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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315
|
// Copyright 2012 The Chromium Authors
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#ifndef NET_BASE_IO_BUFFER_H_
#define NET_BASE_IO_BUFFER_H_
#include <stddef.h>
#include <stdint.h>
#include <memory>
#include <string>
#include "base/compiler_specific.h"
#include "base/containers/heap_array.h"
#include "base/containers/span.h"
#include "base/memory/free_deleter.h"
#include "base/memory/raw_span.h"
#include "base/memory/ref_counted.h"
#include "net/base/net_export.h"
namespace base {
class Pickle;
}
namespace net {
// IOBuffers are reference counted data buffers used for easier asynchronous
// IO handling.
//
// They are often used as the destination buffers for Read() operations, or as
// the source buffers for Write() operations.
//
// IMPORTANT: Never re-use an IOBuffer after cancelling the IO operation that
// was using it, since this may lead to memory corruption!
//
// -----------------------
// Ownership of IOBuffers:
// -----------------------
//
// Although IOBuffers are RefCountedThreadSafe, they are not intended to be
// used as a shared buffer, nor should they be used simultaneously across
// threads. The fact that they are reference counted is an implementation
// detail for allowing them to outlive cancellation of asynchronous
// operations.
//
// Instead, think of the underlying |char*| buffer contained by the IOBuffer
// as having exactly one owner at a time.
//
// Whenever you call an asynchronous operation that takes an IOBuffer,
// ownership is implicitly transferred to the called function, until the
// operation has completed (at which point it transfers back to the caller).
//
// ==> The IOBuffer's data should NOT be manipulated, destroyed, or read
// until the operation has completed.
//
// ==> Cancellation does NOT count as completion. If an operation using
// an IOBuffer is cancelled, the caller should release their
// reference to this IOBuffer at the time of cancellation since
// they can no longer use it.
//
// For instance, if you were to call a Read() operation on some class which
// takes an IOBuffer, and then delete that class (which generally will
// trigger cancellation), the IOBuffer which had been passed to Read() should
// never be re-used.
//
// This usage contract is assumed by any API which takes an IOBuffer, even
// though it may not be explicitly mentioned in the function's comments.
//
// -----------------------
// Motivation
// -----------------------
//
// The motivation for transferring ownership during cancellation is
// to make it easier to work with un-cancellable operations.
//
// For instance, let's say under the hood your API called out to the
// operating system's synchronous ReadFile() function on a worker thread.
// When cancelling through our asynchronous interface, we have no way of
// actually aborting the in progress ReadFile(). We must let it keep running,
// and hence the buffer it was reading into must remain alive. Using
// reference counting we can add a reference to the IOBuffer and make sure
// it is not destroyed until after the synchronous operation has completed.
// Base class, never instantiated, does not own the buffer.
class NET_EXPORT IOBuffer : public base::RefCountedThreadSafe<IOBuffer> {
public:
// Returns the length from bytes() to the end of the buffer. Many methods that
// take an IOBuffer also take a size indicated the number of IOBuffer bytes to
// use from the start of bytes(). That number must be no more than the size()
// of the passed in IOBuffer.
int size() const {
// SetSpan() ensures this fits in an int.
return static_cast<int>(span_.size());
}
char* data() { return reinterpret_cast<char*>(bytes()); }
const char* data() const { return reinterpret_cast<const char*>(bytes()); }
uint8_t* bytes() { return span_.data(); }
const uint8_t* bytes() const { return span_.data(); }
base::span<uint8_t> span() { return span_; }
base::span<const uint8_t> span() const {
// Converts a const base::span<uint8_t> to a base::span<const uint8_t>.
return base::as_byte_span(span_);
}
// Convenience methods for accessing the buffer as a span.
base::span<uint8_t> first(size_t count) { return span().first(count); }
base::span<const uint8_t> first(size_t count) const {
return span().first(count);
}
protected:
friend class base::RefCountedThreadSafe<IOBuffer>;
static void AssertValidBufferSize(size_t size);
IOBuffer();
explicit IOBuffer(base::span<char> span);
explicit IOBuffer(base::span<uint8_t> span);
virtual ~IOBuffer();
// Sets `span_` to `span`. CHECKs if its size is too big to fit in an int.
void SetSpan(base::span<uint8_t> span);
// Like SetSpan(base::span<uint8_t>()), but without a size check. Particularly
// useful to call in the destructor of subclasses, to avoid failing raw
// reference checks.
void ClearSpan();
private:
base::raw_span<uint8_t> span_;
};
// Class which owns its buffer and manages its destruction.
class NET_EXPORT IOBufferWithSize : public IOBuffer {
public:
IOBufferWithSize();
explicit IOBufferWithSize(size_t size);
protected:
~IOBufferWithSize() override;
private:
base::HeapArray<uint8_t> storage_;
};
// This is like IOBufferWithSize, except its constructor takes a vector.
// IOBufferWithSize uses a HeapArray instead of a vector so that it can avoid
// initializing its data. VectorIOBuffer is primarily useful useful for writing
// data, while IOBufferWithSize is primarily useful for reading data.
class NET_EXPORT VectorIOBuffer : public IOBuffer {
public:
explicit VectorIOBuffer(std::vector<uint8_t> vector);
explicit VectorIOBuffer(base::span<const uint8_t> span);
private:
~VectorIOBuffer() override;
std::vector<uint8_t> vector_;
};
// This is a read only IOBuffer. The data is stored in a string and
// the IOBuffer interface does not provide a proper way to modify it.
class NET_EXPORT StringIOBuffer : public IOBuffer {
public:
explicit StringIOBuffer(std::string s);
private:
~StringIOBuffer() override;
std::string string_data_;
};
// This version wraps an existing IOBuffer and provides convenient functions
// to progressively read all the data. The values returned by size() and bytes()
// are updated as bytes are consumed from the buffer.
//
// DrainableIOBuffer is useful when you have an IOBuffer that contains data
// to be written progressively, and Write() function takes an IOBuffer rather
// than char*. DrainableIOBuffer can be used as follows:
//
// // payload is the IOBuffer containing the data to be written.
// buf = base::MakeRefCounted<DrainableIOBuffer>(payload, payload_size);
//
// while (buf->BytesRemaining() > 0) {
// // Write() takes an IOBuffer. If it takes char*, we could
// // simply use the regular IOBuffer like payload->data() + offset.
// int bytes_written = Write(buf, buf->BytesRemaining());
// buf->DidConsume(bytes_written);
// }
//
class NET_EXPORT DrainableIOBuffer : public IOBuffer {
public:
// `base` should be treated as exclusively owned by the DrainableIOBuffer as
// long as the latter exists. Specifically, the span pointed to by `base`,
// including its size, must not change, as the `DrainableIOBuffer` maintains a
// copy of them internally.
DrainableIOBuffer(scoped_refptr<IOBuffer> base, size_t size);
// DidConsume() changes the |data_| pointer so that |data_| always points
// to the first unconsumed byte.
void DidConsume(int bytes);
// Returns the number of unconsumed bytes.
int BytesRemaining() const;
// Returns the number of consumed bytes.
int BytesConsumed() const;
// Seeks to an arbitrary point in the buffer. The notion of bytes consumed
// and remaining are updated appropriately.
void SetOffset(int bytes);
private:
~DrainableIOBuffer() override;
scoped_refptr<IOBuffer> base_;
int used_ = 0;
};
// This version provides a resizable buffer and a changeable offset. The values
// returned by size() and bytes() are updated whenever the offset of the buffer
// is set, or the buffer's capacity is changed.
//
// GrowableIOBuffer is useful when you read data progressively without
// knowing the total size in advance. GrowableIOBuffer can be used as
// follows:
//
// buf = base::MakeRefCounted<GrowableIOBuffer>();
// buf->SetCapacity(1024); // Initial capacity.
//
// while (!some_stream->IsEOF()) {
// // Double the capacity if the remaining capacity is empty.
// if (buf->RemainingCapacity() == 0)
// buf->SetCapacity(buf->capacity() * 2);
// int bytes_read = some_stream->Read(buf, buf->RemainingCapacity());
// buf->set_offset(buf->offset() + bytes_read);
// }
//
class NET_EXPORT GrowableIOBuffer : public IOBuffer {
public:
GrowableIOBuffer();
// realloc memory to the specified capacity.
void SetCapacity(int capacity);
int capacity() { return capacity_; }
// `offset` moves the `data_` pointer, allowing "seeking" in the data.
void set_offset(int offset);
int offset() { return offset_; }
// Advances the offset by `bytes`. It's equivalent to `set_offset(offset() +
// bytes)`, though does not accept negative values, as they likely indicate a
// bug.
void DidConsume(int bytes);
int RemainingCapacity();
// Returns the entire buffer, including the bytes before the `offset()`.
//
// The `span()` method in the base class only gives the part of the buffer
// after `offset()`.
base::span<uint8_t> everything();
base::span<const uint8_t> everything() const;
// Return a span before the `offset()`.
base::span<uint8_t> span_before_offset();
base::span<const uint8_t> span_before_offset() const;
private:
~GrowableIOBuffer() override;
// TODO(329476354): Convert to std::vector, use reserve()+resize() to make
// exact reallocs, and remove `capacity_`. Possibly with an allocator the
// default-initializes, if it's important to not initialize the new memory?
std::unique_ptr<uint8_t, base::FreeDeleter> real_data_;
int capacity_ = 0;
int offset_ = 0;
};
// This version allows a Pickle to be used as the storage for a write-style
// operation, avoiding an extra data copy.
class NET_EXPORT PickledIOBuffer : public IOBuffer {
public:
explicit PickledIOBuffer(std::unique_ptr<const base::Pickle> pickle);
private:
~PickledIOBuffer() override;
const std::unique_ptr<const base::Pickle> pickle_;
};
// This class allows the creation of a temporary IOBuffer that doesn't really
// own the underlying buffer. Please use this class only as a last resort.
// A good example is the buffer for a synchronous operation, where we can be
// sure that nobody is keeping an extra reference to this object so the lifetime
// of the buffer can be completely managed by its intended owner.
// This is now nearly the same as the base IOBuffer class, except that it
// accepts const data as constructor arguments.
class NET_EXPORT WrappedIOBuffer : public IOBuffer {
public:
explicit WrappedIOBuffer(base::span<const char> data);
explicit WrappedIOBuffer(base::span<const uint8_t> data);
protected:
~WrappedIOBuffer() override;
};
} // namespace net
#endif // NET_BASE_IO_BUFFER_H_
|