From 5d885d7fb21414d903d57ffe2b95bcc62c098d6a Mon Sep 17 00:00:00 2001
From: Dean Deng <deandeng@google.com>
Date: Mon, 13 Apr 2020 13:01:28 -0700
Subject: Port socket-related syscalls to VFS2.

Note that most kinds of sockets are not yet supported in VFS2
(only Unix sockets are partially supported at the moment), so
these syscalls will still generally fail. Enabling them allows
us to begin running socket tests for VFS2 as more features are
ported over.

Updates #1476, #1478, #1484, #1485.

PiperOrigin-RevId: 306292294
---
 pkg/sentry/kernel/fd_table.go | 55 +++++++++++++++++++++++++++++++++++++++++++
 pkg/sentry/kernel/task.go     |  9 +++++++
 2 files changed, 64 insertions(+)

(limited to 'pkg/sentry/kernel')

diff --git a/pkg/sentry/kernel/fd_table.go b/pkg/sentry/kernel/fd_table.go
index d09d97825..ed40b5303 100644
--- a/pkg/sentry/kernel/fd_table.go
+++ b/pkg/sentry/kernel/fd_table.go
@@ -307,6 +307,61 @@ func (f *FDTable) NewFDs(ctx context.Context, fd int32, files []*fs.File, flags
 	return fds, nil
 }
 
+// NewFDsVFS2 allocates new FDs guaranteed to be the lowest number available
+// greater than or equal to the fd parameter. All files will share the set
+// flags. Success is guaranteed to be all or none.
+func (f *FDTable) NewFDsVFS2(ctx context.Context, fd int32, files []*vfs.FileDescription, flags FDFlags) (fds []int32, err error) {
+	if fd < 0 {
+		// Don't accept negative FDs.
+		return nil, syscall.EINVAL
+	}
+
+	// Default limit.
+	end := int32(math.MaxInt32)
+
+	// Ensure we don't get past the provided limit.
+	if limitSet := limits.FromContext(ctx); limitSet != nil {
+		lim := limitSet.Get(limits.NumberOfFiles)
+		if lim.Cur != limits.Infinity {
+			end = int32(lim.Cur)
+		}
+		if fd >= end {
+			return nil, syscall.EMFILE
+		}
+	}
+
+	f.mu.Lock()
+	defer f.mu.Unlock()
+
+	// From f.next to find available fd.
+	if fd < f.next {
+		fd = f.next
+	}
+
+	// Install all entries.
+	for i := fd; i < end && len(fds) < len(files); i++ {
+		if d, _, _ := f.getVFS2(i); d == nil {
+			f.setVFS2(i, files[len(fds)], flags) // Set the descriptor.
+			fds = append(fds, i)                 // Record the file descriptor.
+		}
+	}
+
+	// Failure? Unwind existing FDs.
+	if len(fds) < len(files) {
+		for _, i := range fds {
+			f.setVFS2(i, nil, FDFlags{}) // Zap entry.
+		}
+		return nil, syscall.EMFILE
+	}
+
+	if fd == f.next {
+		// Update next search start position.
+		f.next = fds[len(fds)-1] + 1
+	}
+
+	return fds, nil
+}
+
 // NewFDVFS2 allocates a file descriptor greater than or equal to minfd for
 // the given file description. If it succeeds, it takes a reference on file.
 func (f *FDTable) NewFDVFS2(ctx context.Context, minfd int32, file *vfs.FileDescription, flags FDFlags) (int32, error) {
diff --git a/pkg/sentry/kernel/task.go b/pkg/sentry/kernel/task.go
index d6546735e..e5d133d6c 100644
--- a/pkg/sentry/kernel/task.go
+++ b/pkg/sentry/kernel/task.go
@@ -777,6 +777,15 @@ func (t *Task) NewFDs(fd int32, files []*fs.File, flags FDFlags) ([]int32, error
 	return t.fdTable.NewFDs(t, fd, files, flags)
 }
 
+// NewFDsVFS2 is a convenience wrapper for t.FDTable().NewFDsVFS2.
+//
+// This automatically passes the task as the context.
+//
+// Precondition: same as FDTable.
+func (t *Task) NewFDsVFS2(fd int32, files []*vfs.FileDescription, flags FDFlags) ([]int32, error) {
+	return t.fdTable.NewFDsVFS2(t, fd, files, flags)
+}
+
 // NewFDFrom is a convenience wrapper for t.FDTable().NewFDs with a single file.
 //
 // This automatically passes the task as the context.
-- 
cgit v1.2.3