From 97ffc1e0139e124b7e36b5d1a62b90300f231118 Mon Sep 17 00:00:00 2001
From: Eelco Dolstra <edolstra@gmail.com>
Date: Wed, 7 Oct 2020 22:46:01 +0200
Subject: [PATCH] Dynamically disable user namespaces if CLONE_NEWUSER fails

This makes builds work inside nixos-enter.

Fixes #3145.
---
 src/libstore/build.cc | 45 ++++++++++++++++++++-----------------------
 1 file changed, 21 insertions(+), 24 deletions(-)

diff --git a/src/libstore/build.cc b/src/libstore/build.cc
index a645c429d..05a9ef088 100644
--- a/src/libstore/build.cc
+++ b/src/libstore/build.cc
@@ -924,8 +924,8 @@ private:
        result. */
     std::map<Path, ValidPathInfo> prevInfos;
 
-    uid_t sandboxUid = 1000;
-    gid_t sandboxGid = 100;
+    uid_t sandboxUid() { return usingUserNamespace ? 1000 : buildUser->getUID(); }
+    gid_t sandboxGid() { return usingUserNamespace ?  100 : buildUser->getGID(); }
 
     const static Path homeDir;
 
@@ -2428,15 +2428,14 @@ void DerivationGoal::startBuilder()
                 "root:x:0:0:Nix build user:%3%:/noshell\n"
                 "nixbld:x:%1%:%2%:Nix build user:%3%:/noshell\n"
                 "nobody:x:65534:65534:Nobody:/:/noshell\n",
-                sandboxUid, sandboxGid, settings.sandboxBuildDir));
+                sandboxUid(), sandboxGid(), settings.sandboxBuildDir));
 
         /* Declare the build user's group so that programs get a consistent
            view of the system (e.g., "id -gn"). */
         writeFile(chrootRootDir + "/etc/group",
-            (format(
-                "root:x:0:\n"
+            fmt("root:x:0:\n"
                 "nixbld:!:%1%:\n"
-                "nogroup:x:65534:\n") % sandboxGid).str());
+                "nogroup:x:65534:\n", sandboxGid()));
 
         /* Create /etc/hosts with localhost entry. */
         if (!(derivationIsImpure(derivationType)))
@@ -2640,17 +2639,6 @@ void DerivationGoal::startBuilder()
 
         usingUserNamespace = userNamespacesEnabled;
 
-        if (usingUserNamespace) {
-            sandboxUid = 1000;
-            sandboxGid = 100;
-        } else {
-            debug("note: not using a user namespace");
-            if (!buildUser)
-                throw Error("cannot perform a sandboxed build because user namespaces are not enabled; check /proc/sys/user/max_user_namespaces");
-            sandboxUid = buildUser->getUID();
-            sandboxGid = buildUser->getGID();
-        }
-
         Pid helper = startProcess([&]() {
 
             /* Drop additional groups here because we can't do it
@@ -2682,11 +2670,12 @@ void DerivationGoal::startBuilder()
                 flags &= ~CLONE_NEWPID;
                 child = clone(childEntry, stack + stackSize, flags, this);
             }
-            if (child == -1 && (errno == EPERM || errno == EINVAL)) {
+            if (usingUserNamespace && child == -1 && (errno == EPERM || errno == EINVAL)) {
                 /* Some distros patch Linux to not allow unprivileged
                  * user namespaces. If we get EPERM or EINVAL, try
                  * without CLONE_NEWUSER and see if that works.
                  */
+                usingUserNamespace = false;
                 flags &= ~CLONE_NEWUSER;
                 child = clone(childEntry, stack + stackSize, flags, this);
             }
@@ -2697,7 +2686,8 @@ void DerivationGoal::startBuilder()
                 _exit(1);
             if (child == -1) throw SysError("cloning builder process");
 
-            writeFull(builderOut.writeSide.get(), std::to_string(child) + "\n");
+            writeFull(builderOut.writeSide.get(),
+                fmt("%d %d\n", usingUserNamespace, child));
             _exit(0);
         }, options);
 
@@ -2718,7 +2708,10 @@ void DerivationGoal::startBuilder()
         });
 
         pid_t tmp;
-        if (!string2Int<pid_t>(readLine(builderOut.readSide.get()), tmp)) abort();
+        auto ss = tokenizeString<std::vector<std::string>>(readLine(builderOut.readSide.get()));
+        assert(ss.size() == 2);
+        usingUserNamespace = ss[0] == "1";
+        if (!string2Int<pid_t>(ss[1], tmp)) abort();
         pid = tmp;
 
         if (usingUserNamespace) {
@@ -2729,12 +2722,16 @@ void DerivationGoal::startBuilder()
             uid_t hostGid = buildUser ? buildUser->getGID() : getgid();
 
             writeFile("/proc/" + std::to_string(pid) + "/uid_map",
-                (format("%d %d 1") % sandboxUid % hostUid).str());
+                fmt("%d %d 1", sandboxUid(), hostUid));
 
             writeFile("/proc/" + std::to_string(pid) + "/setgroups", "deny");
 
             writeFile("/proc/" + std::to_string(pid) + "/gid_map",
-                (format("%d %d 1") % sandboxGid % hostGid).str());
+                fmt("%d %d 1", sandboxGid(), hostGid));
+        } else {
+            debug("note: not using a user namespace");
+            if (!buildUser)
+                throw Error("cannot perform a sandboxed build because user namespaces are not enabled; check /proc/sys/user/max_user_namespaces");
         }
 
         /* Save the mount namespace of the child. We have to do this
@@ -3595,9 +3592,9 @@ void DerivationGoal::runChild()
             /* Switch to the sandbox uid/gid in the user namespace,
                which corresponds to the build user or calling user in
                the parent namespace. */
-            if (setgid(sandboxGid) == -1)
+            if (setgid(sandboxGid()) == -1)
                 throw SysError("setgid failed");
-            if (setuid(sandboxUid) == -1)
+            if (setuid(sandboxUid()) == -1)
                 throw SysError("setuid failed");
 
             setUser = false;