[tor-commits] [tor/master] Add a sandbox workaround for Glibc 2.33
nickm at torproject.org
nickm at torproject.org
Fri May 7 17:08:35 UTC 2021
commit f5acfe67238a331bf8a6e94715163949999f27e7
Author: Nick Mathewson <nickm at torproject.org>
Date: Fri May 7 12:09:41 2021 -0400
Add a sandbox workaround for Glibc 2.33
This change permits the newfstatat() system call, and fixes issues
40382 (and 40381).
This isn't a free change. From the commit:
// Libc 2.33 uses this syscall to implement both fstat() and stat().
//
// The trouble is that to implement fstat(fd, &st), it calls:
// newfstatat(fs, "", &st, AT_EMPTY_PATH)
// We can't detect this usage in particular, because "" is a pointer
// we don't control. And we can't just look for AT_EMPTY_PATH, since
// AT_EMPTY_PATH only has effect when the path string is empty.
//
// So our only solution seems to be allowing all fstatat calls, which
// means that an attacker can stat() anything on the filesystem. That's
// not a great solution, but I can't find a better one.
---
changes/ticket40382 | 6 ++++++
src/lib/sandbox/sandbox.c | 22 ++++++++++++++++++++++
2 files changed, 28 insertions(+)
diff --git a/changes/ticket40382 b/changes/ticket40382
new file mode 100644
index 0000000000..5ac1b771b9
--- /dev/null
+++ b/changes/ticket40382
@@ -0,0 +1,6 @@
+ o Minor features (compatibility, Linux seccomp sandbox):
+ - Add a workaround to enable the Linux sandbox to work correctly
+ on systems running Glibc 2.33. These versions have started
+ using the fstatat() system call, which previously our sandbox did not
+ allow.
+ Closes ticket 40382; see the ticket for a discussion of tradeoffs.
diff --git a/src/lib/sandbox/sandbox.c b/src/lib/sandbox/sandbox.c
index 168dfd943c..fc90dbe062 100644
--- a/src/lib/sandbox/sandbox.c
+++ b/src/lib/sandbox/sandbox.c
@@ -1608,6 +1608,28 @@ add_noparam_filter(scmp_filter_ctx ctx)
}
}
+ if (is_libc_at_least(2, 33)) {
+#ifdef __NR_newfstatat
+ // Libc 2.33 uses this syscall to implement both fstat() and stat().
+ //
+ // The trouble is that to implement fstat(fd, &st), it calls:
+ // newfstatat(fs, "", &st, AT_EMPTY_PATH)
+ // We can't detect this usage in particular, because "" is a pointer
+ // we don't control. And we can't just look for AT_EMPTY_PATH, since
+ // AT_EMPTY_PATH only has effect when the path string is empty.
+ //
+ // So our only solution seems to be allowing all fstatat calls, which
+ // means that an attacker can stat() anything on the filesystem. That's
+ // not a great solution, but I can't find a better one.
+ rc = seccomp_rule_add_0(ctx, SCMP_ACT_ALLOW, SCMP_SYS(newfstatat));
+ if (rc != 0) {
+ log_err(LD_BUG,"(Sandbox) failed to add newfstatat() syscall; "
+ "received libseccomp error %d", rc);
+ return rc;
+ }
+#endif
+ }
+
return 0;
}
More information about the tor-commits
mailing list