Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
readahead: add trace points
Add a couple of trace points to make debugging readahead logic easier.
[jack@suse.cz: v2]
Link: https://lkml.kernel.org/r/20250909145849.5090-2-jack@suse.cz
Link: https://lkml.kernel.org/r/20250908145533.31528-2-jack@suse.cz
Signed-off-by: Jan Kara <jack@suse.cz>
Tested-by: Pankaj Raghav <p.raghav@samsung.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
authored by
Jan Kara
and committed by
Andrew Morton
60283726
e1831e8d
+140
+132
include/trace/events/readahead.h
+132
include/trace/events/readahead.h
···
1
+
/* SPDX-License-Identifier: GPL-2.0 */
2
+
#undef TRACE_SYSTEM
3
+
#define TRACE_SYSTEM readahead
4
+
5
+
#if !defined(_TRACE_FILEMAP_H) || defined(TRACE_HEADER_MULTI_READ)
6
+
#define _TRACE_READAHEAD_H
7
+
8
+
#include <linux/types.h>
9
+
#include <linux/tracepoint.h>
10
+
#include <linux/mm.h>
11
+
#include <linux/fs.h>
12
+
#include <linux/pagemap.h>
13
+
14
+
TRACE_EVENT(page_cache_ra_unbounded,
15
+
TP_PROTO(struct inode *inode, pgoff_t index, unsigned long nr_to_read,
16
+
unsigned long lookahead_size),
17
+
18
+
TP_ARGS(inode, index, nr_to_read, lookahead_size),
19
+
20
+
TP_STRUCT__entry(
21
+
__field(unsigned long, i_ino)
22
+
__field(dev_t, s_dev)
23
+
__field(pgoff_t, index)
24
+
__field(unsigned long, nr_to_read)
25
+
__field(unsigned long, lookahead_size)
26
+
),
27
+
28
+
TP_fast_assign(
29
+
__entry->i_ino = inode->i_ino;
30
+
__entry->s_dev = inode->i_sb->s_dev;
31
+
__entry->index = index;
32
+
__entry->nr_to_read = nr_to_read;
33
+
__entry->lookahead_size = lookahead_size;
34
+
),
35
+
36
+
TP_printk(
37
+
"dev=%d:%d ino=%lx index=%lu nr_to_read=%lu lookahead_size=%lu",
38
+
MAJOR(__entry->s_dev), MINOR(__entry->s_dev), __entry->i_ino,
39
+
__entry->index, __entry->nr_to_read, __entry->lookahead_size
40
+
)
41
+
);
42
+
43
+
TRACE_EVENT(page_cache_ra_order,
44
+
TP_PROTO(struct inode *inode, pgoff_t index, struct file_ra_state *ra),
45
+
46
+
TP_ARGS(inode, index, ra),
47
+
48
+
TP_STRUCT__entry(
49
+
__field(unsigned long, i_ino)
50
+
__field(dev_t, s_dev)
51
+
__field(pgoff_t, index)
52
+
__field(unsigned int, order)
53
+
__field(unsigned int, size)
54
+
__field(unsigned int, async_size)
55
+
__field(unsigned int, ra_pages)
56
+
),
57
+
58
+
TP_fast_assign(
59
+
__entry->i_ino = inode->i_ino;
60
+
__entry->s_dev = inode->i_sb->s_dev;
61
+
__entry->index = index;
62
+
__entry->order = ra->order;
63
+
__entry->size = ra->size;
64
+
__entry->async_size = ra->async_size;
65
+
__entry->ra_pages = ra->ra_pages;
66
+
),
67
+
68
+
TP_printk(
69
+
"dev=%d:%d ino=%lx index=%lu order=%u size=%u async_size=%u ra_pages=%u",
70
+
MAJOR(__entry->s_dev), MINOR(__entry->s_dev), __entry->i_ino,
71
+
__entry->index, __entry->order, __entry->size,
72
+
__entry->async_size, __entry->ra_pages
73
+
)
74
+
);
75
+
76
+
DECLARE_EVENT_CLASS(page_cache_ra_op,
77
+
TP_PROTO(struct inode *inode, pgoff_t index, struct file_ra_state *ra,
78
+
unsigned long req_count),
79
+
80
+
TP_ARGS(inode, index, ra, req_count),
81
+
82
+
TP_STRUCT__entry(
83
+
__field(unsigned long, i_ino)
84
+
__field(dev_t, s_dev)
85
+
__field(pgoff_t, index)
86
+
__field(unsigned int, order)
87
+
__field(unsigned int, size)
88
+
__field(unsigned int, async_size)
89
+
__field(unsigned int, ra_pages)
90
+
__field(unsigned int, mmap_miss)
91
+
__field(loff_t, prev_pos)
92
+
__field(unsigned long, req_count)
93
+
),
94
+
95
+
TP_fast_assign(
96
+
__entry->i_ino = inode->i_ino;
97
+
__entry->s_dev = inode->i_sb->s_dev;
98
+
__entry->index = index;
99
+
__entry->order = ra->order;
100
+
__entry->size = ra->size;
101
+
__entry->async_size = ra->async_size;
102
+
__entry->ra_pages = ra->ra_pages;
103
+
__entry->mmap_miss = ra->mmap_miss;
104
+
__entry->prev_pos = ra->prev_pos;
105
+
__entry->req_count = req_count;
106
+
),
107
+
108
+
TP_printk(
109
+
"dev=%d:%d ino=%lx index=%lu req_count=%lu order=%u size=%u async_size=%u ra_pages=%u mmap_miss=%u prev_pos=%lld",
110
+
MAJOR(__entry->s_dev), MINOR(__entry->s_dev), __entry->i_ino,
111
+
__entry->index, __entry->req_count, __entry->order,
112
+
__entry->size, __entry->async_size, __entry->ra_pages,
113
+
__entry->mmap_miss, __entry->prev_pos
114
+
)
115
+
);
116
+
117
+
DEFINE_EVENT(page_cache_ra_op, page_cache_sync_ra,
118
+
TP_PROTO(struct inode *inode, pgoff_t index, struct file_ra_state *ra,
119
+
unsigned long req_count),
120
+
TP_ARGS(inode, index, ra, req_count)
121
+
);
122
+
123
+
DEFINE_EVENT(page_cache_ra_op, page_cache_async_ra,
124
+
TP_PROTO(struct inode *inode, pgoff_t index, struct file_ra_state *ra,
125
+
unsigned long req_count),
126
+
TP_ARGS(inode, index, ra, req_count)
127
+
);
128
+
129
+
#endif /* _TRACE_FILEMAP_H */
130
+
131
+
/* This part must be outside protection */
132
+
#include <trace/define_trace.h>
+8
mm/readahead.c
+8
mm/readahead.c
···
129
129
#include <linux/fadvise.h>
130
130
#include <linux/sched/mm.h>
131
131
132
+
#define CREATE_TRACE_POINTS
133
+
#include <trace/events/readahead.h>
134
+
132
135
#include "internal.h"
133
136
134
137
/*
···
228
225
*/
229
226
unsigned int nofs = memalloc_nofs_save();
230
227
228
+
trace_page_cache_ra_unbounded(mapping->host, index, nr_to_read,
229
+
lookahead_size);
231
230
filemap_invalidate_lock_shared(mapping);
232
231
index = mapping_align_index(mapping, index);
233
232
···
475
470
gfp_t gfp = readahead_gfp_mask(mapping);
476
471
unsigned int new_order = ra->order;
477
472
473
+
trace_page_cache_ra_order(mapping->host, start, ra);
478
474
if (!mapping_large_folio_support(mapping)) {
479
475
ra->order = 0;
480
476
goto fallback;
···
560
554
unsigned long max_pages, contig_count;
561
555
pgoff_t prev_index, miss;
562
556
557
+
trace_page_cache_sync_ra(ractl->mapping->host, index, ra, req_count);
563
558
/*
564
559
* Even if readahead is disabled, issue this request as readahead
565
560
* as we'll need it to satisfy the requested range. The forced
···
645
638
if (folio_test_writeback(folio))
646
639
return;
647
640
641
+
trace_page_cache_async_ra(ractl->mapping->host, index, ra, req_count);
648
642
folio_clear_readahead(folio);
649
643
650
644
if (blk_cgroup_congested())