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
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
|
/*
* Ceph - scalable distributed file system
*
* Copyright (C) 2020 SUSE LINUX GmbH
*
* This is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License version 2.1, as published by the Free Software
* Foundation. See file COPYING.
*
*/
#ifndef WNBD_HANDLER_H
#define WNBD_HANDLER_H
#include <wnbd.h>
#include "common/admin_socket.h"
#include "common/ceph_context.h"
#include "common/Thread.h"
#include "include/rbd/librbd.hpp"
#include "include/xlist.h"
#include "global/global_context.h"
// TODO: make this configurable.
#define RBD_WNBD_MAX_TRANSFER 2 * 1024 * 1024
#define SOFT_REMOVE_RETRY_INTERVAL 2
#define DEFAULT_SOFT_REMOVE_TIMEOUT 15
#define DEFAULT_IO_WORKER_COUNT 4
// Not defined by mingw.
#ifndef SCSI_ADSENSE_UNRECOVERED_ERROR
#define SCSI_ADSENSE_UNRECOVERED_ERROR 0x11
#endif
// The following will be assigned to the "Owner" field of the WNBD
// parameters, which can be used to determine the application managing
// a disk. We'll ignore other disks.
#define RBD_WNBD_OWNER_NAME "ceph-rbd-wnbd"
class WnbdHandler;
class WnbdAdminHook : public AdminSocketHook {
WnbdHandler *m_handler;
public:
explicit WnbdAdminHook(WnbdHandler *handler) :
m_handler(handler) {
g_ceph_context->get_admin_socket()->register_command(
"wnbd stats", this, "get WNBD stats");
}
~WnbdAdminHook() override {
g_ceph_context->get_admin_socket()->unregister_commands(this);
}
int call(std::string_view command, const cmdmap_t& cmdmap,
Formatter *f, std::ostream& errss, bufferlist& out) override;
};
class WnbdHandler
{
private:
librbd::Image ℑ
std::string instance_name;
uint64_t block_count;
uint32_t block_size;
bool readonly;
bool rbd_cache_enabled;
uint32_t io_req_workers;
uint32_t io_reply_workers;
WnbdAdminHook* admin_hook;
boost::asio::thread_pool* reply_tpool;
public:
WnbdHandler(librbd::Image& _image, std::string _instance_name,
uint64_t _block_count, uint32_t _block_size,
bool _readonly, bool _rbd_cache_enabled,
uint32_t _io_req_workers,
uint32_t _io_reply_workers)
: image(_image)
, instance_name(_instance_name)
, block_count(_block_count)
, block_size(_block_size)
, readonly(_readonly)
, rbd_cache_enabled(_rbd_cache_enabled)
, io_req_workers(_io_req_workers)
, io_reply_workers(_io_reply_workers)
{
admin_hook = new WnbdAdminHook(this);
// Instead of relying on librbd's own thread pool, we're going to use a
// separate one. This allows us to make assumptions on the threads that
// are going to send the IO replies and thus be able to cache Windows
// OVERLAPPED structures.
reply_tpool = new boost::asio::thread_pool(_io_reply_workers);
}
int start();
// Wait for the handler to stop, which normally happens when the driver
// passes the "Disconnect" request.
int wait();
void shutdown();
int dump_stats(Formatter *f);
~WnbdHandler();
static VOID LogMessage(
WnbdLogLevel LogLevel,
const char* Message,
const char* FileName,
UINT32 Line,
const char* FunctionName);
private:
ceph::mutex shutdown_lock = ceph::make_mutex("WnbdHandler::DisconnectLocker");
bool started = false;
bool terminated = false;
WNBD_DISK* wnbd_disk = nullptr;
struct IOContext
{
xlist<IOContext*>::item item;
WnbdHandler *handler = nullptr;
WNBD_STATUS wnbd_status = {0};
WnbdRequestType req_type = WnbdReqTypeUnknown;
uint64_t req_handle = 0;
uint32_t err_code = 0;
size_t req_size;
uint64_t req_from;
bufferlist data;
IOContext()
: item(this)
{}
void set_sense(uint8_t sense_key, uint8_t asc, uint64_t info);
void set_sense(uint8_t sense_key, uint8_t asc);
};
friend std::ostream &operator<<(std::ostream &os, const IOContext &ctx);
void send_io_response(IOContext *ctx);
static void aio_callback(librbd::completion_t cb, void *arg);
// WNBD IO entry points
static void Read(
PWNBD_DISK Disk,
UINT64 RequestHandle,
PVOID Buffer,
UINT64 BlockAddress,
UINT32 BlockCount,
BOOLEAN ForceUnitAccess);
static void Write(
PWNBD_DISK Disk,
UINT64 RequestHandle,
PVOID Buffer,
UINT64 BlockAddress,
UINT32 BlockCount,
BOOLEAN ForceUnitAccess);
static void Flush(
PWNBD_DISK Disk,
UINT64 RequestHandle,
UINT64 BlockAddress,
UINT32 BlockCount);
static void Unmap(
PWNBD_DISK Disk,
UINT64 RequestHandle,
PWNBD_UNMAP_DESCRIPTOR Descriptors,
UINT32 Count);
static constexpr WNBD_INTERFACE RbdWnbdInterface =
{
Read,
Write,
Flush,
Unmap,
};
};
std::ostream &operator<<(std::ostream &os, const WnbdHandler::IOContext &ctx);
#endif // WNBD_HANDLER_H
|
