Slight revision to this patch...use min_t() instead of conditional
assignment. Also, remove the FIXME comment and replace it with the
explanation that Steve gave earlier.

After receiving a packet, we currently check the header. If it's no
good, then we toss it out and continue the loop, leaving the caller
waiting on that response.

In cases where the packet has length inconsistencies, but the MID is
valid, this leads to unneeded delays. That's especially problematic now
that the client waits indefinitely for responses.

Instead, don't immediately discard the packet if checkSMB fails. Try to
find a matching mid_q_entry, mark it as having a malformed response and
issue the callback.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

In order to determine whether an SMBEcho request can be sent
we need to know that the socket is established (server tcpStatus == CifsGood)
AND that an SMB NegotiateProtocol has been sent (server maxBuf != 0).
Without the second check we can send an Echo request during reconnection
before the server can accept it.

CC: JG <jg@cms.ac>
Reviewed-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

Teach cifs about network namespaces, so mounting uses adresses/routing
visible from the container rather than from init context.

A container is a chroot on steroids that changes more than just the root
filesystem the new processes see.  One thing containers can isolate is
"network namespaces", meaning each container can have its own set of
ethernet interfaces, each with its own own IP address and routing to the
outside world.  And if you open a socket in _userspace_ from processes
within such a container, this works fine.

But sockets opened from within the kernel still use a single global
networking context in a lot of places, meaning the new socket's address
and routing are correct for PID 1 on the host, but are _not_ what
userspace processes in the container get to use.

So when you mount a network filesystem from within in a container, the
mount code in the CIFS driver uses the host's networking context and not
the container's networking context, so it gets the wrong address, uses
the wrong routing, and may even try to go out an interface that the
container can't even access...  Bad stuff.

This patch copies the mount process's network context into the CIFS
structure that stores the rest of the server information for that mount
point, and changes the socket open code to use the saved network context
instead of the global network context.  I.E. "when you attempt to use
these addresses, do so relative to THIS set of network interfaces and
routing rules, not the old global context from back before we supported
containers".

The big long HOWTO sets up a test environment on the assumption you've
never used ocntainers before.  It basically says:

1) configure and build a new kernel that has container support
2) build a new root filesystem that includes the userspace container
control package (LXC)
3) package/run them under KVM (so you don't have to mess up your host
system in order to play with containers).
4) set up some containers under the KVM system
5) set up contradictory routing in the KVM system and the container so
that the host and the container see different things for the same address
6) try to mount a CIFS share from both contexts so you can both force it
to work and force it to fail.

For a long drawn out test reproduction sequence, see:

  http://landley.livejournal.com/47024.html
  http://landley.livejournal.com/47205.html
  http://landley.livejournal.com/47476.htmlSigned-off-by: NRob Landley <rlandley@parallels.com>
Reviewed-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

...and remove length qualifiers from bools.

Before:

	/* size: 1176, cachelines: 19, members: 13 */
	/* sum members: 1165, holes: 2, sum holes: 11 */
	/* bit holes: 1, sum bit holes: 4 bits */
	/* last cacheline: 24 bytes */

After:

	/* size: 1168, cachelines: 19, members: 13 */
	/* last cacheline: 16 bytes */

...savings of 8 bytes per inode.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Reviewed-by: NPavel Shilovsky <piastryyy@gmail.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

Remove fields that are completely unused, and rearrange struct
according to recommendations by "pahole".

Before:

	/* size: 1112, cachelines: 18, members: 49 */
	/* sum members: 1086, holes: 8, sum holes: 26 */
	/* bit holes: 1, sum bit holes: 7 bits */
	/* last cacheline: 24 bytes */

After:

	/* size: 1072, cachelines: 17, members: 42 */
	/* sum members: 1065, holes: 3, sum holes: 7 */
	/* last cacheline: 48 bytes */

...savings of 40 bytes per struct on x86_64. 21 bytes by field removal,
and 19 by reorganizing to eliminate holes.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

Since we don't time out individual requests anymore, remove the code
that we used to use for setting timeouts on different requests.
Reviewed-by: NPavel Shilovsky <piastryyy@gmail.com>
Reviewed-by: NSuresh Jayaraman <sjayaraman@suse.de>
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

If the server isn't responding to echoes, we don't want to leave tasks
hung waiting for it to reply. At that point, we'll want to reconnect
so that soft mounts can return an error to userspace quickly.

If the client hasn't received a reply after a specified number of echo
intervals, assume that the transport is down and attempt to reconnect
the socket.

The number of echo_intervals to wait before attempting to reconnect is
tunable via a module parameter. Setting it to 0, means that the client
will never attempt to reconnect. The default is 5.
Signed-off-by: NJeff Layton <jlayton@redhat.com>

Reviewed-by: NSuresh Jayaraman <sjayaraman@suse.de>
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

In order to incorporate async requests, we need to allow for a more
general way to do things on receive, rather than just waking up a
process.

Turn the task pointer in the mid_q_entry into a callback function and a
generic data pointer. When a response comes in, or the socket is
reconnected, cifsd can call the callback function in order to wake up
the process.

The default is to just wake up the current process which should mean no
change in behavior for existing code.

Also, clean up the locking in cifs_reconnect. There doesn't seem to be
any need to hold both the srv_mutex and GlobalMid_Lock when walking the
list of mids.
Reviewed-by: NSuresh Jayaraman <sjayaraman@suse.de>
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

Reduce false inode collisions by using the CreationTime like an
i_generation field. This way, even if the server ends up reusing
a uniqueid after a delete/create cycle, we can avoid matching
the inode incorrectly.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

The server->sequence_number is already protected by the srv_mutex. The
GlobalMid_lock is unneeded here.
Reviewed-by: NShirish Pargaonkar <shirishpargaonkar@gmail.com>
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

Make connect logic more ip-protocol independent and move RFC1001 stuff into
a separate function. Also replace union addr in TCP_Server_Info structure
with sockaddr_storage.
Signed-off-by: NPavel Shilovsky <piastryyy@gmail.com>
Reviewed-and-Tested-by: NJeff Layton <jlayton@samba.org>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

...this string is zeroed out and nothing ever changes it.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

Currently, the attribute cache timeout for CIFS is hardcoded to 1 second. This
means that the client might have to issue a QPATHINFO/QFILEINFO call every 1
second to verify if something has changes, which seems too expensive. On the
other hand, if the timeout is hardcoded to a higher value, workloads that
expect strict cache coherency might see unexpected results.

Making attribute cache timeout as a tunable will allow us to make a tradeoff
between performance and cache metadata correctness depending on the
application/workload needs.

Add 'actimeo' tunable that can be used to tune the attribute cache timeout.
The default timeout is set to 1 second. Also, display actimeo option value in
/proc/mounts.

It appears to me that 'actimeo' and the proposed (but not yet merged)
'strictcache' option cannot coexist, so care must be taken that we reset the
other option if one of them is set.

Changes since last post:
   - fix option parsing and handle possible values correcly
Reviewed-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NSuresh Jayaraman <sjayaraman@suse.de>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

Radix trees are ideal when you want to track a bunch of pointers and
can't embed a tracking structure within the target of those pointers.
The tradeoff is an increase in memory, particularly if the tree is
sparse.

In CIFS, we use the tlink_tree to track tcon_link structs. A tcon_link
can never be in more than one tlink_tree, so there's no impediment to
using a rb_tree here instead of a radix tree.

Convert the new multiuser mount code to use a rb_tree instead. This
should reduce the memory required to manage the tlink_tree.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

Removed following fields from smb session structure
 cryptkey, ntlmv2_hash, tilen, tiblob
and ntlmssp_auth structure is allocated dynamically only if the auth mech
in NTLMSSP.

response field within a session_key structure is used to initially store the
target info (either plucked from type 2 challenge packet in case of NTLMSSP
or fabricated in case of NTLMv2 without extended security) and then to store
Message Authentication Key (mak) (session key + client response).

Server challenge or cryptkey needed during a NTLMSSP authentication
is now part of ntlmssp_auth structure which gets allocated and freed
once authenticaiton process is done.
Signed-off-by: NShirish Pargaonkar <shirishpargaonkar@gmail.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

Need to have cryptkey or server challenge in smb connection
(struct TCP_Server_Info) for ntlm and ntlmv2 auth types for which
cryptkey (Encryption Key) is supplied just once in Negotiate Protocol
response during an smb connection setup for all the smb sessions over
that smb connection.

For ntlmssp, cryptkey or server challenge is provided for every
smb session in type 2 packet of ntlmssp negotiation, the cryptkey
provided during Negotiation Protocol response before smb connection
does not count.

Rename cryptKey to cryptkey and related changes.
Signed-off-by: NShirish Pargaonkar <shirishpargaonkar@gmail.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

Minor cleanup - Fix spelling mistake, make meaningful (goto) label

In function setup_ntlmv2_rsp(), do not return 0 and leak memory,
let the tiblob get freed.

For function find_domain_name(), pass already available nls table pointer
instead of loading and unloading the table again in this function.

For ntlmv2, the case sensitive password length is the length of the
response, so subtract session key length (16 bytes) from the .len.
Signed-off-by: NShirish Pargaonkar <shirishpargaonkar@gmail.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

Mark dependency on crypto modules in Kconfig.

Defining per structures sdesc and cifs_secmech which are used to store
crypto hash functions and contexts.  They are stored per smb connection
and used for all auth mechs to genereate hash values and signatures.

Allocate crypto hashing functions, security descriptiors, and respective
contexts when a smb/tcp connection is established.
Release them when a tcp/smb connection is taken down.

md5 and hmac-md5 are two crypto hashing functions that are used
throught the life of an smb/tcp connection by various functions that
calcualte signagure and ntlmv2 hash, HMAC etc.

structure ntlmssp_auth is defined as per smb connection.

ntlmssp_auth holds ciphertext which is genereated by rc4/arc4 encryption of
secondary key, a nonce using ntlmv2 session key and sent in the session key
field of the type 3 message sent by the client during ntlmssp
negotiation/exchange

A key is exchanged with the server if client indicates so in flags in
type 1 messsage and server agrees in flag in type 2 message of ntlmssp
negotiation.  If both client and agree, a key sent by client in
type 3 message of ntlmssp negotiation in the session key field.
The key is a ciphertext generated off of secondary key, a nonce, using
ntlmv2 hash via rc4/arc4.

Signing works for ntlmssp in this patch. The sequence number within
the server structure needs to be zero until session is established
i.e. till type 3 packet of ntlmssp exchange of a to be very first
smb session on that smb connection is sent.
Acked-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NShirish Pargaonkar <shirishpargaonkar@gmail.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

Start calculating auth response within a session.  Move/Add pertinet
data structures like session key, server challenge and ntlmv2_hash in
a session structure.  We should do the calculations within a session
before copying session key and response over to server data
structures because a session setup can fail.

Only after a very first smb session succeeds, it copy/make its
session key, session key of smb connection.  This key stays with
the smb connection throughout its life.
sequence_number within server is set to 0x2.

The authentication Message Authentication Key (mak) which consists
of session key followed by client response within structure session_key
is now dynamic.  Every authentication type allocates the key + response
sized memory within its session structure and later either assigns or
frees it once the client response is sent and if session's session key
becomes connetion's session key.

ntlm/ntlmi authentication functions are rearranged.  A function
named setup_ntlm_resp(), similar to setup_ntlmv2_resp(), replaces
function cifs_calculate_session_key().

size of CIFS_SESS_KEY_SIZE is changed to 16, to reflect the byte size
of the key it holds.
Reviewed-by: NJeff Layton <jlayton@samba.org>
Signed-off-by: NShirish Pargaonkar <shirishpargaonkar@gmail.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

GlobalSMBSesLock is now cifs_file_list_lock. Update comments to reflect this.
Signed-off-by: NSuresh Jayaraman <sjayaraman@suse.de>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

write_behind_rc is redundant and just adds complexity to the code. What
we really want to do instead is to use mapping_set_error to reset the
flags on the mapping when we find a writeback error and can't report it
to userspace yet.

For cifs_flush and cifs_fsync, we shouldn't reset the flags since errors
returned there do get reported to userspace.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Reviewed-by: NSuresh Jayaraman <sjayaraman@suse.de>
Reviewed-by: NDave Kleikamp <shaggy@linux.vnet.ibm.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

Signed-off-by: NSteve French <sfrench@us.ibm.com>

The count for cifsFileInfo is currently an atomic, but that just adds
complexity for little value. We generally need to hold cifs_file_list_lock
to traverse the lists anyway so we might as well make this counter
non-atomic and simply use the cifs_file_list_lock to protect it.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Reviewed-by: NSuresh Jayaraman <sjayaraman@suse.de>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

Now that it's feasible for a cifsFileInfo to outlive the filp under
which it was created, move the close processing into cifsFileInfo_put.

This means that the last user of the filehandle always does the actual
on the wire close call. This also allows us to get rid of the closePend
flag from cifsFileInfo. If we have an active reference to the file
then it's never going to have a close pending.

cifs_close is converted to simply put the filehandle.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Reviewed-by: NSuresh Jayaraman <sjayaraman@suse.de>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

cifs_tcp_ses_lock is a rwlock with protects the cifs_tcp_ses_list,
server->smb_ses_list and the ses->tcon_list. It also protects a few
ref counters in server, ses and tcon. In most cases the critical section
doesn't seem to be large, in a few cases where it is slightly large, there
seem to be really no benefit from concurrent access. I briefly considered RCU
mechanism but it appears to me that there is no real need.

Replace it with a spinlock and get rid of the last rwlock in the cifs code.
Signed-off-by: NSuresh Jayaraman <sjayaraman@suse.de>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

...and make it non-inlined in preparation for the move of most of
cifs_close to it.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Reviewed-by: NSuresh Jayaraman <sjayaraman@suse.de>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

Convert this lock to a regular spinlock

A rwlock_t offers little value here. It's more expensive than a regular
spinlock unless you have a fairly large section of code that runs under
the read lock and can benefit from the concurrency.

Additionally, we need to ensure that the refcounting for files isn't
racy and to do that we need to lock areas that can increment it for
write. That means that the areas that can actually use a read_lock are
very few and relatively infrequently used.

While we're at it, change the name to something easier to type, and fix
a bug in find_writable_file. cifsFileInfo_put can sleep and shouldn't be
called while holding the lock.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Reviewed-by: NSuresh Jayaraman <sjayaraman@suse.de>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

All the remaining users of cifsFileInfo->pfile just use it to get
at the f_flags/f_mode. Now that we store that separately in the
cifsFileInfo, there's no need to consult the pfile at all from
a cifsFileInfo pointer.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Reviewed-by: NSuresh Jayaraman <sjayaraman@suse.de>
Acked-by: NDave Kleikamp <shaggy@linux.vnet.ibm.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

Add a f_flags field that holds the f_flags field from the filp. We'll
need this info in case the filp ever goes away before the cifsFileInfo
does. Have cifs_reopen_file use that value instead of filp->f_flags
too and have it take a cifsFileInfo arg instead of a filp.

While we're at it, get rid of some bogus cargo-cult NULL pointer
checks in that function and reduce the level of indentation.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Reviewed-by: NSuresh Jayaraman <sjayaraman@suse.de>
Acked-by: NDave Kleikamp <shaggy@linux.vnet.ibm.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

Start calculation auth response within a session.  Move/Add pertinet
data structures like session key, server challenge and ntlmv2_hash in
a session structure.  We should do the calculations within a session
before copying session key and response over to server data
structures because a session setup can fail.

Only after a very first smb session succeeds, it copies/makes its
session key, session key of smb connection.  This key stays with
the smb connection throughout its life.
Signed-off-by: NShirish Pargaonkar <shirishpargaonkar@gmail.com>
Reviewed-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

Filesystems aren't really supposed to do anything with a vfsmount. It's
considered a layering violation since vfsmounts are entirely managed at
the VFS layer.

CIFS currently keeps an active reference to a vfsmount in order to
prevent the superblock vanishing before an oplock break has completed.
What we really want to do instead is to keep sb->s_active high until the
oplock break has completed. This patch borrows the scheme that NFS uses
for handling sillyrenames.

An atomic_t is added to the cifs_sb_info. When it transitions from 0 to
1, an extra reference to the superblock is taken (by bumping the
s_active value). When it transitions from 1 to 0, that reference is
dropped and a the superblock teardown may proceed if there are no more
references to it.

Also, the vfsmount pointer is removed from cifsFileInfo and from
cifs_new_fileinfo, and some bogus forward declarations are removed from
cifsfs.h.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Reviewed-by: NSuresh Jayaraman <sjayaraman@suse.de>
Acked-by: NDave Kleikamp <shaggy@linux.vnet.ibm.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

cifsFileInfo is a bit problematic. It contains a reference back to the
struct file itself. This makes it difficult for a cifsFileInfo to exist
without a corresponding struct file.

It would be better instead of the cifsFileInfo just held info pertaining
to the open file on the server instead without any back refrences to the
struct file. This would allow it to exist after the filp to which it was
originally attached was closed.

Much of the use of the file pointer in this struct is to get at the
dentry.  Begin divorcing the cifsFileInfo from the struct file by
keeping a reference to the dentry. Since the dentry will have a
reference to the inode, we can eliminate the "pInode" field too and
convert the igrab/iput to dget/dput.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Reviewed-by: NSuresh Jayaraman <sjayaraman@suse.de>
Acked-by: NDave Kleikamp <shaggy@linux.vnet.ibm.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

Signed-off-by: NSteve French <sfrench@us.ibm.com>

This patch is rather large, but it's a bit difficult to do piecemeal...

For non-multiuser mounts, everything will basically work as it does
today. A call to cifs_sb_tlink will return the "master" tcon link.

Turn the tcon pointer in the cifs_sb into a radix tree that uses the
fsuid of the process as a key. The value is a new "tcon_link" struct
that contains info about a tcon that's under construction.

When a new process needs a tcon, it'll call cifs_sb_tcon. That will
then look up the tcon_link in the radix tree. If it exists and is
valid, it's returned.

If it doesn't exist, then we stuff a new tcon_link into the tree and
mark it as pending and then go and try to build the session/tcon.
If that works, the tcon pointer in the tcon_link is updated and the
pending flag is cleared.

If the construction fails, then we set the tcon pointer to an ERR_PTR
and clear the pending flag.

If the radix tree is searched and the tcon_link is marked pending
then we go to sleep and wait for the pending flag to be cleared.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

cifsFileInfo needs a pointer to a tcon, but it doesn't currently hold a
reference to it. Change it to keep a pointer to a tcon_link instead and
hold a reference to it.

That will keep the tcon from being freed until the file is closed.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

Eventually, we'll need to track the use of tcons on a per-sb basis, so that
we know when it's ok to tear them down. Begin this conversion by adding a
new "tcon_link" struct and accessors that get it. For now, the core data
structures are untouched -- cifs_sb still just points to a single tcon and
the pointers are just cast to deal with the accessor functions. A later
patch will flesh this out.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

At mount time, we'll always need to create a tcon that will serve as a
template for others that are associated with the mount. This tcon is
known as the "master" tcon.

In some cases, we'll need to use that tcon regardless of who's accessing
the mount. Add an accessor function for the master tcon and go ahead and
switch the appropriate places to use it.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

When we convert cifs to do multiple sessions per mount, we'll need more
than one tcon per superblock. At that point "cifs_sb->tcon" will make
no sense. Add a new accessor function that gets a tcon given a cifs_sb.
For now, it just returns cifs_sb->tcon. Later it'll do more.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>