Date: Sun, 17 Apr 2011 02:44:52 +0000 (UTC) From: Rick Macklem <rmacklem@FreeBSD.org> To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: svn commit: r220735 - head/sys/fs/nfsclient Message-ID: <201104170244.p3H2iqUX024548@svn.freebsd.org>
next in thread | raw e-mail | index | archive | help
Author: rmacklem Date: Sun Apr 17 02:44:51 2011 New Revision: 220735 URL: http://svn.freebsd.org/changeset/base/220735 Log: Fix readdirplus in the experimental NFS client so that it skips over ".." to avoid a LOR race with nfs_lookup(). This fix is analagous to r138256 in the regular NFS client. MFC after: 2 weeks Modified: head/sys/fs/nfsclient/nfs_clrpcops.c Modified: head/sys/fs/nfsclient/nfs_clrpcops.c ============================================================================== --- head/sys/fs/nfsclient/nfs_clrpcops.c Sat Apr 16 23:38:35 2011 (r220734) +++ head/sys/fs/nfsclient/nfs_clrpcops.c Sun Apr 17 02:44:51 2011 (r220735) @@ -2942,7 +2942,7 @@ nfsrpc_readdirplus(vnode_t vp, struct ui nfsquad_t cookie, ncookie; int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1; int attrflag, tryformoredirs = 1, eof = 0, gotmnton = 0; - int unlocknewvp = 0; + int isdotdot = 0, unlocknewvp = 0; long dotfileid, dotdotfileid = 0, fileno = 0; char *cp; nfsattrbit_t attrbits, dattrbits; @@ -3192,6 +3192,11 @@ nfsrpc_readdirplus(vnode_t vp, struct ui *cp = '\0'; cp += tlen; /* points to cookie storage */ tl2 = (u_int32_t *)cp; + if (len == 2 && cnp->cn_nameptr[0] == '.' && + cnp->cn_nameptr[1] == '.') + isdotdot = 1; + else + isdotdot = 0; uio_iov_base_add(uiop, (tlen + NFSX_HYPER)); uio_iov_len_add(uiop, -(tlen + NFSX_HYPER)); uio_uio_resid_add(uiop, -(tlen + NFSX_HYPER)); @@ -3269,6 +3274,22 @@ nfsrpc_readdirplus(vnode_t vp, struct ui unlocknewvp = 0; FREE((caddr_t)nfhp, M_NFSFH); np = dnp; + } else if (isdotdot != 0) { + /* + * Skip doing a nfscl_nget() call for "..". + * There's a race between acquiring the nfs + * node here and lookups that look for the + * directory being read (in the parent). + * It would try to get a lock on ".." here, + * owning the lock on the directory being + * read. Lookup will hold the lock on ".." + * and try to acquire the lock on the + * directory being read. + * If the directory is unlocked/relocked, + * then there is a LOR with the buflock + * vp is relocked. + */ + free(nfhp, M_NFSFH); } else { error = nfscl_nget(vnode_mount(vp), vp, nfhp, cnp, p, &np, NULL, LK_EXCLUSIVE);
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201104170244.p3H2iqUX024548>