From a19b4785d9eec9b7915a3944f279d73db80ee8f8 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 3 Aug 2020 18:32:28 -0400 Subject: NFS: Report the stateid + status in trace_nfs4_layoutreturn_on_close() Ensure we correctly report the stateid and status in the layoutreturn on close tracepoint. Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/nfs/pnfs.c') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index dd2e14f5875d..d8cdb94c6668 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1549,12 +1549,12 @@ void pnfs_roc_release(struct nfs4_layoutreturn_args *args, default: arg_stateid = &args->stateid; } + trace_nfs4_layoutreturn_on_close(args->inode, &args->stateid, ret); pnfs_layoutreturn_free_lsegs(lo, arg_stateid, &args->range, res_stateid); if (ld_private && ld_private->ops && ld_private->ops->free) ld_private->ops->free(ld_private); pnfs_put_layout_hdr(lo); - trace_nfs4_layoutreturn_on_close(args->inode, 0); } bool pnfs_wait_on_layoutreturn(struct inode *ino, struct rpc_task *task) -- cgit From ff041727e9e029845857cac41aae118ead5e261b Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 4 Aug 2020 16:30:30 -0400 Subject: NFS: Don't move layouts to plh_return_segs list while in use If the layout segment is still in use for a read or a write, we should not move it to the layout plh_return_segs list. If we do, we can end up returning the layout while I/O is still in progress. Fixes: e0b7d420f72a ("pNFS: Don't discard layout segments that are marked for return") Cc: stable@vger.kernel.org # v4.19+ Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 12 +----------- 1 file changed, 1 insertion(+), 11 deletions(-) (limited to 'fs/nfs/pnfs.c') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index d8cdb94c6668..262ce01c7abe 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -2392,16 +2392,6 @@ out_forget: return ERR_PTR(-EAGAIN); } -static int -mark_lseg_invalid_or_return(struct pnfs_layout_segment *lseg, - struct list_head *tmp_list) -{ - if (!mark_lseg_invalid(lseg, tmp_list)) - return 0; - pnfs_cache_lseg_for_layoutreturn(lseg->pls_layout, lseg); - return 1; -} - /** * pnfs_mark_matching_lsegs_return - Free or return matching layout segments * @lo: pointer to layout header @@ -2438,7 +2428,7 @@ pnfs_mark_matching_lsegs_return(struct pnfs_layout_hdr *lo, lseg, lseg->pls_range.iomode, lseg->pls_range.offset, lseg->pls_range.length); - if (mark_lseg_invalid_or_return(lseg, tmp_list)) + if (mark_lseg_invalid(lseg, tmp_list)) continue; remaining++; set_bit(NFS_LSEG_LAYOUTRETURN, &lseg->pls_flags); -- cgit From d474f96104bd4377573526ebae2ee212205a6839 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 5 Aug 2020 09:03:56 -0400 Subject: NFS: Don't return layout segments that are in use If the NFS_LAYOUT_RETURN_REQUESTED flag is set, we want to return the layout as soon as possible, meaning that the affected layout segments should be marked as invalid, and should no longer be in use for I/O. Fixes: f0b429819b5f ("pNFS: Ignore non-recalled layouts in pnfs_layout_need_return()") Cc: stable@vger.kernel.org # v4.19+ Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 34 +++++++++++++++------------------- 1 file changed, 15 insertions(+), 19 deletions(-) (limited to 'fs/nfs/pnfs.c') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 262ce01c7abe..b5baf36d4de5 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1226,31 +1226,27 @@ out: return status; } +static bool +pnfs_layout_segments_returnable(struct pnfs_layout_hdr *lo, + enum pnfs_iomode iomode, + u32 seq) +{ + struct pnfs_layout_range recall_range = { + .length = NFS4_MAX_UINT64, + .iomode = iomode, + }; + return pnfs_mark_matching_lsegs_return(lo, &lo->plh_return_segs, + &recall_range, seq) != -EBUSY; +} + /* Return true if layoutreturn is needed */ static bool pnfs_layout_need_return(struct pnfs_layout_hdr *lo) { - struct pnfs_layout_segment *s; - enum pnfs_iomode iomode; - u32 seq; - if (!test_bit(NFS_LAYOUT_RETURN_REQUESTED, &lo->plh_flags)) return false; - - seq = lo->plh_return_seq; - iomode = lo->plh_return_iomode; - - /* Defer layoutreturn until all recalled lsegs are done */ - list_for_each_entry(s, &lo->plh_segs, pls_list) { - if (seq && pnfs_seqid_is_newer(s->pls_seq, seq)) - continue; - if (iomode != IOMODE_ANY && s->pls_range.iomode != iomode) - continue; - if (test_bit(NFS_LSEG_LAYOUTRETURN, &s->pls_flags)) - return false; - } - - return true; + return pnfs_layout_segments_returnable(lo, lo->plh_return_iomode, + lo->plh_return_seq); } static void pnfs_layoutreturn_before_put_layout_hdr(struct pnfs_layout_hdr *lo) -- cgit From 563c53e73b8b6ec842828736f77e633f7b0911e9 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 11 Aug 2020 13:36:32 -0400 Subject: NFS: Fix flexfiles read failover The current mirrored read failover code is correctly resetting the mirror index between failed reads, however it is not able to actually flip the RPC call over to the next RPC client. The end result is that we keep resending the RPC call to the same client over and over. The fix is to use the pnfs_read_resend_pnfs() mechanism to schedule a new RPC call, but we need to add the ability to pass in a mirror index so that we always retry the next mirror in the list. Fixes: 166bd5b889ac ("pNFS/flexfiles: Fix layoutstats handling during read failovers") Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'fs/nfs/pnfs.c') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index b5baf36d4de5..40332c758d84 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -2939,7 +2939,8 @@ pnfs_try_to_read_data(struct nfs_pgio_header *hdr, } /* Resend all requests through pnfs. */ -void pnfs_read_resend_pnfs(struct nfs_pgio_header *hdr) +void pnfs_read_resend_pnfs(struct nfs_pgio_header *hdr, + unsigned int mirror_idx) { struct nfs_pageio_descriptor pgio; @@ -2950,6 +2951,7 @@ void pnfs_read_resend_pnfs(struct nfs_pgio_header *hdr) nfs_pageio_init_read(&pgio, hdr->inode, false, hdr->completion_ops); + pgio.pg_mirror_idx = mirror_idx; hdr->task.tk_status = nfs_pageio_resend(&pgio, hdr); } } -- cgit