From 67bbceedc9bb8ad48993a8bd6486054756d711f4 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 4 Jan 2021 13:35:46 -0500 Subject: pNFS: Mark layout for return if return-on-close was not sent If the layout return-on-close failed because the layoutreturn was never sent, then we should mark the layout for return again. Fixes: 9c47b18cf722 ("pNFS: Ensure we do clear the return-on-close layout stateid on fatal errors") Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'fs/nfs/pnfs.c') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 07f59dc8cb2e..ccc89fab1802 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1560,12 +1560,18 @@ void pnfs_roc_release(struct nfs4_layoutreturn_args *args, int ret) { struct pnfs_layout_hdr *lo = args->layout; + struct inode *inode = args->inode; const nfs4_stateid *arg_stateid = NULL; const nfs4_stateid *res_stateid = NULL; struct nfs4_xdr_opaque_data *ld_private = args->ld_private; switch (ret) { case -NFS4ERR_NOMATCHING_LAYOUT: + spin_lock(&inode->i_lock); + if (pnfs_layout_is_valid(lo) && + nfs4_stateid_match_other(&args->stateid, &lo->plh_stateid)) + pnfs_set_plh_return_info(lo, args->range.iomode, 0); + spin_unlock(&inode->i_lock); break; case 0: if (res->lrs_present) -- cgit v1.2.3 From 078000d02d57f02dde61de4901f289672e98c8bc Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 4 Jan 2021 13:18:03 -0500 Subject: pNFS: We want return-on-close to complete when evicting the inode If the inode is being evicted, it should be safe to run return-on-close, so we should do it to ensure we don't inadvertently leak layout segments. Fixes: 1c5bd76d17cc ("pNFS: Enable layoutreturn operation for return-on-close") Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 26 ++++++++++---------------- fs/nfs/pnfs.c | 8 +++----- fs/nfs/pnfs.h | 8 +++----- 3 files changed, 16 insertions(+), 26 deletions(-) (limited to 'fs/nfs/pnfs.c') diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 14acd2f79107..2f4679a62712 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -3536,10 +3536,8 @@ static void nfs4_close_done(struct rpc_task *task, void *data) trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status); /* Handle Layoutreturn errors */ - if (pnfs_roc_done(task, calldata->inode, - &calldata->arg.lr_args, - &calldata->res.lr_res, - &calldata->res.lr_ret) == -EAGAIN) + if (pnfs_roc_done(task, &calldata->arg.lr_args, &calldata->res.lr_res, + &calldata->res.lr_ret) == -EAGAIN) goto out_restart; /* hmm. we are done with the inode, and in the process of freeing @@ -6384,10 +6382,8 @@ static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata) trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status); /* Handle Layoutreturn errors */ - if (pnfs_roc_done(task, data->inode, - &data->args.lr_args, - &data->res.lr_res, - &data->res.lr_ret) == -EAGAIN) + if (pnfs_roc_done(task, &data->args.lr_args, &data->res.lr_res, + &data->res.lr_ret) == -EAGAIN) goto out_restart; switch (task->tk_status) { @@ -6441,10 +6437,10 @@ static void nfs4_delegreturn_release(void *calldata) struct nfs4_delegreturndata *data = calldata; struct inode *inode = data->inode; + if (data->lr.roc) + pnfs_roc_release(&data->lr.arg, &data->lr.res, + data->res.lr_ret); if (inode) { - if (data->lr.roc) - pnfs_roc_release(&data->lr.arg, &data->lr.res, - data->res.lr_ret); nfs_post_op_update_inode_force_wcc(inode, &data->fattr); nfs_iput_and_deactive(inode); } @@ -6520,16 +6516,14 @@ static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, nfs_fattr_init(data->res.fattr); data->timestamp = jiffies; data->rpc_status = 0; - data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, cred); data->inode = nfs_igrab_and_active(inode); - if (data->inode) { + if (data->inode || issync) { + data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res, + cred); if (data->lr.roc) { data->args.lr_args = &data->lr.arg; data->res.lr_res = &data->lr.res; } - } else if (data->lr.roc) { - pnfs_roc_release(&data->lr.arg, &data->lr.res, 0); - data->lr.roc = false; } task_setup_data.callback_data = data; diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index ccc89fab1802..a18b1992b2fb 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1509,10 +1509,8 @@ out_noroc: return false; } -int pnfs_roc_done(struct rpc_task *task, struct inode *inode, - struct nfs4_layoutreturn_args **argpp, - struct nfs4_layoutreturn_res **respp, - int *ret) +int pnfs_roc_done(struct rpc_task *task, struct nfs4_layoutreturn_args **argpp, + struct nfs4_layoutreturn_res **respp, int *ret) { struct nfs4_layoutreturn_args *arg = *argpp; int retval = -EAGAIN; @@ -1545,7 +1543,7 @@ int pnfs_roc_done(struct rpc_task *task, struct inode *inode, return 0; case -NFS4ERR_OLD_STATEID: if (!nfs4_layout_refresh_old_stateid(&arg->stateid, - &arg->range, inode)) + &arg->range, arg->inode)) break; *ret = -NFS4ERR_NOMATCHING_LAYOUT; return -EAGAIN; diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h index bbd3de1025f2..d810ae674f4e 100644 --- a/fs/nfs/pnfs.h +++ b/fs/nfs/pnfs.h @@ -297,10 +297,8 @@ bool pnfs_roc(struct inode *ino, struct nfs4_layoutreturn_args *args, struct nfs4_layoutreturn_res *res, const struct cred *cred); -int pnfs_roc_done(struct rpc_task *task, struct inode *inode, - struct nfs4_layoutreturn_args **argpp, - struct nfs4_layoutreturn_res **respp, - int *ret); +int pnfs_roc_done(struct rpc_task *task, struct nfs4_layoutreturn_args **argpp, + struct nfs4_layoutreturn_res **respp, int *ret); void pnfs_roc_release(struct nfs4_layoutreturn_args *args, struct nfs4_layoutreturn_res *res, int ret); @@ -772,7 +770,7 @@ pnfs_roc(struct inode *ino, } static inline int -pnfs_roc_done(struct rpc_task *task, struct inode *inode, +pnfs_roc_done(struct rpc_task *task, struct nfs4_layoutreturn_args **argpp, struct nfs4_layoutreturn_res **respp, int *ret) -- cgit v1.2.3 From c18d1e17ba2f6a1c9257b0b5d2882a6e3f772673 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 4 Jan 2021 15:01:18 -0500 Subject: pNFS: Clean up pnfs_layoutreturn_free_lsegs() Remove the check for whether or not the stateid is NULL, and fix up the callers. Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-) (limited to 'fs/nfs/pnfs.c') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index a18b1992b2fb..16a37214aba9 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1152,7 +1152,7 @@ void pnfs_layoutreturn_free_lsegs(struct pnfs_layout_hdr *lo, LIST_HEAD(freeme); spin_lock(&inode->i_lock); - if (!pnfs_layout_is_valid(lo) || !arg_stateid || + if (!pnfs_layout_is_valid(lo) || !nfs4_stateid_match_other(&lo->plh_stateid, arg_stateid)) goto out_unlock; if (stateid) { @@ -1559,7 +1559,6 @@ void pnfs_roc_release(struct nfs4_layoutreturn_args *args, { struct pnfs_layout_hdr *lo = args->layout; struct inode *inode = args->inode; - const nfs4_stateid *arg_stateid = NULL; const nfs4_stateid *res_stateid = NULL; struct nfs4_xdr_opaque_data *ld_private = args->ld_private; @@ -1569,6 +1568,7 @@ void pnfs_roc_release(struct nfs4_layoutreturn_args *args, if (pnfs_layout_is_valid(lo) && nfs4_stateid_match_other(&args->stateid, &lo->plh_stateid)) pnfs_set_plh_return_info(lo, args->range.iomode, 0); + pnfs_clear_layoutreturn_waitbit(lo); spin_unlock(&inode->i_lock); break; case 0: @@ -1576,11 +1576,10 @@ void pnfs_roc_release(struct nfs4_layoutreturn_args *args, res_stateid = &res->stateid; fallthrough; default: - arg_stateid = &args->stateid; + pnfs_layoutreturn_free_lsegs(lo, &args->stateid, &args->range, + res_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); -- cgit v1.2.3 From 2c8d5fc37fe2384a9bdb6965443ab9224d46f704 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 5 Jan 2021 06:43:45 -0500 Subject: pNFS: Stricter ordering of layoutget and layoutreturn If a layout return is in progress, we should wait for it to complete, in case the layout segment we are picking up gets returned too. Fixes: 30cb3ee299cb ("pNFS: Handle NFS4ERR_OLD_STATEID on layoutreturn by bumping the state seqid") Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 43 +++++++++++++++++++++---------------------- 1 file changed, 21 insertions(+), 22 deletions(-) (limited to 'fs/nfs/pnfs.c') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 16a37214aba9..fc13a3c8bc48 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -2018,6 +2018,27 @@ lookup_again: goto lookup_again; } + /* + * Because we free lsegs when sending LAYOUTRETURN, we need to wait + * for LAYOUTRETURN. + */ + if (test_bit(NFS_LAYOUT_RETURN, &lo->plh_flags)) { + spin_unlock(&ino->i_lock); + dprintk("%s wait for layoutreturn\n", __func__); + lseg = ERR_PTR(pnfs_prepare_to_retry_layoutget(lo)); + if (!IS_ERR(lseg)) { + pnfs_put_layout_hdr(lo); + dprintk("%s retrying\n", __func__); + trace_pnfs_update_layout(ino, pos, count, iomode, lo, + lseg, + PNFS_UPDATE_LAYOUT_RETRY); + goto lookup_again; + } + trace_pnfs_update_layout(ino, pos, count, iomode, lo, lseg, + PNFS_UPDATE_LAYOUT_RETURN); + goto out_put_layout_hdr; + } + lseg = pnfs_find_lseg(lo, &arg, strict_iomode); if (lseg) { trace_pnfs_update_layout(ino, pos, count, iomode, lo, lseg, @@ -2070,28 +2091,6 @@ lookup_again: nfs4_stateid_copy(&stateid, &lo->plh_stateid); } - /* - * Because we free lsegs before sending LAYOUTRETURN, we need to wait - * for LAYOUTRETURN even if first is true. - */ - if (test_bit(NFS_LAYOUT_RETURN, &lo->plh_flags)) { - spin_unlock(&ino->i_lock); - dprintk("%s wait for layoutreturn\n", __func__); - lseg = ERR_PTR(pnfs_prepare_to_retry_layoutget(lo)); - if (!IS_ERR(lseg)) { - if (first) - pnfs_clear_first_layoutget(lo); - pnfs_put_layout_hdr(lo); - dprintk("%s retrying\n", __func__); - trace_pnfs_update_layout(ino, pos, count, iomode, lo, - lseg, PNFS_UPDATE_LAYOUT_RETRY); - goto lookup_again; - } - trace_pnfs_update_layout(ino, pos, count, iomode, lo, lseg, - PNFS_UPDATE_LAYOUT_RETURN); - goto out_put_layout_hdr; - } - if (pnfs_layoutgets_blocked(lo)) { trace_pnfs_update_layout(ino, pos, count, iomode, lo, lseg, PNFS_UPDATE_LAYOUT_BLOCKED); -- cgit v1.2.3 From cb2856c5971723910a86b7d1d0cf623d6919cbc4 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 6 Jan 2021 14:13:22 -0500 Subject: NFS/pNFS: Fix a leak of the layout 'plh_outstanding' counter If we exit _lgopen_prepare_attached() without setting a layout, we will currently leak the plh_outstanding counter. Fixes: 411ae722d10a ("pNFS: Wait for stale layoutget calls to complete in pnfs_update_layout()") Signed-off-by: Trond Myklebust --- fs/nfs/pnfs.c | 1 + 1 file changed, 1 insertion(+) (limited to 'fs/nfs/pnfs.c') diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index fc13a3c8bc48..4f274f21c4ab 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -2244,6 +2244,7 @@ static void _lgopen_prepare_attached(struct nfs4_opendata *data, &rng, GFP_KERNEL); if (!lgp) { pnfs_clear_first_layoutget(lo); + nfs_layoutget_end(lo); pnfs_put_layout_hdr(lo); return; } -- cgit v1.2.3