Merge: pNFS/flexfiles: don't attempt pnfs on fatal DS errors

MR: https://gitlab.com/redhat/rhel/src/kernel/rhel-10/-/merge_requests/287

JIRA: https://issues.redhat.com/browse/RHEL-110294

commit f06bedf
Author: Tigran Mkrtchyan <tigran.mkrtchyan@desy.de>
Date:   Fri Jun 27 09:17:51 2025 +0200

    pNFS/flexfiles: don't attempt pnfs on fatal DS errors

    When an applications get killed (SIGTERM/SIGINT) while pNFS client performs a connection
    to DS, client ends in an infinite loop of connect-disconnect. This
    source of the issue, it that flexfilelayoutdev#nfs4_ff_layout_prepare_ds gets an error
    on nfs4_pnfs_ds_connect with status ERESTARTSYS, which is set by rpc_signal_task, but
    the error is treated as transient, thus retried.

    The issue is reproducible with Ctrl+C the following script(there should be ~1000 files in
    a directory, client should must not have any connections to DSes):

    ```
    echo 3 > /proc/sys/vm/drop_caches

    for i in *
    do
       head -1 $i
    done
    ```

    The change aims to propagate the nfs4_ff_layout_prepare_ds error state
    to the caller that can decide whatever this is a retryable error or not.

    Signed-off-by: Tigran Mkrtchyan <tigran.mkrtchyan@desy.de>
    Link: https://lore.kernel.org/r/20250627071751.189663-1-tigran.mkrtchyan@desy.de
    Fixes: 260f32a ("pNFS/flexfiles: Check the result of nfs4_pnfs_ds_connect")
    Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>

Signed-off-by: Benjamin Coddington <bcodding@redhat.com>

Approved-by: Paulo Alcantara <paalcant@redhat.com>
Approved-by: Scott Mayhew <smayhew@redhat.com>
Approved-by: Olga Kornievskaia <okorniev@redhat.com>
Approved-by: CKI KWF Bot <cki-ci-bot+kwf-gitlab-com@redhat.com>

Merged-by: Julio Faracco <jfaracco@redhat.com>

Author: jcfaracco

fs/nfs/flexfilelayout/flexfilelayout.c

@@ -762,25 +762,28 @@ ff_layout_choose_ds_for_read(struct pnfs_layout_segment *lseg,
 {
 	struct nfs4_ff_layout_segment *fls = FF_LAYOUT_LSEG(lseg);
 	struct nfs4_ff_layout_mirror *mirror;
-	struct nfs4_pnfs_ds *ds;
+	struct nfs4_pnfs_ds *ds = ERR_PTR(-EAGAIN);
 	u32 idx;
 
 	/* mirrors are initially sorted by efficiency */
 	for (idx = start_idx; idx < fls->mirror_array_cnt; idx++) {
 		mirror = FF_LAYOUT_COMP(lseg, idx);
 		ds = nfs4_ff_layout_prepare_ds(lseg, mirror, false);
-		if (!ds)
+		if (IS_ERR(ds))
 			continue;
 
 		if (check_device &&
-		    nfs4_test_deviceid_unavailable(&mirror->mirror_ds->id_node))
+		    nfs4_test_deviceid_unavailable(&mirror->mirror_ds->id_node)) {
+			// reinitialize the error state in case if this is the last iteration
+			ds = ERR_PTR(-EINVAL);
 			continue;
+		}
 
 		*best_idx = idx;
-		return ds;
+		break;
 	}
 
-	return NULL;
+	return ds;
 }
 
 static struct nfs4_pnfs_ds *
@@ -804,7 +807,7 @@ ff_layout_choose_best_ds_for_read(struct pnfs_layout_segment *lseg,
 	struct nfs4_pnfs_ds *ds;
 
 	ds = ff_layout_choose_valid_ds_for_read(lseg, start_idx, best_idx);
-	if (ds)
+	if (!IS_ERR(ds))
 		return ds;
 	return ff_layout_choose_any_ds_for_read(lseg, start_idx, best_idx);
 }
@@ -818,7 +821,7 @@ ff_layout_get_ds_for_read(struct nfs_pageio_descriptor *pgio,
 
 	ds = ff_layout_choose_best_ds_for_read(lseg, pgio->pg_mirror_idx,
 					       best_idx);
-	if (ds || !pgio->pg_mirror_idx)
+	if (!IS_ERR(ds) || !pgio->pg_mirror_idx)
 		return ds;
 	return ff_layout_choose_best_ds_for_read(lseg, 0, best_idx);
 }
@@ -868,7 +871,7 @@ ff_layout_pg_init_read(struct nfs_pageio_descriptor *pgio,
 	req->wb_nio = 0;
 
 	ds = ff_layout_get_ds_for_read(pgio, &ds_idx);
-	if (!ds) {
+	if (IS_ERR(ds)) {
 		if (!ff_layout_no_fallback_to_mds(pgio->pg_lseg))
 			goto out_mds;
 		pnfs_generic_pg_cleanup(pgio);
@@ -942,7 +945,7 @@ ff_layout_pg_init_write(struct nfs_pageio_descriptor *pgio,
 	for (i = 0; i < pgio->pg_mirror_count; i++) {
 		mirror = FF_LAYOUT_COMP(pgio->pg_lseg, i);
 		ds = nfs4_ff_layout_prepare_ds(pgio->pg_lseg, mirror, true);
-		if (!ds) {
+		if (IS_ERR(ds)) {
 			if (!ff_layout_no_fallback_to_mds(pgio->pg_lseg))
 				goto out_mds;
 			pnfs_generic_pg_cleanup(pgio);
@@ -1072,11 +1075,13 @@ static void ff_layout_resend_pnfs_read(struct nfs_pgio_header *hdr)
 {
 	u32 idx = hdr->pgio_mirror_idx + 1;
 	u32 new_idx = 0;
+	struct nfs4_pnfs_ds *ds;
 
-	if (ff_layout_choose_any_ds_for_read(hdr->lseg, idx, &new_idx))
-		ff_layout_send_layouterror(hdr->lseg);
-	else
+	ds = ff_layout_choose_any_ds_for_read(hdr->lseg, idx, &new_idx);
+	if (IS_ERR(ds))
 		pnfs_error_mark_layout_for_return(hdr->inode, hdr->lseg);
+	else
+		ff_layout_send_layouterror(hdr->lseg);
 	pnfs_read_resend_pnfs(hdr, new_idx);
 }
 
@@ -1867,15 +1872,18 @@ ff_layout_read_pagelist(struct nfs_pgio_header *hdr)
 	u32 idx = hdr->pgio_mirror_idx;
 	int vers;
 	struct nfs_fh *fh;
+	bool ds_fatal_error = false;
 
 	dprintk("--> %s ino %lu pgbase %u req %zu@%llu\n",
 		__func__, hdr->inode->i_ino,
 		hdr->args.pgbase, (size_t)hdr->args.count, offset);
 
 	mirror = FF_LAYOUT_COMP(lseg, idx);
 	ds = nfs4_ff_layout_prepare_ds(lseg, mirror, false);
-	if (!ds)
+	if (IS_ERR(ds)) {
+		ds_fatal_error = nfs_error_is_fatal(PTR_ERR(ds));
 		goto out_failed;
+	}
 
 	ds_clnt = nfs4_ff_find_or_create_ds_client(mirror, ds->ds_clp,
 						   hdr->inode);
@@ -1923,7 +1931,7 @@ ff_layout_read_pagelist(struct nfs_pgio_header *hdr)
 	return PNFS_ATTEMPTED;
 
 out_failed:
-	if (ff_layout_avoid_mds_available_ds(lseg))
+	if (ff_layout_avoid_mds_available_ds(lseg) && !ds_fatal_error)
 		return PNFS_TRY_AGAIN;
 	trace_pnfs_mds_fallback_read_pagelist(hdr->inode,
 			hdr->args.offset, hdr->args.count,
@@ -1945,11 +1953,14 @@ ff_layout_write_pagelist(struct nfs_pgio_header *hdr, int sync)
 	int vers;
 	struct nfs_fh *fh;
 	u32 idx = hdr->pgio_mirror_idx;
+	bool ds_fatal_error = false;
 
 	mirror = FF_LAYOUT_COMP(lseg, idx);
 	ds = nfs4_ff_layout_prepare_ds(lseg, mirror, true);
-	if (!ds)
+	if (IS_ERR(ds)) {
+		ds_fatal_error = nfs_error_is_fatal(PTR_ERR(ds));
 		goto out_failed;
+	}
 
 	ds_clnt = nfs4_ff_find_or_create_ds_client(mirror, ds->ds_clp,
 						   hdr->inode);
@@ -2000,7 +2011,7 @@ ff_layout_write_pagelist(struct nfs_pgio_header *hdr, int sync)
 	return PNFS_ATTEMPTED;
 
 out_failed:
-	if (ff_layout_avoid_mds_available_ds(lseg))
+	if (ff_layout_avoid_mds_available_ds(lseg) && !ds_fatal_error)
 		return PNFS_TRY_AGAIN;
 	trace_pnfs_mds_fallback_write_pagelist(hdr->inode,
 			hdr->args.offset, hdr->args.count,
@@ -2043,7 +2054,7 @@ static int ff_layout_initiate_commit(struct nfs_commit_data *data, int how)
 	idx = calc_ds_index_from_commit(lseg, data->ds_commit_index);
 	mirror = FF_LAYOUT_COMP(lseg, idx);
 	ds = nfs4_ff_layout_prepare_ds(lseg, mirror, true);
-	if (!ds)
+	if (IS_ERR(ds))
 		goto out_err;
 
 	ds_clnt = nfs4_ff_find_or_create_ds_client(mirror, ds->ds_clp,

fs/nfs/flexfilelayout/flexfilelayoutdev.c

@@ -370,11 +370,11 @@ nfs4_ff_layout_prepare_ds(struct pnfs_layout_segment *lseg,
 			  struct nfs4_ff_layout_mirror *mirror,
 			  bool fail_return)
 {
-	struct nfs4_pnfs_ds *ds = NULL;
+	struct nfs4_pnfs_ds *ds;
 	struct inode *ino = lseg->pls_layout->plh_inode;
 	struct nfs_server *s = NFS_SERVER(ino);
 	unsigned int max_payload;
-	int status;
+	int status = -EAGAIN;
 
 	if (!ff_layout_init_mirror_ds(lseg->pls_layout, mirror))
 		goto noconnect;
@@ -418,7 +418,7 @@ nfs4_ff_layout_prepare_ds(struct pnfs_layout_segment *lseg,
 	ff_layout_send_layouterror(lseg);
 	if (fail_return || !ff_layout_has_available_ds(lseg))
 		pnfs_error_mark_layout_for_return(ino, lseg);
-	ds = NULL;
+	ds = ERR_PTR(status);
 out:
 	return ds;
 }