summaryrefslogtreecommitdiffstats
path: root/fs
diff options
context:
space:
mode:
authorDavid Howells <dhowells@redhat.com>2021-06-29 23:31:48 +0200
committerDavid Howells <dhowells@redhat.com>2023-12-28 10:45:22 +0100
commit0e0f2dfe880fb19e4b15a7ca468623eb0b4ba586 (patch)
tree3b80b9d42d124c68d32c352c6140fdc788bc1d37 /fs
parentnetfs: Prep to use folio->private for write grouping and streaming write (diff)
downloadlinux-0e0f2dfe880fb19e4b15a7ca468623eb0b4ba586.tar.xz
linux-0e0f2dfe880fb19e4b15a7ca468623eb0b4ba586.zip
netfs: Dispatch write requests to process a writeback slice
Dispatch one or more write reqeusts to process a writeback slice, where a slice is tailored more to logical block divisions within the file (such as crypto blocks, an object layout or cache granules) than the protocol RPC maximum capacity. The dispatch doesn't happen until throttling allows, at which point the entire writeback slice is processed and queued. A slice may be written to multiple destinations (one or more servers and the local cache) and the writes to each destination might be split up along different lines. The writeback slice holds the required folios pinned. An iov_iter is provided in netfs_write_request that describes the buffer to be used. This may be part of the pagecache, may have auxiliary padding pages attached or may be a bounce buffer resulting from crypto or compression. Consequently, the filesystem must not twiddle the folio markings directly. The following API is available to the filesystem: (1) The ->create_write_requests() method is called to ask the filesystem to create the requests it needs. This is passed the writeback slice to be processed. (2) The filesystem should then call netfs_create_write_request() to create the requests it needs. (3) Once a request is initialised, netfs_queue_write_request() can be called to dispatch it asynchronously, if not completed immediately. (4) netfs_write_request_completed() should be called to note the completion of a request. (5) netfs_get_write_request() and netfs_put_write_request() are provided to refcount a request. These take constants from the netfs_wreq_trace enum for logging into ftrace. (6) The ->free_write_request is method is called to ask the filesystem to clean up a request. Signed-off-by: David Howells <dhowells@redhat.com> Reviewed-by: Jeff Layton <jlayton@kernel.org> cc: linux-cachefs@redhat.com cc: linux-fsdevel@vger.kernel.org cc: linux-mm@kvack.org
Diffstat (limited to 'fs')
-rw-r--r--fs/netfs/Makefile3
-rw-r--r--fs/netfs/internal.h6
-rw-r--r--fs/netfs/output.c363
3 files changed, 371 insertions, 1 deletions
diff --git a/fs/netfs/Makefile b/fs/netfs/Makefile
index cf3fc847b8ac..c69c6775b8ac 100644
--- a/fs/netfs/Makefile
+++ b/fs/netfs/Makefile
@@ -7,7 +7,8 @@ netfs-y := \
locking.o \
main.o \
misc.o \
- objects.o
+ objects.o \
+ output.o
netfs-$(CONFIG_NETFS_STATS) += stats.o
diff --git a/fs/netfs/internal.h b/fs/netfs/internal.h
index d72292e40f9b..0f20587f5a9b 100644
--- a/fs/netfs/internal.h
+++ b/fs/netfs/internal.h
@@ -89,6 +89,12 @@ static inline void netfs_see_request(struct netfs_io_request *rreq,
}
/*
+ * output.c
+ */
+int netfs_begin_write(struct netfs_io_request *wreq, bool may_wait,
+ enum netfs_write_trace what);
+
+/*
* stats.c
*/
#ifdef CONFIG_NETFS_STATS
diff --git a/fs/netfs/output.c b/fs/netfs/output.c
new file mode 100644
index 000000000000..2ad0fd8c32be
--- /dev/null
+++ b/fs/netfs/output.c
@@ -0,0 +1,363 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/* Network filesystem high-level write support.
+ *
+ * Copyright (C) 2023 Red Hat, Inc. All Rights Reserved.
+ * Written by David Howells (dhowells@redhat.com)
+ */
+
+#include <linux/fs.h>
+#include <linux/mm.h>
+#include <linux/pagemap.h>
+#include <linux/slab.h>
+#include <linux/writeback.h>
+#include <linux/pagevec.h>
+#include "internal.h"
+
+/**
+ * netfs_create_write_request - Create a write operation.
+ * @wreq: The write request this is storing from.
+ * @dest: The destination type
+ * @start: Start of the region this write will modify
+ * @len: Length of the modification
+ * @worker: The worker function to handle the write(s)
+ *
+ * Allocate a write operation, set it up and add it to the list on a write
+ * request.
+ */
+struct netfs_io_subrequest *netfs_create_write_request(struct netfs_io_request *wreq,
+ enum netfs_io_source dest,
+ loff_t start, size_t len,
+ work_func_t worker)
+{
+ struct netfs_io_subrequest *subreq;
+
+ subreq = netfs_alloc_subrequest(wreq);
+ if (subreq) {
+ INIT_WORK(&subreq->work, worker);
+ subreq->source = dest;
+ subreq->start = start;
+ subreq->len = len;
+ subreq->debug_index = wreq->subreq_counter++;
+
+ switch (subreq->source) {
+ case NETFS_UPLOAD_TO_SERVER:
+ netfs_stat(&netfs_n_wh_upload);
+ break;
+ case NETFS_WRITE_TO_CACHE:
+ netfs_stat(&netfs_n_wh_write);
+ break;
+ default:
+ BUG();
+ }
+
+ subreq->io_iter = wreq->io_iter;
+ iov_iter_advance(&subreq->io_iter, subreq->start - wreq->start);
+ iov_iter_truncate(&subreq->io_iter, subreq->len);
+
+ trace_netfs_sreq_ref(wreq->debug_id, subreq->debug_index,
+ refcount_read(&subreq->ref),
+ netfs_sreq_trace_new);
+ atomic_inc(&wreq->nr_outstanding);
+ list_add_tail(&subreq->rreq_link, &wreq->subrequests);
+ trace_netfs_sreq(subreq, netfs_sreq_trace_prepare);
+ }
+
+ return subreq;
+}
+EXPORT_SYMBOL(netfs_create_write_request);
+
+/*
+ * Process a completed write request once all the component operations have
+ * been completed.
+ */
+static void netfs_write_terminated(struct netfs_io_request *wreq, bool was_async)
+{
+ struct netfs_io_subrequest *subreq;
+ struct netfs_inode *ctx = netfs_inode(wreq->inode);
+
+ _enter("R=%x[]", wreq->debug_id);
+
+ trace_netfs_rreq(wreq, netfs_rreq_trace_write_done);
+
+ list_for_each_entry(subreq, &wreq->subrequests, rreq_link) {
+ if (!subreq->error)
+ continue;
+ switch (subreq->source) {
+ case NETFS_UPLOAD_TO_SERVER:
+ /* Depending on the type of failure, this may prevent
+ * writeback completion unless we're in disconnected
+ * mode.
+ */
+ if (!wreq->error)
+ wreq->error = subreq->error;
+ break;
+
+ case NETFS_WRITE_TO_CACHE:
+ /* Failure doesn't prevent writeback completion unless
+ * we're in disconnected mode.
+ */
+ if (subreq->error != -ENOBUFS)
+ ctx->ops->invalidate_cache(wreq);
+ break;
+
+ default:
+ WARN_ON_ONCE(1);
+ if (!wreq->error)
+ wreq->error = -EIO;
+ return;
+ }
+ }
+
+ wreq->cleanup(wreq);
+
+ _debug("finished");
+ trace_netfs_rreq(wreq, netfs_rreq_trace_wake_ip);
+ clear_bit_unlock(NETFS_RREQ_IN_PROGRESS, &wreq->flags);
+ wake_up_bit(&wreq->flags, NETFS_RREQ_IN_PROGRESS);
+
+ netfs_clear_subrequests(wreq, was_async);
+ netfs_put_request(wreq, was_async, netfs_rreq_trace_put_complete);
+}
+
+/*
+ * Deal with the completion of writing the data to the cache.
+ */
+void netfs_write_subrequest_terminated(void *_op, ssize_t transferred_or_error,
+ bool was_async)
+{
+ struct netfs_io_subrequest *subreq = _op;
+ struct netfs_io_request *wreq = subreq->rreq;
+ unsigned int u;
+
+ _enter("%x[%x] %zd", wreq->debug_id, subreq->debug_index, transferred_or_error);
+
+ switch (subreq->source) {
+ case NETFS_UPLOAD_TO_SERVER:
+ netfs_stat(&netfs_n_wh_upload_done);
+ break;
+ case NETFS_WRITE_TO_CACHE:
+ netfs_stat(&netfs_n_wh_write_done);
+ break;
+ case NETFS_INVALID_WRITE:
+ break;
+ default:
+ BUG();
+ }
+
+ if (IS_ERR_VALUE(transferred_or_error)) {
+ subreq->error = transferred_or_error;
+ trace_netfs_failure(wreq, subreq, transferred_or_error,
+ netfs_fail_write);
+ goto failed;
+ }
+
+ if (WARN(transferred_or_error > subreq->len - subreq->transferred,
+ "Subreq excess write: R%x[%x] %zd > %zu - %zu",
+ wreq->debug_id, subreq->debug_index,
+ transferred_or_error, subreq->len, subreq->transferred))
+ transferred_or_error = subreq->len - subreq->transferred;
+
+ subreq->error = 0;
+ subreq->transferred += transferred_or_error;
+
+ if (iov_iter_count(&subreq->io_iter) != subreq->len - subreq->transferred)
+ pr_warn("R=%08x[%u] ITER POST-MISMATCH %zx != %zx-%zx %x\n",
+ wreq->debug_id, subreq->debug_index,
+ iov_iter_count(&subreq->io_iter), subreq->len,
+ subreq->transferred, subreq->io_iter.iter_type);
+
+ if (subreq->transferred < subreq->len)
+ goto incomplete;
+
+ __clear_bit(NETFS_SREQ_NO_PROGRESS, &subreq->flags);
+out:
+ trace_netfs_sreq(subreq, netfs_sreq_trace_terminated);
+
+ /* If we decrement nr_outstanding to 0, the ref belongs to us. */
+ u = atomic_dec_return(&wreq->nr_outstanding);
+ if (u == 0)
+ netfs_write_terminated(wreq, was_async);
+ else if (u == 1)
+ wake_up_var(&wreq->nr_outstanding);
+
+ netfs_put_subrequest(subreq, was_async, netfs_sreq_trace_put_terminated);
+ return;
+
+incomplete:
+ if (transferred_or_error == 0) {
+ if (__test_and_set_bit(NETFS_SREQ_NO_PROGRESS, &subreq->flags)) {
+ subreq->error = -ENODATA;
+ goto failed;
+ }
+ } else {
+ __clear_bit(NETFS_SREQ_NO_PROGRESS, &subreq->flags);
+ }
+
+ __set_bit(NETFS_SREQ_SHORT_IO, &subreq->flags);
+ set_bit(NETFS_RREQ_INCOMPLETE_IO, &wreq->flags);
+ goto out;
+
+failed:
+ switch (subreq->source) {
+ case NETFS_WRITE_TO_CACHE:
+ netfs_stat(&netfs_n_wh_write_failed);
+ set_bit(NETFS_RREQ_INCOMPLETE_IO, &wreq->flags);
+ break;
+ case NETFS_UPLOAD_TO_SERVER:
+ netfs_stat(&netfs_n_wh_upload_failed);
+ set_bit(NETFS_RREQ_FAILED, &wreq->flags);
+ wreq->error = subreq->error;
+ break;
+ default:
+ break;
+ }
+ goto out;
+}
+EXPORT_SYMBOL(netfs_write_subrequest_terminated);
+
+static void netfs_write_to_cache_op(struct netfs_io_subrequest *subreq)
+{
+ struct netfs_io_request *wreq = subreq->rreq;
+ struct netfs_cache_resources *cres = &wreq->cache_resources;
+
+ trace_netfs_sreq(subreq, netfs_sreq_trace_submit);
+
+ cres->ops->write(cres, subreq->start, &subreq->io_iter,
+ netfs_write_subrequest_terminated, subreq);
+}
+
+static void netfs_write_to_cache_op_worker(struct work_struct *work)
+{
+ struct netfs_io_subrequest *subreq =
+ container_of(work, struct netfs_io_subrequest, work);
+
+ netfs_write_to_cache_op(subreq);
+}
+
+/**
+ * netfs_queue_write_request - Queue a write request for attention
+ * @subreq: The write request to be queued
+ *
+ * Queue the specified write request for processing by a worker thread. We
+ * pass the caller's ref on the request to the worker thread.
+ */
+void netfs_queue_write_request(struct netfs_io_subrequest *subreq)
+{
+ if (!queue_work(system_unbound_wq, &subreq->work))
+ netfs_put_subrequest(subreq, false, netfs_sreq_trace_put_wip);
+}
+EXPORT_SYMBOL(netfs_queue_write_request);
+
+/*
+ * Set up a op for writing to the cache.
+ */
+static void netfs_set_up_write_to_cache(struct netfs_io_request *wreq)
+{
+ struct netfs_cache_resources *cres;
+ struct netfs_io_subrequest *subreq;
+ struct netfs_inode *ctx = netfs_inode(wreq->inode);
+ struct fscache_cookie *cookie = netfs_i_cookie(ctx);
+ loff_t start = wreq->start;
+ size_t len = wreq->len;
+ int ret;
+
+ if (!fscache_cookie_enabled(cookie)) {
+ clear_bit(NETFS_RREQ_WRITE_TO_CACHE, &wreq->flags);
+ return;
+ }
+
+ _debug("write to cache");
+ subreq = netfs_create_write_request(wreq, NETFS_WRITE_TO_CACHE, start, len,
+ netfs_write_to_cache_op_worker);
+ if (!subreq)
+ return;
+
+ cres = &wreq->cache_resources;
+ ret = fscache_begin_read_operation(cres, cookie);
+ if (ret < 0) {
+ netfs_write_subrequest_terminated(subreq, ret, false);
+ return;
+ }
+
+ ret = cres->ops->prepare_write(cres, &start, &len, i_size_read(wreq->inode),
+ true);
+ if (ret < 0) {
+ netfs_write_subrequest_terminated(subreq, ret, false);
+ return;
+ }
+
+ netfs_queue_write_request(subreq);
+}
+
+/*
+ * Begin the process of writing out a chunk of data.
+ *
+ * We are given a write request that holds a series of dirty regions and
+ * (partially) covers a sequence of folios, all of which are present. The
+ * pages must have been marked as writeback as appropriate.
+ *
+ * We need to perform the following steps:
+ *
+ * (1) If encrypting, create an output buffer and encrypt each block of the
+ * data into it, otherwise the output buffer will point to the original
+ * folios.
+ *
+ * (2) If the data is to be cached, set up a write op for the entire output
+ * buffer to the cache, if the cache wants to accept it.
+ *
+ * (3) If the data is to be uploaded (ie. not merely cached):
+ *
+ * (a) If the data is to be compressed, create a compression buffer and
+ * compress the data into it.
+ *
+ * (b) For each destination we want to upload to, set up write ops to write
+ * to that destination. We may need multiple writes if the data is not
+ * contiguous or the span exceeds wsize for a server.
+ */
+int netfs_begin_write(struct netfs_io_request *wreq, bool may_wait,
+ enum netfs_write_trace what)
+{
+ struct netfs_inode *ctx = netfs_inode(wreq->inode);
+
+ _enter("R=%x %llx-%llx f=%lx",
+ wreq->debug_id, wreq->start, wreq->start + wreq->len - 1,
+ wreq->flags);
+
+ trace_netfs_write(wreq, what);
+ if (wreq->len == 0 || wreq->iter.count == 0) {
+ pr_err("Zero-sized write [R=%x]\n", wreq->debug_id);
+ return -EIO;
+ }
+
+ wreq->io_iter = wreq->iter;
+
+ /* ->outstanding > 0 carries a ref */
+ netfs_get_request(wreq, netfs_rreq_trace_get_for_outstanding);
+ atomic_set(&wreq->nr_outstanding, 1);
+
+ /* Start the encryption/compression going. We can do that in the
+ * background whilst we generate a list of write ops that we want to
+ * perform.
+ */
+ // TODO: Encrypt or compress the region as appropriate
+
+ /* We need to write all of the region to the cache */
+ if (test_bit(NETFS_RREQ_WRITE_TO_CACHE, &wreq->flags))
+ netfs_set_up_write_to_cache(wreq);
+
+ /* However, we don't necessarily write all of the region to the server.
+ * Caching of reads is being managed this way also.
+ */
+ if (test_bit(NETFS_RREQ_UPLOAD_TO_SERVER, &wreq->flags))
+ ctx->ops->create_write_requests(wreq, wreq->start, wreq->len);
+
+ if (atomic_dec_and_test(&wreq->nr_outstanding))
+ netfs_write_terminated(wreq, false);
+
+ if (!may_wait)
+ return -EIOCBQUEUED;
+
+ wait_on_bit(&wreq->flags, NETFS_RREQ_IN_PROGRESS,
+ TASK_UNINTERRUPTIBLE);
+ return wreq->error;
+}