+++ This bug was initially created as a clone of Bug #1223889 +++ Description of problem: No issue using nfs only with fuse. For 32bits applications on 64bits systems, ino32 is mandadory but the readdirp optimisation still return 64 bits inodes when getdent syscall is triggerd. Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: 1. Have a 32 bits application listing directories on a 64 bits system 2. Run it 3. Cry Actual results: open("/usr/global/lib/", O_RDONLY|O_NONBLOCK|O_LARGEFILE|O_DIRECTORY|O_CLOEXEC) = 5 getdents(5, 0x9b27308, 32768) = -1 EOVERFLOW (Value too large for defined data type) Expected results: open("/usr/global/lib/", O_RDONLY|O_NONBLOCK|O_LARGEFILE|O_DIRECTORY|O_CLOEXEC) = 5 getdents(5, /* 1 entries */, 32768) = 88 Additional info: use use-readdirp=off when mounting the volume as a workaround but impact performaces. --- Additional comment from Niels de Vos on 2015-05-20 22:48:16 CEST --- Compile tested only, needs real functional test and maybe a test-case too. This bug needs to get cloned for the mainline version and the patch needs review+merging in the master branch before it can get included in 3.5. --- Additional comment from Cyril Peponnet on 2015-05-21 01:22:27 CEST --- I gave a try with RPMs generated through http://koji.fedoraproject.org/koji/taskinfo?taskID=9810286 Seems to work fine with enable-ino32 to mount options and use 32bit binary to call getdents syscall. Thanks ! For the record to test it you can compile: /* * * List directories using getdents() because ls, find and Python libraries * * use readdir() which is slower (but uses getdents() underneath. * * * * Compile with * * ]$ gcc getdents.c -o getdents * */ #define _GNU_SOURCE #include <dirent.h> /* Defines DT_* constants */ #include <fcntl.h> #include <stdio.h> #include <unistd.h> #include <stdlib.h> #include <sys/stat.h> #include <sys/syscall.h> #define handle_error(msg) \ do { perror(msg); exit(EXIT_FAILURE); } while (0) struct linux_dirent { long d_ino; off_t d_off; unsigned short d_reclen; char d_name[]; }; #define BUF_SIZE 1024*1024*5 int main(int argc, char *argv[]) { int fd, nread; char buf[BUF_SIZE]; struct with gcc -m32 getdents getdents.c And run it like ./getdents /path/ Before the patch: strace -e getdents ./getdents /mnt/test/ [ Process PID=15906 runs in 32 bit mode. ] getdents(3, /* 1 entries */, 5242880) = 16 getdents(3, 0xff8f65bf, 5242880) = -1 EOVERFLOW (Value too large for defined data type) getdents: Value too large for defined data type +++ exited with 1 +++
REVIEW: http://review.gluster.org/10883 (fuse: squash 64-bit inodes in readdirp when enable-ino32 is set) posted (#1) for review on release-3.6 by Niels de Vos (ndevos)
REVIEW: http://review.gluster.org/10883 (fuse: squash 64-bit inodes in readdirp when enable-ino32 is set) posted (#2) for review on release-3.6 by Niels de Vos (ndevos)
This bug is getting closed because a release has been made available that should address the reported issue. In case the problem is still not fixed with glusterfs-v3.6.4, please open a new bug report. glusterfs-v3.6.4 has been announced on the Gluster mailinglists [1], packages for several distributions should become available in the near future. Keep an eye on the Gluster Users mailinglist [2] and the update infrastructure for your distribution. [1] https://www.gluster.org/pipermail/gluster-users/2015-July/022826.html [2] http://thread.gmane.org/gmane.comp.file-systems.gluster.user