On Wed, 2011-04-20 at 20:27 +0300, Benny Halevy wrote:
> Non-rpc layout driver such as for objects and blocks
> implement their own I/O path and error handling logic.
> Therefore bypass NFS-based error handling for these layout drivers.
>
> Signed-off-by: Benny Halevy <bhalevy@xxxxxxxxxxx>
> ---
> fs/nfs/internal.h | 2 +
> fs/nfs/nfs4filelayout.c | 1 +
> fs/nfs/nfs4proc.c | 14 +++++++++++-
> fs/nfs/pnfs.c | 48 +++++++++++++++++++++++++++++++++++++++++++++++
> fs/nfs/pnfs.h | 7 +++++-
> include/linux/nfs_xdr.h | 2 +
> 6 files changed, 71 insertions(+), 3 deletions(-)
>
> diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h
> index ce118ce..1914d2f 100644
> --- a/fs/nfs/internal.h
> +++ b/fs/nfs/internal.h
> @@ -310,6 +310,8 @@ extern int nfs_migrate_page(struct address_space *,
> #endif
>
> /* nfs4proc.c */
> +extern void __nfs4_read_done_cb(struct nfs_read_data *);
> +extern void __nfs4_write_done_cb(struct nfs_write_data *);
> extern void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data);
> extern int nfs4_init_client(struct nfs_client *clp,
> const struct rpc_timeout *timeparms,
> diff --git a/fs/nfs/nfs4filelayout.c b/fs/nfs/nfs4filelayout.c
> index 2feab7f..e67a0d4 100644
> --- a/fs/nfs/nfs4filelayout.c
> +++ b/fs/nfs/nfs4filelayout.c
> @@ -859,6 +859,7 @@ static struct pnfs_layoutdriver_type filelayout_type = {
> .id = LAYOUT_NFSV4_1_FILES,
> .name = "LAYOUT_NFSV4_1_FILES",
> .owner = THIS_MODULE,
> + .flags = PNFS_USE_RPC_CODE,
This isn't being used anywhere, so why do I need it in this patch?
> .alloc_lseg = filelayout_alloc_lseg,
> .free_lseg = filelayout_free_lseg,
> .pg_test = filelayout_pg_test,
> diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
> index d0eb50b..cc2cdcd 100644
> --- a/fs/nfs/nfs4proc.c
> +++ b/fs/nfs/nfs4proc.c
> @@ -3149,6 +3149,11 @@ static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
> return err;
> }
>
> +void __nfs4_read_done_cb(struct nfs_read_data *data)
^^^^^^^^^^^^^^^^^^^^^^^^^^^ why the wrapper?
> +{
> + nfs_invalidate_atime(data->inode);
> +}
> +
> static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
> {
> struct nfs_server *server = NFS_SERVER(data->inode);
> @@ -3158,7 +3163,7 @@ static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_read_data *data)
> return -EAGAIN;
> }
>
> - nfs_invalidate_atime(data->inode);
> + __nfs4_read_done_cb(data);
> if (task->tk_status > 0)
> renew_lease(server, data->timestamp);
> return 0;
> @@ -3198,6 +3203,11 @@ void nfs4_reset_read(struct rpc_task *task, struct nfs_read_data *data)
> }
> EXPORT_SYMBOL_GPL(nfs4_reset_read);
>
> +void __nfs4_write_done_cb(struct nfs_write_data *data)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Again, why the wrapper?
> +{
> + nfs_post_op_update_inode_force_wcc(data->inode, data->res.fattr);
> +}
> +
> static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data)
> {
> struct inode *inode = data->inode;
> @@ -3208,7 +3218,7 @@ static int nfs4_write_done_cb(struct rpc_task *task, struct nfs_write_data *data
> }
> if (task->tk_status >= 0) {
> renew_lease(NFS_SERVER(inode), data->timestamp);
> - nfs_post_op_update_inode_force_wcc(inode, data->res.fattr);
> + __nfs4_write_done_cb(data);
> }
> return 0;
> }
> diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
> index a5050d2..18ae397 100644
> --- a/fs/nfs/pnfs.c
> +++ b/fs/nfs/pnfs.c
> @@ -1130,6 +1130,30 @@ pnfs_pageio_init_write(struct nfs_pageio_descriptor *pgio, struct inode *inode)
> pgio->pg_test = (ld && ld->pg_test) ? pnfs_write_pg_test : NULL;
> }
>
> +/*
> + * Called by non rpc-based layout drivers
> + */
> +int
> +pnfs_write_done(struct nfs_write_data *data)
^^^^^^^^^^^^^^^^^^ If this is not generic to all pnfs layout drivers,
then why the apparently generic name?
Why isn't this being introduced together with a driver that actually
uses the functionality? There is no way to review it outside of that
context.
> +{
> + int status;
> +
> + put_lseg(data->lseg);
> + data->lseg = NULL;
> + if (!data->pnfs_error) {
> + __nfs4_write_done_cb(data);
> + data->mds_ops->rpc_call_done(NULL, data);
> + data->mds_ops->rpc_release(data);
> + return 0;
> + }
> +
> + dprintk("%s: pnfs_error=%d, retry via MDS\n", __func__,
> + data->pnfs_error);
> + status = nfs_initiate_write(data, NFS_CLIENT(data->inode), data->mds_ops, NFS_FILE_SYNC);
> + return status ? : -EAGAIN;
> +}
> +EXPORT_SYMBOL_GPL(pnfs_write_done);
> +
> enum pnfs_try_status
> pnfs_try_to_write_data(struct nfs_write_data *wdata,
> const struct rpc_call_ops *call_ops, int how)
> @@ -1155,6 +1179,30 @@ pnfs_try_to_write_data(struct nfs_write_data *wdata,
> }
>
> /*
> + * Called by non rpc-based layout drivers
> + */
> +int
> +pnfs_read_done(struct nfs_read_data *data)
> +{
> + int status;
> +
> + put_lseg(data->lseg);
> + data->lseg = NULL;
> + if (!data->pnfs_error) {
> + __nfs4_read_done_cb(data);
> + data->mds_ops->rpc_call_done(NULL, data);
> + data->mds_ops->rpc_release(data);
> + return 0;
> + }
> +
> + dprintk("%s: pnfs_error=%d, retry via MDS\n", __func__,
> + data->pnfs_error);
> + status = nfs_initiate_read(data, NFS_CLIENT(data->inode), data->mds_ops);
> + return status ? : -EAGAIN;
> +}
> +EXPORT_SYMBOL_GPL(pnfs_read_done);
> +
> +/*
> * Call the appropriate parallel I/O subsystem read function.
> */
> enum pnfs_try_status
> diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h
> index 9f8e970..18b84ce 100644
> --- a/fs/nfs/pnfs.h
> +++ b/fs/nfs/pnfs.h
> @@ -65,8 +65,11 @@ enum {
> };
>
> enum layoutdriver_policy_flags {
> + /* Should the full nfs rpc cleanup code be used after io */
> + PNFS_USE_RPC_CODE = 1 << 0,
> +
> /* Should the pNFS client commit and return the layout upon a setattr */
> - PNFS_LAYOUTRET_ON_SETATTR = 1 << 0,
> + PNFS_LAYOUTRET_ON_SETATTR = 1 << 1,
> };
>
> /* Per-layout driver specific registration structure */
> @@ -182,6 +185,8 @@ bool pnfs_roc_drain(struct inode *ino, u32 *barrier);
> void pnfs_set_layoutcommit(struct nfs_write_data *wdata);
> int pnfs_layoutcommit_inode(struct inode *inode, bool sync);
> int _pnfs_return_layout(struct inode *, struct pnfs_layout_range *, bool wait);
> +int pnfs_write_done(struct nfs_write_data *);
> +int pnfs_read_done(struct nfs_read_data *);
>
> static inline int lo_fail_bit(u32 iomode)
> {
> diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h
> index 01eb1ae..41f896a 100644
> --- a/include/linux/nfs_xdr.h
> +++ b/include/linux/nfs_xdr.h
> @@ -1108,6 +1108,7 @@ struct nfs_read_data {
> const struct rpc_call_ops *mds_ops;
> int (*read_done_cb) (struct rpc_task *task, struct nfs_read_data *data);
> __u64 mds_offset;
> + int pnfs_error;
> struct page *page_array[NFS_PAGEVEC_SIZE];
> };
>
> @@ -1133,6 +1134,7 @@ struct nfs_write_data {
> unsigned long timestamp; /* For lease renewal */
> #endif
> __u64 mds_offset; /* Filelayout dense stripe */
> + int pnfs_error;
> struct page *page_array[NFS_PAGEVEC_SIZE];
> };
>
--
Trond Myklebust
Linux NFS client maintainer
NetApp
Trond.Myklebust@xxxxxxxxxx
www.netapp.com
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
