd_node(node, CVMX_SSO_SL_PPX_TAG(core));
result.index = sl_ppx_tag.s.index;
result.tag_type = sl_ppx_tag.s.tt;
result.tag = sl_ppx_tag.s.tag;
/* Get native XGRP value */
xgrp.xgrp = sl_ppx_tag.s.grp;
/* Return legacy style group 0..15 */
result.grp = xgrp.group;
} else if (octeon_has_feature(OCTEON_FEATURE_CN68XX_WQE)) {
cvmx_pow_sl_tag_resp_t load_resp;
load_addr.u64 = 0;
load_addr.sstatus_cn68xx.mem_region = CVMX_IO_SEG;
load_addr.sstatus_cn68xx.is_io = 1;
load_addr.sstatus_cn68xx.did = CVMX_OCT_DID_TAG_TAG5;
load_addr.sstatus_cn68xx.coreid = cvmx_get_core_num();
load_addr.sstatus_cn68xx.opcode = 3;
load_resp.u64 = csr_rd(load_addr.u64);
result.grp = load_resp.s.grp;
result.index = load_resp.s.index;
result.tag_type = load_resp.s.tag_type;
result.tag = load_resp.s.tag;
} else {
cvmx_pow_tag_load_resp_t load_resp;
load_addr.u64 = 0;
load_addr.sstatus.mem_region = CVMX_IO_SEG;
load_addr.sstatus.is_io = 1;
load_addr.sstatus.did = CVMX_OCT_DID_TAG_TAG1;
load_addr.sstatus.coreid = cvmx_get_core_num();
load_addr.sstatus.get_cur = 1;
load_resp.u64 = csr_rd(load_addr.u64);
result.grp = load_resp.s_sstatus2.grp;
result.index = load_resp.s_sstatus2.index;
result.tag_type = load_resp.s_sstatus2.tag_type;
result.tag = load_resp.s_sstatus2.tag;
}
return result;
}
/**
* Get the POW WQE for this core. This returns the work queue
* entry currently associated with this core.
*
* Return: WQE pointer
*/
static inline cvmx_wqe_t *cvmx_pow_get_current_wqp(void)
{
cvmx_pow_load_addr_t load_addr;
cvmx_pow_tag_load_resp_t load_resp;
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
cvmx_sso_sl_ppx_wqp_t sso_wqp;
int node = cvmx_get_node_num();
int core = cvmx_get_local_core_num();
sso_wqp.u64 = csr_rd_node(node, CVMX_SSO_SL_PPX_WQP(core));
if (sso_wqp.s.wqp)
return (cvmx_wqe_t *)cvmx_phys_to_ptr(sso_wqp.s.wqp);
return (cvmx_wqe_t *)0;
}
if (octeon_has_feature(OCTEON_FEATURE_CN68XX_WQE)) {
load_addr.u64 = 0;
load_addr.sstatus_cn68xx.mem_region = CVMX_IO_SEG;
load_addr.sstatus_cn68xx.is_io = 1;
load_addr.sstatus_cn68xx.did = CVMX_OCT_DID_TAG_TAG5;
load_addr.sstatus_cn68xx.coreid = cvmx_get_core_num();
load_addr.sstatus_cn68xx.opcode = 4;
load_resp.u64 = csr_rd(load_addr.u64);
if (load_resp.s_sstatus3_cn68xx.wqp)
return (cvmx_wqe_t *)cvmx_phys_to_ptr(load_resp.s_sstatus3_cn68xx.wqp);
else
return (cvmx_wqe_t *)0;
} else {
load_addr.u64 = 0;
load_addr.sstatus.mem_region = CVMX_IO_SEG;
load_addr.sstatus.is_io = 1;
load_addr.sstatus.did = CVMX_OCT_DID_TAG_TAG1;
load_addr.sstatus.coreid = cvmx_get_core_num();
load_addr.sstatus.get_cur = 1;
load_addr.sstatus.get_wqp = 1;
load_resp.u64 = csr_rd(load_addr.u64);
return (cvmx_wqe_t *)cvmx_phys_to_ptr(load_resp.s_sstatus4.wqp);
}
}
/**
* @INTERNAL
* Print a warning if a tag switch is pending for this core
*
* @param function Function name checking for a pending tag switch
*/
static inline void __cvmx_pow_warn_if_pending_switch(const char *function)
{
u64 switch_complete;
CVMX_MF_CHORD(switch_complete);
cvmx_warn_if(!switch_complete, "%s called with tag switch in progress\n", function);
}
/**
* Waits for a tag switch to complete by polling the completion bit.
* Note that switches to NULL complete immediately and do not need
* to be waited for.
*/
static inline void cvmx_pow_tag_sw_wait(void)
{
const u64 TIMEOUT_MS = 10; /* 10ms timeout */
u64 switch_complete;
u64 start_cycle;
if (CVMX_ENABLE_POW_CHECKS)
start_cycle = get_timer(0);
while (1) {
CVMX_MF_CHORD(switch_complete);
if (cvmx_likely(switch_complete))
break;
if (CVMX_ENABLE_POW_CHECKS) {
if (cvmx_unlikely(get_timer(start_cycle) > TIMEOUT_MS)) {
debug("WARNING: %s: Tag switch is taking a long time, possible deadlock\n",
__func__);
}
}
}
}
/**
* Synchronous work request. Requests work from the POW.
* This function does NOT wait for previous tag switches to complete,
* so the caller must ensure that there is not a pending tag switch.
*
* @param wait When set, call stalls until work becomes available, or
* times out. If not set, returns immediately.
*
* Return: Returns the WQE pointer from POW. Returns NULL if no work was
* available.
*/
static inline cvmx_wqe_t *cvmx_pow_work_request_sync_nocheck(cvmx_pow_wait_t wait)
{
cvmx_pow_load_addr_t ptr;
cvmx_pow_tag_load_resp_t result;
if (CVMX_ENABLE_POW_CHECKS)
__cvmx_pow_warn_if_pending_switch(__func__);
ptr.u64 = 0;
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
ptr.swork_78xx.node = cvmx_get_node_num();
ptr.swork_78xx.mem_region = CVMX_IO_SEG;
ptr.swork_78xx.is_io = 1;
ptr.swork_78xx.did = CVMX_OCT_DID_TAG_SWTAG;
ptr.swork_78xx.wait = wait;
} else {
ptr.swork.mem_region = CVMX_IO_SEG;
ptr.swork.is_io = 1;
ptr.swork.did = CVMX_OCT_DID_TAG_SWTAG;
ptr.swork.wait = wait;
}
result.u64 = csr_rd(ptr.u64);
if (result.s_work.no_work)
return NULL;
else
return (cvmx_wqe_t *)cvmx_phys_to_ptr(result.s_work.addr);
}
/**
* Synchronous work request. Requests work from the POW.
* This function waits for any previous tag switch to complete before
* requesting the new work.
*
* @param wait When set, call stalls until work becomes available, or
* times out. If not set, returns immediately.
*
* Return: Returns the WQE pointer from POW. Returns NULL if no work was
* available.
*/
static inline cvmx_wqe_t *cvmx_pow_work_request_sync(cvmx_pow_wait_t wait)
{
/* Must not have a switch pending when requesting work */
cvmx_pow_tag_sw_wait();
return (cvmx_pow_work_request_sync_nocheck(wait));
}
/**
* Synchronous null_rd request. Requests a switch out of NULL_NULL POW state.
* This function waits for any previous tag switch to complete before
* requesting the null_rd.
*
* Return: Returns the POW state of type cvmx_pow_tag_type_t.
*/
static inline cvmx_pow_tag_type_t cvmx_pow_work_request_null_rd(void)
{
cvmx_pow_load_addr_t ptr;
cvmx_pow_tag_load_resp_t result;
/* Must not have a switch pending when requesting work */
cvmx_pow_tag_sw_wait();
ptr.u64 = 0;
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
ptr.swork_78xx.mem_region = CVMX_IO_SEG;
ptr.swork_78xx.is_io = 1;
ptr.swork_78xx.did = CVMX_OCT_DID_TAG_NULL_RD;
ptr.swork_78xx.node = cvmx_get_node_num();
} else {
ptr.snull_rd.mem_region = CVMX_IO_SEG;
ptr.snull_rd.is_io = 1;
ptr.snull_rd.did = CVMX_OCT_DID_TAG_NULL_RD;
}
result.u64 = csr_rd(ptr.u64);
return (cvmx_pow_tag_type_t)result.s_null_rd.state;
}
/**
* Asynchronous work request.
* Work is requested from the POW unit, and should later be checked with
* function cvmx_pow_work_response_async.
* This function does NOT wait for previous tag switches to complete,
* so the caller must ensure that there is not a pending tag switch.
*
* @param scr_addr Scratch memory address that response will be returned to,
* which is either a valid WQE, or a response with the invalid bit set.
* Byte address, must be 8 byte aligned.
* @param wait 1 to cause response to wait for work to become available
* (or timeout)
* 0 to cause response to return immediately
*/
static inline void cvmx_pow_work_request_async_nocheck(int scr_addr, cvmx_pow_wait_t wait)
{
cvmx_pow_iobdma_store_t data;
if (CVMX_ENABLE_POW_CHECKS)
__cvmx_pow_warn_if_pending_switch(__func__);
/* scr_addr must be 8 byte aligned */
data.u64 = 0;
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
data.s_cn78xx.node = cvmx_get_node_num();
data.s_cn78xx.scraddr = scr_addr >> 3;
data.s_cn78xx.len = 1;
data.s_cn78xx.did = CVMX_OCT_DID_TAG_SWTAG;
data.s_cn78xx.wait = wait;
} else {
data.s.scraddr = scr_addr >> 3;
data.s.len = 1;
data.s.did = CVMX_OCT_DID_TAG_SWTAG;
data.s.wait = wait;
}
cvmx_send_single(data.u64);
}
/**
* Asynchronous work request.
* Work is requested from the POW unit, and should later be checked with
* function cvmx_pow_work_response_async.
* This function waits for any previous tag switch to complete before
* requesting the new work.
*
* @param scr_addr Scratch memory address that response will be returned to,
* which is either a valid WQE, or a response with the invalid bit set.
* Byte address, must be 8 byte aligned.
* @param wait 1 to cause response to wait for work to become available
* (or timeout)
* 0 to cause response to return immediately
*/
static inline void cvmx_pow_work_request_async(int scr_addr, cvmx_pow_wait_t wait)
{
/* Must not have a switch pending when requesting work */
cvmx_pow_tag_sw_wait();
cvmx_pow_work_request_async_nocheck(scr_addr, wait);
}
/**
* Gets result of asynchronous work request. Performs a IOBDMA sync
* to wait for the response.
*
* @param scr_addr Scratch memory address to get result from
* Byte address, must be 8 byte aligned.
* Return: Returns the WQE from the scratch register, or NULL if no work was
* available.
*/
static inline cvmx_wqe_t *cvmx_pow_work_response_async(int scr_addr)
{
cvmx_pow_tag_load_resp_t result;
CVMX_SYNCIOBDMA;
result.u64 = cvmx_scratch_read64(scr_addr);
if (result.s_work.no_work)
return NULL;
else
return (cvmx_wqe_t *)cvmx_phys_to_ptr(result.s_work.addr);
}
/**
* Checks if a work queue entry pointer returned by a work
* request is valid. It may be invalid due to no work
* being available or due to a timeout.
*
* @param wqe_ptr pointer to a work queue entry returned by the POW
*
* Return: 0 if pointer is valid
* 1 if invalid (no work was returned)
*/
static inline u64 cvmx_pow_work_invalid(cvmx_wqe_t *wqe_ptr)
{
return (!wqe_ptr); /* FIXME: improve */
}
/**
* Starts a tag switch to the provided tag value and tag type. Completion for
* the tag switch must be checked for separately.
* This function does NOT update the
* work queue entry in dram to match tag value and type, so the application must
* keep track of these if they are important to the application.
* This tag switch command must not be used for switches to NULL, as the tag
* switch pending bit will be set by the switch request, but never cleared by
* the hardware.
*
* NOTE: This should not be used when switching from a NULL tag. Use
* cvmx_pow_tag_sw_full() instead.
*
* This function does no checks, so the caller must ensure that any previous tag
* switch has completed.
*
* @param tag new tag value
* @param tag_type new tag type (ordered or atomic)
*/
static inline void cvmx_pow_tag_sw_nocheck(u32 tag, cvmx_pow_tag_type_t tag_type)
{
union cvmx_pow_tag_req_addr ptr;
cvmx_pow_tag_req_t tag_req;
if (CVMX_ENABLE_POW_CHECKS) {
cvmx_pow_tag_info_t current_tag;
__cvmx_pow_warn_if_pending_switch(__func__);
current_tag = cvmx_pow_get_current_tag();
cvmx_warn_if(current_tag.tag_type == CVMX_POW_TAG_TYPE_NULL_NULL,
"%s called with NULL_NULL tag\n", __func__);
cvmx_warn_if(current_tag.tag_type == CVMX_POW_TAG_TYPE_NULL,
"%s called with NULL tag\n", __func__);
cvmx_warn_if((current_tag.tag_type == tag_type) && (current_tag.tag == tag),
"%s called to perform a tag switch to the same tag\n", __func__);
cvmx_warn_if(
tag_type == CVMX_POW_TAG_TYPE_NULL,
"%s called to perform a tag switch to NULL. Use cvmx_pow_tag_sw_null() instead\n",
__func__);
}
/*
* Note that WQE in DRAM is not updated here, as the POW does not read
* from DRAM once the WQE is in flight. See hardware manual for
* complete details.
* It is the application's responsibility to keep track of the
* current tag value if that is important.
*/
tag_req.u64 = 0;
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
tag_req.s_cn78xx_other.op = CVMX_POW_TAG_OP_SWTAG;
tag_req.s_cn78xx_other.type = tag_type;
} else if (octeon_has_feature(OCTEON_FEATURE_CN68XX_WQE)) {
tag_req.s_cn68xx_other.op = CVMX_POW_TAG_OP_SWTAG;
tag_req.s_cn68xx_other.tag = tag;
tag_req.s_cn68xx_other.type = tag_type;
} else {
tag_req.s_cn38xx.op = CVMX_POW_TAG_OP_SWTAG;
tag_req.s_cn38xx.tag = tag;
tag_req.s_cn38xx.type = tag_type;
}
ptr.u64 = 0;
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
ptr.s_cn78xx.mem_region = CVMX_IO_SEG;
ptr.s_cn78xx.is_io = 1;
ptr.s_cn78xx.did = CVMX_OCT_DID_TAG_SWTAG;
ptr.s_cn78xx.node = cvmx_get_node_num();
ptr.s_cn78xx.tag = tag;
} else {
ptr.s.mem_region = CVMX_IO_SEG;
ptr.s.is_io = 1;
ptr.s.did = CVMX_OCT_DID_TAG_SWTAG;
}
/* Once this store arrives at POW, it will attempt the switch
software must wait for the switch to complete separately */
cvmx_write_io(ptr.u64, tag_req.u64);
}
/**
* Starts a tag switch to the provided tag value and tag type. Completion for
* the tag switch must be checked for separately.
* This function does NOT update the
* work queue entry in dram to match tag value and type, so the application must
* keep track of these if they are important to the application.
* This tag switch command must not be used for switches to NULL, as the tag
* switch pending bit will be set by the switch request, but never cleared by
* the hardware.
*
* NOTE: This should not be used when switching from a NULL tag. Use
* cvmx_pow_tag_sw_full() instead.
*
* This function waits for any previous tag switch to complete, and also
* displays an error on tag switches to NULL.
*
* @param tag new tag value
* @param tag_type new tag type (ordered or atomic)
*/
static inline void cvmx_pow_tag_sw(u32 tag, cvmx_pow_tag_type_t tag_type)
{
/*
* Note that WQE in DRAM is not updated here, as the POW does not read
* from DRAM once the WQE is in flight. See hardware manual for
* complete details. It is the application's responsibility to keep
* track of the current tag value if that is important.
*/
/*
* Ensure that there is not a pending tag switch, as a tag switch
* cannot be started if a previous switch is still pending.
*/
cvmx_pow_tag_sw_wait();
cvmx_pow_tag_sw_nocheck(tag, tag_type);
}
/**
* Starts a tag switch to the provided tag value and tag type. Completion for
* the tag switch must be checked for separately.
* This function does NOT update the
* work queue entry in dram to match tag value and type, so the application must
* keep track of these if they are important to the application.
* This tag switch command must not be used for switches to NULL, as the tag
* switch pending bit will be set by the switch request, but never cleared by
* the hardware.
*
* This function must be used for tag switches from NULL.
*
* This function does no checks, so the caller must ensure that any previous tag
* switch has completed.
*
* @param wqp pointer to work queue entry to submit. This entry is
* updated to match the other parameters
* @param tag tag value to be assigned to work queue entry
* @param tag_type type of tag
* @param group group value for the work queue entry.
*/
static inline void cvmx_pow_tag_sw_full_nocheck(cvmx_wqe_t *wqp, u32 tag,
cvmx_pow_tag_type_t tag_type, u64 group)
{
union cvmx_pow_tag_req_addr ptr;
cvmx_pow_tag_req_t tag_req;
unsigned int node = cvmx_get_node_num();
u64 wqp_phys = cvmx_ptr_to_phys(wqp);
if (CVMX_ENABLE_POW_CHECKS) {
cvmx_pow_tag_info_t current_tag;
__cvmx_pow_warn_if_pending_switch(__func__);
current_tag = cvmx_pow_get_current_tag();
cvmx_warn_if(current_tag.tag_type == CVMX_POW_TAG_TYPE_NULL_NULL,
"%s called with NULL_NULL tag\n", __func__);
cvmx_warn_if((current_tag.tag_type == tag_type) && (current_tag.tag == tag),
"%s called to perform a tag switch to the same tag\n", __func__);
cvmx_warn_if(
tag_type == CVMX_POW_TAG_TYPE_NULL,
"%s called to perform a tag switch to NULL. Use cvmx_pow_tag_sw_null() instead\n",
__func__);
if ((wqp != cvmx_phys_to_ptr(0x80)) && cvmx_pow_get_current_wqp())
cvmx_warn_if(wqp != cvmx_pow_get_current_wqp(),
"%s passed WQE(%p) doesn't match the address in the POW(%p)\n",
__func__, wqp, cvmx_pow_get_current_wqp());
}
/*
* Note that WQE in DRAM is not updated here, as the POW does not
* read from DRAM once the WQE is in flight. See hardware manual
* for complete details. It is the application's responsibility to
* keep track of the current tag value if that is important.
*/
tag_req.u64 = 0;
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
unsigned int xgrp;
if (wqp_phys != 0x80) {
/* If WQE is valid, use its XGRP:
* WQE GRP is 10 bits, and is mapped
* to legacy GRP + QoS, includes node number.
*/
xgrp = wqp->word1.cn78xx.grp;
/* Use XGRP[node] too */
node = xgrp >> 8;
/* Modify XGRP with legacy group # from arg */
xgrp &= ~0xf8;
xgrp |= 0xf8 & (group << 3);
} else {
/* If no WQE, build XGRP with QoS=0 and current node */
xgrp = group << 3;
xgrp |= node << 8;
}
tag_req.s_cn78xx_other.op = CVMX_POW_TAG_OP_SWTAG_FULL;
tag_req.s_cn78xx_other.type = tag_type;
tag_req.s_cn78xx_other.grp = xgrp;
tag_req.s_cn78xx_other.wqp = wqp_phys;
} else if (octeon_has_feature(OCTEON_FEATURE_CN68XX_WQE)) {
tag_req.s_cn68xx_other.op = CVMX_POW_TAG_OP_SWTAG_FULL;
tag_req.s_cn68xx_other.tag = tag;
tag_req.s_cn68xx_other.type = tag_type;
tag_req.s_cn68xx_other.grp = group;
} else {
tag_req.s_cn38xx.op = CVMX_POW_TAG_OP_SWTAG_FULL;
tag_req.s_cn38xx.tag = tag;
tag_req.s_cn38xx.type = tag_type;
tag_req.s_cn38xx.grp = group;
}
ptr.u64 = 0;
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
ptr.s_cn78xx.mem_region = CVMX_IO_SEG;
ptr.s_cn78xx.is_io = 1;
ptr.s_cn78xx.did = CVMX_OCT_DID_TAG_SWTAG;
ptr.s_cn78xx.node = node;
ptr.s_cn78xx.tag = tag;
} else {
ptr.s.mem_region = CVMX_IO_SEG;
ptr.s.is_io = 1;
ptr.s.did = CVMX_OCT_DID_TAG_SWTAG;
ptr.s.addr = wqp_phys;
}
/* Once this store arrives at POW, it will attempt the switch
software must wait for the switch to complete separately */
cvmx_write_io(ptr.u64, tag_req.u64);
}
/**
* Starts a tag switch to the provided tag value and tag type.
* Completion for the tag switch must be checked for separately.
* This function does NOT update the work queue entry in dram to match tag value
* and type, so the application must keep track of these if they are important
* to the application. This tag switch command must not be used for switches
* to NULL, as the tag switch pending bit will be set by the switch request,
* but never cleared by the hardware.
*
* This function must be used for tag switches from NULL.
*
* This function waits for any pending tag switches to complete
* before requesting the tag switch.
*
* @param wqp Pointer to work queue entry to submit.
* This entry is updated to match the other parameters
* @param tag Tag value to be assigned to work queue entry
* @param tag_type Type of tag
* @param group Group value for the work queue entry.
*/
static inline void cvmx_pow_tag_sw_full(cvmx_wqe_t *wqp, u32 tag, cvmx_pow_tag_type_t tag_type,
u64 group)
{
/*
* Ensure that there is not a pending tag switch, as a tag switch cannot
* be started if a previous switch is still pending.
*/
cvmx_pow_tag_sw_wait();
cvmx_pow_tag_sw_full_nocheck(wqp, tag, tag_type, group);
}
/**
* Switch to a NULL tag, which ends any ordering or
* synchronization provided by the POW for the current
* work queue entry. This operation completes immediately,
* so completion should not be waited for.
* This function does NOT wait for previous tag switches to complete,
* so the caller must ensure that any previous tag switches have completed.
*/
static inline void cvmx_pow_tag_sw_null_nocheck(void)
{
union cvmx_pow_tag_req_addr ptr;
cvmx_pow_tag_req_t tag_req;
if (CVMX_ENABLE_POW_CHECKS) {
cvmx_pow_tag_info_t current_tag;
__cvmx_pow_warn_if_pending_switch(__func__);
current_tag = cvmx_pow_get_current_tag();
cvmx_warn_if(current_tag.tag_type == CVMX_POW_TAG_TYPE_NULL_NULL,
"%s called with NULL_NULL tag\n", __func__);
cvmx_warn_if(current_tag.tag_type == CVMX_POW_TAG_TYPE_NULL,
"%s called when we already have a NULL tag\n", __func__);
}
tag_req.u64 = 0;
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
tag_req.s_cn78xx_other.op = CVMX_POW_TAG_OP_SWTAG;
tag_req.s_cn78xx_other.type = CVMX_POW_TAG_TYPE_NULL;
} else if (octeon_has_feature(OCTEON_FEATURE_CN68XX_WQE)) {
tag_req.s_cn68xx_other.op = CVMX_POW_TAG_OP_SWTAG;
tag_req.s_cn68xx_other.type = CVMX_POW_TAG_TYPE_NULL;
} else {
tag_req.s_cn38xx.op = CVMX_POW_TAG_OP_SWTAG;
tag_req.s_cn38xx.type = CVMX_POW_TAG_TYPE_NULL;
}
ptr.u64 = 0;
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
ptr.s_cn78xx.mem_region = CVMX_IO_SEG;
ptr.s_cn78xx.is_io = 1;
ptr.s_cn78xx.did = CVMX_OCT_DID_TAG_TAG1;
ptr.s_cn78xx.node = cvmx_get_node_num();
} else {
ptr.s.mem_region = CVMX_IO_SEG;
ptr.s.is_io = 1;
ptr.s.did = CVMX_OCT_DID_TAG_TAG1;
}
cvmx_write_io(ptr.u64, tag_req.u64);
}
/**
* Switch to a NULL tag, which ends any ordering or
* synchronization provided by the POW for the current
* work queue entry. This operation completes immediately,
* so completion should not be waited for.
* This function waits for any pending tag switches to complete
* before requesting the switch to NULL.
*/
static inline void cvmx_pow_tag_sw_null(void)
{
/*
* Ensure that there is not a pending tag switch, as a tag switch cannot
* be started if a previous switch is still pending.
*/
cvmx_pow_tag_sw_wait();
cvmx_pow_tag_sw_null_nocheck();
}
/**
* Submits work to an input queue.
* This function updates the work queue entry in DRAM to match the arguments given.
* Note that the tag provided is for the work queue entry submitted, and
* is unrelated to the tag that the core currently holds.
*
* @param wqp pointer to work queue entry to submit.
* This entry is updated to match the other parameters
* @param tag tag value to be assigned to work queue entry
* @param tag_type type of tag
* @param qos Input queue to add to.
* @param grp group value for the work queue entry.
*/
static inline void cvmx_pow_work_submit(cvmx_wqe_t *wqp, u32 tag, cvmx_pow_tag_type_t tag_type,
u64 qos, u64 grp)
{
union cvmx_pow_tag_req_addr ptr;
cvmx_pow_tag_req_t tag_req;
tag_req.u64 = 0;
ptr.u64 = 0;
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
unsigned int node = cvmx_get_node_num();
unsigned int xgrp;
xgrp = (grp & 0x1f) << 3;
xgrp |= (qos & 7);
xgrp |= 0x300 & (node << 8);
wqp->word1.cn78xx.rsvd_0 = 0;
wqp->word1.cn78xx.rsvd_1 = 0;
wqp->word1.cn78xx.tag = tag;
wqp->word1.cn78xx.tag_type = tag_type;
wqp->word1.cn78xx.grp = xgrp;
CVMX_SYNCWS;
tag_req.s_cn78xx_other.op = CVMX_POW_TAG_OP_ADDWQ;
tag_req.s_cn78xx_other.type = tag_type;
tag_req.s_cn78xx_other.wqp = cvmx_ptr_to_phys(wqp);
tag_req.s_cn78xx_other.grp = xgrp;
ptr.s_cn78xx.did = 0x66; // CVMX_OCT_DID_TAG_TAG6;
ptr.s_cn78xx.mem_region = CVMX_IO_SEG;
ptr.s_cn78xx.is_io = 1;
ptr.s_cn78xx.node = node;
ptr.s_cn78xx.tag = tag;
} else if (octeon_has_feature(OCTEON_FEATURE_CN68XX_WQE)) {
/* Reset all reserved bits */
wqp->word1.cn68xx.zero_0 = 0;
wqp->word1.cn68xx.zero_1 = 0;
wqp->word1.cn68xx.zero_2 = 0;
wqp->word1.cn68xx.qos = qos;
wqp->word1.cn68xx.grp = grp;
wqp->word1.tag = tag;
wqp->word1.tag_type = tag_type;
tag_req.s_cn68xx_add.op = CVMX_POW_TAG_OP_ADDWQ;
tag_req.s_cn68xx_add.type = tag_type;
tag_req.s_cn68xx_add.tag = tag;
tag_req.s_cn68xx_add.qos = qos;
tag_req.s_cn68xx_add.grp = grp;
ptr.s.mem_region = CVMX_IO_SEG;
ptr.s.is_io = 1;
ptr.s.did = CVMX_OCT_DID_TAG_TAG1;
ptr.s.addr = cvmx_ptr_to_phys(wqp);
} else {
/* Reset all reserved bits */
wqp->word1.cn38xx.zero_2 = 0;
wqp->word1.cn38xx.qos = qos;
wqp->word1.cn38xx.grp = grp;
wqp->word1.tag = tag;
wqp->word1.tag_type = tag_type;
tag_req.s_cn38xx.op = CVMX_POW_TAG_OP_ADDWQ;
tag_req.s_cn38xx.type = tag_type;
tag_req.s_cn38xx.tag = tag;
tag_req.s_cn38xx.qos = qos;
tag_req.s_cn38xx.grp = grp;
ptr.s.mem_region = CVMX_IO_SEG;
ptr.s.is_io = 1;
ptr.s.did = CVMX_OCT_DID_TAG_TAG1;
ptr.s.addr = cvmx_ptr_to_phys(wqp);
}
/* SYNC write to memory before the work submit.
* This is necessary as POW may read values from DRAM at this time */
CVMX_SYNCWS;
cvmx_write_io(ptr.u64, tag_req.u64);
}
/**
* This function sets the group mask for a core. The group mask
* indicates which groups each core will accept work from. There are
* 16 groups.
*
* @param core_num core to apply mask to
* @param mask Group mask, one bit for up to 64 groups.
* Each 1 bit in the mask enables the core to accept work from
* the corresponding group.
* The CN68XX supports 64 groups, earlier models only support
* 16 groups.
*
* The CN78XX in backwards compatibility mode allows up to 32 groups,
* so the 'mask' argument has one bit for every of the legacy
* groups, and a '1' in the mask causes a total of 8 groups
* which share the legacy group numbher and 8 qos levels,
* to be enabled for the calling processor core.
* A '0' in the mask will disable the current core
* from receiving work from the associated group.
*/
static inline void cvmx_pow_set_group_mask(u64 core_num, u64 mask)
{
u64 valid_mask;
int num_groups = cvmx_pow_num_groups();
if (num_groups >= 64)
valid_mask = ~0ull;
else
valid_mask = (1ull << num_groups) - 1;
if ((mask & valid_mask) == 0) {
printf("ERROR: %s empty group mask disables work on core# %llu, ignored.\n",
__func__, (unsigned long long)core_num);
return;
}
cvmx_warn_if(mask & (~valid_mask), "%s group number range exceeded: %#llx\n", __func__,
(unsigned long long)mask);
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
unsigned int mask_set;
cvmx_sso_ppx_sx_grpmskx_t grp_msk;
unsigned int core, node;
unsigned int rix; /* Register index */
unsigned int grp; /* Legacy group # */
unsigned int bit; /* bit index */
unsigned int xgrp; /* native group # */
node = cvmx_coremask_core_to_node(core_num);
core = cvmx_coremask_core_on_node(core_num);
/* 78xx: 256 groups divided into 4 X 64 bit registers */
/* 73xx: 64 groups are in one register */
for (rix = 0; rix < (cvmx_sso_num_xgrp() >> 6); rix++) {
grp_msk.u64 = 0;
for (bit = 0; bit < 64; bit++) {
/* 8-bit native XGRP number */
xgrp = (rix << 6) | bit;
/* Legacy 5-bit group number */
grp = (xgrp >> 3) & 0x1f;
/* Inspect legacy mask by legacy group */
if (mask & (1ull << grp))
grp_msk.s.grp_msk |= 1ull << bit;
/* Pre-set to all 0's */
}
for (mask_set = 0; mask_set < cvmx_sso_num_maskset(); mask_set++) {
csr_wr_node(node, CVMX_SSO_PPX_SX_GRPMSKX(core, mask_set, rix),
grp_msk.u64);
}
}
} else if (octeon_has_feature(OCTEON_FEATURE_CN68XX_WQE)) {
cvmx_sso_ppx_grp_msk_t grp_msk;
grp_msk.s.grp_msk = mask;
csr_wr(CVMX_SSO_PPX_GRP_MSK(core_num), grp_msk.u64);
} else {
cvmx_pow_pp_grp_mskx_t grp_msk;
grp_msk.u64 = csr_rd(CVMX_POW_PP_GRP_MSKX(core_num));
grp_msk.s.grp_msk = mask & 0xffff;
csr_wr(CVMX_POW_PP_GRP_MSKX(core_num), grp_msk.u64);
}
}
/**
* This function gets the group mask for a core. The group mask
* indicates which groups each core will accept work from.
*
* @param core_num core to apply mask to
* Return: Group mask, one bit for up to 64 groups.
* Each 1 bit in the mask enables the core to accept work from
* the corresponding group.
* The CN68XX supports 64 groups, earlier models only support
* 16 groups.
*
* The CN78XX in backwards compatibility mode allows up to 32 groups,
* so the 'mask' argument has one bit for every of the legacy
* groups, and a '1' in the mask causes a total of 8 groups
* which share the legacy group numbher and 8 qos levels,
* to be enabled for the calling processor core.
* A '0' in the mask will disable the current core
* from receiving work from the associated group.
*/
static inline u64 cvmx_pow_get_group_mask(u64 core_num)
{
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
cvmx_sso_ppx_sx_grpmskx_t grp_msk;
unsigned int core, node, i;
int rix; /* Register index */
u64 mask = 0;
node = cvmx_coremask_core_to_node(core_num);
core = cvmx_coremask_core_on_node(core_num);
/* 78xx: 256 groups divided into 4 X 64 bit registers */
/* 73xx: 64 groups are in one register */
for (rix = (cvmx_sso_num_xgrp() >> 6) - 1; rix >= 0; rix--) {
/* read only mask_set=0 (both 'set' was written same) */
grp_msk.u64 = csr_rd_node(node, CVMX_SSO_PPX_SX_GRPMSKX(core, 0, rix));
/* ASSUME: (this is how mask bits got written) */
/* grp_mask[7:0]: all bits 0..7 are same */
/* grp_mask[15:8]: all bits 8..15 are same, etc */
/* DO: mask[7:0] = grp_mask.u64[56,48,40,32,24,16,8,0] */
for (i = 0; i < 8; i++)
mask |= (grp_msk.u64 & ((u64)1 << (i * 8))) >> (7 * i);
/* we collected 8 MSBs in mask[7:0], <<=8 and continue */
if (cvmx_likely(rix != 0))
mask <<= 8;
}
return mask & 0xFFFFFFFF;
} else if (octeon_has_feature(OCTEON_FEATURE_CN68XX_WQE)) {
cvmx_sso_ppx_grp_msk_t grp_msk;
grp_msk.u64 = csr_rd(CVMX_SSO_PPX_GRP_MSK(core_num));
return grp_msk.u64;
} else {
cvmx_pow_pp_grp_mskx_t grp_msk;
grp_msk.u64 = csr_rd(CVMX_POW_PP_GRP_MSKX(core_num));
return grp_msk.u64 & 0xffff;
}
}
/*
* Returns 0 if 78xx(73xx,75xx) is not programmed in legacy compatible mode
* Returns 1 if 78xx(73xx,75xx) is programmed in legacy compatible mode
* Returns 1 if octeon model is not 78xx(73xx,75xx)
*/
static inline u64 cvmx_pow_is_legacy78mode(u64 core_num)
{
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
cvmx_sso_ppx_sx_grpmskx_t grp_msk0, grp_msk1;
unsigned int core, node, i;
int rix; /* Register index */
u64 mask = 0;
node = cvmx_coremask_core_to_node(core_num);
core = cvmx_coremask_core_on_node(core_num);
/* 78xx: 256 groups divided into 4 X 64 bit registers */
/* 73xx: 64 groups are in one register */
/* 1) in order for the 78_SSO to be in legacy compatible mode
* the both mask_sets should be programmed the same */
for (rix = (cvmx_sso_num_xgrp() >> 6) - 1; rix >= 0; rix--) {
/* read mask_set=0 (both 'set' was written same) */
grp_msk0.u64 = csr_rd_node(node, CVMX_SSO_PPX_SX_GRPMSKX(core, 0, rix));
grp_msk1.u64 = csr_rd_node(node, CVMX_SSO_PPX_SX_GRPMSKX(core, 1, rix));
if (grp_msk0.u64 != grp_msk1.u64) {
return 0;
}
/* (this is how mask bits should be written) */
/* grp_mask[7:0]: all bits 0..7 are same */
/* grp_mask[15:8]: all bits 8..15 are same, etc */
/* 2) in order for the 78_SSO to be in legacy compatible
* mode above should be true (test only mask_set=0 */
for (i = 0; i < 8; i++) {
mask = (grp_msk0.u64 >> (i << 3)) & 0xFF;
if (!(mask == 0 || mask == 0xFF)) {
return 0;
}
}
}
/* if we come here, the 78_SSO is in legacy compatible mode */
}
return 1; /* the SSO/POW is in legacy (or compatible) mode */
}
/**
* This function sets POW static priorities for a core. Each input queue has
* an associated priority value.
*
* @param core_num core to apply priorities to
* @param priority Vector of 8 priorities, one per POW Input Queue (0-7).
* Highest priority is 0 and lowest is 7. A priority value
* of 0xF instructs POW to skip the Input Queue when
* scheduling to this specific core.
* NOTE: priorities should not have gaps in values, meaning
* {0,1,1,1,1,1,1,1} is a valid configuration while
* {0,2,2,2,2,2,2,2} is not.
*/
static inline void cvmx_pow_set_priority(u64 core_num, const u8 priority[])
{
/* Detect gaps between priorities and flag error */
if (!octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
int i;
u32 prio_mask = 0;
for (i = 0; i < 8; i++)
if (priority[i] != 0xF)
prio_mask |= 1 << priority[i];
if (prio_mask ^ ((1 << cvmx_pop(prio_mask)) - 1)) {
debug("ERROR: POW static priorities should be contiguous (0x%llx)\n",
(unsigned long long)prio_mask);
return;
}
}
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
unsigned int group;
unsigned int node = cvmx_get_node_num();
cvmx_sso_grpx_pri_t grp_pri;
/*grp_pri.s.weight = 0x3f; these will be anyway overwritten */
/*grp_pri.s.affinity = 0xf; by the next csr_rd_node(..), */
for (group = 0; group < cvmx_sso_num_xgrp(); group++) {
grp_pri.u64 = csr_rd_node(node, CVMX_SSO_GRPX_PRI(group));
grp_pri.s.pri = priority[group & 0x7];
csr_wr_node(node, CVMX_SSO_GRPX_PRI(group), grp_pri.u64);
}
} else if (octeon_has_feature(OCTEON_FEATURE_CN68XX_WQE)) {
cvmx_sso_ppx_qos_pri_t qos_pri;
qos_pri.u64 = csr_rd(CVMX_SSO_PPX_QOS_PRI(core_num));
qos_pri.s.qos0_pri = priority[0];
qos_pri.s.qos1_pri = priority[1];
qos_pri.s.qos2_pri = priority[2];
qos_pri.s.qos3_pri = priority[3];
qos_pri.s.qos4_pri = priority[4];
qos_pri.s.qos5_pri = priority[5];
qos_pri.s.qos6_pri = priority[6];
qos_pri.s.qos7_pri = priority[7];
csr_wr(CVMX_SSO_PPX_QOS_PRI(core_num), qos_pri.u64);
} else {
/* POW priorities on CN5xxx .. CN66XX */
cvmx_pow_pp_grp_mskx_t grp_msk;
grp_msk.u64 = csr_rd(CVMX_POW_PP_GRP_MSKX(core_num));
grp_msk.s.qos0_pri = priority[0];
grp_msk.s.qos1_pri = priority[1];
grp_msk.s.qos2_pri = priority[2];
grp_msk.s.qos3_pri = priority[3];
grp_msk.s.qos4_pri = priority[4];
grp_msk.s.qos5_pri = priority[5];
grp_msk.s.qos6_pri = priority[6];
grp_msk.s.qos7_pri = priority[7];
csr_wr(CVMX_POW_PP_GRP_MSKX(core_num), grp_msk.u64);
}
}
/**
* This function gets POW static priorities for a core. Each input queue has
* an associated priority value.
*
* @param[in] core_num core to get priorities for
* @param[out] priority Pointer to u8[] where to return priorities
* Vector of 8 priorities, one per POW Input Queue (0-7).
* Highest priority is 0 and lowest is 7. A priority value
* of 0xF instructs POW to skip the Input Queue when
* scheduling to this specific core.
* NOTE: priorities should not have gaps in values, meaning
* {0,1,1,1,1,1,1,1} is a valid configuration while
* {0,2,2,2,2,2,2,2} is not.
*/
static inline void cvmx_pow_get_priority(u64 core_num, u8 priority[])
{
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
unsigned int group;
unsigned int node = cvmx_get_node_num();
cvmx_sso_grpx_pri_t grp_pri;
/* read priority only from the first 8 groups */
/* the next groups are programmed the same (periodicaly) */
for (group = 0; group < 8 /*cvmx_sso_num_xgrp() */; group++) {
grp_pri.u64 = csr_rd_node(node, CVMX_SSO_GRPX_PRI(group));
priority[group /* & 0x7 */] = grp_pri.s.pri;
}
} else if (octeon_has_feature(OCTEON_FEATURE_CN68XX_WQE)) {
cvmx_sso_ppx_qos_pri_t qos_pri;
qos_pri.u64 = csr_rd(CVMX_SSO_PPX_QOS_PRI(core_num));
priority[0] = qos_pri.s.qos0_pri;
priority[1] = qos_pri.s.qos1_pri;
priority[2] = qos_pri.s.qos2_pri;
priority[3] = qos_pri.s.qos3_pri;
priority[4] = qos_pri.s.qos4_pri;
priority[5] = qos_pri.s.qos5_pri;
priority[6] = qos_pri.s.qos6_pri;
priority[7] = qos_pri.s.qos7_pri;
} else {
/* POW priorities on CN5xxx .. CN66XX */
cvmx_pow_pp_grp_mskx_t grp_msk;
grp_msk.u64 = csr_rd(CVMX_POW_PP_GRP_MSKX(core_num));
priority[0] = grp_msk.s.qos0_pri;
priority[1] = grp_msk.s.qos1_pri;
priority[2] = grp_msk.s.qos2_pri;
priority[3] = grp_msk.s.qos3_pri;
priority[4] = grp_msk.s.qos4_pri;
priority[5] = grp_msk.s.qos5_pri;
priority[6] = grp_msk.s.qos6_pri;
priority[7] = grp_msk.s.qos7_pri;
}
/* Detect gaps between priorities and flag error - (optional) */
if (!octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
int i;
u32 prio_mask = 0;
for (i = 0; i < 8; i++)
if (priority[i] != 0xF)
prio_mask |= 1 << priority[i];
if (prio_mask ^ ((1 << cvmx_pop(prio_mask)) - 1)) {
debug("ERROR:%s: POW static priorities should be contiguous (0x%llx)\n",
__func__, (unsigned long long)prio_mask);
return;
}
}
}
static inline void cvmx_sso_get_group_priority(int node, cvmx_xgrp_t xgrp, int *priority,
int *weight, int *affinity)
{
cvmx_sso_grpx_pri_t grp_pri;
if (!octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
debug("ERROR: %s is not supported on this chip)\n", __func__);
return;
}
grp_pri.u64 = csr_rd_node(node, CVMX_SSO_GRPX_PRI(xgrp.xgrp));
*affinity = grp_pri.s.affinity;
*priority = grp_pri.s.pri;
*weight = grp_pri.s.weight;
}
/**
* Performs a tag switch and then an immediate deschedule. This completes
* immediately, so completion must not be waited for. This function does NOT
* update the wqe in DRAM to match arguments.
*
* This function does NOT wait for any prior tag switches to complete, so the
* calling code must do this.
*
* Note the following CAVEAT of the Octeon HW behavior when
* re-scheduling DE-SCHEDULEd items whose (next) state is
* ORDERED:
* - If there are no switches pending at the time that the
* HW executes the de-schedule, the HW will only re-schedule
* the head of the FIFO associated with the given tag. This
* means that in many respects, the HW treats this ORDERED
* tag as an ATOMIC tag. Note that in the SWTAG_DESCH
* case (to an ORDERED tag), the HW will do the switch
* before the deschedule whenever it is possible to do
* the switch immediately, so it may often look like
* this case.
* - If there is a pending switch to ORDERED at the time
* the HW executes the de-schedule, the HW will perform
* the switch at the time it re-schedules, and will be
* able to reschedule any/all of the entries with the
* same tag.
* Due to this behavior, the RECOMMENDATION to software is
* that they have a (next) state of ATOMIC when they
* DE-SCHEDULE. If an ORDERED tag is what was really desired,
* SW can choose to immediately switch to an ORDERED tag
* after the work (that has an ATOMIC tag) is re-scheduled.
* Note that since there are never any tag switches pending
* when the HW re-schedules, this switch can be IMMEDIATE upon
* the reception of the pointer during the re-schedule.
*
* @param tag New tag value
* @param tag_type New tag type
* @param group New group value
* @param no_sched Control whether this work queue entry will be rescheduled.
* - 1 : don't schedule this work
* - 0 : allow this work to be scheduled.
*/
static inline void cvmx_pow_tag_sw_desched_nocheck(u32 tag, cvmx_pow_tag_type_t tag_type, u64 group,
u64 no_sched)
{
union cvmx_pow_tag_req_addr ptr;
cvmx_pow_tag_req_t tag_req;
if (CVMX_ENABLE_POW_CHECKS) {
cvmx_pow_tag_info_t current_tag;
__cvmx_pow_warn_if_pending_switch(__func__);
current_tag = cvmx_pow_get_current_tag();
cvmx_warn_if(current_tag.tag_type == CVMX_POW_TAG_TYPE_NULL_NULL,
"%s called with NULL_NULL tag\n", __func__);
cvmx_warn_if(current_tag.tag_type == CVMX_POW_TAG_TYPE_NULL,
"%s called with NULL tag. Deschedule not allowed from NULL state\n",
__func__);
cvmx_warn_if((current_tag.tag_type != CVMX_POW_TAG_TYPE_ATOMIC) &&
(tag_type != CVMX_POW_TAG_TYPE_ATOMIC),
"%s called where neither the before or after tag is ATOMIC\n",
__func__);
}
tag_req.u64 = 0;
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
cvmx_wqe_t *wqp = cvmx_pow_get_current_wqp();
if (!wqp) {
debug("ERROR: Failed to get WQE, %s\n", __func__);
return;
}
group &= 0x1f;
wqp->word1.cn78xx.tag = tag;
wqp->word1.cn78xx.tag_type = tag_type;
wqp->word1.cn78xx.grp = group << 3;
CVMX_SYNCWS;
tag_req.s_cn78xx_other.op = CVMX_POW_TAG_OP_SWTAG_DESCH;
tag_req.s_cn78xx_other.type = tag_type;
tag_req.s_cn78xx_other.grp = group << 3;
tag_req.s_cn78xx_other.no_sched = no_sched;
} else if (octeon_has_feature(OCTEON_FEATURE_CN68XX_WQE)) {
group &= 0x3f;
tag_req.s_cn68xx_other.op = CVMX_POW_TAG_OP_SWTAG_DESCH;
tag_req.s_cn68xx_other.tag = tag;
tag_req.s_cn68xx_other.type = tag_type;
tag_req.s_cn68xx_other.grp = group;
tag_req.s_cn68xx_other.no_sched = no_sched;
} else {
group &= 0x0f;
tag_req.s_cn38xx.op = CVMX_POW_TAG_OP_SWTAG_DESCH;
tag_req.s_cn38xx.tag = tag;
tag_req.s_cn38xx.type = tag_type;
tag_req.s_cn38xx.grp = group;
tag_req.s_cn38xx.no_sched = no_sched;
}
ptr.u64 = 0;
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
ptr.s.mem_region = CVMX_IO_SEG;
ptr.s.is_io = 1;
ptr.s.did = CVMX_OCT_DID_TAG_TAG3;
ptr.s_cn78xx.node = cvmx_get_node_num();
ptr.s_cn78xx.tag = tag;
} else {
ptr.s.mem_region = CVMX_IO_SEG;
ptr.s.is_io = 1;
ptr.s.did = CVMX_OCT_DID_TAG_TAG3;
}
cvmx_write_io(ptr.u64, tag_req.u64);
}
/**
* Performs a tag switch and then an immediate deschedule. This completes
* immediately, so completion must not be waited for. This function does NOT
* update the wqe in DRAM to match arguments.
*
* This function waits for any prior tag switches to complete, so the
* calling code may call this function with a pending tag switch.
*
* Note the following CAVEAT of the Octeon HW behavior when
* re-scheduling DE-SCHEDULEd items whose (next) state is
* ORDERED:
* - If there are no switches pending at the time that the
* HW executes the de-schedule, the HW will only re-schedule
* the head of the FIFO associated with the given tag. This
* means that in many respects, the HW treats this ORDERED
* tag as an ATOMIC tag. Note that in the SWTAG_DESCH
* case (to an ORDERED tag), the HW will do the switch
* before the deschedule whenever it is possible to do
* the switch immediately, so it may often look like
* this case.
* - If there is a pending switch to ORDERED at the time
* the HW executes the de-schedule, the HW will perform
* the switch at the time it re-schedules, and will be
* able to reschedule any/all of the entries with the
* same tag.
* Due to this behavior, the RECOMMENDATION to software is
* that they have a (next) state of ATOMIC when they
* DE-SCHEDULE. If an ORDERED tag is what was really desired,
* SW can choose to immediately switch to an ORDERED tag
* after the work (that has an ATOMIC tag) is re-scheduled.
* Note that since there are never any tag switches pending
* when the HW re-schedules, this switch can be IMMEDIATE upon
* the reception of the pointer during the re-schedule.
*
* @param tag New tag value
* @param tag_type New tag type
* @param group New group value
* @param no_sched Control whether this work queue entry will be rescheduled.
* - 1 : don't schedule this work
* - 0 : allow this work to be scheduled.
*/
static inline void cvmx_pow_tag_sw_desched(u32 tag, cvmx_pow_tag_type_t tag_type, u64 group,
u64 no_sched)
{
/* Need to make sure any writes to the work queue entry are complete */
CVMX_SYNCWS;
/* Ensure that there is not a pending tag switch, as a tag switch cannot be started
* if a previous switch is still pending. */
cvmx_pow_tag_sw_wait();
cvmx_pow_tag_sw_desched_nocheck(tag, tag_type, group, no_sched);
}
/**
* Descchedules the current work queue entry.
*
* @param no_sched no schedule flag value to be set on the work queue entry.
* If this is set the entry will not be rescheduled.
*/
static inline void cvmx_pow_desched(u64 no_sched)
{
union cvmx_pow_tag_req_addr ptr;
cvmx_pow_tag_req_t tag_req;
if (CVMX_ENABLE_POW_CHECKS) {
cvmx_pow_tag_info_t current_tag;
__cvmx_pow_warn_if_pending_switch(__func__);
current_tag = cvmx_pow_get_current_tag();
cvmx_warn_if(current_tag.tag_type == CVMX_POW_TAG_TYPE_NULL_NULL,
"%s called with NULL_NULL tag\n", __func__);
cvmx_warn_if(current_tag.tag_type == CVMX_POW_TAG_TYPE_NULL,
"%s called with NULL tag. Deschedule not expected from NULL state\n",
__func__);
}
/* Need to make sure any writes to the work queue entry are complete */
CVMX_SYNCWS;
tag_req.u64 = 0;
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
tag_req.s_cn78xx_other.op = CVMX_POW_TAG_OP_DESCH;
tag_req.s_cn78xx_other.no_sched = no_sched;
} else if (octeon_has_feature(OCTEON_FEATURE_CN68XX_WQE)) {
tag_req.s_cn68xx_other.op = CVMX_POW_TAG_OP_DESCH;
tag_req.s_cn68xx_other.no_sched = no_sched;
} else {
tag_req.s_cn38xx.op = CVMX_POW_TAG_OP_DESCH;
tag_req.s_cn38xx.no_sched = no_sched;
}
ptr.u64 = 0;
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
ptr.s_cn78xx.mem_region = CVMX_IO_SEG;
ptr.s_cn78xx.is_io = 1;
ptr.s_cn78xx.did = CVMX_OCT_DID_TAG_TAG3;
ptr.s_cn78xx.node = cvmx_get_node_num();
} else {
ptr.s.mem_region = CVMX_IO_SEG;
ptr.s.is_io = 1;
ptr.s.did = CVMX_OCT_DID_TAG_TAG3;
}
cvmx_write_io(ptr.u64, tag_req.u64);
}
/******************************************************************************/
/* OCTEON3-specific functions. */
/******************************************************************************/
/**
* This function sets the the affinity of group to the cores in 78xx.
* It sets up all the cores in core_mask to accept work from the specified group.
*
* @param xgrp Group to accept work from, 0 - 255.
* @param core_mask Mask of all the cores which will accept work from this group
* @param mask_set Every core has set of 2 masks which can be set to accept work
* from 256 groups. At the time of get_work, cores can choose which mask_set
* to get work from. 'mask_set' values range from 0 to 3, where each of the
* two bits represents a mask set. Cores will be added to the mask set with
* corresponding bit set, and removed from the mask set with corresponding
* bit clear.
* Note: cores can only accept work from SSO groups on the same node,
* so the node number for the group is derived from the core number.
*/
static inline void cvmx_sso_set_group_core_affinity(cvmx_xgrp_t xgrp,
const struct cvmx_coremask *core_mask,
u8 mask_set)
{
cvmx_sso_ppx_sx_grpmskx_t grp_msk;
int core;
int grp_index = xgrp.xgrp >> 6;
int bit_pos = xgrp.xgrp % 64;
if (!octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
debug("ERROR: %s is not supported on this chip)\n", __func__);
return;
}
cvmx_coremask_for_each_core(core, core_mask)
{
unsigned int node, ncore;
u64 reg_addr;
node = cvmx_coremask_core_to_node(core);
ncore = cvmx_coremask_core_on_node(core);
reg_addr = CVMX_SSO_PPX_SX_GRPMSKX(ncore, 0, grp_index);
grp_msk.u64 = csr_rd_node(node, reg_addr);
if (mask_set & 1)
grp_msk.s.grp_msk |= (1ull << bit_pos);
else
grp_msk.s.grp_msk &= ~(1ull << bit_pos);
csr_wr_node(node, reg_addr, grp_msk.u64);
reg_addr = CVMX_SSO_PPX_SX_GRPMSKX(ncore, 1, grp_index);
grp_msk.u64 = csr_rd_node(node, reg_addr);
if (mask_set & 2)
grp_msk.s.grp_msk |= (1ull << bit_pos);
else
grp_msk.s.grp_msk &= ~(1ull << bit_pos);
csr_wr_node(node, reg_addr, grp_msk.u64);
}
}
/**
* This function sets the priority and group affinity arbitration for each group.
*
* @param node Node number
* @param xgrp Group 0 - 255 to apply mask parameters to
* @param priority Priority of the group relative to other groups
* 0x0 - highest priority
* 0x7 - lowest priority
* @param weight Cross-group arbitration weight to apply to this group.
* valid values are 1-63
* h/w default is 0x3f
* @param affinity Processor affinity arbitration weight to apply to this group.
* If zero, affinity is disabled.
* valid values are 0-15
* h/w default which is 0xf.
* @param modify_mask mask of the parameters which needs to be modified.
* enum cvmx_sso_group_modify_mask
* to modify only priority -- set bit0
* to modify only weight -- set bit1
* to modify only affinity -- set bit2
*/
static inline void cvmx_sso_set_group_priority(int node, cvmx_xgrp_t xgrp, int priority, int weight,
int affinity,
enum cvmx_sso_group_modify_mask modify_mask)
{
cvmx_sso_grpx_pri_t grp_pri;
if (!octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
debug("ERROR: %s is not supported on this chip)\n", __func__);
return;
}
if (weight <= 0)
weight = 0x3f; /* Force HW default when out of range */
grp_pri.u64 = csr_rd_node(node, CVMX_SSO_GRPX_PRI(xgrp.xgrp));
if (grp_pri.s.weight == 0)
grp_pri.s.weight = 0x3f;
if (modify_mask & CVMX_SSO_MODIFY_GROUP_PRIORITY)
grp_pri.s.pri = priority;
if (modify_mask & CVMX_SSO_MODIFY_GROUP_WEIGHT)
grp_pri.s.weight = weight;
if (modify_mask & CVMX_SSO_MODIFY_GROUP_AFFINITY)
grp_pri.s.affinity = affinity;
csr_wr_node(node, CVMX_SSO_GRPX_PRI(xgrp.xgrp), grp_pri.u64);
}
/**
* Asynchronous work request.
* Only works on CN78XX style SSO.
*
* Work is requested from the SSO unit, and should later be checked with
* function cvmx_pow_work_response_async.
* This function does NOT wait for previous tag switches to complete,
* so the caller must ensure that there is not a pending tag switch.
*
* @param scr_addr Scratch memory address that response will be returned to,
* which is either a valid WQE, or a response with the invalid bit set.
* Byte address, must be 8 byte aligned.
* @param xgrp Group to receive work for (0-255).
* @param wait
* 1 to cause response to wait for work to become available (or timeout)
* 0 to cause response to return immediately
*/
static inline void cvmx_sso_work_request_grp_async_nocheck(int scr_addr, cvmx_xgrp_t xgrp,
cvmx_pow_wait_t wait)
{
cvmx_pow_iobdma_store_t data;
unsigned int node = cvmx_get_node_num();
if (CVMX_ENABLE_POW_CHECKS) {
__cvmx_pow_warn_if_pending_switch(__func__);
cvmx_warn_if(!octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE), "Not CN78XX");
}
/* scr_addr must be 8 byte aligned */
data.u64 = 0;
data.s_cn78xx.scraddr = scr_addr >> 3;
data.s_cn78xx.len = 1;
data.s_cn78xx.did = CVMX_OCT_DID_TAG_SWTAG;
data.s_cn78xx.grouped = 1;
data.s_cn78xx.index_grp_mask = (node << 8) | xgrp.xgrp;
data.s_cn78xx.wait = wait;
data.s_cn78xx.node = node;
cvmx_send_single(data.u64);
}
/**
* Synchronous work request from the node-local SSO without verifying
* pending tag switch. It requests work from a specific SSO group.
*
* @param lgrp The local group number (within the SSO of the node of the caller)
* from which to get the work.
* @param wait When set, call stalls until work becomes available, or times out.
* If not set, returns immediately.
*
* Return: Returns the WQE pointer from SSO.
* Returns NULL if no work was available.
*/
static inline void *cvmx_sso_work_request_grp_sync_nocheck(unsigned int lgrp, cvmx_pow_wait_t wait)
{
cvmx_pow_load_addr_t ptr;
cvmx_pow_tag_load_resp_t result;
unsigned int node = cvmx_get_node_num() & 3;
if (CVMX_ENABLE_POW_CHECKS) {
__cvmx_pow_warn_if_pending_switch(__func__);
cvmx_warn_if(!octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE), "Not CN78XX");
}
ptr.u64 = 0;
ptr.swork_78xx.mem_region = CVMX_IO_SEG;
ptr.swork_78xx.is_io = 1;
ptr.swork_78xx.did = CVMX_OCT_DID_TAG_SWTAG;
ptr.swork_78xx.node = node;
ptr.swork_78xx.grouped = 1;
ptr.swork_78xx.index = (lgrp & 0xff) | node << 8;
ptr.swork_78xx.wait = wait;
result.u64 = csr_rd(ptr.u64);
if (result.s_work.no_work)
return NULL;
else
return cvmx_phys_to_ptr(result.s_work.addr);
}
/**
* Synchronous work request from the node-local SSO.
* It requests work from a specific SSO group.
* This function waits for any previous tag switch to complete before
* requesting the new work.
*
* @param lgrp The node-local group number from which to get the work.
* @param wait When set, call stalls until work becomes available, or times out.
* If not set, returns immediately.
*
* Return: The WQE pointer or NULL, if work is not available.
*/
static inline void *cvmx_sso_work_request_grp_sync(unsigned int lgrp, cvmx_pow_wait_t wait)
{
cvmx_pow_tag_sw_wait();
return cvmx_sso_work_request_grp_sync_nocheck(lgrp, wait);
}
/**
* This function sets the group mask for a core. The group mask bits
* indicate which groups each core will accept work from.
*
* @param core_num Processor core to apply mask to.
* @param mask_set 7XXX has 2 sets of masks per core.
* Bit 0 represents the first mask set, bit 1 -- the second.
* @param xgrp_mask Group mask array.
* Total number of groups is divided into a number of
* 64-bits mask sets. Each bit in the mask, if set, enables
* the core to accept work from the corresponding group.
*
* NOTE: Each core can be configured to accept work in accordance to both
* mask sets, with the first having higher precedence over the second,
* or to accept work in accordance to just one of the two mask sets.
* The 'core_num' argument represents a processor core on any node
* in a coherent multi-chip system.
*
* If the 'mask_set' argument is 3, both mask sets are configured
* with the same value (which is not typically the intention),
* so keep in mind the function needs to be called twice
* to set a different value into each of the mask sets,
* once with 'mask_set=1' and second time with 'mask_set=2'.
*/
static inline void cvmx_pow_set_xgrp_mask(u64 core_num, u8 mask_set, const u64 xgrp_mask[])
{
unsigned int grp, node, core;
u64 reg_addr;
if (!octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
debug("ERROR: %s is not supported on this chip)\n", __func__);
return;
}
if (CVMX_ENABLE_POW_CHECKS) {
cvmx_warn_if(((mask_set < 1) || (mask_set > 3)), "Invalid mask set");
}
if ((mask_set < 1) || (mask_set > 3))
mask_set = 3;
node = cvmx_coremask_core_to_node(core_num);
core = cvmx_coremask_core_on_node(core_num);
for (grp = 0; grp < (cvmx_sso_num_xgrp() >> 6); grp++) {
if (mask_set & 1) {
reg_addr = CVMX_SSO_PPX_SX_GRPMSKX(core, 0, grp),
csr_wr_node(node, reg_addr, xgrp_mask[grp]);
}
if (mask_set & 2) {
reg_addr = CVMX_SSO_PPX_SX_GRPMSKX(core, 1, grp),
csr_wr_node(node, reg_addr, xgrp_mask[grp]);
}
}
}
/**
* This function gets the group mask for a core. The group mask bits
* indicate which groups each core will accept work from.
*
* @param core_num Processor core to apply mask to.
* @param mask_set 7XXX has 2 sets of masks per core.
* Bit 0 represents the first mask set, bit 1 -- the second.
* @param xgrp_mask Provide pointer to u64 mask[8] output array.
* Total number of groups is divided into a number of
* 64-bits mask sets. Each bit in the mask represents
* the core accepts work from the corresponding group.
*
* NOTE: Each core can be configured to accept work in accordance to both
* mask sets, with the first having higher precedence over the second,
* or to accept work in accordance to just one of the two mask sets.
* The 'core_num' argument represents a processor core on any node
* in a coherent multi-chip system.
*/
static inline void cvmx_pow_get_xgrp_mask(u64 core_num, u8 mask_set, u64 *xgrp_mask)
{
cvmx_sso_ppx_sx_grpmskx_t grp_msk;
unsigned int grp, node, core;
u64 reg_addr;
if (!octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
debug("ERROR: %s is not supported on this chip)\n", __func__);
return;
}
if (CVMX_ENABLE_POW_CHECKS) {
cvmx_warn_if(mask_set != 1 && mask_set != 2, "Invalid mask set");
}
node = cvmx_coremask_core_to_node(core_num);
core = cvmx_coremask_core_on_node(core_num);
for (grp = 0; grp < cvmx_sso_num_xgrp() >> 6; grp++) {
if (mask_set & 1) {
reg_addr = CVMX_SSO_PPX_SX_GRPMSKX(core, 0, grp),
grp_msk.u64 = csr_rd_node(node, reg_addr);
xgrp_mask[grp] = grp_msk.s.grp_msk;
}
if (mask_set & 2) {
reg_addr = CVMX_SSO_PPX_SX_GRPMSKX(core, 1, grp),
grp_msk.u64 = csr_rd_node(node, reg_addr);
xgrp_mask[grp] = grp_msk.s.grp_msk;
}
}
}
/**
* Executes SSO SWTAG command.
* This is similar to cvmx_pow_tag_sw() function, but uses linear
* (vs. integrated group-qos) group index.
*/
static inline void cvmx_pow_tag_sw_node(cvmx_wqe_t *wqp, u32 tag, cvmx_pow_tag_type_t tag_type,
int node)
{
union cvmx_pow_tag_req_addr ptr;
cvmx_pow_tag_req_t tag_req;
if (cvmx_unlikely(!octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE))) {
debug("ERROR: %s is supported on OCTEON3 only\n", __func__);
return;
}
CVMX_SYNCWS;
cvmx_pow_tag_sw_wait();
if (CVMX_ENABLE_POW_CHECKS) {
cvmx_pow_tag_info_t current_tag;
__cvmx_pow_warn_if_pending_switch(__func__);
current_tag = cvmx_pow_get_current_tag();
cvmx_warn_if(current_tag.tag_type == CVMX_POW_TAG_TYPE_NULL_NULL,
"%s called with NULL_NULL tag\n", __func__);
cvmx_warn_if(current_tag.tag_type == CVMX_POW_TAG_TYPE_NULL,
"%s called with NULL tag\n", __func__);
cvmx_warn_if((current_tag.tag_type == tag_type) && (current_tag.tag == tag),
"%s called to perform a tag switch to the same tag\n", __func__);
cvmx_warn_if(
tag_type == CVMX_POW_TAG_TYPE_NULL,
"%s called to perform a tag switch to NULL. Use cvmx_pow_tag_sw_null() instead\n",
__func__);
}
wqp->word1.cn78xx.tag = tag;
wqp->word1.cn78xx.tag_type = tag_type;
CVMX_SYNCWS;
tag_req.u64 = 0;
tag_req.s_cn78xx_other.op = CVMX_POW_TAG_OP_SWTAG;
tag_req.s_cn78xx_other.type = tag_type;
ptr.u64 = 0;
ptr.s_cn78xx.mem_region = CVMX_IO_SEG;
ptr.s_cn78xx.is_io = 1;
ptr.s_cn78xx.did = CVMX_OCT_DID_TAG_SWTAG;
ptr.s_cn78xx.node = node;
ptr.s_cn78xx.tag = tag;
cvmx_write_io(ptr.u64, tag_req.u64);
}
/**
* Executes SSO SWTAG_FULL command.
* This is similar to cvmx_pow_tag_sw_full() function, but
* uses linear (vs. integrated group-qos) group index.
*/
static inline void cvmx_pow_tag_sw_full_node(cvmx_wqe_t *wqp, u32 tag, cvmx_pow_tag_type_t tag_type,
u8 xgrp, int node)
{
union cvmx_pow_tag_req_addr ptr;
cvmx_pow_tag_req_t tag_req;
u16 gxgrp;
if (cvmx_unlikely(!octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE))) {
debug("ERROR: %s is supported on OCTEON3 only\n", __func__);
return;
}
/* Ensure that there is not a pending tag switch, as a tag switch cannot be
* started, if a previous switch is still pending. */
CVMX_SYNCWS;
cvmx_pow_tag_sw_wait();
if (CVMX_ENABLE_POW_CHECKS) {
cvmx_pow_tag_info_t current_tag;
__cvmx_pow_warn_if_pending_switch(__func__);
current_tag = cvmx_pow_get_current_tag();
cvmx_warn_if(current_tag.tag_type == CVMX_POW_TAG_TYPE_NULL_NULL,
"%s called with NULL_NULL tag\n", __func__);
cvmx_warn_if((current_tag.tag_type == tag_type) && (current_tag.tag == tag),
"%s called to perform a tag switch to the same tag\n", __func__);
cvmx_warn_if(
tag_type == CVMX_POW_TAG_TYPE_NULL,
"%s called to perform a tag switch to NULL. Use cvmx_pow_tag_sw_null() instead\n",
__func__);
if ((wqp != cvmx_phys_to_ptr(0x80)) && cvmx_pow_get_current_wqp())
cvmx_warn_if(wqp != cvmx_pow_get_current_wqp(),
"%s passed WQE(%p) doesn't match the address in the POW(%p)\n",
__func__, wqp, cvmx_pow_get_current_wqp());
}
gxgrp = node;
gxgrp = gxgrp << 8 | xgrp;
wqp->word1.cn78xx.grp = gxgrp;
wqp->word1.cn78xx.tag = tag;
wqp->word1.cn78xx.tag_type = tag_type;
CVMX_SYNCWS;
tag_req.u64 = 0;
tag_req.s_cn78xx_other.op = CVMX_POW_TAG_OP_SWTAG_FULL;
tag_req.s_cn78xx_other.type = tag_type;
tag_req.s_cn78xx_other.grp = gxgrp;
tag_req.s_cn78xx_other.wqp = cvmx_ptr_to_phys(wqp);
ptr.u64 = 0;
ptr.s_cn78xx.mem_region = CVMX_IO_SEG;
ptr.s_cn78xx.is_io = 1;
ptr.s_cn78xx.did = CVMX_OCT_DID_TAG_SWTAG;
ptr.s_cn78xx.node = node;
ptr.s_cn78xx.tag = tag;
cvmx_write_io(ptr.u64, tag_req.u64);
}
/**
* Submits work to an SSO group on any OCI node.
* This function updates the work queue entry in DRAM to match
* the arguments given.
* Note that the tag provided is for the work queue entry submitted,
* and is unrelated to the tag that the core currently holds.
*
* @param wqp pointer to work queue entry to submit.
* This entry is updated to match the other parameters
* @param tag tag value to be assigned to work queue entry
* @param tag_type type of tag
* @param xgrp native CN78XX group in the range 0..255
* @param node The OCI node number for the target group
*
* When this function is called on a model prior to CN78XX, which does
* not support OCI nodes, the 'node' argument is ignored, and the 'xgrp'
* parameter is converted into 'qos' (the lower 3 bits) and 'grp' (the higher
* 5 bits), following the backward-compatibility scheme of translating
* between new and old style group numbers.
*/
static inline void cvmx_pow_work_submit_node(cvmx_wqe_t *wqp, u32 tag, cvmx_pow_tag_type_t tag_type,
u8 xgrp, u8 node)
{
union cvmx_pow_tag_req_addr ptr;
cvmx_pow_tag_req_t tag_req;
u16 group;
if (cvmx_unlikely(!octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE))) {
debug("ERROR: %s is supported on OCTEON3 only\n", __func__);
return;
}
group = node;
group = group << 8 | xgrp;
wqp->word1.cn78xx.tag = tag;
wqp->word1.cn78xx.tag_type = tag_type;
wqp->word1.cn78xx.grp = group;
CVMX_SYNCWS;
tag_req.u64 = 0;
tag_req.s_cn78xx_other.op = CVMX_POW_TAG_OP_ADDWQ;
tag_req.s_cn78xx_other.type = tag_type;
tag_req.s_cn78xx_other.wqp = cvmx_ptr_to_phys(wqp);
tag_req.s_cn78xx_other.grp = group;
ptr.u64 = 0;
ptr.s_cn78xx.did = 0x66; // CVMX_OCT_DID_TAG_TAG6;
ptr.s_cn78xx.mem_region = CVMX_IO_SEG;
ptr.s_cn78xx.is_io = 1;
ptr.s_cn78xx.node = node;
ptr.s_cn78xx.tag = tag;
/* SYNC write to memory before the work submit. This is necessary
** as POW may read values from DRAM at this time */
CVMX_SYNCWS;
cvmx_write_io(ptr.u64, tag_req.u64);
}
/**
* Executes the SSO SWTAG_DESCHED operation.
* This is similar to the cvmx_pow_tag_sw_desched() function, but
* uses linear (vs. unified group-qos) group index.
*/
static inline void cvmx_pow_tag_sw_desched_node(cvmx_wqe_t *wqe, u32 tag,
cvmx_pow_tag_type_t tag_type, u8 xgrp, u64 no_sched,
u8 node)
{
union cvmx_pow_tag_req_addr ptr;
cvmx_pow_tag_req_t tag_req;
u16 group;
if (cvmx_unlikely(!octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE))) {
debug("ERROR: %s is supported on OCTEON3 only\n", __func__);
return;
}
/* Need to make sure any writes to the work queue entry are complete */
CVMX_SYNCWS;
/*
* Ensure that there is not a pending tag switch, as a tag switch cannot
* be started if a previous switch is still pending.
*/
cvmx_pow_tag_sw_wait();
if (CVMX_ENABLE_POW_CHECKS) {
cvmx_pow_tag_info_t current_tag;
__cvmx_pow_warn_if_pending_switch(__func__);
current_tag = cvmx_pow_get_current_tag();
cvmx_warn_if(current_tag.tag_type == CVMX_POW_TAG_TYPE_NULL_NULL,
"%s called with NULL_NULL tag\n", __func__);
cvmx_warn_if(current_tag.tag_type == CVMX_POW_TAG_TYPE_NULL,
"%s called with NULL tag. Deschedule not allowed from NULL state\n",
__func__);
cvmx_warn_if((current_tag.tag_type != CVMX_POW_TAG_TYPE_ATOMIC) &&
(tag_type != CVMX_POW_TAG_TYPE_ATOMIC),
"%s called where neither the before or after tag is ATOMIC\n",
__func__);
}
group = node;
group = group << 8 | xgrp;
wqe->word1.cn78xx.tag = tag;
wqe->word1.cn78xx.tag_type = tag_type;
wqe->word1.cn78xx.grp = group;
CVMX_SYNCWS;
tag_req.u64 = 0;
tag_req.s_cn78xx_other.op = CVMX_POW_TAG_OP_SWTAG_DESCH;
tag_req.s_cn78xx_other.type = tag_type;
tag_req.s_cn78xx_other.grp = group;
tag_req.s_cn78xx_other.no_sched = no_sched;
ptr.u64 = 0;
ptr.s.mem_region = CVMX_IO_SEG;
ptr.s.is_io = 1;
ptr.s.did = CVMX_OCT_DID_TAG_TAG3;
ptr.s_cn78xx.node = node;
ptr.s_cn78xx.tag = tag;
cvmx_write_io(ptr.u64, tag_req.u64);
}
/* Executes the UPD_WQP_GRP SSO operation.
*
* @param wqp Pointer to the new work queue entry to switch to.
* @param xgrp SSO group in the range 0..255
*
* NOTE: The operation can be performed only on the local node.
*/
static inline void cvmx_sso_update_wqp_group(cvmx_wqe_t *wqp, u8 xgrp)
{
union cvmx_pow_tag_req_addr addr;
cvmx_pow_tag_req_t data;
int node = cvmx_get_node_num();
int group = node << 8 | xgrp;
if (!octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
debug("ERROR: %s is not supported on this chip)\n", __func__);
return;
}
wqp->word1.cn78xx.grp = group;
CVMX_SYNCWS;
data.u64 = 0;
data.s_cn78xx_other.op = CVMX_POW_TAG_OP_UPDATE_WQP_GRP;
data.s_cn78xx_other.grp = group;
data.s_cn78xx_other.wqp = cvmx_ptr_to_phys(wqp);
addr.u64 = 0;
addr.s_cn78xx.mem_region = CVMX_IO_SEG;
addr.s_cn78xx.is_io = 1;
addr.s_cn78xx.did = CVMX_OCT_DID_TAG_TAG1;
addr.s_cn78xx.node = node;
cvmx_write_io(addr.u64, data.u64);
}
/******************************************************************************/
/* Define usage of bits within the 32 bit tag values. */
/******************************************************************************/
/*
* Number of bits of the tag used by software. The SW bits
* are always a contiguous block of the high starting at bit 31.
* The hardware bits are always the low bits. By default, the top 8 bits
* of the tag are reserved for software, and the low 24 are set by the IPD unit.
*/
#define CVMX_TAG_SW_BITS (8)
#define CVMX_TAG_SW_SHIFT (32 - CVMX_TAG_SW_BITS)
/* Below is the list of values for the top 8 bits of the tag. */
/*
* Tag values with top byte of this value are reserved for internal executive
* uses
*/
#define CVMX_TAG_SW_BITS_INTERNAL 0x1
/*
* The executive divides the remaining 24 bits as follows:
* the upper 8 bits (bits 23 - 16 of the tag) define a subgroup
* the lower 16 bits (bits 15 - 0 of the tag) define are the value with
* the subgroup. Note that this section describes the format of tags generated
* by software - refer to the hardware documentation for a description of the
* tags values generated by the packet input hardware.
* Subgroups are defined here
*/
/* Mask for the value portion of the tag */
#define CVMX_TAG_SUBGROUP_MASK 0xFFFF
#define CVMX_TAG_SUBGROUP_SHIFT 16
#define CVMX_TAG_SUBGROUP_PKO 0x1
/* End of executive tag subgroup definitions */
/* The remaining values software bit values 0x2 - 0xff are available
* for application use */
/**
* This function creates a 32 bit tag value from the two values provided.
*
* @param sw_bits The upper bits (number depends on configuration) are set
* to this value. The remainder of bits are set by the hw_bits parameter.
* @param hw_bits The lower bits (number depends on configuration) are set
* to this value. The remainder of bits are set by the sw_bits parameter.
*
* Return: 32 bit value of the combined hw and sw bits.
*/
static inline u32 cvmx_pow_tag_compose(u64 sw_bits, u64 hw_bits)
{
return (((sw_bits & cvmx_build_mask(CVMX_TAG_SW_BITS)) << CVMX_TAG_SW_SHIFT) |
(hw_bits & cvmx_build_mask(32 - CVMX_TAG_SW_BITS)));
}
/**
* Extracts the bits allocated for software use from the tag
*
* @param tag 32 bit tag value
*
* Return: N bit software tag value, where N is configurable with
* the CVMX_TAG_SW_BITS define
*/
static inline u32 cvmx_pow_tag_get_sw_bits(u64 tag)
{
return ((tag >> (32 - CVMX_TAG_SW_BITS)) & cvmx_build_mask(CVMX_TAG_SW_BITS));
}
/**
*
* Extracts the bits allocated for hardware use from the tag
*
* @param tag 32 bit tag value
*
* Return: (32 - N) bit software tag value, where N is configurable with
* the CVMX_TAG_SW_BITS define
*/
static inline u32 cvmx_pow_tag_get_hw_bits(u64 tag)
{
return (tag & cvmx_build_mask(32 - CVMX_TAG_SW_BITS));
}
static inline u64 cvmx_sso3_get_wqe_count(int node)
{
cvmx_sso_grpx_aq_cnt_t aq_cnt;
unsigned int grp = 0;
u64 cnt = 0;
for (grp = 0; grp < cvmx_sso_num_xgrp(); grp++) {
aq_cnt.u64 = csr_rd_node(node, CVMX_SSO_GRPX_AQ_CNT(grp));
cnt += aq_cnt.s.aq_cnt;
}
return cnt;
}
static inline u64 cvmx_sso_get_total_wqe_count(void)
{
if (octeon_has_feature(OCTEON_FEATURE_CN78XX_WQE)) {
int node = cvmx_get_node_num();
return cvmx_sso3_get_wqe_count(node);
} else if (OCTEON_IS_MODEL(OCTEON_CN68XX)) {
cvmx_sso_iq_com_cnt_t sso_iq_com_cnt;
sso_iq_com_cnt.u64 = csr_rd(CVMX_SSO_IQ_COM_CNT);
return (sso_iq_com_cnt.s.iq_cnt);
} else {
cvmx_pow_iq_com_cnt_t pow_iq_com_cnt;
pow_iq_com_cnt.u64 = csr_rd(CVMX_POW_IQ_COM_CNT);
return (pow_iq_com_cnt.s.iq_cnt);
}
}
/**
* Store the current POW internal state into the supplied
* buffer. It is recommended that you pass a buffer of at least
* 128KB. The format of the capture may change based on SDK
* version and Octeon chip.
*
* @param buffer Buffer to store capture into
* @param buffer_size The size of the supplied buffer
*
* Return: Zero on success, negative on failure
*/
int cvmx_pow_capture(void *buffer, int buffer_size);
/**
* Dump a POW capture to the console in a human readable format.
*
* @param buffer POW capture from cvmx_pow_capture()
* @param buffer_size Size of the buffer
*/
void cvmx_pow_display(void *buffer, int buffer_size);
/**
* Return the number of POW entries supported by this chip
*
* Return: Number of POW entries
*/
int cvmx_pow_get_num_entries(void);
int cvmx_pow_get_dump_size(void);
/**
* This will allocate count number of SSO groups on the specified node to the
* calling application. These groups will be for exclusive use of the
* application until they are freed.
* @param node The numa node for the allocation.
* @param base_group Pointer to the initial group, -1 to allocate anywhere.
* @param count The number of consecutive groups to allocate.
* Return: 0 on success and -1 on failure.
*/
int cvmx_sso_reserve_group_range(int node, int *base_group, int count);
#define cvmx_sso_allocate_group_range cvmx_sso_reserve_group_range
int cvmx_sso_reserve_group(int node);
#define cvmx_sso_allocate_group cvmx_sso_reserve_group
int cvmx_sso_release_group_range(int node, int base_group, int count);
int cvmx_sso_release_group(int node, int group);
/**
* Show integrated SSO configuration.
*
* @param node node number
*/
int cvmx_sso_config_dump(unsigned int node);
/**
* Show integrated SSO statistics.
*
* @param node node number
*/
int cvmx_sso_stats_dump(unsigned int node);
/**
* Clear integrated SSO statistics.
*
* @param node node number
*/
int cvmx_sso_stats_clear(unsigned int node);
/**
* Show SSO core-group affinity and priority per node (multi-node systems)
*/
void cvmx_pow_mask_priority_dump_node(unsigned int node, struct cvmx_coremask *avail_coremask);
/**
* Show POW/SSO core-group affinity and priority (legacy, single-node systems)
*/
static inline void cvmx_pow_mask_priority_dump(struct cvmx_coremask *avail_coremask)
{
cvmx_pow_mask_priority_dump_node(0 /*node */, avail_coremask);
}
/**
* Show SSO performance counters (multi-node systems)
*/
void cvmx_pow_show_perf_counters_node(unsigned int node);
/**
* Show POW/SSO performance counters (legacy, single-node systems)
*/
static inline void cvmx_pow_show_perf_counters(void)
{
cvmx_pow_show_perf_counters_node(0 /*node */);
}
#endif /* __CVMX_POW_H__ */
Reference in a new issue View git blame Copy permalink