mounting cifs filesystems not possible
I discovered that this is a slow_path issue, which is tainting the kernel, I already filed this as a gentoo bug, but since this is kernel related, I think here's the right place. The server I'm trying to mount is a Windows 2003 Server, I can inquire more if neccessary.
PS: please CC me, I not on the list anymore (too much traffic, just how I was warned)
Gerhard
A message that is not in the archive (sent off-list or not yet received): [email protected]
Well this is indeed strange. First I thought the problem was solved because I had an old version of mount installed, but the successful mount with latest version of mount.cifs was only by chance. I get 1 successful mount out of ~ 30 tries (each time unloading the cifs module). And looking at the source in export.c I discovered, that reexporting cifs- mount points is still not supported :-( : This seems to be very tricky. If you need more information to track the tainting problem down, pls give me a note.
PS: please CC [email protected], I'm not on the list
Quoted 26 lines · Andrew Morton
Quoted 7 lines · On Fri, 17 Jul 2009 10:22:42 +0200 (CEST) "Gerhard Hintermayer"…
I discovered that this is a
slow_pathissue, which is tainting the kernel, I already filed this as a gentoo bug, but since this is kernel related, I think here's the right place. The server I'm trying to mount is a Windows 2003 Server, I can inquire more if neccessary.
PS: please CC me, I not on the list anymore (too much traffic, just how I was warned)please do cc the CIFS maintainer and list on future issues:
COMMON INTERNET FILE SYSTEM (CIFS)
P: Steve French M: [email protected] L: [email protected] L: [email protected] W: http://linux-cifs.samba.org/ T: git git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6.git S: Supported F: Documentation/filesystems/cifs.txt F: fs/cifs/Thanks.
Quoted 8 lines · Gerhard Hintermayer
Well this is indeed strange. First I thought the problem was solved because I had an old version of mount installed, but the successful mount with latest version of mount.cifs was only by chance. I get 1 successful mount out of ~ 30 tries (each time unloading the cifs module). And looking at the source in
export.cI discovered, that reexporting cifs- mount points is still not supported :-( : This seems to be very tricky. If you need more information to track the tainting problem down, pls give me a note.
What is kernel version?
What is symptom?
Have you opened a bug against the project bugzilla (bugzilla.samba.org)?
The dmesg (Linux message log) will show certain serious errors, but enabling additional cifs debugging flags can increase the debug output. Setting /proc/fs/cifs/DebugData to 1 will cause additional cifs informational messages to be logged, and setting
it to 3 (“echo 3 > /proc/fs/cifs/cifsFYI”) will also log return codes from most cifs entry points to dmesg.
One of the most powerful tools for analyzing networking or network filesystem problems is “Wireshark."
http://wiki.samba.org/index.php/Capture_Packets shows instructions how to do wireshark captures.
Signature · 3 lines
Thanks, Steve
See my original post to LKML at http://lkml.org/lkml/2009/7/15/217.
You can see kernel version, and dmesg output (including cifsFYI set to
7) there.
I did file a bug report to gentoo bug list and lkml. Shall I also do so for bugzilla.samba.org ? I think it's more kernel related.
The DebugData of the share I'm trying to mount (fortunaltely I got a successful mount after several tries):
brklev2b ~ # cat /proc/fs/cifs/DebugData
Display Internal CIFS Data Structures for Debugging
---------------------------------------------------
CIFS Version 1.57
Active VFS Requests: 0
Servers:
1) Name: 10.5.12.105 Domain: EUROPE Uses: 1 OS: Windows Server 2003 R2 3790 Service Pack 1
NOS: Windows Server 2003 R2 5.2 Capability: 0x1f3fd
SMB session status: 1 TCP status: 1
Local Users To Server: 1 SecMode: 0x3 Req On Wire: 0 Shares:
1) \\BRKVN05\AtlasPM Mounts: 1 Type: NTFS DevInfo: 0x20
Attributes: 0x700ff
PathComponentMax: 255 Status: 0x1 type: DISK DISCONNECTED
MIDs:
Gerhard
Quoted 32 lines · Steve French
Quoted 8 lines · Gerhard Hintermayer
Well this is indeed strange. First I thought the problem was solved because I had an old version of mount installed, but the successful mount with latest version of mount.cifs was only by chance. I get 1 successful mount out of ~ 30 tries (each time unloading the cifs module). And looking at the source in
export.cI discovered, that reexporting cifs- mount points is still not supported :-( : This seems to be very tricky. If you need more information to track the tainting problem down, pls give me a note.What is kernel version?
What is symptom?Have you opened a bug against the project bugzilla (bugzilla.samba.org)?
The dmesg (Linux message log) will show certain serious errors, but enabling additional cifs debugging flags can increase the debug output. Setting
/proc/fs/cifs/DebugDatato 1 will cause additional cifs informational messages to be logged, and setting
it to 3 (“echo 3 > /proc/fs/cifs/cifsFYI”) will also log return codes from most cifs entry points to dmesg.
One of the most powerful tools for analyzing networking or network filesystem problems is “Wireshark."
http://wiki.samba.org/index.php/Capture_Packets shows instructions how to do wireshark captures.--
Thanks,Steve
Quoted 9 lines · Gerhard Hintermayer
See my original post to LKML at http://lkml.org/lkml/2009/7/15/217.
You can see kernel version, and dmesg output (including cifsFYI set to
7) there.
I did file a bug report to gentoo bug list and lkml. Shall I also do so for bugzilla.samba.org ? I think it's more kernel related.The DebugData of the share I'm trying to mount (fortunaltely I got a
mount -t cifs -o user=aprol,password=*****,ip=10.5.12.105 //BRKVN05/AtlasPM
/mnt/AtlasPM
The mount syntax you listed in your post is a little strange with the mount options preceding rather than following the UNC name and device, ie rather than the usual:
mount.cifs {//unc-name} {mount-point} [-o options]
It looks like the first 3/4 of mount succeeded (setting up a session, and connecting to the share) to the point of doing the lookup of the root directory of the share (which I don't see in your trace). Typically there have been the following reasons for that to fail:
1) the user has no permission on the root of the share
or
2) the root of the share is a dfs referral to another server (this should work now, but was probably not supported a few kernel releases ago)
or
3) the server reports that the root directory is not a directory (ie is a file for example)
It may be that an earlier mount failed perhaps oopsing in the vfs, and that could have left an existing cifs session and tree connection which made it seem like a later mount worked (since it found an existing session) but that seems unlikely.
It would be helpful to see if the client did a network request (in this case SMB transact2 Query Path Info) on the root directory of the share, so if you have a wireshark trace that would be useful.
Signature · 3 lines
Thanks, Steve
Sorry, wireshark not installed.
The strange thing is, that sometimes mounting succeeds. So it should'nt be an issue with the root share (unless the server doesn't always answer in the same way)
Here's a dmesg output with a successful mount. (had to try for several times, unloading/loading of module cifs is not necessary) The only diffs which might be the cause are different flags (64 vs 0 when unable to mount) in line 1 and a different Xid in line 2.
fs/cifs/cifsfs.c: Devname: //BRKVN05/AtlasPM flags: 64 fs/cifs/connect.c: CIFS VFS: in cifs_mount as Xid: 97 with uid: 0 fs/cifs/connect.c: Username: aprol
fs/cifs/connect.c: UNC: \\BRKVN05\AtlasPM ip: 10.5.12.105 fs/cifs/connect.c: Socket created
fs/cifs/connect.c: sndbuf 16384 rcvbuf 87380 rcvtimeo 0x2bc fs/cifs/connect.c: Existing smb sess not found fs/cifs/cifssmb.c: secFlags 0x7
fs/cifs/transport.c: For smb_command 114
fs/cifs/transport.c: Sending smb: total_len 51 fs/cifs/connect.c: Demultiplex PID: 31969
fs/cifs/connect.c: rfc1002 length 0x6f
fs/cifs/cifssmb.c: Dialect: 0
fs/cifs/cifssmb.c: negprot rc 0
fs/cifs/connect.c: Security Mode: 0x3 Capabilities: 0x1f3fd TimeAdjust: -7200 fs/cifs/sess.c: sess setup type 2
fs/cifs/transport.c: For smb_command 115
fs/cifs/transport.c: Sending smb: total_len 242 fs/cifs/connect.c: rfc1002 length 0xc7
fs/cifs/misc.c: Null buffer passed to cifs_small_buf_release fs/cifs/sess.c: ssetup rc from sendrecv2 is 0
fs/cifs/sess.c: UID = 20482
fs/cifs/sess.c: bleft 153
fs/cifs/sess.c: words left: 0
fs/cifs/sess.c: ssetup freeing small buf f031ae40 fs/cifs/connect.c: CIFS Session Established successfully fs/cifs/connect.c: file mode: 0x5f7 dir mode: 0x1ff fs/cifs/transport.c: For smb_command 117
fs/cifs/transport.c: Sending smb: total_len 90 fs/cifs/connect.c: rfc1002 length 0x42
fs/cifs/connect.c: disk share connection
fs/cifs/connect.c: Tcon flags: 0x1
fs/cifs/connect.c: CIFS Tcon rc = 0
fs/cifs/cifssmb.c: In QFSDeviceInfo
fs/cifs/transport.c: For smb_command 50
fs/cifs/transport.c: Sending smb: total_len 72 fs/cifs/connect.c: rfc1002 length 0x44
fs/cifs/cifssmb.c: In QFSAttributeInfo
fs/cifs/transport.c: For smb_command 50
fs/cifs/transport.c: Sending smb: total_len 72 fs/cifs/connect.c: rfc1002 length 0x50
fs/cifs/connect.c: CIFS VFS: leaving cifs_mount (xid = 97) rc = 0 fs/cifs/inode.c: CIFS VFS: in cifs_iget as Xid: 98 with uid: 0 fs/cifs/inode.c: Getting info on
fs/cifs/transport.c: For smb_command 50
fs/cifs/transport.c: Sending smb: total_len 78 fs/cifs/connect.c: rfc1002 length 0xb8
fs/cifs/inode.c: Old time 0
fs/cifs/inode.c: New time 67264957
Gerhard
Quoted 43 lines · Steve French
Quoted 9 lines · Gerhard Hintermayer
See my original post to LKML at http://lkml.org/lkml/2009/7/15/217.
You can see kernel version, and dmesg output (including cifsFYI set to
7) there.
I did file a bug report to gentoo bug list and lkml. Shall I also do so for bugzilla.samba.org ? I think it's more kernel related.The DebugData of the share I'm trying to mount (fortunaltely I got a
mount -t cifs -o user=aprol,password=*****,ip=10.5.12.105 //BRKVN05/AtlasPM
/mnt/AtlasPMThe mount syntax you listed in your post is a little strange with the mount options preceding rather than following the UNC name and device, ie rather than the usual:
mount.cifs {//unc-name} {mount-point} [-o options]It looks like the first 3/4 of mount succeeded (setting up a session, and connecting to the share) to the point of doing the lookup of the root directory of the share (which I don't see in your trace). Typically there have been the following reasons for that to fail:
1) the user has no permission on the root of the share
or
2) the root of the share is a dfs referral to another server (this should work now, but was probably not supported a few kernel releases ago)
or
3) the server reports that the root directory is not a directory (ie is a file for example)It may be that an earlier mount failed perhaps oopsing in the vfs, and that could have left an existing cifs session and tree connection which made it seem like a later mount worked (since it found an existing session) but that seems unlikely.
It would be helpful to see if the client did a network request (in this case SMB transact2 Query Path Info) on the root directory of the share, so if you have a wireshark trace that would be useful. --
Thanks,Steve
Quoted 4 lines · Gerhard Hintermayer
Sorry, wireshark not installed.
The strange thing is, that sometimes mounting succeeds. So it should'nt be an issue with the root share (unless the server doesn't always answer in the same way)
I am looking for the equivalent lines in dmesg in the failing trace to those that
follow (in your success trace) the line:
fs/cifs/inode.c: CIFS VFS: in cifs_iget
If we do the QFSAttributeInfo (which is optional), the next (and final) place to fail
is getting information on the root inode
Signature · 3 lines
Thanks, Steve