As explained in bug #14796, there are situations where the seccomp based
sandboxing in openssh can get confused, leading to connection issues.
As explained by Thomas in the bug report:
glibc does not care about the kernel headers when deciding whether to
try the clock_gettime64() syscall or not: it always use it, and if that
fails at runtime, it falls back to clock_gettime(). This is how glibc
ends up using clock_gettime64() even if your kernel does not support it.
On the other hand, the OpenSSL seccomp code relies on kernel headers to
decide whether the clock_gettime64() syscall should be in the allowed
list of syscalls or not.
So when you are in a situation where glibc is recent, but your kernel is
older, you get into precisely the problem you have: glibc tries to use
clock_gettime64, but OpenSSH seccomp configuration prevents that, which
does not allow glibc to gracefully fallback to clock_gettime (as seccomp
is configured to kill the process on filter violations).
As a workaround, add a _OPENSSH_SANDBOX option (defaulting to y) to
decide if sandboxing should be used or not.
--with-sandbox expects the type of sandboxing to use, and if not
specified, will use the first one available in a list: pledge, systrace,
darwin, seccomp, capsicum, rlimit. On Linux, only seccomp and rlimit are
available, and rlimit probably does not bring much security-wise, so in
all practical matters, on Linux, sandboxing uses seccomp or there is no
sandboxing, so let's just disable sandboxing when we do not want to use
seccomp, and let configure detect seccomp when we request sandboxing.
Fixes (works around) #14796
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
[yann.morin.1998@free.fr: add § about sandboxing types]
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
f204766b8f