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
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
//! linux_raw syscalls supporting `rustix::event`.
//!
//! # Safety
//!
//! See the `rustix::backend` module documentation for details.
#![allow(unsafe_code, clippy::undocumented_unsafe_blocks)]
use crate::backend::conv::{
by_ref, c_int, c_uint, opt_mut, opt_ref, pass_usize, ret, ret_c_int, ret_error, ret_owned_fd,
ret_usize, size_of, slice_mut, zero,
};
use crate::event::{epoll, EventfdFlags, FdSetElement, PollFd, Timespec};
use crate::fd::{BorrowedFd, OwnedFd};
use crate::io;
use core::ptr::null_mut;
use linux_raw_sys::general::{kernel_sigset_t, EPOLL_CTL_ADD, EPOLL_CTL_DEL, EPOLL_CTL_MOD};
#[inline]
pub(crate) fn poll(fds: &mut [PollFd<'_>], timeout: Option<&Timespec>) -> io::Result<usize> {
let (fds_addr_mut, fds_len) = slice_mut(fds);
#[cfg(target_pointer_width = "32")]
unsafe {
// If we don't have Linux 5.1, and the timeout fits in a
// `__kernel_old_timespec`, use plain `ppoll`.
//
// We do this unconditionally, rather than trying `ppoll_time64` and
// falling back on `Errno::NOSYS`, because seccomp configurations will
// sometimes abort the process on syscalls they don't recognize.
#[cfg(not(feature = "linux_5_1"))]
{
use linux_raw_sys::general::__kernel_old_timespec;
// If we don't have a timeout, or if we can convert the timeout to
// a `__kernel_old_timespec`, the use `__NR_ppoll`.
fn convert(timeout: &Timespec) -> Option<__kernel_old_timespec> {
Some(__kernel_old_timespec {
tv_sec: timeout.tv_sec.try_into().ok()?,
tv_nsec: timeout.tv_nsec.try_into().ok()?,
})
}
let old_timeout = if let Some(timeout) = timeout {
match convert(timeout) {
// Could not convert timeout.
None => None,
// Could convert timeout. Ok!
Some(old_timeout) => Some(Some(old_timeout)),
}
} else {
// No timeout. Ok!
Some(None)
};
if let Some(mut old_timeout) = old_timeout {
// Call `ppoll`.
//
// Linux's `ppoll` mutates the timeout argument. Our public
// interface does not do this, because it's not portable to other
// platforms, so we create a temporary value to hide this behavior.
return ret_usize(syscall!(
__NR_ppoll,
fds_addr_mut,
fds_len,
opt_mut(old_timeout.as_mut()),
zero(),
size_of::<kernel_sigset_t, _>()
));
}
}
// We either have Linux 5.1 or the timeout didn't fit in
// `__kernel_old_timespec` so `__NR_ppoll_time64` will either
// succeed or fail due to our having no other options.
// Call `ppoll_time64`.
//
// Linux's `ppoll_time64` mutates the timeout argument. Our public
// interface does not do this, because it's not portable to other
// platforms, so we create a temporary value to hide this behavior.
ret_usize(syscall!(
__NR_ppoll_time64,
fds_addr_mut,
fds_len,
opt_mut(timeout.copied().as_mut()),
zero(),
size_of::<kernel_sigset_t, _>()
))
}
#[cfg(target_pointer_width = "64")]
unsafe {
// Call `ppoll`.
//
// Linux's `ppoll` mutates the timeout argument. Our public interface
// does not do this, because it's not portable to other platforms, so
// we create a temporary value to hide this behavior.
ret_usize(syscall!(
__NR_ppoll,
fds_addr_mut,
fds_len,
opt_mut(timeout.copied().as_mut()),
zero(),
size_of::<kernel_sigset_t, _>()
))
}
}
pub(crate) unsafe fn select(
nfds: i32,
readfds: Option<&mut [FdSetElement]>,
writefds: Option<&mut [FdSetElement]>,
exceptfds: Option<&mut [FdSetElement]>,
timeout: Option<&crate::timespec::Timespec>,
) -> io::Result<i32> {
let len = crate::event::fd_set_num_elements_for_bitvector(nfds);
let readfds = match readfds {
Some(readfds) => {
assert!(readfds.len() >= len);
readfds.as_mut_ptr()
}
None => null_mut(),
};
let writefds = match writefds {
Some(writefds) => {
assert!(writefds.len() >= len);
writefds.as_mut_ptr()
}
None => null_mut(),
};
let exceptfds = match exceptfds {
Some(exceptfds) => {
assert!(exceptfds.len() >= len);
exceptfds.as_mut_ptr()
}
None => null_mut(),
};
#[cfg(target_pointer_width = "32")]
{
// If we don't have Linux 5.1, and the timeout fits in a
// `__kernel_old_timespec`, use plain `pselect6`.
//
// We do this unconditionally, rather than trying `pselect6_time64` and
// falling back on `Errno::NOSYS`, because seccomp configurations will
// sometimes abort the process on syscalls they don't recognize.
#[cfg(not(feature = "linux_5_1"))]
{
use linux_raw_sys::general::__kernel_old_timespec;
// If we don't have a timeout, or if we can convert the timeout to
// a `__kernel_old_timespec`, the use `__NR_pselect6`.
fn convert(timeout: &Timespec) -> Option<__kernel_old_timespec> {
Some(__kernel_old_timespec {
tv_sec: timeout.tv_sec.try_into().ok()?,
tv_nsec: timeout.tv_nsec.try_into().ok()?,
})
}
let old_timeout = if let Some(timeout) = timeout {
match convert(timeout) {
// Could not convert timeout.
None => None,
// Could convert timeout. Ok!
Some(old_timeout) => Some(Some(old_timeout)),
}
} else {
// No timeout. Ok!
Some(None)
};
if let Some(mut old_timeout) = old_timeout {
// Call `pselect6`.
//
// Linux's `pselect6` mutates the timeout argument. Our public
// interface does not do this, because it's not portable to other
// platforms, so we create a temporary value to hide this behavior.
return ret_c_int(syscall!(
__NR_pselect6,
c_int(nfds),
readfds,
writefds,
exceptfds,
opt_mut(old_timeout.as_mut()),
zero()
));
}
}
// We either have Linux 5.1 or the timeout didn't fit in
// `__kernel_old_timespec` so `__NR_pselect6_time64` will either
// succeed or fail due to our having no other options.
// Call `pselect6_time64`.
//
// Linux's `pselect6_time64` mutates the timeout argument. Our public
// interface does not do this, because it's not portable to other
// platforms, so we create a temporary value to hide this behavior.
ret_c_int(syscall!(
__NR_pselect6_time64,
c_int(nfds),
readfds,
writefds,
exceptfds,
opt_mut(timeout.copied().as_mut()),
zero()
))
}
#[cfg(target_pointer_width = "64")]
{
// Call `pselect6`.
//
// Linux's `pselect6` mutates the timeout argument. Our public interface
// does not do this, because it's not portable to other platforms, so we
// create a temporary value to hide this behavior.
ret_c_int(syscall!(
__NR_pselect6,
c_int(nfds),
readfds,
writefds,
exceptfds,
opt_mut(timeout.copied().as_mut()),
zero()
))
}
}
#[inline]
pub(crate) fn epoll_create(flags: epoll::CreateFlags) -> io::Result<OwnedFd> {
// SAFETY: `__NR_epoll_create1` doesn't access any user memory.
unsafe { ret_owned_fd(syscall_readonly!(__NR_epoll_create1, flags)) }
}
#[inline]
pub(crate) fn epoll_add(
epfd: BorrowedFd<'_>,
fd: BorrowedFd<'_>,
event: &epoll::Event,
) -> io::Result<()> {
// SAFETY: `__NR_epoll_ctl` with `EPOLL_CTL_ADD` doesn't modify any user
// memory, and it only reads from `event`.
unsafe {
ret(syscall_readonly!(
__NR_epoll_ctl,
epfd,
c_uint(EPOLL_CTL_ADD),
fd,
by_ref(event)
))
}
}
#[inline]
pub(crate) fn epoll_mod(
epfd: BorrowedFd<'_>,
fd: BorrowedFd<'_>,
event: &epoll::Event,
) -> io::Result<()> {
// SAFETY: `__NR_epoll_ctl` with `EPOLL_CTL_MOD` doesn't modify any user
// memory, and it only reads from `event`.
unsafe {
ret(syscall_readonly!(
__NR_epoll_ctl,
epfd,
c_uint(EPOLL_CTL_MOD),
fd,
by_ref(event)
))
}
}
#[inline]
pub(crate) fn epoll_del(epfd: BorrowedFd<'_>, fd: BorrowedFd<'_>) -> io::Result<()> {
// SAFETY: `__NR_epoll_ctl` with `EPOLL_CTL_DEL` doesn't access any user
// memory.
unsafe {
ret(syscall_readonly!(
__NR_epoll_ctl,
epfd,
c_uint(EPOLL_CTL_DEL),
fd,
zero()
))
}
}
#[inline]
pub(crate) unsafe fn epoll_wait(
epfd: BorrowedFd<'_>,
events: (*mut crate::event::epoll::Event, usize),
timeout: Option<&Timespec>,
) -> io::Result<usize> {
// If we don't have Linux 5.1, and the timeout fits in an `i32`, use plain
// `epoll_pwait`.
//
// We do this unconditionally, rather than trying `epoll_pwait2` and
// falling back on `Errno::NOSYS`, because seccomp configurations will
// sometimes abort the process on syscalls they don't recognize.
#[cfg(not(feature = "linux_5_11"))]
{
// If we don't have a timeout, or if we can convert the timeout to an
// `i32`, the use `__NR_epoll_pwait`.
let old_timeout = if let Some(timeout) = timeout {
// Try to convert the timeout; if this is `Some`, we're ok!
timeout.as_c_int_millis()
} else {
// No timeout. Ok!
Some(-1)
};
if let Some(old_timeout) = old_timeout {
// Call `epoll_pwait`.
return ret_usize(syscall!(
__NR_epoll_pwait,
epfd,
events.0,
pass_usize(events.1),
c_int(old_timeout),
zero()
));
}
}
// Call `epoll_pwait2`.
//
// We either have Linux 5.1 or the timeout didn't fit in an `i32`, so
// `__NR_epoll_pwait2` will either succeed or fail due to our having no
// other options.
ret_usize(syscall!(
__NR_epoll_pwait2,
epfd,
events.0,
pass_usize(events.1),
opt_ref(timeout),
zero()
))
}
#[inline]
pub(crate) fn eventfd(initval: u32, flags: EventfdFlags) -> io::Result<OwnedFd> {
unsafe { ret_owned_fd(syscall_readonly!(__NR_eventfd2, c_uint(initval), flags)) }
}
#[inline]
pub(crate) fn pause() {
unsafe {
#[cfg(any(target_arch = "aarch64", target_arch = "riscv64"))]
let error = ret_error(syscall_readonly!(
__NR_ppoll,
zero(),
zero(),
zero(),
zero()
));
#[cfg(not(any(target_arch = "aarch64", target_arch = "riscv64")))]
let error = ret_error(syscall_readonly!(__NR_pause));
debug_assert_eq!(error, io::Errno::INTR);
}
}