podman exec --preserve-fds=2: panic

[edsantiago]

I know, I know, this is undefined, but it probably shouldn't panic:

$ ./bin/podman run --name test -d alpine top
3e96d0911deb459b8549fcc31b0124a13d18340b5dfbcc7c68791723acc0696c
$ ./bin/podman exec --preserve-fds=2 -i test date
runtime: epollwait on fd 4 failed with 9
fatal error: runtime: netpoll failed

runtime stack:
runtime.throw(0x1a568ef, 0x17)
        /usr/lib/golang/src/runtime/panic.go:1116 +0x72
runtime.netpoll(0x0, 0x0)
        /usr/lib/golang/src/runtime/netpoll_epoll.go:123 +0x363
runtime.findrunnable(0xc00005f800, 0x0)
        /usr/lib/golang/src/runtime/proc.go:2126 +0xc60
runtime.schedule()
        /usr/lib/golang/src/runtime/proc.go:2520 +0x2fc
runtime.park_m(0xc0005a2300)
        /usr/lib/golang/src/runtime/proc.go:2690 +0x9d
runtime.mcall(0x800000)
        /usr/lib/golang/src/runtime/asm_amd64.s:318 +0x5b
[lots and lots more]

With just 1, behavior is inconsistent:

$ ./bin/podman exec --preserve-fds=1 -i test date
Wed Jun 17 17:08:50 UTC 2020
Error: read /home/esm/.local/share/containers/storage/overlay-containers/3e96d0911deb459b8549fcc31b0124a13d18340b5dfbcc7c68791723acc0696c/userdata/bb1ef4c6c8a7b4ab51338ec78fe8f5696cdf3d36a3b00665d5bb229cb2801933/exit/3e96d0911deb459b8549fcc31b0124a13d18340b5dfbcc7c68791723acc0696c: bad file descriptor

or

$ ./bin/podman exec --preserve-fds=1 -i test date
Wed Jun 17 17:09:09 UTC 2020
2020/06/17 11:09:09 bolt.Close(): funlock error: bad file descriptor
ERRO[0000] failed to close libpod db: "db file close: close /home/esm/.local/share/containers/storage/libpod/bolt_state.db: bad file descriptor"

or sometimes even no error at all

$ ./bin/podman exec --preserve-fds=1 -i test date
Wed Jun 17 17:09:10 UTC 2020

master @ f293606

[QiWang19]

Is the same issue appearing again? #5769

[github-actions]

A friendly reminder that this issue had no activity for 30 days.

[rhatdan]

@edsantiago @QiWang19 Is this still an issue?

[edsantiago]

Still present in 17f9b80

[edsantiago]

Still present

[edsantiago]

Even simpler:

$ ./bin/podman run --preserve-fds=2 alpine date
runtime: epollwait on fd 4 failed with 9
fatal error: runtime: netpoll failed

[rhatdan]

ls /proc/self/fd
0 1 2 3

int fd_is_valid(int fd)
{
return fcntl(fd, F_GETFD) != -1 || errno != EBADF;
}