Kernfs should not try to rename a file to itself.

One precondition of VFS.PrepareRenameAt is that the `from` and `to` dentries
are not the same. Kernfs was not checking this, which could lead to a deadlock.

PiperOrigin-RevId: 359385974

1 files changed, 14 insertions, 0 deletions

diff --git a/test/syscalls/linux/rename.cc b/test/syscalls/linux/rename.cc
index 22c8c19cf..b1a813de0 100644
--- a/test/syscalls/linux/rename.cc
+++ b/test/syscalls/linux/rename.cc
@@ -424,6 +424,20 @@ TEST(RenameTest, SysfsPathEndingWithDots) {
               SyscallFailsWithErrno(EBUSY));
 }
 
+TEST(RenameTest, SysfsFileToSelf) {
+  // If a non-root user tries to rename inside /sys then we get EPERM.
+  SKIP_IF(geteuid() != 0);
+  std::string const path = "/sys/devices/system/cpu/online";
+  EXPECT_THAT(rename(path.c_str(), path.c_str()), SyscallSucceeds());
+}
+
+TEST(RenameTest, SysfsDirectoryToSelf) {
+  // If a non-root user tries to rename inside /sys then we get EPERM.
+  SKIP_IF(geteuid() != 0);
+  std::string const path = "/sys/devices";
+  EXPECT_THAT(rename(path.c_str(), path.c_str()), SyscallSucceeds());
+}
+
 }  // namespace
 
 }  // namespace testing