// SPDX-License-Identifier: GPL-2.0
/*
* Copyright (C) 2011 STRATO. All rights reserved.
*/
#include <linux/sched.h>
#include <linux/pagemap.h>
#include <linux/writeback.h>
#include <linux/blkdev.h>
#include <linux/slab.h>
#include <linux/workqueue.h>
#include "ctree.h"
#include "volumes.h"
#include "disk-io.h"
#include "transaction.h"
#include "dev-replace.h"
#include "block-group.h"
#undef DEBUG
/*
* This is the implementation for the generic read ahead framework.
*
* To trigger a readahead, btrfs_reada_add must be called. It will start
* a read ahead for the given range [start, end) on tree root. The returned
* handle can either be used to wait on the readahead to finish
* (btrfs_reada_wait), or to send it to the background (btrfs_reada_detach).
*
* The read ahead works as follows:
* On btrfs_reada_add, the root of the tree is inserted into a radix_tree.
* reada_start_machine will then search for extents to prefetch and trigger
* some reads. When a read finishes for a node, all contained node/leaf
* pointers that lie in the given range will also be enqueued. The reads will
* be triggered in sequential order, thus giving a big win over a naive
* enumeration. It will also make use of multi-device layouts. Each disk
* will have its on read pointer and all disks will by utilized in parallel.
* Also will no two disks read both sides of a mirror simultaneously, as this
* would waste seeking capacity. Instead both disks will read different parts
* of the filesystem.
* Any number of readaheads can be started in parallel. The read order will be
* determined globally, i.e. 2 parallel readaheads will normally finish faster
* than the 2 started one after another.
*/
#define MAX_IN_FLIGHT 6
struct reada_extctl {
struct list_head list;
struct reada_control *rc;
u64 generation;
};
struct reada_extent {
u64 logical;
struct btrfs_key top;
struct list_head extctl;
int refcnt;
spinlock_t lock;
struct reada_zone *zones[BTRFS_MAX_MIRRORS];
int nzones;
int scheduled;
};
struct reada_zone {
u64 start;
u64 end;
u64 elems;
struct list_head list;
spinlock_t lock;
int locked;
struct btrfs_device *device;
struct btrfs_device *devs[BTRFS_MAX_MIRRORS]; /* full list, incl
* self */
int ndevs;
struct kref refcnt;
};
struct reada_machine_work {
struct btrfs_work work;
struct btrfs_fs_info *fs_info;
};
static void reada_extent_put(struct btrfs_fs_info *, struct reada_extent *);
static void reada_control_release(struct kref *kref);
static void reada_zone_release(struct kref *kref);
static void reada_start_machine(struct btrfs_fs_info *fs_info);
static void __reada_start_machine(struct btrfs_fs_info *fs_info);
static int reada_add_block(struct reada_control *rc, u64 logical,
struct btrfs_key *top, u64 generation);
/* recurses */
/* in case of err, eb might be NULL */
static void __readahead_hook(struct btrfs_fs_info *fs_info,
struct reada_extent *re, struct extent_buffer *eb,
int err)
{
int nritems;
int i;
u64 bytenr;
u64 generation;
struct list_head list;
spin_lock(&re->lock);
/*
* just take the full list from the extent. afterwards we
* don't need the lock anymore
*/
list_replace_init(&re->extctl, &list);
re->scheduled = 0;
spin_unlock(&re->lock);
/*
* this is the error case, the extent buffer has not been
* read correctly. We won't access anything from it and
* just cleanup our data structures. Effectively this will
* cut the branch below this node from read ahead.
*/
if (err)
goto cleanup;
/*
* FIXME: currently we just set nritems to 0 if this is a leaf,
* effectively ignoring the content. In a next step we could
* trigger more readahead depending from the content, e.g.
* fetch the checksums for the extents in the leaf.
*/
if (!btrfs_header_level(eb))
goto cleanup;
nritems = btrfs_header_nritems(eb);
generation = btrfs_header_generation(eb);
for (i = 0; i < nritems; i++) {
struct reada_extctl *rec;
u64 n_gen;
struct btrfs_key key;
struct btrfs_key next_key;
btrfs_node_key_to_cpu(eb, &key, i);
if (i + 1 < nritems)
btrfs_node_key_to_cpu(eb, &next_key, i + 1);
else
next_key = re->top;
bytenr = btrfs_node_blockptr(eb, i);
n_gen = btrfs_node_ptr_generation(eb, i);
list_for_each_entry(rec, &list, list) {
struct reada_control *rc = rec->rc;
/*
* if the generation doesn't match, just ignore this
* extctl. This will probably cut off a branch from
* prefetch. Alternatively one could start a new (sub-)
* prefetch for this branch, starting again from root.
* FIXME: move the generation check out of this loop
*/
#ifdef DEBUG
if (rec->generation != generation) {
btrfs_debug(fs_info,
"generation mismatch for (%llu,%d,%llu) %llu != %llu",
key.objectid, key.type, key.offset,
rec->generation, generation);
}
#endif
if (rec->generation == generation &&
btrfs_comp_cpu_keys(&key, &rc->key_end) < 0 &&
btrfs_comp_cpu_keys(&