The branch main has been updated by rmacklem:
URL:
https://cgit.FreeBSD.org/src/commit/?id=24af0fcdfc4983fd3cef10f9d949aca79476c2c2
commit 24af0fcdfc4983fd3cef10f9d949aca79476c2c2
Author: Rick Macklem <rmack...@freebsd.org>
AuthorDate: 2021-10-13 22:48:54 +0000
Commit: Rick Macklem <rmack...@freebsd.org>
CommitDate: 2021-10-13 22:48:54 +0000
nfscl: Make nfscl_getlayout() acquire the correct pNFS layout
Without this patch, if a pNFS read layout has already been acquired
for a file, writes would be redirected to the Metadata Server (MDS),
because nfscl_getlayout() would not acquire a read/write layout for
the file. This happened because there was no "mode" argument to
nfscl_getlayout() to indicate whether reading or writing was being done.
Since doing I/O through the Metadata Server is not encouraged for some
pNFS servers, it is preferable to get a read/write layout for writes
instead of redirecting the write to the MDS.
This patch adds a access mode argument to nfscl_getlayout() and
nfsrpc_getlayout(), so that nfscl_getlayout() knows to acquire a read/write
layout for writing, even if a read layout has already been acquired.
This patch only affects NFSv4.1/4.2 client behaviour when pNFS ("pnfs" mount
option against a server that supports pNFS) is in use.
This problem was detected during a recent NFSv4 interoperability
testing event held by the IETF working group.
MFC after: 2 week
---
sys/fs/nfs/nfs_var.h | 2 +-
sys/fs/nfsclient/nfs_clrpcops.c | 14 +++++++-------
sys/fs/nfsclient/nfs_clstate.c | 7 ++++---
3 files changed, 12 insertions(+), 11 deletions(-)
diff --git a/sys/fs/nfs/nfs_var.h b/sys/fs/nfs/nfs_var.h
index 459860078c30..c5d164bd1b93 100644
--- a/sys/fs/nfs/nfs_var.h
+++ b/sys/fs/nfs/nfs_var.h
@@ -637,7 +637,7 @@ int nfscl_layout(struct nfsmount *, vnode_t, u_int8_t *,
int, nfsv4stateid_t *,
int, int, struct nfsclflayouthead *, struct nfscllayout **, struct ucred *,
NFSPROC_T *);
struct nfscllayout *nfscl_getlayout(struct nfsclclient *, uint8_t *, int,
- uint64_t, struct nfsclflayout **, int *);
+ uint64_t, uint32_t, struct nfsclflayout **, int *);
void nfscl_dserr(uint32_t, uint32_t, struct nfscldevinfo *,
struct nfscllayout *, struct nfsclds *);
void nfscl_cancelreqs(struct nfsclds *);
diff --git a/sys/fs/nfsclient/nfs_clrpcops.c b/sys/fs/nfsclient/nfs_clrpcops.c
index 66e093505ca3..b550ae832ff4 100644
--- a/sys/fs/nfsclient/nfs_clrpcops.c
+++ b/sys/fs/nfsclient/nfs_clrpcops.c
@@ -150,7 +150,7 @@ static int nfsrpc_locku(struct nfsrv_descript *, struct
nfsmount *,
static int nfsrpc_setaclrpc(vnode_t, struct ucred *, NFSPROC_T *,
struct acl *, nfsv4stateid_t *, void *);
static int nfsrpc_getlayout(struct nfsmount *, vnode_t, struct nfsfh *, int,
- uint32_t *, nfsv4stateid_t *, uint64_t, struct nfscllayout **,
+ uint32_t, uint32_t *, nfsv4stateid_t *, uint64_t, struct nfscllayout **,
struct ucred *, NFSPROC_T *);
static int nfsrpc_fillsa(struct nfsmount *, struct sockaddr_in *,
struct sockaddr_in6 *, sa_family_t, int, int, struct nfsclds **,
@@ -5684,8 +5684,8 @@ nfsmout:
*/
static int
nfsrpc_getlayout(struct nfsmount *nmp, vnode_t vp, struct nfsfh *nfhp,
- int iomode, uint32_t *notifybitsp, nfsv4stateid_t *stateidp, uint64_t off,
- struct nfscllayout **lypp, struct ucred *cred, NFSPROC_T *p)
+ int iomode, uint32_t rw, uint32_t *notifybitsp, nfsv4stateid_t *stateidp,
+ uint64_t off, struct nfscllayout **lypp, struct ucred *cred, NFSPROC_T *p)
{
struct nfscllayout *lyp;
struct nfsclflayout *flp;
@@ -5705,7 +5705,7 @@ nfsrpc_getlayout(struct nfsmount *nmp, vnode_t vp, struct
nfsfh *nfhp,
* flp == NULL.
*/
lyp = nfscl_getlayout(nmp->nm_clp, nfhp->nfh_fh, nfhp->nfh_len,
- off, &flp, &recalled);
+ off, rw, &flp, &recalled);
islocked = 0;
if (lyp == NULL || flp == NULL) {
if (recalled != 0)
@@ -6036,7 +6036,7 @@ nfscl_doiods(vnode_t vp, struct uio *uiop, int *iomode,
int *must_commit,
/* Search for a layout for this file. */
off = uiop->uio_offset;
layp = nfscl_getlayout(nmp->nm_clp, np->n_fhp->nfh_fh,
- np->n_fhp->nfh_len, off, &rflp, &recalled);
+ np->n_fhp->nfh_len, off, rwaccess, &rflp, &recalled);
if (layp == NULL || rflp == NULL) {
if (recalled != 0) {
NFSFREECRED(newcred);
@@ -6054,7 +6054,7 @@ nfscl_doiods(vnode_t vp, struct uio *uiop, int *iomode,
int *must_commit,
else
iolaymode = NFSLAYOUTIOMODE_READ;
error = nfsrpc_getlayout(nmp, vp, np->n_fhp, iolaymode,
- NULL, &stateid, off, &layp, newcred, p);
+ rwaccess, NULL, &stateid, off, &layp, newcred, p);
if (error != 0) {
NFSLOCKNODE(np);
np->n_flag |= NNOLAYOUT;
@@ -7740,7 +7740,7 @@ nfsrpc_getopenlayout(struct nfsmount *nmp, vnode_t vp,
u_int8_t *nfhp,
* on it, iff flp != NULL or a lock (exclusive lock) on it iff
* flp == NULL.
*/
- lyp = nfscl_getlayout(nmp->nm_clp, newfhp, newfhlen, 0, &flp,
+ lyp = nfscl_getlayout(nmp->nm_clp, newfhp, newfhlen, 0, mode, &flp,
&recalled);
NFSCL_DEBUG(4, "nfsrpc_getopenlayout nfscl_getlayout lyp=%p\n", lyp);
if (lyp == NULL)
diff --git a/sys/fs/nfsclient/nfs_clstate.c b/sys/fs/nfsclient/nfs_clstate.c
index 9a7a879fffd3..8fca7734c197 100644
--- a/sys/fs/nfsclient/nfs_clstate.c
+++ b/sys/fs/nfsclient/nfs_clstate.c
@@ -5141,7 +5141,8 @@ nfscl_layout(struct nfsmount *nmp, vnode_t vp, u_int8_t
*fhp, int fhlen,
*/
struct nfscllayout *
nfscl_getlayout(struct nfsclclient *clp, uint8_t *fhp, int fhlen,
- uint64_t off, struct nfsclflayout **retflpp, int *recalledp)
+ uint64_t off, uint32_t rwaccess, struct nfsclflayout **retflpp,
+ int *recalledp)
{
struct nfscllayout *lyp;
mount_t mp;
@@ -5157,8 +5158,8 @@ nfscl_getlayout(struct nfsclclient *clp, uint8_t *fhp,
int fhlen,
TAILQ_REMOVE(&clp->nfsc_layout, lyp, nfsly_list);
TAILQ_INSERT_HEAD(&clp->nfsc_layout, lyp, nfsly_list);
lyp->nfsly_timestamp = NFSD_MONOSEC + 120;
- error = nfscl_findlayoutforio(lyp, off,
- NFSV4OPEN_ACCESSREAD, retflpp);
+ error = nfscl_findlayoutforio(lyp, off, rwaccess,
+ retflpp);
if (error == 0)
nfsv4_getref(&lyp->nfsly_lock, NULL,
NFSCLSTATEMUTEXPTR, mp);
