Migrate from using io.ReadSeeker to io.ReaderAt.

This provides the following benefits:
- We can now use pkg/fd package which does not take ownership
  of the file descriptor. So it does not close the fd when garbage collected.
  This reduces scope of errors from unexpected garbage collection of io.File.
- It enforces the offset parameter in every read call.
  It does not affect the fd offset nor is it affected by it. Hence reducing
  scope of error of using stale offsets when reading.
- We do not need to serialize the usage of any global file descriptor anymore.
  So this drops the mutual exclusion req hence reducing complexity and
  congestion.

PiperOrigin-RevId: 260635174

1 files changed, 2 insertions, 4 deletions

diff --git a/pkg/sentry/fs/ext/inode.go b/pkg/sentry/fs/ext/inode.go
index 7d2a445fb..00e022953 100644
--- a/pkg/sentry/fs/ext/inode.go
+++ b/pkg/sentry/fs/ext/inode.go
@@ -75,7 +75,7 @@ func (in *inode) tryIncRef() bool {
 // decRef decrements the inode ref count and releases the inode resources if
 // the ref count hits 0.
 //
-// Preconditions: Must have locked fs.mu.
+// Precondition: Must have locked fs.mu.
 func (in *inode) decRef(fs *filesystem) {
 	if refs := atomic.AddInt64(&in.refs, -1); refs == 0 {
 		delete(fs.inodeCache, in.inodeNum)
@@ -86,9 +86,7 @@ func (in *inode) decRef(fs *filesystem) {
 
 // newInode is the inode constructor. Reads the inode off disk. Identifies
 // inodes based on the absolute inode number on disk.
-//
-// Preconditions: Must hold the mutex of the filesystem containing dev.
-func newInode(ctx context.Context, dev io.ReadSeeker, sb disklayout.SuperBlock, bgs []disklayout.BlockGroup, inodeNum uint32) (*inode, error) {
+func newInode(ctx context.Context, dev io.ReaderAt, sb disklayout.SuperBlock, bgs []disklayout.BlockGroup, inodeNum uint32) (*inode, error) {
 	if inodeNum == 0 {
 		panic("inode number 0 on ext filesystems is not possible")
 	}