Description of problem: There must be an unexpected limit in the cifs kernel module which breaks listening of directories containing a lot of long file names. Version-Release number of selected component (if applicable): Client: kernel-2.6.18-1.2200.fc5 Server: samba-3.0.23c-1.fc5 How reproducible: Always, if directory is filled up Steps to Reproduce: 1. create a directory on a samba server 2. create file names like e.g.: $ rm -rf *.txt; filename="`perl -e "print 't' x 80";`"; for i in `seq 1 37`; do touch $filename.$i.txt; done (36 will work) or $ rm -rf *.txt; filename="`perl -e "print 't' x 70";`"; for i in `seq 1 40`; do touch $filename.$i.txt; done (39 will work) 3. mount this directory via cifs on a client 4. try to list directory: $ ls /mnt/autofs/store/test/ Actual results: ls: reading directory /mnt/autofs/store/test/ Invalid argument Expected results: Listing of directory works Additional info: must be a client side (cifs module problem), smbclient works well.
Same happen on another network, here client is also FC6, but server is RHEL4 (samba-3.0.10-1.4E.9)
Same here. The server is a Windows (2003?) box. RHEL4.4's samba can mount the folder in question, but FC6's samba-common-3.0.23c-2 complains about 'Invalid argument'. Removing some files from the folder makes the folder readable again even to the FC6 samba.
I had similar error "cifs mount can't list directory" in FC6 x86_x64 listing shares from a samba(3.0.14a-ubuntu'breezy')server even when smbclient can list it This drived me crazy.because fc6 kernel have no support for smbfs I get stuck. I can't figure how to compile smbfs module the supplied fc6 kernel. So make new big partition 8 gigas at list and download last kernel(2.6.18.2) and make gconf,make bzImage,make modules,make modules_install,etc...and make a new boot option, grub ,mkinitrd..etc (only for stupid avanced users who like freeze his computer):) and ... cool smbfs don't work :( but cifs is ok :)) no more problems so it's seem a kernel or kernel module problem to me
Could someone update the target version so that this doesn't fall off the radar? Unless there are two separate bugs, this still persists in FC6.
Changed target to FC6
Changed priority because no one of Red Hat cares about...btw. RHEL5 beta is also based on 2.6.18, perhaps this bug exists also here.
Whoa! WORKSFORME ever since the 2.6.18-1.2849.fc6 kernel update.
Can confirm #7