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
|
linux-user: fails to deliver signals raised during pselect
Description of problem:
When run via qemu a program which blocks signals but unmasks them during `pselect` does not catch these signals when returning from `pselect`.
Used as reference on expected behavior: [The new pselect() system call](https://lwn.net/Articles/176911/)
Steps to reproduce:
A minimal test case below mimics behavior as encountered in the test suite of `p11-kit` ([link](https://github.com/p11-glue/p11-kit)) (which attempts to catch `SIGTERM` in a similar way and results in lingering processes after running the test suite).
```C
#include <stdio.h>
#include <unistd.h>
#include <signal.h>
#include <sys/select.h>
static void handler(int sig)
{
puts("SIGNAL");
}
int main(int argc, char *argv[])
{
struct sigaction sa;
fd_set rfds;
sigset_t emptyset, blockset;
sigemptyset (&blockset);
sigemptyset (&emptyset);
sigaddset (&blockset, SIGUSR1);
sa.sa_handler = handler;
sigemptyset(&sa.sa_mask);
sa.sa_flags = 0;
sigaction(SIGUSR1, &sa, NULL);
sigprocmask (SIG_BLOCK, &blockset, NULL);
FD_ZERO(&rfds);
while(1) {
pselect(0, &rfds, NULL, NULL, NULL, &emptyset);
}
return 0;
}
```
Running this without qemu should print _SIGNAL_ when sent `SIGUSR1`:
```
$ ./a.out &
[1] 1683587
$ kill -USR1 %1
$ SIGNAL
```
When run with `qemu-x86_64` however, it does not (also qemu's `-strace` confirms the signal isn't received whereas a strace of qemu shows it's in fact delivered).
The pselect call itself _is_ interrupted, but the signal goes missing.
|