Return-Path: Received: from rcsinet15.oracle.com ([148.87.113.117]:40072 "EHLO rcsinet15.oracle.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756414Ab1H3UBJ convert rfc822-to-8bit (ORCPT ); Tue, 30 Aug 2011 16:01:09 -0400 Subject: Re: [PATCH] NFSD: Add a cache for fs_locations information Content-Type: text/plain; charset=us-ascii From: Chuck Lever In-Reply-To: <20110830151845.4dec5b40@tlielax.poochiereds.net> Date: Tue, 30 Aug 2011 16:00:52 -0400 Cc: "J. Bruce Fields" , Trond Myklebust , Linux NFS Mailing List Message-Id: <852F14A9-7826-47A0-BCD6-D5D46E38A39C@oracle.com> References: <20110829185115.3413.92459.stgit@matisse.1015granger.net> <20110830151845.4dec5b40@tlielax.poochiereds.net> To: Jeff Layton Sender: linux-nfs-owner@vger.kernel.org List-ID: MIME-Version: 1.0 On Aug 30, 2011, at 3:18 PM, Jeff Layton wrote: > On Mon, 29 Aug 2011 14:51:15 -0400 > Chuck Lever wrote: > >> From: Trond Myklebust >> >> Signed-off-by: Trond Myklebust >> [ cel: since this is server-side, use nfsd4_ prefix instead of nfs4_ prefix. ] >> Signed-off-by: Chuck Lever >> --- >> >> fs/nfsd/nfsd.h | 7 +++++++ >> fs/nfsd/vfs.c | 15 +++++++++++++++ >> 2 files changed, 22 insertions(+), 0 deletions(-) >> >> diff --git a/fs/nfsd/nfsd.h b/fs/nfsd/nfsd.h >> index 7ecfa24..d314812 100644 >> --- a/fs/nfsd/nfsd.h >> +++ b/fs/nfsd/nfsd.h >> @@ -335,6 +335,13 @@ static inline u32 nfsd_suppattrs2(u32 minorversion) >> #define NFSD_SUPPATTR_EXCLCREAT_WORD2 \ >> NFSD_WRITEABLE_ATTRS_WORD2 >> >> +extern int nfsd4_is_junction(struct dentry *dentry); >> +#else >> +static inline int nfsd4_is_junction(struct dentry *dentry) >> +{ >> + return 0; >> +} >> + >> #endif /* CONFIG_NFSD_V4 */ >> >> #endif /* LINUX_NFSD_NFSD_H */ >> diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c >> index fd0acca..5dac667 100644 >> --- a/fs/nfsd/vfs.c >> +++ b/fs/nfsd/vfs.c >> @@ -168,6 +168,8 @@ int nfsd_mountpoint(struct dentry *dentry, struct svc_export *exp) >> { >> if (d_mountpoint(dentry)) >> return 1; >> + if (nfsd4_is_junction(dentry)) >> + return 1; >> if (!(exp->ex_flags & NFSEXP_V4ROOT)) >> return 0; >> return dentry->d_inode != NULL; >> @@ -592,6 +594,19 @@ nfsd4_get_nfs4_acl(struct svc_rqst *rqstp, struct dentry *dentry, struct nfs4_ac >> return error; >> } >> >> +#define NFSD_XATTR_JUNCTION_PREFIX XATTR_TRUSTED_PREFIX "junction." >> +#define NFSD_XATTR_JUNCTION_TYPE NFSD_XATTR_JUNCTION_PREFIX "type" >> +int nfsd4_is_junction(struct dentry *dentry) >> +{ >> + ssize_t ret; >> + >> + if (dentry->d_inode != NULL) { >> + ret = vfs_getxattr(dentry, NFSD_XATTR_JUNCTION_TYPE, NULL, 0); >> + if (ret > 0) >> + return 1; >> + } >> + return 0; >> +} > > Won't the above check be rather expensive? You'll need to do a > getxattr call on almost every path component of every lookup, right? > > I may be misremembering your talk from connectathon, but I thought you > were planning to use a well-known mode for junctions that would cut > down on the number of unnecessary getxattrs... Yes, that's the plan. To reduce overhead, the S_ISVTX bit must be set before NFSD does the expensive xattr test. However, I don't think this kind of filtering was ever implemented. I got the patch from here: http://git.linux-nfs.org/?p=trondmy/nfs-2.6.git;a=shortlog;h=refs/heads/fedfs-for-2.6.34 and that doesn't seem to have it either. I can implement something and repost these. >> #endif /* defined(CONFIG_NFSD_V4) */ >> >> #ifdef CONFIG_NFSD_V3 >> >> -- >> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in >> the body of a message to majordomo@vger.kernel.org >> More majordomo info at http://vger.kernel.org/majordomo-info.html > > > -- > Jeff Layton -- Chuck Lever chuck[dot]lever[at]oracle[dot]com