NFS: Charge unstable writes by request size, not folio size

nfs_folio_mark_unstable() and nfs_folio_clear_commit() charge and
uncharge NR_WRITEBACK/WB_WRITEBACK by folio_nr_pages(folio) once per
*request* added to or removed from a commit list. This is correct only
when a folio has a single associated request. When pg_test splits a
folio into N sub-folio requests (e.g. pNFS flexfiles striping with a
stripe unit smaller than the folio size, or plain wsize-limited
splitting), each of the N requests independently charges the whole
folio's page count, inflating the accounting by a factor of N per
folio. With large folios and small stripe units this reaches multiple
orders of magnitude: a 2 MiB folio split into 512 4 KiB requests can
charge up to 512x its real size, pushing global dirty+writeback
accounting past the system's dirty threshold and forcing every
buffered writer on the host into the hard-throttle path, including
unrelated in-kernel NFS server threads sharing the box.

Charge each request only for the pages it actually covers.

Fixes: 0c493b5cf16e ("NFS: Convert buffered writes to use folios")
Cc: stable@vger.kernel.org
Signed-off-by: Benjamin Coddington <bcodding@hammerspace.com>
Assisted-By: Claude Sonnet 5 <noreply@anthropic.com>
Signed-off-by: Anna Schumaker <anna.schumaker@hammerspace.com>

diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h
index acaeff7ddfdf70f4d6126566b248c19d55f08302..e4533f5836321d8634a3604a6e0d10cb3b11d3aa 100644 (file)
--- a/fs/nfs/internal.h
+++ b/fs/nfs/internal.h
@@ -851,17 +851,19 @@ void nfs_super_set_maxbytes(struct super_block *sb, __u64 maxfilesize)
 }
 
 /*
- * Record the page as unstable (an extra writeback period) and mark its
- * inode as dirty.
+ * Record the request's range as unstable (an extra writeback period) and
+ * mark its inode as dirty.
  */
-static inline void nfs_folio_mark_unstable(struct folio *folio,
+static inline void nfs_folio_mark_unstable(struct nfs_page *req,
                                           struct nfs_commit_info *cinfo)
 {
+       struct folio *folio = nfs_page_to_folio(req);
+
        if (folio && !cinfo->dreq) {
                struct inode *inode = folio->mapping->host;
-               long nr = folio_nr_pages(folio);
+               long nr = DIV_ROUND_UP(req->wb_bytes, PAGE_SIZE);
 
-               /* This page is really still in write-back - just that the
+               /* This range is really still in write-back - just that the
                 * writeback is happening on the server now.
                 */
                node_stat_mod_folio(folio, NR_WRITEBACK, nr);

diff --git a/fs/nfs/pnfs_nfs.c b/fs/nfs/pnfs_nfs.c
index 0ff43dbcb7cd77aeab5fe0048182cce6cf724199..648c95b78eeab73ab4481e2531ed6f303879b34c 100644 (file)
--- a/fs/nfs/pnfs_nfs.c
+++ b/fs/nfs/pnfs_nfs.c
@@ -1199,7 +1199,7 @@ pnfs_layout_mark_request_commit(struct nfs_page *req,
 
        nfs_request_add_commit_list_locked(req, list, cinfo);
        mutex_unlock(&NFS_I(cinfo->inode)->commit_mutex);
-       nfs_folio_mark_unstable(nfs_page_to_folio(req), cinfo);
+       nfs_folio_mark_unstable(req, cinfo);
        return;
 out_resched:
        mutex_unlock(&NFS_I(cinfo->inode)->commit_mutex);

diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index fcffb8c9e9df70d318d0fbd7136f1bd060cf11d1..d2b03ceaeb4f195fe165bd7ec794e16799c7041a 100644 (file)
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -807,7 +807,7 @@ nfs_request_add_commit_list(struct nfs_page *req, struct nfs_commit_info *cinfo)
        mutex_lock(&NFS_I(cinfo->inode)->commit_mutex);
        nfs_request_add_commit_list_locked(req, &cinfo->mds->list, cinfo);
        mutex_unlock(&NFS_I(cinfo->inode)->commit_mutex);
-       nfs_folio_mark_unstable(nfs_page_to_folio(req), cinfo);
+       nfs_folio_mark_unstable(req, cinfo);
 }
 EXPORT_SYMBOL_GPL(nfs_request_add_commit_list);
 
@@ -866,10 +866,12 @@ nfs_mark_request_commit(struct nfs_page *req, struct pnfs_layout_segment *lseg,
        nfs_request_add_commit_list(req, cinfo);
 }
 
-static void nfs_folio_clear_commit(struct folio *folio)
+static void nfs_folio_clear_commit(struct nfs_page *req)
 {
+       struct folio *folio = nfs_page_to_folio(req);
+
        if (folio) {
-               long nr = folio_nr_pages(folio);
+               long nr = DIV_ROUND_UP(req->wb_bytes, PAGE_SIZE);
 
                node_stat_mod_folio(folio, NR_WRITEBACK, -nr);
                bdi_wb_stat_mod(folio->mapping->host, WB_WRITEBACK, -nr);
@@ -889,7 +891,7 @@ static void nfs_clear_request_commit(struct nfs_commit_info *cinfo,
                        nfs_request_remove_commit_list(req, cinfo);
                }
                mutex_unlock(&NFS_I(inode)->commit_mutex);
-               nfs_folio_clear_commit(nfs_page_to_folio(req));
+               nfs_folio_clear_commit(req);
        }
 }
 
@@ -1741,7 +1743,7 @@ void nfs_retry_commit(struct list_head *page_list,
                req = nfs_list_entry(page_list->next);
                nfs_list_remove_request(req);
                nfs_mark_request_commit(req, lseg, cinfo, ds_commit_idx);
-               nfs_folio_clear_commit(nfs_page_to_folio(req));
+               nfs_folio_clear_commit(req);
                nfs_unlock_and_release_request(req);
        }
 }
@@ -1813,7 +1815,7 @@ static void nfs_commit_release_pages(struct nfs_commit_data *data)
                req = nfs_list_entry(data->pages.next);
                nfs_list_remove_request(req);
                folio = nfs_page_to_folio(req);
-               nfs_folio_clear_commit(folio);
+               nfs_folio_clear_commit(req);
 
                dprintk("NFS:       commit (%s/%llu %d@%lld)",
                        nfs_req_openctx(req)->dentry->d_sb->s_id,