Date: Fri, 6 Nov 2009 22:29:46 +0000 (UTC) From: Attilio Rao <attilio@FreeBSD.org> To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: svn commit: r199007 - head/sys/fs/fifofs Message-ID: <200911062229.nA6MTld2030248@svn.freebsd.org>
next in thread | raw e-mail | index | archive | help
Author: attilio Date: Fri Nov 6 22:29:46 2009 New Revision: 199007 URL: http://svn.freebsd.org/changeset/base/199007 Log: - Improve comments about locking of the "struct fifoinfo" which is a bit unclear. - Fix a memory leak [0] [0] Diagnosed by: Dorr H. Clark <dclark at engr dot scu dot edu> MFC: 1 week Modified: head/sys/fs/fifofs/fifo_vnops.c Modified: head/sys/fs/fifofs/fifo_vnops.c ============================================================================== --- head/sys/fs/fifofs/fifo_vnops.c Fri Nov 6 22:13:39 2009 (r199006) +++ head/sys/fs/fifofs/fifo_vnops.c Fri Nov 6 22:29:46 2009 (r199007) @@ -78,6 +78,10 @@ struct fileops fifo_ops_f = { /* * This structure is associated with the FIFO vnode and stores * the state associated with the FIFO. + * Notes about locking: + * - fi_readsock and fi_writesock are invariant since init time. + * - fi_readers and fi_writers are vnode lock protected. + * - fi_wgen is fif_mtx lock protected. */ struct fifoinfo { struct socket *fi_readsock; @@ -223,14 +227,9 @@ fail1: } /* - * General access to fi_readers and fi_writers is protected using - * the vnode lock. - * - * Protect the increment of fi_readers and fi_writers and the - * associated calls to wakeup() with the fifo mutex in addition - * to the vnode lock. This allows the vnode lock to be dropped - * for the msleep() calls below, and using the fifo mutex with - * msleep() prevents the wakeup from being missed. + * Use the fifo_mtx lock here, in addition to the vnode lock, + * in order to allow vnode lock dropping before msleep() calls + * and still avoiding missed wakeups. */ mtx_lock(&fifo_mtx); if (ap->a_mode & FREAD) { @@ -249,6 +248,8 @@ fail1: if (ap->a_mode & FWRITE) { if ((ap->a_mode & O_NONBLOCK) && fip->fi_readers == 0) { mtx_unlock(&fifo_mtx); + if (fip->fi_writers == 0) + fifo_cleanup(vp); return (ENXIO); } fip->fi_writers++;
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200911062229.nA6MTld2030248>