@@ -13,10 +13,13 @@
*
* Assume that the value LONG_MAX is an invalid pointer.
*
- * aarch64 linux versions v5.3 up to v5.6-rc1 may incorrectly report
- * EINVAL instead of ENOMEM, see:
+ * aarch64 linux versions v5.4-rc1 up to v5.6-rc3 may incorrectly report
+ * EINVAL instead of ENOMEM, the fix patch see:
* 597399d0cb91 ("arm64: tags: Preserve tags for addresses translated via TTBR1")
* d0022c0ef29b ("arm64: memory: Add missing brackets to untagged_addr() macro")
+ *
+ * this bug was introduced because of the following commit, see:
+ * 057d3389108e ("mm: untag user pointers passed to memory syscalls")
*/
#include <sys/mman.h>
@@ -13,10 +13,13 @@
*
* Assume that the value LONG_MAX is an invalid pointer.
*
- * aarch64 linux versions v5.3 up to v5.6-rc1 may incorrectly report
- * EINVAL instead of ENOMEM, see:
+ * aarch64 linux versions v5.4-rc1 up to v5.6-rc3 may incorrectly report
+ * EINVAL instead of ENOMEM, the fix patch see:
* 597399d0cb91 ("arm64: tags: Preserve tags for addresses translated via TTBR1")
* d0022c0ef29b ("arm64: memory: Add missing brackets to untagged_addr() macro")
+ *
+ * this bug was introduced because of the following commit, see:
+ * 057d3389108e ("mm: untag user pointers passed to memory syscalls")
*/
#include <sys/mman.h>
The previous patch f52839ba25 ("openposix/mlock_8-1,munlock_10-1: document known aarch64 issue") has documented the fix patch, I have verified the introdueced patch is 057d3389108ed ("mm: untag user pointers passed to memory syscalls"). So add it in comment and make testers known whether it is a regression test or a new bug when hitting the EINVAL error. Also modify the kernel version scope of the problem. Signed-off-by: Yang Xu <xuyang2018.jy@fujitsu.com> --- .../conformance/interfaces/mlock/8-1.c | 7 +++++-- .../conformance/interfaces/munlock/10-1.c | 7 +++++-- 2 files changed, 10 insertions(+), 4 deletions(-)