1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
|
/*
*
* Copyright (c) 2018 Huawei Technologies Co.,Ltd.
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at:
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef _NSFW_RING_DATA_H_
#define _NSFW_RING_DATA_H_
#include <stdint.h>
#include "types.h"
#include "common_mem_api.h"
#define VALUE_LEN 40
/*
Ring Data has two part; Ver&Data
val is a pointer offset base on rte_perf_ring::Addrbase, this struct support 1TB, it's enough now;
future __int128 maybe used, this type can perfectly solve the version & address rang problem.
*/
union RingData_U
{
struct RingData_S
{
/*
value of data, indeed it's a pointer offset base on rte_perf_ring::Addrbase;40bit is enough for user space addr
ver must using 24bit, so val using 40bit; a CAS now just support 64bit; in future, we may using __int128,now __int128 not support well.
*/
volatile unsigned long long val:VALUE_LEN;
/*
version of data, using 16b store version flg is more suitable for Address save, but using 16b version is too short, it's value range is [0-65535];
between two cpu schedule time (TM-SPACE) of one process/thread, other processes/threads do N times queue oper. if N > 65535, still have a chance of ABA.
if using a 24bit save version flg, if ABA happened, 16777216 times queue oper need done in one TM-SPACE, it's impossible for today cpu.
*/
volatile unsigned long long ver:(64 - VALUE_LEN);
} data_s;
u64 data_l;
};
/*
this high perf Ring rely on the init value of Ring Slot;
Ring Must init using PerfRingInit, Pool Must init using PerfPoolInit
the addrbase is base addr for all element; now we support 1024G offset;
for nstack the Ring element is from hugepage, and the addr is in stack space.
1. not support a ring who's element space range bigger than 1024GB
[if one element from heep, one from stack, range will bigger than 1024GB, we not support]
2. one more thing addr from mmap is in stack
3. rte_perf_ring must create by rte_perf_ring_create/rte_perf_pool_create
*/
struct nsfw_mem_ring
{
u8 memtype; //shared, no shared
u8 ringflag; //scmp, scsp, mcsp,mcmp
u16 reserv; //reserv data
u32 size; //size of the Ring, must 2^n
u32 eltsize; //for s-pool, it is the size of per buf, if is ring, eltsize is zero.
u32 mask; //mask of the Ring, used mask mod Head/Tail to get real pos, must 2^n-1
void *Addrbase; /*Cause the Addr we support just 40b(1024G), we using a basAddr+offset to get the real addr; ring[x].data_s.val just store offset;
* not used when no shared mode
*/
volatile u32_t prodhflag; //for nshmem fork recover
volatile u32_t prodtflag; //for nshmem fork recover
volatile u32_t conshflag; //for nshmem fork recover
volatile u32_t constflag; //for nshmem fork recover
nsfw_res res_chk;
struct
{
volatile u32 head; //Head of the Ring, used to indicate pos where to pull a val
volatile u32 tail; //for nshmem, shmem not used.
} prod;
struct
{
volatile u32 head; //for nshmem, shmem not used.
volatile u32 tail; //Tail of the Ring, used to indicate pos where to push a val
} cons;
u32 uireserv[4]; //reserved for update
union RingData_U ring[0]; //Value of Ring
};
#define PERFRING_ADDR_RANGE (0xFFFFFFFFFFL)
#endif /*_NSFW_RING_DATA_H_*/
|
