nfs-utils.git
8 years agomountd: Update mountd/exportfs man pages to reflect IPv6 changes
Chuck Lever [Mon, 27 Sep 2010 14:14:34 +0000 (10:14 -0400)]
mountd: Update mountd/exportfs man pages to reflect IPv6 changes

Document IPv6 support in rpc.mountd and exportfs, and clarify existing
language in the man page.

Clean up: Use bold consistently for program names, and italics
consistently for file names.  Use "rpc.mountd" consistently as the
name of the mountd daemon.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agomountd: Use MNT status values instead of NFSERR
Chuck Lever [Mon, 27 Sep 2010 14:13:39 +0000 (10:13 -0400)]
mountd: Use MNT status values instead of NFSERR

Clean up:  The MNT protocol has its own enum type defining error
status values.  While the values can be the same as the NFSERR enum
type on some systems, it's not guaranteed to be true everywhere.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agomountd: Fix up version and usage messages
Chuck Lever [Mon, 27 Sep 2010 14:11:18 +0000 (10:11 -0400)]
mountd: Fix up version and usage messages

Clean up: rpc.mountd is no longer known as kmountd.  Use the program's
basename rather than the full pathname for the usage message.  Display
a version message at start up similar to statd's.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agomountd: Unregister mountd if my_svc_run() returns
Chuck Lever [Mon, 27 Sep 2010 14:09:49 +0000 (10:09 -0400)]
mountd: Unregister mountd if my_svc_run() returns

Fix a long standing bug: when my_svc_run() returns, mountd should
unregister itself with the local rpcbind so that it can subsequently
start cleanly.

Log a more helpful error message in this case.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agomountd: Support TI-RPC mountd listener
Chuck Lever [Mon, 27 Sep 2010 14:09:10 +0000 (10:09 -0400)]
mountd: Support TI-RPC mountd listener

If TI-RPC is available, use it to create mountd's svc listener.  If
not, use the old function, rpc_init(), to create mountd's listener.

IPv6 can be supported if TI-RPC is available.  In this case,
/etc/netconfig is searched to determine which transports to advertise.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agomountd: Make NFS version checks more strict
Chuck Lever [Mon, 27 Sep 2010 14:06:35 +0000 (10:06 -0400)]
mountd: Make NFS version checks more strict

Ensure users and programmers specify NFS version numbers correctly.
This also makes the next patch more clean.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agonfs-utils: Fix source code character encoding
Chuck Lever [Mon, 27 Sep 2010 13:50:11 +0000 (09:50 -0400)]
nfs-utils: Fix source code character encoding

Minor clean up.

Most modern Linux distributions set UTF-8 locales.  Standardize the
character encoding of source files on UTF-8, to squelch vim com-
plaints.

I probably missed a few spots.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agolibnfs.a: Remove support/nfs/fstab.c
Steve Dickson [Mon, 27 Sep 2010 13:35:26 +0000 (09:35 -0400)]
libnfs.a: Remove support/nfs/fstab.c

Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agorpc.nfsd: mount up nfsdfs is it doesn't appear to be mounted yet nfs-utils-1-2-3-rc6
Jeff Layton [Thu, 16 Sep 2010 18:34:39 +0000 (14:34 -0400)]
rpc.nfsd: mount up nfsdfs is it doesn't appear to be mounted yet

There's a bit of a chicken and egg problem when nfsd is run the first
time. On Fedora/RHEL at least, /proc/fs/nfsd is mounted up whenever nfsd
is plugged in via a modprobe.conf "install" directive.

If someone runs rpc.nfsd without plugging in nfsd.ko first,
/proc/fs/nfsd won't be mounted and rpc.nfsd will end up using the legacy
nfsctl interface. After that, nfsd will be plugged in and subsequent
rpc.nfsd invocations will use that instead.

This is a problem as some nfsd command-line options are ignored when the
legacy interface is used. It'll also be a problem for people who want
IPv6 enabled servers. The upshot is that we really don't want to use the
legacy interface unless there is no other option.

To avoid this situation, have rpc.nfsd check to see if the "threads"
file is already present. If it's not, then make an attempt to mount
/proc/fs/nfsd.  This is a "best-effort" sort of thing, however so we
just ignore the return code from the mount attempt and fall back to
using nfsctl() if it fails.

Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agolibexport.a: Enable IPv6 support in hostname.c
Chuck Lever [Thu, 16 Sep 2010 18:25:52 +0000 (14:25 -0400)]
libexport.a: Enable IPv6 support in hostname.c

If --enable-ipv6 is specified when building nfs-utils, libexport's
host_foo() helpers can now return both IPv4 and IPv6 addresses.

This means IPv6 presentation addresses and IPv6 DNS resolution
results are handled properly in the mountd cache and /etc/exports,
but does not yet enable IPv6 mountd listeners.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agomountd: Ensure cache downcall can handle IPv6 addresses
Chuck Lever [Thu, 16 Sep 2010 18:21:48 +0000 (14:21 -0400)]
mountd: Ensure cache downcall can handle IPv6 addresses

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agomountd: Handle IPv6 addresses in kernel auth_unix_ip upcalls
Chuck Lever [Thu, 16 Sep 2010 18:21:08 +0000 (14:21 -0400)]
mountd: Handle IPv6 addresses in kernel auth_unix_ip upcalls

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agomountd: clean up cache API
Chuck Lever [Thu, 16 Sep 2010 18:19:19 +0000 (14:19 -0400)]
mountd: clean up cache API

Clean up: Squelch compiler warnings and document public parts of
cache API.

cache.c: At top level:
cache.c:67: warning: no previous prototype for auth_unix_ip
cache.c:123: warning: no previous prototype for auth_unix_gid
cache.c:217: warning: no previous prototype for get_uuid
cache.c:247: warning: no previous prototype for uuid_by_path
cache.c:326: warning: no previous prototype for nfsd_fh
cache.c:745: warning: no previous prototype for nfsd_export
cache.c:820: warning: no previous prototype for cache_open
cache.c:832: warning: no previous prototype for cache_set_fd
cache.c:841: warning: no previous prototype for
cache_process_req
cache.c:921: warning: no previous prototype for cache_export
cache.c:953: warning: no previous prototype for
cache_get_filehandle

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agoexportfs: Enable IPv6 support in matchhostname()
Chuck Lever [Thu, 16 Sep 2010 18:17:08 +0000 (14:17 -0400)]
exportfs: Enable IPv6 support in matchhostname()

To gain IPv6 support in matchhostname(), simply replace the socket
address comparison helpers with the generic versions that can handle
IPv4 and IPv6.

host_addrinfo() (called by matchhostname()) returns IPv6 addresses
only if IPv6 support is enabled.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agomountd: Support IPv6 in mountlist_list()
Chuck Lever [Thu, 16 Sep 2010 17:54:21 +0000 (13:54 -0400)]
mountd: Support IPv6 in mountlist_list()

Replace inet_aton(3) and gethostbyaddr(3) calls in mountlist_list()
with calls to the new host_foo() DNS helpers.

The new functions will support IPv6 without additional changes, once
IPv6 is enabled in the generic hostname helpers.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agomountd: Handle memory exhaustion in mountlist_list()
Chuck Lever [Thu, 16 Sep 2010 17:51:51 +0000 (13:51 -0400)]
mountd: Handle memory exhaustion in mountlist_list()

I'm about to replace inet_aton(3)/gethostbyaddr(3) with
host_pton()/host_canonname() in mountlist_list().

Since host_canonname() returns a string allocated with strdup(3)
instead of xstrdup(), mountlist_list() must now deal with memory
exhaustion properly.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agomountd: Add mountlist_freeall()
Chuck Lever [Thu, 16 Sep 2010 17:48:38 +0000 (13:48 -0400)]
mountd: Add mountlist_freeall()

I'm about to add a second bit of logic that needs to free all
mountlist records, so introduce a helper for freeing them.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agomountd: support IPv6 in mountlist_del_all()
Chuck Lever [Thu, 16 Sep 2010 17:46:26 +0000 (13:46 -0400)]
mountd: support IPv6 in mountlist_del_all()

Replace IPv4-specific code in the mountlist_del_all() path with code
that is address family agnostic.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agomountd: Support IPv6 in mountd's svc routines
Chuck Lever [Thu, 16 Sep 2010 17:44:02 +0000 (13:44 -0400)]
mountd: Support IPv6 in mountd's svc routines

Replace IPv4-specific code with use of our generic hostname helpers
in the routines that handle incoming MNT RPC requests.

These functions will support IPv6 without additional changes, once
IPv6 is enabled in the generic hostname helpers.

As part of this update, I've modified all of mountd's _svc routines
to use a debug message format that is consistent with statd.  It may
be overkill for some of these; if so we can pull them out later.

Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agomountd: add IPv6 support in auth_authenticate()
Chuck Lever [Thu, 16 Sep 2010 17:39:39 +0000 (13:39 -0400)]
mountd: add IPv6 support in auth_authenticate()

Make the entire auth_authenticate() code path address-family agnostic.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agolibnfs.a: Fix API for getfh() & friends
Chuck Lever [Thu, 16 Sep 2010 13:32:52 +0000 (09:32 -0400)]
libnfs.a: Fix API for getfh() & friends

This is more of a clean-up than a behavioral change.

POSIX requires that a "struct sockaddr" is the same size as a "struct
sockaddr_in".  Therefore, a variable or field of type "struct sockaddr"
cannot contain an AF_INET6 address.  However, "struct sockaddr *" is
often used to reference a generic (ie non-address family specific)
socket address, generating some confusion about this.

The nfsctl_arg struct uses a struct sockaddr (not a pointer) to pass
the client's IP address to the kernel.  This means the legacy nfsctl()
kernel API can never support IPv6.  Fortunately for us, this legacy
interface was replaced by a text-based cache interface a few years
back.  We don't need to support non-AF_INET addresses here.

The getfh() functions in nfs-utils provide a handy C API for the
kernel's nfsctl interface.  The getfh() functions still take a struct
sockaddr *, though, and that can imply that a non-IPv4 address can be
passed via this API.  To make it abundantly clear that only IPv4
addresses can be used with this interface, change the synopses of
getfh() and friends to take a struct sockaddr_in * instead of a struct
sockaddr * .

This makes these functions conform with other places in mountd and
exportfs that already grok the difference between a struct sockaddr
and a struct sockaddr_in.

While we're here...

Introduce some nice documenting comments for the get_fh() functions,
and...

Since mountd will support IPv6 in the near future, assert that the
family of client addresses passed to this API is indeed AF_INET, in
order to prevent non-AF_INET addresses from ever being passed to the
legacy nfsctl() interface.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agomount.nfs: Don't do anything fancy if this is a remount
Chuck Lever [Thu, 16 Sep 2010 11:28:18 +0000 (07:28 -0400)]
mount.nfs: Don't do anything fancy if this is a remount

We don't want to append "vers=4" or perform any negotiation if the
"remount" mount option was specified.  It will just end in tears.

This attempts to address

  https://qa.mandriva.com/show_bug.cgi?id=60311

and

  https://bugzilla.linux-nfs.org/show_bug.cgi?id=187

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agomount.nfs: Refactor mount version and protocol autonegotiation
Chuck Lever [Thu, 16 Sep 2010 11:26:07 +0000 (07:26 -0400)]
mount.nfs: Refactor mount version and protocol autonegotiation

Clean up.

I'm beginning to agree with Bruce and Steve's assessment that the
fallthrough switch case in nfs_try_mount() is more difficult to read
and understand than it needs to be.  The logic that manages
negotiating NFS version and protocol settings is getting more complex
over time anyway.

So let's split the autonegotiation piece out of nfs_try_mount().

We can reduce indenting, and use cleaner switch-based logic.  Also,
adding more comments can only help.

Neil also suggested replacing the pre-call "errno = 0" trick.  The
lower-level functions may try to mount several times (given a list of
addresses to try).  errno could be set by any of those.  The mount
request will succeed at some point, and "success" is returned, but
errno is still set to some non-zero value.

The kernel version check in nfs_try_mount() is more or less loop
invariant: it's impossible for the result of that test to change
between retries.  So we should be able to safely move it to the logic
that sets the initial value of mi->version.

This patch is not supposed to cause a behavioral change.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agorpc.idmapd: Type of idmap client should be defined by ic_id not ic_clid.
Bian Naimeng [Thu, 16 Sep 2010 11:12:47 +0000 (07:12 -0400)]
rpc.idmapd: Type of idmap client should be defined by ic_id not ic_clid.

The type of idmap_client is defined by idmap_client.ic_id for nfs,
so nfsd should have the same style.

Signed-off-by: Bian Naimeng <biannm@cn.fujitsu.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agonfsstat: add release_lockowner to client stats
Benny Halevy [Tue, 20 Jul 2010 05:40:34 +0000 (08:40 +0300)]
nfsstat: add release_lockowner to client stats

Signed-off-by: Benny Halevy <bhalevy@panasas.com>
8 years agoCleaned up a warning from commit 44f09b7
Steve Dickson [Thu, 9 Sep 2010 14:47:32 +0000 (10:47 -0400)]
Cleaned up a warning from commit 44f09b7

Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agomount.nfs: Prepare way for "vers=4,rdma" mounts
Chuck Lever [Wed, 8 Sep 2010 17:28:24 +0000 (13:28 -0400)]
mount.nfs: Prepare way for "vers=4,rdma" mounts

At some point, when the kernel starts to support "vers=4,rdma" mounts,
we will want the mount.nfs command to pass "vers=4,rdma" mounts
instead of rejecting them.

Assuming that the kernel will reject these today with EPROTONOSUPPORT,
that would cause the version fallback logic to go to "vers=3,rdma"
automatically.  So the extra check we have now is not needed anyway.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agomount.nfs: Support an "rdma" mount option
Chuck Lever [Wed, 8 Sep 2010 17:27:31 +0000 (13:27 -0400)]
mount.nfs: Support an "rdma" mount option

The kernel NFS client's mount option parser recognizes a stand-alone
"rdma" mount option, similar to the legacy "udp" and "tcp" options.

The mount.nfs command text-based mount option parser used to pass
"rdma" straight to the kernel, but since we've started handling MNT in
the kernel instead of in user space, "rdma" on the command line has
not worked.

Until now, no-one has noticed, especially since an "rdma" mount option
isn't documented in nfs(5).

Support "rdma" in mount.nfs command, and document it in nfs(5).

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agomount.nfs: Use nfs_nfs_protocol() for checking for proto=rdma
Chuck Lever [Wed, 8 Sep 2010 17:26:34 +0000 (13:26 -0400)]
mount.nfs: Use nfs_nfs_protocol() for checking for proto=rdma

Clean up: Now that nfs_get_proto() can recognize "rdma" we can re-use
nfs_nfs_protocol() instead of ad hoc checks for "proto=rdma".

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agogetport: Recognize "rdma" and "rdma6" netid
Chuck Lever [Wed, 8 Sep 2010 17:25:56 +0000 (13:25 -0400)]
getport: Recognize "rdma" and "rdma6" netid

The mount.nfs command must recognize the values of "rdma" and "rdma6"
with the "proto=" mount option.  Typically the mount.nfs command
relies on libtirpc or getprotobyname(3) to recognize netids and
translate them to protocol numbers.

RFCs 5665 and 5666 define the "rdma" and "rdma6" netids.  IANA defines
a specific port number for NFS over RDMA (20049), but has not provided
a protocol name and number for RDMA transports, and is not expected
to.  The best we can do is translate these by hand, as needed, to get
RDMA mount requests to the kernel without erroring out.

Only the forward translation is needed until such time that "rdma" and
"rdma6" start to appear in rpcbind registries.  For now, the version
and transport negotiation logic is skipped, avoiding rpcbind queries
for RDMA mounts.

Note: As of kernel 2.6.36, the kernel's NFS over RDMA transport
capability does not support IPv6.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agoClarification about options supported by different versions
Guillaume Rousse [Thu, 9 Sep 2010 14:33:47 +0000 (10:33 -0400)]
Clarification about options supported by different versions

Signed-off-by: Guillaume Rousse <Guillaume.Rousse@inria.fr>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agolibexport.a: Refactor init_netmask()
Chuck Lever [Thu, 9 Sep 2010 14:27:17 +0000 (10:27 -0400)]
libexport.a: Refactor init_netmask()

Instead of a single function that can handle both AF_INET and AF_INET6
addresses, two separate functions might be cleaner.

The original plan was to keep code redundancy at a minimum, but the
resulting code was cumbersome at best.  I think I've traded a little
extra code for something that will be much easier to read, understand,
and maintain.

I've also eliminated the "#if / #endif" instances inside the functions.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agolibexport.a: IPv6 support in client_check()
Chuck Lever [Tue, 31 Aug 2010 19:31:57 +0000 (15:31 -0400)]
libexport.a: IPv6 support in client_check()

Introduce support for IPv6 in client_check()'s helpers.  The local
addrs_match() twins are no longer needed since we can use
nfs_compare_addrs() now.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agolibexport.a: IPv6 support for client_init_subnet()
Chuck Lever [Tue, 31 Aug 2010 19:30:14 +0000 (15:30 -0400)]
libexport.a: IPv6 support for client_init_subnet()

To parse and store an IPv6 host or subnet address, init_netmask()
needs to handle 128 bit subnet masks.

Unfortunately what once was a pretty simple little function has grown
much larger.  This logic must now not only parse IPv6 addresses
correctly, but must also distinguish between IPv4 and IPv6.

To avoid code duplication, I'm "bending" the cardinal rule of not
using "#ifdef" inside functions.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agolibexport.a: Prepare init_subnetwork() for IPv6 support
Chuck Lever [Tue, 31 Aug 2010 19:29:02 +0000 (15:29 -0400)]
libexport.a: Prepare init_subnetwork() for IPv6 support

Retire the slash32 logic in inet_netmask() in favor of a more generic
netmask parser that can support IPv6 addresses.

If an invalid IP address string is given to inet_addr(3), it returns
INADDR_NONE, which is actually a "valid" address (255.255.255.255).
We're none the wiser to the substitution until something breaks later.

This patch provides better sanity checking of the parsed address, now
that such an error can be reported to client_init()'s callers.
We can also check the prefixlen value a little more carefully as well.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agolibexport.a: Use host helper to parse address in client_init()
Chuck Lever [Tue, 31 Aug 2010 19:27:19 +0000 (15:27 -0400)]
libexport.a: Use host helper to parse address in client_init()

Take the first step towards making it possible to parse either IPv4 or
IPv6 addresses in client_init().  It won't handle IPv6 until
host_pton() has IPv6 support enabled, and it still doesn't deal with
IPv6 netmasks yet.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agolibexport.a: Prepare to recognize IPv6 addresses in client_gettype()
Chuck Lever [Tue, 31 Aug 2010 19:25:35 +0000 (15:25 -0400)]
libexport.a: Prepare to recognize IPv6 addresses in client_gettype()

The current open-coded parsing logic in client_gettype() will be hard
to modify to recognize IPv6 addresses.  Use a more generic mechanism
for detecting IP presentation addresses.

IPv6 will be enabled automatically in client_gettype() when host_pton()
is changed to support IPv6 addresses.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agoexportfs: Use xlog() for error reporting
Chuck Lever [Tue, 24 Aug 2010 11:19:34 +0000 (07:19 -0400)]
exportfs: Use xlog() for error reporting

exportfs already invokes xlog_open() because libexport.a uses xlog()
exclusively for error reporting and debugging messages.  If we can
use xlog() throughout exportfs itself, that enables xlog debugging
messages everywhere in the code path.

In addition, use xlog() instead of fprintf(stderr) for reporting
errors in exportfs.c, to be consistent with libexport.a and other
components of nfs-utils.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agoexportfs: exportfs.c no longer needs #include "xmalloc.h"
Chuck Lever [Tue, 24 Aug 2010 11:18:04 +0000 (07:18 -0400)]
exportfs: exportfs.c no longer needs #include "xmalloc.h"

Clean up:  No calls to xmalloc() or xstrdup() here.  No need for the
double #include of xmalloc.h.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agostatd: statd fails to monitor if no reverse mapping of mon_name exists
Chuck Lever [Tue, 24 Aug 2010 11:16:10 +0000 (07:16 -0400)]
statd: statd fails to monitor if no reverse mapping of mon_name exists

Commit 8ce130c4 switched in the new statd_canonical_name() function
that constructs a "unique" name statd can use to uniquely identify a
monitor record.

The legacy statd would monitor a client that sent an IP address with
no reverse map as its caller_name.  To remain bug-for-bug compatible,
allow this case in the new statd.

This shouldn't be a problem: statd_canonical_name() needs to create
a unique name for the monitored host so it can keep track of monitor
requests from the same remote.  The IP address itself should work as
well as the host's canonical name, in case there is no reverse
mapping.

We still enforce the requirement that a mon_name that is a DNS name
must have a forward map to an IP address.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agolibexport: Fix IP address check in check_netgroup()
Chuck Lever [Tue, 24 Aug 2010 11:13:54 +0000 (07:13 -0400)]
libexport: Fix IP address check in check_netgroup()

Neil Brown reports that recent changes to replace
gethostby{addr,name}(3) with get{addr,info}name(3) may have
inadvertently broken netgroup support.

There used to be a gethostbyaddr(3) call in the third paragraph in
check_netgroup().  The reason for that gethostbyaddr(3) call was that
the first innetgr(3) call has already confirmed that hname is not a
member of the netgroup.  We also need to confirm that, if hname
happens to be an IP address, the hostname bound to that IP address is
not a member of the netgroup, either.

Fix this by restoring appropriate address to hostname mapping of hname
before retrying the innetgr(3) call.

See http://marc.info/?l=linux-nfs&m=128084830214653&w=2 .

Introduced by commit 0509d3428f523776ddd9d6e9fa318587d3ec7d84.

Reviewed-by: Neil Brown <neilb@suse.de>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
8 years agobexport: Add a common exit label to check_netgroup()
Chuck Lever [Tue, 24 Aug 2010 11:06:14 +0000 (07:06 -0400)]
bexport: Add a common exit label to check_netgroup()

check_netgroup() is going to be changed to free dynamically
allocated resources before it returns, so a common
exit point is needed.

Reviewed-by: Neil Brown <neilb@suse.de>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoFix the description of nfsversion mount option in the man nfs page
Guillaume Rousse [Tue, 17 Aug 2010 21:47:38 +0000 (17:47 -0400)]
Fix the description of nfsversion mount option in the man nfs page

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoRemove warnings from nfsctl.c nfs-utils-1-2-3-rc5
Steve Dickson [Mon, 19 Jul 2010 16:12:52 +0000 (12:12 -0400)]
Remove warnings from nfsctl.c

nfsctl.c: In function 'expsetup':
nfsctl.c:112: warning: signed and unsigned type in conditional expression

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoRemoved warnings from mountd.c
Steve Dickson [Wed, 4 Aug 2010 14:48:17 +0000 (10:48 -0400)]
Removed warnings from mountd.c

mountd.c: In function 'mount_null_1_svc':
mountd.c:195: warning: unused parameter 'rqstp'
mountd.c:195: warning: unused parameter 'argp'
mountd.c:195: warning: unused parameter 'resp'
mountd.c: In function 'mount_dump_1_svc':
mountd.c:213: warning: unused parameter 'argp'
mountd.c: In function 'mount_umnt_1_svc':
mountd.c:224: warning: unused parameter 'resp'
mountd.c: In function 'mount_umntall_1_svc':
mountd.c:248: warning: unused parameter 'argp'
mountd.c:248: warning: unused parameter 'resp'
mountd.c: In function 'mount_export_1_svc':
mountd.c:258: warning: unused parameter 'argp'
mountd.c: In function 'mount_exportall_1_svc':
mountd.c:269: warning: unused parameter 'argp'
mountd.c: In function 'mount_dump_1_svc':
mountd.c:216: warning: unused parameter 'argp'
mountd.c: In function 'mount_umnt_1_svc':
mountd.c:227: warning: unused parameter 'resp'
mountd.c: In function 'mount_umntall_1_svc':
mountd.c:251: warning: unused parameter 'argp'
mountd.c:251: warning: unused parameter 'resp'
mountd.c: In function 'mount_export_1_svc':
mountd.c:261: warning: unused parameter 'argp'
mountd.c: In function 'mount_exportall_1_svc':
mountd.c:272: warning: unused parameter 'argp'

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoRemoved warnings from idmapd.c
Steve Dickson [Wed, 4 Aug 2010 14:41:18 +0000 (10:41 -0400)]
Removed warnings from idmapd.c

idmapd.c:120: warning: missing initializer
idmapd.c:120: warning: (near initialization for 'nfsd_ic[0].ic_event')
idmapd.c:121: warning: missing initializer
idmapd.c:121: warning: (near initialization for 'nfsd_ic[1].ic_event')
idmapd.c: In function 'flush_nfsd_cache':
idmapd.c:173: warning: comparison between signed and unsigned integer expressions
idmapd.c: In function 'dirscancb':
idmapd.c:384: warning: unused parameter 'fd'
idmapd.c:384: warning: unused parameter 'which'
idmapd.c: In function 'svrreopen':
idmapd.c:468: warning: unused parameter 'fd'
idmapd.c:468: warning: unused parameter 'which'
idmapd.c:468: warning: unused parameter 'data'
idmapd.c: In function 'clntscancb':
idmapd.c:474: warning: unused parameter 'fd'
idmapd.c:474: warning: unused parameter 'which'
idmapd.c: In function 'nfsdcb':
idmapd.c:488: warning: unused parameter 'fd'
idmapd.c: In function 'nfscb':
idmapd.c:663: warning: unused parameter 'fd'
idmapd.c: In function 'validateascii':
idmapd.c:850: warning: comparison between signed and unsigned integer expressions
idmapd.c:858: warning: comparison between signed and unsigned integer expressions
idmapd.c: In function 'getfield':
idmapd.c:916: warning: comparison between signed and unsigned integer expressions

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoRemoved warnings from configfile.c
Steve Dickson [Wed, 4 Aug 2010 14:29:28 +0000 (10:29 -0400)]
Removed warnings from configfile.c

configfile.c:195: warning: 'inline' is not at beginning of declaration
configfile.c:232: warning: 'inline' is not at beginning of declaration

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoRemoved warnings from svcgssd_proc.c
Steve Dickson [Wed, 4 Aug 2010 14:26:13 +0000 (10:26 -0400)]
Removed warnings from svcgssd_proc.c

svcgssd_proc.c: In function 'send_response':
svcgssd_proc.c:135: warning: unused parameter 'f'
svcgssd_proc.c: In function 'handle_nullreq':
svcgssd_proc.c:434: warning: comparison of unsigned expression < 0 is always false

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoRemoved warnings from network.c
Steve Dickson [Wed, 4 Aug 2010 14:22:13 +0000 (10:22 -0400)]
Removed warnings from network.c

network.c: In function 'nfs_verify_family':
network.c:1366: warning: unused parameter 'family'

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoRemoved warnings from nfs4mount.c
Steve Dickson [Tue, 20 Jul 2010 22:43:46 +0000 (18:43 -0400)]
Removed warnings from nfs4mount.c

nfs4mount.c: In function 'fill_ipv4_sockaddr':
nfs4mount.c:149: warning: comparison between signed and unsigned integer expressions

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoRemoved warnings from nfsmount.c
Steve Dickson [Tue, 20 Jul 2010 22:40:31 +0000 (18:40 -0400)]
Removed warnings from nfsmount.c

nfsmount.c: In function 'nfsmount':
nfsmount.c:513: warning: missing initializer
nfsmount.c:513: warning: (near initialization for 'mnt_server.saddr')
nfsmount.c:514: warning: missing initializer
nfsmount.c:514: warning: (near initialization for 'nfs_server.saddr')

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoRemoved warnings from svcgssd.c
Steve Dickson [Tue, 20 Jul 2010 22:22:32 +0000 (18:22 -0400)]
Removed warnings from svcgssd.c

svcgssd.c: In function 'sig_hup':
svcgssd.c:160: warning: unused parameter 'signal'

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoRemoved warnings from gssd_proc.c
Steve Dickson [Tue, 20 Jul 2010 22:13:45 +0000 (18:13 -0400)]
Removed warnings from gssd_proc.c

gssd_proc.c: In function 'create_auth_rpc_client':
gssd_proc.c:939: warning: comparison between signed and unsigned integer expressions
gssd_proc.c:939: warning: comparison between signed and unsigned integer expressions
gssd_proc.c: In function 'handle_krb5_upcall':
gssd_proc.c:1164: warning: comparison between signed and unsigned integer expressions
gssd_proc.c: In function 'handle_spkm3_upcall':
gssd_proc.c:1178: warning: comparison between signed and unsigned integer expressions

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoRemoved warnings from krb5_util.c
Steve Dickson [Tue, 20 Jul 2010 22:08:13 +0000 (18:08 -0400)]
Removed warnings from krb5_util.c

krb5_util.c: In function 'realm_and_service_match':
krb5_util.c:617: warning: unused parameter 'context'
krb5_util.c: In function 'limit_krb5_enctypes':
krb5_util.c:1275: warning: unused parameter 'uid'

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoRemoved warnings from gssd_main_loop.c
Steve Dickson [Tue, 20 Jul 2010 21:53:47 +0000 (17:53 -0400)]
Removed warnings from gssd_main_loop.c

gssd_main_loop.c: In function 'dir_notify_handler':
gssd_main_loop.c:64: warning: unused parameter 'sig'
gssd_main_loop.c:64: warning: unused parameter 'si'
gssd_main_loop.c:64: warning: unused parameter 'data'

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoRemoved warnings from gssd.c
Steve Dickson [Tue, 20 Jul 2010 21:45:17 +0000 (17:45 -0400)]
Removed warnings from gssd.c

gssd.c: In function 'sig_hup':
gssd.c:78: warning: unused parameter 'signal'

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoRemoved warnings from atomicio.c
Steve Dickson [Tue, 20 Jul 2010 21:19:20 +0000 (17:19 -0400)]
Removed warnings from atomicio.c

atomicio.c: In function 'atomicio':
atomicio.c:48: warning: comparison between signed and unsigned integer expressions

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoRemoved warnings from nfsstat.c
Steve Dickson [Mon, 19 Jul 2010 18:54:51 +0000 (14:54 -0400)]
Removed warnings from nfsstat.c

nfsstat.c: In function 'print_callstats':
nfsstat.c:797: warning: comparison between signed and unsigned integer
expressions
nfsstat.c:801: warning: comparison between signed and unsigned integer
expressions
nfsstat.c:802: warning: comparison between signed and unsigned integer
expressions
nfsstat.c:805: warning: comparison between signed and unsigned integer
expressions

nfsstat.c: In function 'print_callstats_list':
nfsstat.c:821: warning: comparison between signed and unsigned integer
expressions
nfsstat.c:828: warning: comparison between signed and unsigned integer
expressions

nfsstat.c: In function 'unpause':
nfsstat.c:1111: warning: unused parameter 'sig'
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoRemoved warnings from nfssvc.c
Steve Dickson [Mon, 19 Jul 2010 18:48:37 +0000 (14:48 -0400)]
Removed warnings from nfssvc.c

nfssvc.c:184: warning: comparison between signed and unsigned integer expressions
nfssvc.c: In function 'nfssvc_setvers':
nfssvc.c:254: warning: comparison between signed and unsigned integer expressions
nfssvc.c: In function 'nfssvc_threads':
nfssvc.c:280: warning: comparison between signed and unsigned integer expressions

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoRemoved warnings from cache.c
Steve Dickson [Mon, 19 Jul 2010 18:43:26 +0000 (14:43 -0400)]
Removed warnings from cache.c

cache.c:812: warning: missing initializer
cache.c:812: warning: (near initialization for 'cachelist[0].f')
cache.c:813: warning: missing initializer
cache.c:813: warning: (near initialization for 'cachelist[1].f')
cache.c:814: warning: missing initializer
cache.c:814: warning: (near initialization for 'cachelist[2].f')
cache.c:815: warning: missing initializer
cache.c:815: warning: (near initialization for 'cachelist[3].f')
cache.c:816: warning: missing initializer
cache.c:816: warning: (near initialization for 'cachelist[4].f')
cache.c: In function 'cache_export_ent':
cache.c:887: warning: comparison between signed and unsigned integer expressions
cache.c:907: warning: comparison between signed and unsigned integer expressions

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoRemoved warnings from fsloc.c
Steve Dickson [Mon, 19 Jul 2010 17:07:00 +0000 (13:07 -0400)]
Removed warnings from fsloc.c

fsloc.c: In function 'replicas_lookup':
fsloc.c:149: warning: unused parameter 'key'

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoRemoved warnings from conffile.c
Steve Dickson [Mon, 19 Jul 2010 16:39:52 +0000 (12:39 -0400)]
Removed warnings from conffile.c

conffile.c: In function 'conf_parse_line':
conffile.c:301: warning: comparison between signed and unsigned integer expressions

conffile.c: In function 'conf_load_defaults':
conffile.c:356: warning: unused parameter 'tr'

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoRemove warnings from nfs_mntent.c
Steve Dickson [Mon, 19 Jul 2010 16:34:14 +0000 (12:34 -0400)]
Remove warnings from nfs_mntent.c

nfs_mntent.c: In function 'mangle':
nfs_mntent.c:36: warning: comparison between signed and unsigned integer expressions

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoemove warnings from cacheio.c
Steve Dickson [Mon, 19 Jul 2010 16:31:48 +0000 (12:31 -0400)]
emove warnings from cacheio.c

cacheio.c: In function 'cache_flush':
cacheio.c:352: warning: comparison between signed and unsigned integer expressions

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoRemove warnings from svc_socket.c
Steve Dickson [Mon, 19 Jul 2010 16:28:09 +0000 (12:28 -0400)]
Remove warnings from svc_socket.c

svc_socket.c: In function 'svcudp_socket':
svc_socket.c:160: warning: unused parameter 'reuse'

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoRemoved warnings from rpcdispatch.c
Steve Dickson [Mon, 19 Jul 2010 16:17:24 +0000 (12:17 -0400)]
Removed warnings from rpcdispatch.c

rpcdispatch.c: In function 'rpc_dispatch':
rpcdispatch.c:30: warning: comparison between signed and unsigned
     integer expressions
rpcdispatch.c:35: warning: comparison between signed and unsigned
     integer expressions

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoEnable extra compile warnings (-Wextra) by default.
Steve Dickson [Mon, 19 Jul 2010 16:05:51 +0000 (12:05 -0400)]
Enable extra compile warnings (-Wextra) by default.

Added -Wextra to the CFLAGS which enables more checking
during compilation, which in turn, will hopefully flag
potential problems before they occur.

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agogssd: picking wrong creds
Doug Nazar [Mon, 19 Jul 2010 15:08:46 +0000 (11:08 -0400)]
gssd: picking wrong creds

When not using machine credentials for root, if the machine
credential cache file is newer than the root credential file
the wrong file will get picked. Ignore the machine file in
this case.

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agomountd: Fix memcmp result comparison error for uuid match.
Doug Nazar [Mon, 19 Jul 2010 14:58:47 +0000 (10:58 -0400)]
mountd: Fix memcmp result comparison error for uuid match.

Fixed Small typo in the new fs uuid comparison code

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agogssd: Fix endtime issue
Doug Nazar [Mon, 19 Jul 2010 14:48:34 +0000 (10:48 -0400)]
gssd: Fix endtime issue

Commit 4c5ff6d4 removed the setting of endtime for a few contexts by
accident.

Now to figure out why I get stale handles on submounts.

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agonfs-utils: Fix C aliasing rules violation in nfs_getrpccaller()
Chuck Lever [Mon, 19 Jul 2010 14:09:12 +0000 (10:09 -0400)]
nfs-utils: Fix C aliasing rules violation in nfs_getrpccaller()

Squelch compiler warnings reported with -Wextra:

In file included from statd.c:24:
../../support/include/rpcmisc.h: In function nfs_getrpccaller_in:
../../support/include/rpcmisc.h:58: warning: dereferencing type-punned
pointer might break strict-aliasing rules
../../support/include/rpcmisc.h: In function nfs_getrpccaller:
../../support/include/rpcmisc.h:63: warning: dereferencing type-punned
pointer might break strict-aliasing rules

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agonfs-iostat.py: divide by zero with fresh mount nfs-utils-1-2-3-rc4
Kevin Constantine [Tue, 22 Jun 2010 21:43:19 +0000 (17:43 -0400)]
nfs-iostat.py: divide by zero with fresh mount

When an export is freshly mounted, /proc/self/mountstats displays age = 0.
This causes nfs-iostat.py to divide by zero throwing an error.  When we
have age = 0, other stats are greater than 0, so we'll set age = 1 and
print the relevant stats.

Signed-off-by: Kevin Constantine <kevin.constantine@disneyanimation.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agolibexport.a: Remove unused hostent-based DNS helper functions
Chuck Lever [Tue, 22 Jun 2010 17:31:38 +0000 (13:31 -0400)]
libexport.a: Remove unused hostent-based DNS helper functions

Clean up:  Get rid of hostent-based DNS helper functions in
libexport.a that have been replaced by addrinfo-based DNS helpers.

None of the original code remains, so replace the copyright notice as
well.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agolibexport.a: Replace matchhostname()
Chuck Lever [Tue, 22 Jun 2010 17:06:40 +0000 (13:06 -0400)]
libexport.a: Replace matchhostname()

So that exportfs can eventually support IPv6 addresses, copy statd's
getaddrinfo(3)-based matchhostname to exportfs, with adjustments for
dealing with export wildcards and netgroups.  Until exportfs has full
IPv6 support, however, we want to ensure that IPv6 addresses continue
to remain blocked in the address comparison code used by exportfs.  At
a later point we'll replace much of this with the generic functions
in sockaddr.h.

Since it contains special logic for handling wildcard and netgroups,
this function is specialized for exportfs, and does not belong in
one of the shared libraries.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agolibexport.a: Add documenting comments
Chuck Lever [Tue, 22 Jun 2010 16:51:42 +0000 (12:51 -0400)]
libexport.a: Add documenting comments

Clean up.  Add a few additional documenting comments for globally
visible functions.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agolibexport.a: Make export_add() static
Chuck Lever [Tue, 22 Jun 2010 16:44:44 +0000 (12:44 -0400)]
libexport.a: Make export_add() static

Clean up: export_add() is not called from outside of export.c, so make
it a static helper.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agolibexport.a: Make export_read() return void
Chuck Lever [Tue, 22 Jun 2010 16:43:55 +0000 (12:43 -0400)]
libexport.a: Make export_read() return void

Clean up: export_read()'s return value is always zero, and its only
caller never checks it.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agomountd: Replace "struct hostent" with "struct addrinfo"
Chuck Lever [Tue, 22 Jun 2010 16:43:01 +0000 (12:43 -0400)]
mountd: Replace "struct hostent" with "struct addrinfo"

struct hostent can store either IPv4 or IPv6 addresses, but it can't
store both address families concurrently for the same host.  Neither
can hostent deal with parts of socket addresses that are outside of
the sin{,6}_addr field.

Replace the use of "struct hostent" everywhere in libexport.a, mountd,
and exportfs with "struct addrinfo".  This is a large change, but
there are so many strong dependencies on struct hostent that this
can't easily be broken into smaller pieces.

One benefit of this change is that hostent_dup() is no longer
required, since the results of getaddrinfo(3) are already dynamically
allocated.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoexportfs: Add a common exit label in exportfs()
Chuck Lever [Tue, 22 Jun 2010 16:40:27 +0000 (12:40 -0400)]
exportfs: Add a common exit label in exportfs()

Clean up: Reduce code duplication by introducing a goto label for
freeing hp and exiting.  This will make replacing "struct hostent *"
with "struct addrinfo *" more straightforward in this code.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agolibexport.a: Add helpers to manage DNS lookups
Chuck Lever [Tue, 22 Jun 2010 14:41:03 +0000 (10:41 -0400)]
libexport.a: Add helpers to manage DNS lookups

Introduce DNS query helpers based on getaddrinfo(3) and
getnameinfo(3).  These will eventually replace the existing
hostent-based functions in support/export/hostname.c.

Put some of these new helpers to immediate use, where convenient.

As they are part of libexport.a, I've added the forward declarations
for these new functions in exportfs.h rather than misc.h, where the
hostent-based forward declarations are currently.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoImprove support for exporting btrfs subvolumes
NeilBrown [Tue, 22 Jun 2010 14:37:25 +0000 (10:37 -0400)]
Improve support for exporting btrfs subvolumes

If you export two subvolumes of a btrfs filesystem, they will both be
given the same uuid so lookups will be confused.
blkid cannot differentiate the two, so we must use the fsid from
statfs64 to identify the filesystem.

We cannot tell if blkid or statfs is best without knowing internal
details of the filesystem in question, so we need to encode specific
knowledge of btrfs in mountd.  This is unfortunate.

To ensure smooth handling of this and possible future changes in uuid
generation, we add infrastructure for multiple different uuids to be
recognised on old filehandles, but only the preferred on is used on
new filehandles.

Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agomount.nfs: silently fails when the network protocol is not found
Steve Dickson [Thu, 3 Jun 2010 12:53:22 +0000 (08:53 -0400)]
mount.nfs: silently fails when the network protocol is not found

mount.nfs should display some type of error diagnostics when
the network protocol can not be determined.

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agomount: silently fails when bad option values are given
Steve Dickson [Thu, 3 Jun 2010 12:32:56 +0000 (08:32 -0400)]
mount: silently fails when bad option values are given

mount.nfs should not only fail when an invalid option values
are supplied (as it does), it should also print a diagnostic
message identifying the problem

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agonfsiostat.man: Add linebreak before --version option
Kevin Constantine [Wed, 2 Jun 2010 12:34:14 +0000 (08:34 -0400)]
nfsiostat.man: Add linebreak before --version option

The nfsiostat man file was missing a linebreak before the --verbose option

Signed-off-by: Kevin Constantine <kevin.constantine@disneyanimation.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agomountd: Convert colons in IPv6 presentation addresses tosemicolons
Chuck Lever [Mon, 24 May 2010 10:02:22 +0000 (06:02 -0400)]
mountd: Convert colons in IPv6 presentation addresses tosemicolons

The /var/lib/nfs/rmtab file delineates fields in each of its lines
with a ":".  The first field contains the IP address of a client, in
presentation format.  IPv6 presentation format addresses contain
colons, which screws up the field delineation of rmtab.

Use a simple simple scheme to convert the colons in incoming client
names to some other character, and then convert them back when the
rmtab file is read.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agolibexport.a: Refactor rmtab_read()
Chuck Lever [Mon, 24 May 2010 10:01:22 +0000 (06:01 -0400)]
libexport.a: Refactor rmtab_read()

Clean up: Make it easier to add IPv6 support by refactoring part of
rmtab_read() into a helper function.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoMountd listens on 2 different versions for NFSv2 (MOUNTv1 and MOUNTv2)
NeilBrown [Mon, 24 May 2010 09:55:26 +0000 (05:55 -0400)]
Mountd listens on 2 different versions for NFSv2 (MOUNTv1 and MOUNTv2)
and one for NFSv3 (MOUNTv3)

When --no-nfs-version requests an NFS version to be disabled, the
code actually disabled the MOUNT version. This works is several cases,
but requires --no-nfs-version 1 to completely disable NFSv2, which
is wrong.

So if we do disable 1, 2, and 3. mountd complain and won't run, it
is not possible to run just v4 - i.e. not listening for MOUNT requests
at all (as v4 doesn't need them).

So change the handling of "--no-nfs-version 2" it disable MOUNTv1 as
well as
MOUNTv2, and allow mountd to continue running as long as one of
NFSv2 NFSv3 NFSv4 is enabled.

Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agomountd/exportfs: Make m_addrlist field a nfs_sockaddr nfs-utils-1-2-3-rc3
Chuck Lever [Wed, 5 May 2010 19:41:07 +0000 (15:41 -0400)]
mountd/exportfs: Make m_addrlist field a nfs_sockaddr

To store non-AF_INET addresses in the nfs_client structure, we need to
use more than in_addr for the m_addrlist field.  Make m_addrlist
larger, then add a few helper functions to handle type casting and
array indexing cleanly.

We could treat the nfs_client address list as if all the addresses
in the list were the same family.  This might work for MCL_SUBNETWORK
type nfs_clients.  However, during the transition to IPv6, most hosts
will have at least one IPv4 and one IPv6 address.  For MCL_FQDN, I
think we need to have the ability to store addresses from both
families in one nfs_client.

Additionally, IPv6 scope IDs are not part of struct sin6_addr.  To
support link-local IPv6 addresses and the like, a scope ID must be
stored.

Thus, each slot in the address list needs to be capable of storing an
entire socket address, and not simply the network address part.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agolibexport.a: Clean up client_add()
Chuck Lever [Wed, 5 May 2010 18:42:01 +0000 (14:42 -0400)]
libexport.a: Clean up client_add()

Clean up: client_add()'s current callers never set unknown m_type
values, so the m_type check is unnecessary.

All of client_add()'s callers are in the same source file where it is
defined, so make it a static helper function.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agolibexport.a: Refactor client_init()
Chuck Lever [Wed, 5 May 2010 18:37:13 +0000 (14:37 -0400)]
libexport.a: Refactor client_init()

Clean up:  Move subnet hostname parsing logic out of client_init() to
make it simpler to introduce IPv6 support.

Make client_init() a helper, since it's already static.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoMake capabilities support optional
Mike Frysinger [Fri, 23 Apr 2010 16:22:53 +0000 (12:22 -0400)]
Make capabilities support optional

The new code using libcap is quite minor, so rather than always reqiure
libcap support, make it a normal --enable type flag.  Current default
behavior is retained -- if libcap is found, it is enabled, else it is
disabled like every nfs-utils version in the past.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoset +x on autogen.sh
Mike Frysinger [Fri, 23 Apr 2010 16:18:27 +0000 (12:18 -0400)]
set +x on autogen.sh

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agolibexport.a: Allow malloc(3) failures in client_lookup() and friends
Chuck Lever [Fri, 23 Apr 2010 16:09:35 +0000 (12:09 -0400)]
libexport.a: Allow malloc(3) failures in client_lookup() and friends

Clean up: Use malloc(3) instead of xmalloc() in client_lookup() and
client_dup(), ensuring that a failed memory allocation here doesn't
cause our process to exit suddenly.

Allocation of nfs_client records and the m_hostname string are now
consistently handled with malloc(3), calloc(3), strdup(3), and
free(3).

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agolibexport.a: Allow m_hostname allocation to fail instead of exit
Chuck Lever [Fri, 23 Apr 2010 16:08:49 +0000 (12:08 -0400)]
libexport.a: Allow m_hostname allocation to fail instead of exit

Clean up: Replace xstrdup() with strdup(3) in client_init(), to
prevent the process from exiting if the memory allocation fails.

Note that both of client_init()'s callers set m_hostname equal to NULL
before calling, thus the extra free(3) at the top of client_init() is
unneeded.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agolibexport.a: Allow client_init() to fail instead of exit
Chuck Lever [Fri, 23 Apr 2010 16:02:34 +0000 (12:02 -0400)]
libexport.a: Allow client_init() to fail instead of exit

client_init()'s current callers can now deal correctly with a failure.
Get rid of code that can cause our process to exit in client_init(),
if address mask parsing or memory allocation fails.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agolibexport.a: Add client_free()
Chuck Lever [Fri, 23 Apr 2010 16:00:28 +0000 (12:00 -0400)]
libexport.a: Add client_free()

Clean up: Introduce a helper to free a nfs_client record.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoAdd in autoconf support for mountstats and nfsiostats nfs-utils-1-2-3-rc2
Steve Dickson [Fri, 16 Apr 2010 14:56:29 +0000 (10:56 -0400)]
Add in autoconf support for mountstats and nfsiostats

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoIntroduce man page for the nfsiostats command
Steve Dickson [Fri, 16 Apr 2010 14:49:10 +0000 (10:49 -0400)]
Introduce man page for the nfsiostats command

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoIntroduce man page for the mountstats command
Steve Dickson [Fri, 16 Apr 2010 14:46:41 +0000 (10:46 -0400)]
Introduce man page for the mountstats command

Signed-off-by: Steve Dickson <steved@redhat.com>
9 years agoRemoved some miscellaneous warnings in the new gssd code.
Steve Dickson [Fri, 16 Apr 2010 17:31:57 +0000 (13:31 -0400)]
Removed some miscellaneous warnings in the new gssd code.

Signed-off-by: Steve Dickson <steved@redhat.com>