nfs: Update to version number 2.8.2

nfs: Update to version number 2.8.2

- Update from versionj number 2.7.1 to 2.8.2
- Update of rootfile
- Changelog
    2.8.2
	exports: Fix referrals when --enable-junction=no
	    Commit 15dc0bea ("exportd: Moved cache upcalls routines into
	    libexport.a") caused write_fsloc() to be elided when junction support is
	    disabled. Remove the not needed #ifdef HAVE_JUNCTION_SUPPORT which
	    blocks the referrals from working when --enable-junction=no is set.
	    (only the #ifdef HAVE_JUNCTION_SUPPORT should be around actual
	    junction code).
	    Fixes: 15dc0bea ("exportd: Moved cache upcalls routines into libexport.a")
Link: https://bugs.debian.org/1035908
Link: https://bugs.debian.org/1083098
	nfsidmap(v2): Add guards around [nfsidmap] usages of [sysconf].
	    sysconf(_SC_GETPW_R_SIZE_MAX) and sysconf(_SC_GETGR_R_SIZE_MAX)
	    return -1 on musl, which causes either segmentation faults or ENOMEM
	    errors.
	    Replace all usages of sysconf with dedicated methods that guard against
	    a result of -1.
	libnsm(v2): fix the safer atomic filenames fix
	    Commit 9f7a91b51ffc ("libnsm: safer atomic filenames") messed up the length
	    arguement to snprintf() in nsm_make_temp_pathname such that the length is
	    longer than the computed string.  When compiled with "-O
	    -D_FORTIFY_SOURCE=3", __snprintf_chk will fail and abort statd.
	    The fix is to correct the original size calculation, then pull one from the
	    snprintf length for the final "/".
	Revert "libnsm: fix the safer atomic filenames fix"
	    This reverts commit 8fcddae4437510137baf108f477d116ce345ce80.
	libnsm: fix the safer atomic filenames fix
	    Commit 9f7a91b51ffc ("libnsm: safer atomic filenames") messed up the length
	    arguement to snprintf() in nsm_make_temp_pathname such that the length is
	    longer than the computed string.  When compiled with "-O
	    -D_FORTIFY_SOURCE=3", __snprintf_chk will fail and abort statd.
	    The fix is to correct the original size calculation, then pull one from the
	    snprintf length for the final "/".
	nfsd: dump default number of threads to 16
	    nfsdctl defaults to 16 threads.  Since the nfs-server.service file first
	    tries nfsdctl and then falls back to rpc.nfsd, it would probably be wise
	    to make the default in rpc.nfsd and nfs.conf 16, for the sake of
	    consistency and to avoid surprises.
	autoconf: don't build nfsdcltrack by default
	    Now that we've started the process to remove legacy v4 client tracking
	    methods, let's stop building nfsdcltrack by default.
	nfs(5): Update rsize/wsize options
	    The rsize/wsize values are not multiples of 1024 but multiples of the
	    system's page size or powers of 2 if < system's page size as defined
	    in fs/nfs/internal.h:nfs_io_size().
	nfsdctl: clarify when versions can be set on the man page
	    Attempting to make version changes while there are nfsd threads running
	    fails with -EBUSY, so make note of it on the man page.
	nfsdctl: fix up the help text in version_usage()
	    The help text in version_usage() has examples with a 'v' character in
	    the version string, but the format string in the sscanf() call in
	    version_func() doesn't contain a 'v' character.
	libnsm: safer atomic filenames
	    We've gotten a report of reboot notifications being sent to domains that
	    end in '.new', which can happen if the NSM temporary pathname code leaves a
	    file behind.  Let's fix this up by prepending a single '.' to the temp path
	    which will never be resolvable as a DNS record.
	    https://lore.kernel.org/linux-nfs/04D30B5A-C53E-4920-ADCB-C77F5577669E@oracle.com/T/#t
	nfs-utils: fixup statd testing simulator host arg
	    The getopt setup for the host arg was not expecing a value, update it as
	    expected
	reexport.h: Include unistd.h to compile with musl
	Makefile.am: allow mount.nfs to be writeable by owner
	    On Red Hat-based systems, the debug symbol files are built with a
	    .gdb_index section to speed up gdb initialization.  The gdb-add-index
	    program calls objcopy to merge the index file into the object file.
	    That fails if the object file isn't writeable by the owner.
	mount.nfs: retry NFSv3 mount after NFSv4 failure in auto negotiation
	    The problem happens when a v3 mount fails with ETIMEDOUT after
	    the v4 mount failed with EPROTONOSUPPORT, in mount auto negotiation.
	    It immediately breaks from the "for" loop in nfsmount_fg()
	    or nfsmount_child() due to EPROTONOSUPPORT, never doing the expected
	    retries until timeout.
    2.8.1
	nfs-utils: use getpwuid_r() and getpwnam_r() in gssd
	    gssd uses getpwuid(3) and getpwnam(3) in a pthreads context but
	    these functions are not thread safe.
	nfsdcld: prevent from accessing /var/lib/nfs/nfsdcld in read-only file system during boot
	    I saw a VMWare guest that hit a rare condition during boot;
	    nfsdcld started too early to check access on /var/lib/nfs/nfsdcld which were
	    still in read-only file system as follows:
	      nfsdcld[...]: Unexpected error when checking access on /var/lib/nfs/nfsdcld: Read-only file system
	      systemd[1]: nfsdcld.service: Main process exited, code=exited, status=226/NAMESPACE
	      systemd[1]: nfsdcld.service: Failed with result 'exit-code'.
	    nfsdcld.service needs to wait the root file system to be remounted at least.
	systemd: use nfsdctl to start and stop the nfs server
	    Attempt to use nfsdctl to start and stop the nfs-server. If that fails
	    for any reason, use rpc.nfsd to do it instead.
	nfsdctl: asciidoc source for the manpage
	    Convert to manpage with:
	        asciidoctor -b manpage nfsdctl.adoc
	nfsdctl: add the nfsdctl utility to nfs-utils
	    This tool is based on Lorenzo's original nfsdctl tool [1]. His original
	    tool used getopt_long to indicate the command, but that's somewhat
	    limiting. This converts it to a subcommand-based interface, where each
	    subcommand can take its own options, in the spirit of commands like
	    nmcli or virsh.
	    There are currently 6 different subcommands:
	        pool-mode            get/set current pool mode setting
	        listener             get/set listener info
	        version              get/set supported NFS versions
	        threads              get/set nfsd thread settings
	        status               get current RPC processing info
	        autostart            start server with settings from /etc/nfs.conf
	    Each can take different options, and we can expand this interface later
	    with more commands as necessary.
	    This is based on Lorenzo's original userland tool:
	        https://github.com/LorenzoBianconi/nfsdctl
	rpc.idmapd: nfsopen() failures should not be fatal
	    dirscancb() loops over all clnt* subdirectories of /run/rpc_pipefs/nfs/.
	    Some of these directories contain /idmap files, others don't. nfsopen()
	    returns -1 for the latter; we then want to skip the directory, not abort
	    the entire scan.
	mount.nfs: improve EPROTO error message for RDMA mounts
	    When mounting NFS shares using RDMA, users may encounter this rather
	    unclear error message:
	        mount.nfs: Protocol error
	    Often there are either no RDMA interfaces existing, or that routing is
	    being done via other interfaces. This patch enhances the `mount_error`
	    function to provide a more informative message in such cases.
	support/junction/path.c: Fix buld for musl
	    Fixed:
	    path.c:164:24: error: implicit declaration of function 'strchrnul'; did you mean 'strchr'? [-Wimplicit-function-declaration]
	    [snip]
	    path.c:239:27: error: 'NAME_MAX' undeclared (first use in this function); did you mean 'AF_MAX'?
	support/include/junction.h: Define macros for musl
	    Fixed 1:
	    In file included from cache.c:1217:
	    ../../support/include/junction.h:128:21: error: expected ';' before 'char'
	      128 | __attribute_malloc__
	          |                     ^
	          |                     ;
	      129 | char            **nfs_dup_string_array(char **array);
	    Fixed 2:
	    junction.c: In function 'junction_set_sticky_bit':
	    junction.c:164:39: error: 'ALLPERMS' undeclared (first use in this function)
	      164 |         stb.st_mode &= (unsigned int)~ALLPERMS;
	nfsdcld: don't send null client ids to the kernel
	    It's apparently possible for the sqlite database to get corrupted and
	    cause one or more rows to have null in the id column.
	    The knfsd fix was posted here:
	    https://lore.kernel.org/linux-nfs/20240903111446.659884-1-lilingfeng3@huawei.com/
	    nfsdcld should have a similar fix.  If we encounter a client record with
	    a null id, just skip it instead of sending it to the kernel.

Signed-off-by: Adolf Belka <adolf.belka@ipfire.org>
Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>