Kernel: Add a basic implementation of unveil()

This syscall is a complement to pledge() and adds the same sort of
incremental relinquishing of capabilities for filesystem access.

The first call to unveil() will "drop a veil" on the process, and from
now on, only unveiled parts of the filesystem are visible to it.

Each call to unveil() specifies a path to either a directory or a file
along with permissions for that path. The permissions are a combination
of the following:

- r: Read access (like the "rpath" promise)
- w: Write access (like the "wpath" promise)
- x: Execute access
- c: Create/remove access (like the "cpath" promise)

Attempts to open a path that has not been unveiled with fail with
ENOENT. If the unveiled path lacks sufficient permissions, it will fail
with EACCES.

Like pledge(), subsequent calls to unveil() with the same path can only
remove permissions, not add them.

Once you call unveil(nullptr, nullptr), the veil is locked, and it's no
longer possible to unveil any more paths for the process, ever.

This concept comes from OpenBSD, and their implementation does various
things differently, I'm sure. This is just a first implementation for
SerenityOS, and we'll keep improving on it as we go. :^)

Kernel/FileSystem/VirtualFileSystem.cpp

@@ -564,7 +564,7 @@ KResult VFS::link(StringView old_path, StringView new_path, Custody& base)
 KResult VFS::unlink(StringView path, Custody& base)
 {
     RefPtr<Custody> parent_custody;
-    auto custody_or_error = resolve_path(path, base, &parent_custody, O_NOFOLLOW_NOERROR);
+    auto custody_or_error = resolve_path(path, base, &parent_custody, O_NOFOLLOW_NOERROR | O_UNLINK_INTERNAL);
     if (custody_or_error.is_error())
         return custody_or_error.error();
     auto& custody = *custody_or_error.value();
@@ -709,8 +709,70 @@ Custody& VFS::root_custody()
     return *m_root_custody;
 }
 
+const UnveiledPath* VFS::find_matching_unveiled_path(StringView path)
+{
+    for (auto& unveiled_path : current->process().unveiled_paths()) {
+        if (path == unveiled_path.path)
+            return &unveiled_path;
+        if (path.starts_with(unveiled_path.path) && path.length() > unveiled_path.path.length() && path[unveiled_path.path.length()] == '/')
+            return &unveiled_path;
+    }
+    return nullptr;
+}
+
+KResult VFS::validate_path_against_process_veil(StringView path, int options)
+{
+    if (current->process().unveil_state() == UnveilState::None)
+        return KSuccess;
+
+    // FIXME: Figure out a nicer way to do this.
+    if (String(path).contains("/.."))
+        return KResult(-EINVAL);
+
+    auto* unveiled_path = find_matching_unveiled_path(path);
+    if (!unveiled_path) {
+        dbg() << *current << " rejecting path '" << path << "' since it hasn't been unveiled.";
+        return KResult(-ENOENT);
+    }
+
+    if (options & O_CREAT) {
+        if (!(unveiled_path->permissions & UnveiledPath::Access::CreateOrRemove)) {
+            dbg() << *current << " rejecting path '" << path << "' since it hasn't been unveiled with 'c' permission.";
+            return KResult(-EACCES);
+        }
+    }
+    if (options & O_UNLINK_INTERNAL) {
+        if (!(unveiled_path->permissions & UnveiledPath::Access::CreateOrRemove)) {
+            dbg() << *current << " rejecting path '" << path << "' for unlink since it hasn't been unveiled with 'c' permission.";
+            return KResult(-EACCES);
+        }
+        return KSuccess;
+    }
+    if ((options & O_RDWR) || (options & O_WRONLY)) {
+        if (!(unveiled_path->permissions & UnveiledPath::Access::Write)) {
+            dbg() << *current << " rejecting path '" << path << "' since it hasn't been unveiled with 'w' permission.";
+            return KResult(-EACCES);
+        }
+    } else if (options & O_EXEC) {
+        if (!(unveiled_path->permissions & UnveiledPath::Access::Execute)) {
+            dbg() << *current << " rejecting path '" << path << "' since it hasn't been unveiled with 'x' permission.";
+            return KResult(-EACCES);
+        }
+    } else {
+        if (!(unveiled_path->permissions & UnveiledPath::Access::Read)) {
+            dbg() << *current << " rejecting path '" << path << "' since it hasn't been unveiled with 'r' permission.";
+            return KResult(-EACCES);
+        }
+    }
+    return KSuccess;
+}
+
 KResultOr<NonnullRefPtr<Custody>> VFS::resolve_path(StringView path, Custody& base, RefPtr<Custody>* out_parent, int options, int symlink_recursion_level)
 {
+    auto result = validate_path_against_process_veil(path, options);
+    if (result.is_error())
+        return result;
+
     if (symlink_recursion_level >= symlink_recursion_limit)
         return KResult(-ELOOP);
 

Kernel/FileSystem/VirtualFileSystem.h

@@ -53,6 +53,7 @@
 #define O_CLOEXEC 02000000
 #define O_DIRECT 04000000
 #define O_NOFOLLOW_NOERROR 0x4000000
+#define O_UNLINK_INTERNAL 0x8000000
 
 #define MS_NODEV 1
 #define MS_NOEXEC 2
@@ -62,6 +63,7 @@
 class Custody;
 class Device;
 class FileDescription;
+class UnveiledPath;
 
 struct UidAndGid {
     uid_t uid;
@@ -134,6 +136,9 @@ public:
 private:
     friend class FileDescription;
 
+    const UnveiledPath* find_matching_unveiled_path(StringView path);
+    KResult validate_path_against_process_veil(StringView path, int options);
+
     RefPtr<Inode> get_inode(InodeIdentifier);
 
     bool is_vfs_root(InodeIdentifier) const;

Kernel/Process.cpp

@@ -1820,6 +1820,10 @@ bool Process::validate(const Syscall::ImmutableBufferArgument<DataType, SizeType
 
 String Process::validate_and_copy_string_from_user(const char* user_characters, size_t user_length) const
 {
+    if (!user_characters)
+        return {};
+    if (user_length == 0)
+        return String::empty();
     if (!validate_read(user_characters, user_length))
         return {};
     SmapDisabler disabler;
@@ -1933,6 +1937,9 @@ int Process::sys$open(const Syscall::SC_open_params* user_params)
     if (options & O_NOFOLLOW_NOERROR)
         return -EINVAL;
 
+    if (options & O_UNLINK_INTERNAL)
+        return -EINVAL;
+
     if ((options & O_RDWR) || (options & O_WRONLY))
         REQUIRE_PROMISE(wpath);
     else
@@ -4588,3 +4595,74 @@ Region& Process::add_region(NonnullOwnPtr<Region> region)
     m_regions.append(move(region));
     return *ptr;
 }
+
+int Process::sys$unveil(const Syscall::SC_unveil_params* user_params)
+{
+    Syscall::SC_unveil_params params;
+    if (!validate_read_and_copy_typed(&params, user_params))
+        return -EFAULT;
+
+    if (!params.path.characters && !params.permissions.characters) {
+        m_unveil_state = UnveilState::VeilLocked;
+        return 0;
+    }
+
+    if (m_unveil_state == UnveilState::VeilLocked)
+        return -EPERM;
+
+    if (!params.path.characters || !params.permissions.characters)
+        return -EINVAL;
+
+    if (params.permissions.length > 4)
+        return -EINVAL;
+
+    auto path = get_syscall_path_argument(params.path);
+    if (path.is_error())
+        return path.error();
+
+    if (path.value().is_empty() || path.value().characters()[0] != '/')
+        return -EINVAL;
+
+    auto permissions = validate_and_copy_string_from_user(params.permissions);
+    if (permissions.is_null())
+        return -EFAULT;
+
+    unsigned new_permissions = 0;
+    for (size_t i = 0; i < permissions.length(); ++i) {
+        switch (permissions[i]) {
+        case 'r':
+            new_permissions |= UnveiledPath::Access::Read;
+            break;
+        case 'w':
+            new_permissions |= UnveiledPath::Access::Write;
+            break;
+        case 'x':
+            new_permissions |= UnveiledPath::Access::Execute;
+            break;
+        case 'c':
+            new_permissions |= UnveiledPath::Access::CreateOrRemove;
+            break;
+        default:
+            return -EINVAL;
+        }
+    }
+
+    for (int i = 0; i < m_unveiled_paths.size(); ++i) {
+        auto& unveiled_path = m_unveiled_paths[i];
+        if (unveiled_path.path == path.value()) {
+            if (new_permissions & ~unveiled_path.permissions)
+                return -EPERM;
+            if (!new_permissions) {
+                m_unveiled_paths.remove(i);
+                return 0;
+            }
+            unveiled_path.permissions = new_permissions;
+            return 0;
+        }
+    }
+
+    m_unveiled_paths.append({ path.value(), new_permissions });
+    ASSERT(m_unveil_state != UnveilState::VeilLocked);
+    m_unveil_state = UnveilState::VeilDropped;
+    return 0;
+}

Kernel/Process.h

@@ -82,6 +82,24 @@ enum class Pledge : u32 {
 #undef __ENUMERATE_PLEDGE_PROMISE
 };
 
+enum class UnveilState {
+    None,
+    VeilDropped,
+    VeilLocked,
+};
+
+struct UnveiledPath {
+    enum Access {
+        Read = 1,
+        Write = 2,
+        Execute = 4,
+        CreateOrRemove = 8,
+    };
+
+    String path;
+    unsigned permissions { 0 };
+};
+
 class Process : public InlineLinkedListNode<Process>
     , public Weakable<Process> {
     friend class InlineLinkedListNode<Process>;
@@ -282,6 +300,7 @@ public:
     int sys$set_process_boost(pid_t, int amount);
     int sys$chroot(const char* path, size_t path_length, int mount_flags);
     int sys$pledge(const Syscall::SC_pledge_params*);
+    int sys$unveil(const Syscall::SC_unveil_params*);
 
     static void initialize();
 
@@ -380,6 +399,9 @@ public:
     bool has_promises() const { return m_promises; }
     bool has_promised(Pledge pledge) const { return m_promises & (1u << (u32)pledge); }
 
+    UnveilState unveil_state() const { return m_unveil_state; }
+    const Vector<UnveiledPath>& unveiled_paths() const { return m_unveiled_paths; }
+
 private:
     friend class MemoryManager;
     friend class Scheduler;
@@ -481,6 +503,9 @@ private:
     u32 m_promises { 0 };
     u32 m_execpromises { 0 };
 
+    UnveilState m_unveil_state { UnveilState::None };
+    Vector<UnveiledPath> m_unveiled_paths;
+
     WaitQueue& futex_queue(i32*);
     HashMap<u32, OwnPtr<WaitQueue>> m_futex_queues;
 };

Kernel/Syscall.h

@@ -174,7 +174,8 @@ typedef u32 socklen_t;
     __ENUMERATE_SYSCALL(set_thread_boost)           \
     __ENUMERATE_SYSCALL(set_process_boost)          \
     __ENUMERATE_SYSCALL(chroot)                     \
-    __ENUMERATE_SYSCALL(pledge)
+    __ENUMERATE_SYSCALL(pledge)                     \
+    __ENUMERATE_SYSCALL(unveil)
 
 namespace Syscall {
 
@@ -385,6 +386,11 @@ struct SC_pledge_params {
     StringArgument execpromises;
 };
 
+struct SC_unveil_params {
+    StringArgument path;
+    StringArgument permissions;
+};
+
 void initialize();
 int sync();
 

Libraries/LibC/unistd.cpp

@@ -694,5 +694,13 @@ int pledge(const char* promises, const char* execpromises)
     __RETURN_WITH_ERRNO(rc, rc, -1);
 }
 
+int unveil(const char* path, const char* permissions)
+{
+    Syscall::SC_unveil_params params {
+        { path, path ? strlen(path) : 0 },
+        { permissions, permissions ? strlen(permissions) : 0 }
+    };
+    int rc = syscall(SC_unveil, &params);
+    __RETURN_WITH_ERRNO(rc, rc, -1);
+}
 }
-

Libraries/LibC/unistd.h

@@ -141,6 +141,7 @@ int reboot();
 int mount(const char* source, const char* target, const char* fs_type, int flags);
 int umount(const char* mountpoint);
 int pledge(const char* promises, const char* execpromises);
+int unveil(const char* path, const char* permissions);
 
 enum {
     _PC_NAME_MAX,