From: Jeff Layton <jlayton@redhat.com> Date: Mon, 26 Jul 2010 11:15:55 -0400 Subject: [fs] nfs: fix file create failure with HPUX client Message-id: <1280142955-8734-1-git-send-email-jlayton@redhat.com> Patchwork-id: 27091 O-Subject: [RHEL5.6 PATCH] BZ#605720: nfsd(v2/v3): fix the failure of creation from HPUX client Bugzilla: 605720 RH-Acked-by: Steve Dickson <SteveD@redhat.com> From: wengang wang <wen.gang.wang@oracle.com> I'm not aware of any customers having reported this problem, but Oracle is apparently carrying a patch for OEL to fix this problem. The problem is apparently that the HPUX NFS client sets the mode of the file oddly on create. nfsd then attempts to do a follow on setattr to truncate the file, but that fails on some filesystems because of the mode. This patch fixes this by making the server skip the truncate if we're setting the size to zero and it was just created by nfsd. I don't have a client that behaves like this, but I've tested this patch from RHEL clients and it seems to be OK. The OEL patch for this seems to be based on an earlier version submitted by Wengang, but this one more closely matches what finally went upstream. Original patch description follows: -------------------------------[snip]---------------------------- sometimes HPUX nfs client sends a create request to linux nfs server(v2/v3). the dump of the request is like: obj_attributes mode: value follows set_it: value follows (1) mode: 00 uid: no value set_it: no value (0) gid: value follows set_it: value follows (1) gid: 8030 size: value follows set_it: value follows (1) size: 0 atime: don't change set_it: don't change (0) mtime: don't change set_it: don't change (0) note that mode is 00(havs no rwx privilege even for the owner) and it requires to set size to 0. as current nfsd(v2/v3) implementation, the server does mainly 2 steps: 1) creates the file in mode specified by calling vfs_create(). 2) sets attributes for the file by calling nfsd_setattr(). at step 2), it finally calls file system specific setattr() function which may fail when checking permission because changing size needs WRITE privilege but it has none since mode is 000. for this case, a new file created, we may simply ignore the request of setting size to 0, so that WRITE privilege is not needed and the open succeeds. Signed-off-by: Wengang Wang <wen.gang.wang@oracle.com> -- vfs.c | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+) Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu> diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c index 9f636e2..bc4c754 100644 --- a/fs/nfsd/vfs.c +++ b/fs/nfsd/vfs.c @@ -1115,6 +1115,21 @@ nfsd_commit(struct svc_rqst *rqstp, struct svc_fh *fhp, } #endif /* CONFIG_NFSD_V3 */ +/* HPUX client sometimes creates a file in mode 000, and sets size to 0. + * setting size to 0 may fail for some specific file systems by the permission + * checking which requires WRITE permission but the mode is 000. + * we ignore the resizing(to 0) on the just new created file, since the size is + * 0 after file created. + * + * call this only after vfs_create() is called. + * */ +static void +nfsd_check_ignore_resizing(struct iattr *iap) +{ + if ((iap->ia_valid & ATTR_SIZE) && (iap->ia_size == 0)) + iap->ia_valid &= ~ATTR_SIZE; +} + /* * Create a file (regular, directory, device, fifo); UNIX sockets * not yet implemented. @@ -1197,6 +1212,8 @@ nfsd_create(struct svc_rqst *rqstp, struct svc_fh *fhp, switch (type) { case S_IFREG: err = vfs_create(dirp, dchild, iap->ia_mode, NULL); + if (!err) + nfsd_check_ignore_resizing(iap); break; case S_IFDIR: err = vfs_mkdir(dirp, dchild, iap->ia_mode); @@ -1350,6 +1367,8 @@ nfsd_create_v3(struct svc_rqst *rqstp, struct svc_fh *fhp, /* setattr will sync the child (or not) */ } + nfsd_check_ignore_resizing(iap); + if (createmode == NFS3_CREATE_EXCLUSIVE) { /* Cram the verifier into atime/mtime */ iap->ia_valid = ATTR_MTIME|ATTR_ATIME