netfs: zero the tail of a short DIO/unbuffered readgrafted

The buffered read collector zero-fills the tail of a short read that
stops below the inode's i_size (netfs_clear_unread()), so a read that
races an extending write still returns the expected number of bytes.
The non-buffered collector path does no such thing: it just records
how much was transferred.

Add the same zero-fill for the non-buffered case, gated on
NETFS_SREQ_CLEAR_TAIL: a subreq's source sets that flag when a short
result from it is known to be safe to treat as a hole, as opposed to
NETFS_SREQ_HIT_EOF, which means the read genuinely ran off the end of
the file and should be reported short as-is. Only CLEAR_TAIL should
zero-fill here; a real EOF must stay a real short read.

No source currently sets CLEAR_TAIL on an unbuffered/DIO subrequest,
so this is inert on its own -- a following change teaches cifs to set
it in the one case that needs it.

Fixes: e2d46f2ec332 ("netfs: Change the read result collector to only use one work item")
Reviewed-by: David Howells <dhowells@redhat.com>
Reviewed-by: Namjae Jeon <linkinjeon@kernel.org>
Signed-off-by: Paulo Alcantara <pc@manguebit.org>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Ronnie Sahlberg <ronniesahlberg@gmail.com>
Cc: Shyam Prasad N <sprasad@microsoft.com>
Cc: Tom Talpey <tom@talpey.com>
Cc: Bharath SM <bharathsm@microsoft.com>
Cc: stable@vger.kernel.org

1 files changed, 183 insertions, 0 deletions

diff --git a/drivers/interconnect/debugfs-client.c b/drivers/interconnect/debugfs-client.c
new file mode 100644
index 000000000..91f86d923
--- /dev/null
+++ b/drivers/interconnect/debugfs-client.c
@@ -0,0 +1,183 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (c) 2023, Qualcomm Innovation Center, Inc. All rights reserved.
+ */
+#include <linux/debugfs.h>
+#include <linux/interconnect.h>
+#include <linux/platform_device.h>
+
+#include "internal.h"
+
+/*
+ * This can be dangerous, therefore don't provide any real compile time
+ * configuration option for this feature.
+ * People who want to use this will need to modify the source code directly.
+ */
+#undef INTERCONNECT_ALLOW_WRITE_DEBUGFS
+
+#if defined(INTERCONNECT_ALLOW_WRITE_DEBUGFS) && defined(CONFIG_DEBUG_FS)
+
+static LIST_HEAD(debugfs_paths);
+static DEFINE_MUTEX(debugfs_lock);
+
+static struct platform_device *pdev;
+static struct icc_path *cur_path;
+
+static char *src_node;
+static char *dst_node;
+static u32 avg_bw;
+static u32 peak_bw;
+static u32 tag;
+
+struct debugfs_path {
+	const char *src;
+	const char *dst;
+	struct icc_path *path;
+	struct list_head list;
+};
+
+static struct icc_path *get_path(const char *src, const char *dst)
+{
+	struct debugfs_path *path;
+
+	list_for_each_entry(path, &debugfs_paths, list) {
+		if (!strcmp(path->src, src) && !strcmp(path->dst, dst))
+			return path->path;
+	}
+
+	return NULL;
+}
+
+static int icc_get_set(void *data, u64 val)
+{
+	struct debugfs_path *debugfs_path;
+	char *src, *dst;
+	int ret = 0;
+
+	mutex_lock(&debugfs_lock);
+
+	rcu_read_lock();
+	src = rcu_dereference(src_node);
+	dst = rcu_dereference(dst_node);
+
+	/*
+	 * If we've already looked up a path, then use the existing one instead
+	 * of calling icc_get() again. This allows for updating previous BW
+	 * votes when "get" is written to multiple times for multiple paths.
+	 */
+	cur_path = get_path(src, dst);
+	if (cur_path) {
+		rcu_read_unlock();
+		goto out;
+	}
+
+	src = kstrdup(src, GFP_ATOMIC);
+	dst = kstrdup(dst, GFP_ATOMIC);
+	rcu_read_unlock();
+
+	if (!src || !dst) {
+		ret = -ENOMEM;
+		goto err_free;
+	}
+
+	cur_path = icc_get(&pdev->dev, src, dst);
+	if (IS_ERR(cur_path)) {
+		ret = PTR_ERR(cur_path);
+		goto err_free;
+	}
+
+	debugfs_path = kzalloc_obj(*debugfs_path);
+	if (!debugfs_path) {
+		ret = -ENOMEM;
+		goto err_put;
+	}
+
+	debugfs_path->path = cur_path;
+	debugfs_path->src = src;
+	debugfs_path->dst = dst;
+	list_add_tail(&debugfs_path->list, &debugfs_paths);
+
+	goto out;
+
+err_put:
+	icc_put(cur_path);
+err_free:
+	kfree(src);
+	kfree(dst);
+out:
+	mutex_unlock(&debugfs_lock);
+	return ret;
+}
+
+DEFINE_DEBUGFS_ATTRIBUTE(icc_get_fops, NULL, icc_get_set, "%llu\n");
+
+static int icc_commit_set(void *data, u64 val)
+{
+	int ret;
+
+	mutex_lock(&debugfs_lock);
+
+	if (!cur_path) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	if (IS_ERR(cur_path)) {
+		ret = PTR_ERR(cur_path);
+		goto out;
+	}
+
+	icc_set_tag(cur_path, tag);
+	ret = icc_set_bw(cur_path, avg_bw, peak_bw);
+out:
+	mutex_unlock(&debugfs_lock);
+	return ret;
+}
+
+DEFINE_DEBUGFS_ATTRIBUTE(icc_commit_fops, NULL, icc_commit_set, "%llu\n");
+
+int icc_debugfs_client_init(struct dentry *icc_dir)
+{
+	struct dentry *client_dir;
+	int ret;
+
+	pdev = platform_device_alloc("icc-debugfs-client", PLATFORM_DEVID_NONE);
+	if (!pdev)
+		return -ENOMEM;
+
+	ret = platform_device_add(pdev);
+	if (ret) {
+		pr_err("%s: failed to add platform device: %d\n", __func__, ret);
+		platform_device_put(pdev);
+		return ret;
+	}
+
+	src_node = kstrdup("", GFP_KERNEL);
+	dst_node = kstrdup("", GFP_KERNEL);
+	if (!src_node || !dst_node) {
+		kfree(dst_node);
+		kfree(src_node);
+		return -ENOMEM;
+	}
+
+	client_dir = debugfs_create_dir("test_client", icc_dir);
+
+	debugfs_create_str("src_node", 0600, client_dir, &src_node);
+	debugfs_create_str("dst_node", 0600, client_dir, &dst_node);
+	debugfs_create_file("get", 0200, client_dir, NULL, &icc_get_fops);
+	debugfs_create_u32("avg_bw", 0600, client_dir, &avg_bw);
+	debugfs_create_u32("peak_bw", 0600, client_dir, &peak_bw);
+	debugfs_create_u32("tag", 0600, client_dir, &tag);
+	debugfs_create_file("commit", 0200, client_dir, NULL, &icc_commit_fops);
+
+	return 0;
+}
+
+#else
+
+int icc_debugfs_client_init(struct dentry *icc_dir)
+{
+	return 0;
+}
+
+#endif