From bug 2723: * unfsd has some minor stuff: > ==10081== Syscall param open(mode) contains uninitialised byte(s) > ==10081== at 0x3BFB6E62D0: __open_nocancel (in /usr/lib64/libc-2.18.so) > ==10081== by 0x407A14: nfsproc3_create_3_svc (nfs.c:488) > ==10081== by 0x4038D0: nfs3_program_3 (daemon.c:566) > ==10081== by 0x3BFB7252A0: svc_getreq_common (in /usr/lib64/libc-2.18.so) > ==10081== by 0x3BFB7253ED: svc_getreq_poll (in /usr/lib64/libc-2.18.so) > ==10081== by 0x404647: main (daemon.c:828) > ==10081== > ==10081== 344 (336 direct, 8 indirect) bytes in 1 blocks are definitely lost in loss record 10 of 13 > ==10081== at 0x4A0645D: malloc (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so) > ==10081== by 0x3BFB725A8E: svctcp_create (in /usr/lib64/libc-2.18.so) > ==10081== by 0x403D2C: create_tcp_transport (daemon.c:794) > ==10081== by 0x404189: main (daemon.c:934)
Cannot reproduce on CentOS 6, neither with client 4.3.0 nor with 4.5.0rc1. Could be a valgrind/distro problem. Or, it only happens with a specific unfsd use case. Unfortunately this bug report does not mention what was done...
*** Bug 7129 has been marked as a duplicate of this bug. ***
Can reproduce on Fedora 27 with 4.9.0 client.