Hi all! Got it on current master (several mkdir/rmdir reproduces it) against Windows 7. [ 3474.795977] WARNING: at /home/apw/COD/linux/fs/inode.c:290 drop_nlink+0x3b/0x40() [ 3474.795980] Hardware name: MS-7309 [ 3474.795982] Modules linked in: cifs(O) des_generic md4 nls_utf8 binfmt_misc snd_hda_codec_realtek snd_hda_intel snd_hda_codec ppdev snd_hwdep snd_pcm nouveau ttm drm_kms_helper drm lp mxm_wmi video snd_seq_midi snd_rawmidi snd_seq_midi_event snd_seq snd_timer snd_seq_device snd i2c_nforce2 k8temp parport_pc soundcore parport wmi mac_hid snd_page_alloc serio_raw floppy skge pata_amd forcedeth sata_nv [last unloaded: cifs] [ 3474.796056] Pid: 8508, comm: rmdir Tainted: G W O 3.3.0-030300rc2-generic-pae #201201311735 [ 3474.796060] Call Trace: [ 3474.796069] [<c1042792>] warn_slowpath_common+0x72/0xa0 [ 3474.796076] [<c115f63b>] ? drop_nlink+0x3b/0x40 [ 3474.796082] [<c115f63b>] ? drop_nlink+0x3b/0x40 [ 3474.796089] [<c10427e2>] warn_slowpath_null+0x22/0x30 [ 3474.796095] [<c115f63b>] drop_nlink+0x3b/0x40 [ 3474.796108] [<f874bbc7>] cifs_rmdir+0x107/0x1e0 [cifs] [ 3474.796116] [<c1153f80>] vfs_rmdir+0xa0/0xf0 [ 3474.796122] [<c115756f>] do_rmdir+0xef/0x100 [ 3474.796129] [<c1147012>] ? filp_close+0x52/0x80 [ 3474.796135] [<c11575d5>] sys_rmdir+0x15/0x20 [ 3474.796142] [<c15a6c5f>] sysenter_do_call+0x12/0x28 [ 3474.796147] ---[ end trace b449c0b09a984281 ]--- I also, saw something like this for inc_nlink in mkdir call (when we try to inc zero nlink value). The problem is that POSIX filesystems increment nlink value for directory on every mkdir call - Samba with posix extensions enabled does the same thing and return the right value on query info. But as for Windows semantic - Links for directory is ordinary link - the value stay the same and doesn't depend on the number of child directoris. I think we need to do two thing to avoid problems like that: 1) do inc/drop nlink value in posix case only. 2) add if (nlink > 2) check before drop_nlink to work around possible broken servers (we can raise cERROR in zero case). -- Best regards, Pavel Shilovsky. -- To unsubscribe from this list: send the line "unsubscribe linux-cifs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
