From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andreas Ehmanns Date: Mon, 14 Dec 2015 20:24:10 +0100 Subject: [Buildroot] Add config to nfsd to set number of threads In-Reply-To: <20151210201744.GC3597@free.fr> References: <5669D3BF.2070207@gmx.de> <20151210201744.GC3597@free.fr> Message-ID: <566F175A.5080006@gmx.de> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net Hi Yann, Am 10.12.2015 um 21:17 schrieb Yann E. MORIN: > Andreas, All, > > On 2015-12-10 20:34 +0100, Andreas Ehmanns spake thusly: >> Dear All, >> actually the nfs-utils package provides an NFS server together with an init >> script. The number of threads the NFS server shall use is hard-coded in the >> init script. The attached patch moves this parameter to a config file under >> /etc which is read by the init script. > The idea is OK with me. However, I have a few comments... > >> Additionally the return value of the init script in case of an not existing >> file has been changed from 0 to 2 to indicate an error condition. > We don't care about the value of the exit status of a SysV init script. > Whether it returns 0, 1, 2 or whatever else is completely ignored by > /sbin/init . > >> From f1eae31bb357587c3491a1b45891dda8c6c0dbff Mon Sep 17 00:00:00 2001 >> From: Andreas Ehmanns >> Date: Thu, 10 Dec 2015 15:51:23 +0100 >> Subject: [PATCH 1/1] Make number of nfsd threads configurable >> >> Signed-off-by: Andreas Ehmanns >> --- >> package/nfs-utils/S60nfs | 13 ++++++++----- >> package/nfs-utils/nfs-utils.mk | 2 ++ >> package/nfs-utils/nfsd.conf | 8 ++++++++ >> 3 files changed, 18 insertions(+), 5 deletions(-) >> create mode 100644 package/nfs-utils/nfsd.conf >> >> diff --git a/package/nfs-utils/S60nfs b/package/nfs-utils/S60nfs >> index ec7c909..ddc472e 100755 >> --- a/package/nfs-utils/S60nfs >> +++ b/package/nfs-utils/S60nfs >> @@ -3,10 +3,13 @@ >> # nfs This shell script takes care of starting and stopping >> # the NFS services. Stolen from RedHat FC5. >> >> -[ -x /usr/sbin/rpc.statd ] || exit 0 >> -[ -x /usr/sbin/rpc.nfsd ] || exit 0 >> -[ -x /usr/sbin/rpc.mountd ] || exit 0 >> -[ -x /usr/sbin/exportfs ] || exit 0 >> +[ -x /usr/sbin/rpc.statd ] || exit 2 >> +[ -x /usr/sbin/rpc.nfsd ] || exit 2 >> +[ -x /usr/sbin/rpc.mountd ] || exit 2 >> +[ -x /usr/sbin/exportfs ] || exit 2 > Do we really need to check for those files? nfs-utils has no config > option for any of those progs, only for: rpc.lockd and rpc.rquotad, but > we're not using them in this init script. > > This means that we can assume that all the utilities above *are* > installed when the init script is. > >> +[ -r /etc/nfsd.conf ] || exit 2 >> +. /etc/nfsd.conf > Two comments about that: > > - move the file to /etc/default/nfsd > - do not make it mandatory and just provide defaults > - use a variable to store the path > - no need to prefix the variable with 'NFSD' > > (yes, that's four comments! ;-) ) > > So, what about: > > CFG_FILE=/etc/default/nfsd > > NR_THREADS=2 > if [ -f "${CFG_FILE}" ]; then > . "${CFG_FILE}" > fi > >> mkdir -p /var/lock/subsys >> mkdir -p /run/nfs/sm >> @@ -25,7 +28,7 @@ start() { >> echo "done" >> >> printf "Starting NFS daemon: " >> - rpc.nfsd 2 >> + rpc.nfsd $NFSD_NR_OF_THREADS > I have a preference for using ${foo} when expanding variables, so; > > rpc.nfsd ${NR_THREADS} > >> echo "done" >> >> printf "Starting NFS mountd: " >> diff --git a/package/nfs-utils/nfs-utils.mk b/package/nfs-utils/nfs-utils.mk >> index 30f12fd..a26c626 100644 >> --- a/package/nfs-utils/nfs-utils.mk >> +++ b/package/nfs-utils/nfs-utils.mk >> @@ -51,6 +51,8 @@ endif >> define NFS_UTILS_INSTALL_INIT_SYSV >> $(INSTALL) -D -m 0755 package/nfs-utils/S60nfs \ >> $(TARGET_DIR)/etc/init.d/S60nfs >> + $(INSTALL) -m 644 package/nfs-utils/nfsd.conf \ >> + $(TARGET_DIR)/etc/nfsd.conf >> endef >> >> define NFS_UTILS_INSTALL_INIT_SYSTEMD >> diff --git a/package/nfs-utils/nfsd.conf b/package/nfs-utils/nfsd.conf >> new file mode 100644 > [--SNIP--] > > That file is no longer needed now that there is a default in the script. > Just let the user provide one via a rootfs overlay or a post-install > script. > > Care to resubmit a proper patch with the above fixed, please? Thanks! :-) > > Regards, > Yann E. MORIN. > sounds good to me. I will prepare a patch and send via git this week. Regards, Andreas