Native code within the OpenJDK code base attempted to call close() on a file descriptor repeatedly if it returned the error code, EINTR. However, the close() native call is not restartable and this caused the virtual machine to crash. The close() call is now only made once.
Description of problem (from upstream bug):
In java.base/linux/native/libnet/linux_close.c:
do {
if (fd1 < 0) {
rv = close(fd2);
} else {
rv = dup2(fd1, fd2);
}
} while (rv == -1 && errno == EINTR);
However, the close() call is not restartable.
Only dup2() needs to be restarted if interrupted.
Version-Release number of selected component (if applicable):
All versions
Additional info:
- This is a request from one of our customers to backport the patch for https://bugs.openjdk.org/browse/JDK-8205959 to OpenJDK 8. This was identified as the cause of one of their applications crashing by an external vendor.
Comment 3Andrew John Hughes
2023-02-09 17:38:11 UTC
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory (Important: java-1.8.0-openjdk security and bug fix update), and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.
https://access.redhat.com/errata/RHSA-2023:1904
Description of problem (from upstream bug): In java.base/linux/native/libnet/linux_close.c: do { if (fd1 < 0) { rv = close(fd2); } else { rv = dup2(fd1, fd2); } } while (rv == -1 && errno == EINTR); However, the close() call is not restartable. Only dup2() needs to be restarted if interrupted. Version-Release number of selected component (if applicable): All versions Additional info: - This is a request from one of our customers to backport the patch for https://bugs.openjdk.org/browse/JDK-8205959 to OpenJDK 8. This was identified as the cause of one of their applications crashing by an external vendor.