From fe0fe83585f88346557868a803a479dfaaa0688a Mon Sep 17 00:00:00 2001 From: Boaz Harrosh Date: Fri, 6 Jan 2012 09:31:20 +0200 Subject: pnfs-obj: Must return layout on IO error As mandated by the standard. In case of an IO error, a pNFS objects layout driver must return it's layout. This is because all device errors are reported to the server as part of the layout return buffer. This is implemented the same way PNFS_LAYOUTRET_ON_SETATTR is done, through a bit flag on the pnfs_layoutdriver_type->flags member. The flag is set by the layout driver that wants a layout_return preformed at pnfs_ld_{write,read}_done in case of an error. (Though I have not defined a wrapper like pnfs_ld_layoutret_on_setattr because this code is never called outside of pnfs.c and pnfs IO paths) Without this patch 3.[0-2] Kernels leak memory and have an annoying WARN_ON after every IO error utilizing the pnfs-obj driver. [This patch is for 3.2 Kernel. 3.1/0 Kernels need a different patch] CC: Stable Tree Signed-off-by: Boaz Harrosh Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 12 ++++++++++++ 1 file changed, 12 insertions(+) (limited to 'fs/nfs/pnfs.c') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 8e672a2b2d69..f881a6387942 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1178,6 +1178,15 @@ void pnfs_ld_write_done(struct nfs_write_data *data) put_lseg(data->lseg); data->lseg = NULL; dprintk("pnfs write error = %d\n", data->pnfs_error); + if (NFS_SERVER(data->inode)->pnfs_curr_ld->flags & + PNFS_LAYOUTRET_ON_ERROR) { + /* Don't lo_commit on error, Server will needs to + * preform a file recovery. + */ + clear_bit(NFS_INO_LAYOUTCOMMIT, + &NFS_I(data->inode)->flags); + pnfs_return_layout(data->inode); + } } data->mds_ops->rpc_release(data); } @@ -1267,6 +1276,9 @@ static void pnfs_ld_handle_read_error(struct nfs_read_data *data) put_lseg(data->lseg); data->lseg = NULL; dprintk("pnfs write error = %d\n", data->pnfs_error); + if (NFS_SERVER(data->inode)->pnfs_curr_ld->flags & + PNFS_LAYOUTRET_ON_ERROR) + pnfs_return_layout(data->inode); nfs_pageio_init_read_mds(&pgio, data->inode); -- cgit From e2fecb215b321db0e4a5b2597349a63c07bec42f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 6 Jan 2012 08:57:46 -0500 Subject: NFS: Remove pNFS bloat from the generic write path We have no business doing any this in the standard write release path. Get rid of it, and put it in the pNFS layer. Also, while we're at it, get rid of the completely bogus unlock/relock semantics that were present in nfs_writeback_release_full(). It is not only unnecessary, but actually dangerous to release the write lock just in order to take it again in nfs_page_async_flush(). Better just to open code the pgio operations in a pnfs helper. Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 30 ++++++++++++++++++++++++++++-- 1 file changed, 28 insertions(+), 2 deletions(-) (limited to 'fs/nfs/pnfs.c') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index f881a6387942..17149a490065 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1166,6 +1166,33 @@ pnfs_generic_pg_test(struct nfs_pageio_descriptor *pgio, struct nfs_page *prev, } EXPORT_SYMBOL_GPL(pnfs_generic_pg_test); +static int pnfs_write_done_resend_to_mds(struct inode *inode, struct list_head *head) +{ + struct nfs_pageio_descriptor pgio; + LIST_HEAD(failed); + + /* Resend all requests through the MDS */ + nfs_pageio_init_write_mds(&pgio, inode, FLUSH_STABLE); + while (!list_empty(head)) { + struct nfs_page *req = nfs_list_entry(head->next); + + nfs_list_remove_request(req); + if (!nfs_pageio_add_request(&pgio, req)) + nfs_list_add_request(req, &failed); + } + nfs_pageio_complete(&pgio); + + if (!list_empty(&failed)) { + /* For some reason our attempt to resend pages. Mark the + * overall send request as having failed, and let + * nfs_writeback_release_full deal with the error. + */ + list_move(&failed, head); + return -EIO; + } + return 0; +} + /* * Called by non rpc-based layout drivers */ @@ -1175,8 +1202,6 @@ void pnfs_ld_write_done(struct nfs_write_data *data) pnfs_set_layoutcommit(data); data->mds_ops->rpc_call_done(&data->task, data); } else { - put_lseg(data->lseg); - data->lseg = NULL; dprintk("pnfs write error = %d\n", data->pnfs_error); if (NFS_SERVER(data->inode)->pnfs_curr_ld->flags & PNFS_LAYOUTRET_ON_ERROR) { @@ -1187,6 +1212,7 @@ void pnfs_ld_write_done(struct nfs_write_data *data) &NFS_I(data->inode)->flags); pnfs_return_layout(data->inode); } + data->task.tk_status = pnfs_write_done_resend_to_mds(data->inode, &data->pages); } data->mds_ops->rpc_release(data); } -- cgit