Hi Greg,
I just got a report that the proposed fix has a bug in one of the error paths. I am trying to fix that, so you might want to wait with the backport.
Sorry about that, Tigran.
----- Original Message -----
From: "Greg Kroah-Hartman" gregkh@linuxfoundation.org To: "stable" stable@vger.kernel.org Cc: "Greg Kroah-Hartman" gregkh@linuxfoundation.org, patches@lists.linux.dev, "Tigran Mkrtchyan" tigran.mkrtchyan@desy.de, "Trond Myklebust" trond.myklebust@hammerspace.com, "Sasha Levin" sashal@kernel.org Sent: Tuesday, 26 August, 2025 13:06:18 Subject: [PATCH 5.10 168/523] pNFS/flexfiles: dont attempt pnfs on fatal DS errors
5.10-stable review patch. If anyone has any objections, please let me know.
From: Tigran Mkrtchyan tigran.mkrtchyan@desy.de
[ Upstream commit f06bedfa62d57f7b67d44aacd6badad2e13a803f ]
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 doneThe 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: 260f32adb88d ("pNFS/flexfiles: Check the result of nfs4_pnfs_ds_connect") Signed-off-by: Trond Myklebust trond.myklebust@hammerspace.com Signed-off-by: Sasha Levin sashal@kernel.org
fs/nfs/flexfilelayout/flexfilelayout.c | 26 ++++++++++++++--------- fs/nfs/flexfilelayout/flexfilelayoutdev.c | 6 +++--- 2 files changed, 19 insertions(+), 13 deletions(-)
diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index a053dd05057f..57150b27c0fd 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -739,14 +739,14 @@ 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 &&
@@ -754,10 +754,10 @@ ff_layout_choose_ds_for_read(struct pnfs_layout_segment *lseg, continue;
*best_idx = idx;
return ds;
break;
- return NULL;
- return ds;
}
static struct nfs4_pnfs_ds * @@ -933,7 +933,7 @@ 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);@@ -1820,6 +1820,7 @@ 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,
@@ -1827,8 +1828,10 @@ ff_layout_read_pagelist(struct nfs_pgio_header *hdr)
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);
@@ -1869,7 +1872,7 @@ 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,
@@ -1890,11 +1893,14 @@ 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);
@@ -1937,7 +1943,7 @@ 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,
@@ -1979,7 +1985,7 @@ 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,
diff --git a/fs/nfs/flexfilelayout/flexfilelayoutdev.c b/fs/nfs/flexfilelayout/flexfilelayoutdev.c index 4b0cdddce6eb..11777d33a85e 100644 --- a/fs/nfs/flexfilelayout/flexfilelayoutdev.c +++ b/fs/nfs/flexfilelayout/flexfilelayoutdev.c @@ -368,11 +368,11 @@ 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;
@@ -410,7 +410,7 @@ 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; } -- 2.39.5