[PATCH v11 27/36] netfs: Build a list of regions undergoing writeback

[David Howells <dhowells@redhat.com>]

Build a linked list of the regions of file position that are involved in a
writeback whilst processing the folios to be written back.  These will be
used in a subsequent patch to work out which folios need unlocking rather
than walking the bvecq list of folios.

This will mean that the folio list need not be maintained as a single list
of folios, but can instead be made into multiple lists and have various
roundings applied to it - and can be ignored by the collector, except when
various pins are pulled out of it by subrequest cleanup.

Signed-off-by: David Howells <dhowells@redhat.com>
cc: Paulo Alcantara <pc@manguebit.org>
cc: Matthew Wilcox <willy@infradead.org>
cc: Christoph Hellwig <hch@infradead.org>
cc: netfs@lists.linux.dev
cc: linux-fsdevel@vger.kernel.org
---
 fs/netfs/internal.h    |  1 +
 fs/netfs/main.c        |  7 +++++++
 fs/netfs/objects.c     |  7 +++++++
 fs/netfs/write_issue.c | 20 ++++++++++++++++++++
 include/linux/netfs.h  | 12 ++++++++++++
 5 files changed, 47 insertions(+)

diff --git a/fs/netfs/internal.h b/fs/netfs/internal.h
index 8d91748ae257..e6e768ad9648 100644
--- a/fs/netfs/internal.h
+++ b/fs/netfs/internal.h
@@ -44,6 +44,7 @@ extern struct list_head netfs_io_requests;
 extern spinlock_t netfs_proc_lock;
 extern mempool_t netfs_request_pool;
 extern mempool_t netfs_subrequest_pool;
+extern mempool_t netfs_writeback_pool;
 extern mempool_t netfs_bvecq_pool;
 
 #ifdef CONFIG_PROC_FS
diff --git a/fs/netfs/main.c b/fs/netfs/main.c
index 5d8b87f71888..b8da5e85cc67 100644
--- a/fs/netfs/main.c
+++ b/fs/netfs/main.c
@@ -28,6 +28,7 @@ static struct kmem_cache *netfs_request_slab;
 static struct kmem_cache *netfs_subrequest_slab;
 mempool_t netfs_request_pool;
 mempool_t netfs_subrequest_pool;
+mempool_t netfs_writeback_pool;
 mempool_t netfs_bvecq_pool;
 
 #ifdef CONFIG_PROC_FS
@@ -110,6 +111,9 @@ static int __init netfs_init(void)
 
 	if (mempool_init_kmalloc_pool(&netfs_bvecq_pool, 100, BVECQ_STD_SIZE) < 0)
 		goto error_bvecq_pool;
+	if (mempool_init_kmalloc_pool(&netfs_writeback_pool, 100,
+				      sizeof(struct netfs_writeback)) < 0)
+		goto error_writeback_pool;
 
 	netfs_request_slab = kmem_cache_create("netfs_request",
 					       sizeof(struct netfs_io_request), 0,
@@ -163,6 +167,8 @@ static int __init netfs_init(void)
 error_reqpool:
 	kmem_cache_destroy(netfs_request_slab);
 error_req:
+	mempool_exit(&netfs_writeback_pool);
+error_writeback_pool:
 	mempool_exit(&netfs_bvecq_pool);
 error_bvecq_pool:
 	return ret;
@@ -177,6 +183,7 @@ static void __exit netfs_exit(void)
 	kmem_cache_destroy(netfs_subrequest_slab);
 	mempool_exit(&netfs_request_pool);
 	kmem_cache_destroy(netfs_request_slab);
+	mempool_exit(&netfs_writeback_pool);
 	mempool_exit(&netfs_bvecq_pool);
 }
 module_exit(netfs_exit);
diff --git a/fs/netfs/objects.c b/fs/netfs/objects.c
index 740971955198..763be168d6b1 100644
--- a/fs/netfs/objects.c
+++ b/fs/netfs/objects.c
@@ -151,6 +151,13 @@ static void netfs_deinit_request(struct netfs_io_request *rreq)
 	bvecq_pos_unset(&rreq->dispatch_cursor);
 	bvecq_pos_unset(&rreq->collect_cursor);
 	bvecq_put(rreq->spare);
+	while (rreq->writebacks) {
+		struct netfs_writeback *wback = rreq->writebacks;
+
+		rreq->writebacks = wback->next;
+		mempool_free(wback, &netfs_bvecq_pool);
+		
+	}
 
 	if (atomic_dec_and_test(&ictx->io_count))
 		wake_up_var(&ictx->io_count);
diff --git a/fs/netfs/write_issue.c b/fs/netfs/write_issue.c
index 60a417026e10..ace790b127ca 100644
--- a/fs/netfs/write_issue.c
+++ b/fs/netfs/write_issue.c
@@ -335,6 +335,7 @@ static int netfs_write_folio(struct netfs_io_request *wreq,
 	struct netfs_io_stream *upload = &wreq->io_streams[0];
 	struct netfs_io_stream *cache  = &wreq->io_streams[1];
 	struct netfs_io_stream *stream;
+	struct netfs_writeback *wback;
 	struct netfs_group *fgroup; /* TODO: Use this with ceph */
 	struct netfs_folio *finfo;
 	struct bvecq *queue = wreq->load_cursor.bvecq;
@@ -433,6 +434,25 @@ static int netfs_write_folio(struct netfs_io_request *wreq,
 	folio_start_writeback(folio);
 	folio_unlock(folio);
 
+	/* Keep track of what we will need to unlock. */
+	wback = wreq->writebacks_tail;
+	if (!wback || fpos != wback->start + wback->len || wback->len > LONG_MAX) {
+		wback = mempool_alloc(&netfs_writeback_pool, wreq->gfp);
+		wback->next = NULL;
+		wback->start = fpos;
+		wback->len = fsize;
+
+		if (wreq->writebacks)
+			/* Order write of next after last write of len in old tail. */
+			smp_store_release(&wreq->writebacks_tail->next, wback);
+		else
+			wreq->writebacks = wback;
+		wreq->writebacks_tail = wback;
+	} else {
+		/* Order update of len after setting pointer. */
+		smp_store_release(&wback->len, wback->len + fsize);
+	}
+ 
 	if (fgroup == NETFS_FOLIO_COPY_TO_CACHE) {
 		if (!cache->avail) {
 			trace_netfs_folio(folio, netfs_folio_trace_cancel_copy);
diff --git a/include/linux/netfs.h b/include/linux/netfs.h
index 8d5f548a77c3..70eb32f073f8 100644
--- a/include/linux/netfs.h
+++ b/include/linux/netfs.h
@@ -134,6 +134,16 @@ enum netfs_cache_collect {
 	NETFS_CACHE_COLLECT_WRITE_CANCEL,	/* Currently collecting cancelled writes */
 };
 
+/*
+ * Record of a contiguous region undergoing writeback.  The tail region (ie. if
+ * next is NULL) may be extended dynamically.
+ */
+struct netfs_writeback {
+	struct netfs_writeback	*next;		/* Next extent in list */
+	uoff_t			start;		/* Start position */
+	size_t			len;		/* Total size (can increase) */
+};
+
 /*
  * Stream of I/O subrequests going to a particular destination, such as the
  * server or the local cache.  This is mainly intended for writing where we may
@@ -249,6 +259,8 @@ struct netfs_io_request {
 #endif
 	struct netfs_io_stream	io_streams[2];	/* Streams of parallel I/O operations */
 #define NR_IO_STREAMS 2 //wreq->nr_io_streams
+	struct netfs_writeback	*writebacks;	/* List of regions undergoing writeback */
+	struct netfs_writeback	*writebacks_tail; /* Tail of region list */
 	struct netfs_group	*group;		/* Writeback group being written back */
 	struct bvecq		*spare;		/* Advance allocation of bvecq */
 	struct bvecq_pos	load_cursor;	/* Point at which new folios are loaded in */