1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
4 * Copyright (C) 2002 Cluster File Systems, Inc.
6 * This file is part of Lustre, http://www.lustre.org.
8 * Lustre is free software; you can redistribute it and/or
9 * modify it under the terms of version 2 of the GNU General Public
10 * License as published by the Free Software Foundation.
12 * Lustre is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
17 * You should have received a copy of the GNU General Public License
18 * along with Lustre; if not, write to the Free Software
19 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
22 #define DEBUG_SUBSYSTEM S_LLITE
24 #include <linux/version.h>
25 #include <linux/lustre_lite.h>
26 #include <linux/lprocfs_status.h>
27 #include <linux/seq_file.h>
28 #include <linux/obd_support.h>
30 #include "llite_internal.h"
32 /* /proc/lustre/llite mount point registration */
33 struct proc_dir_entry *proc_lustre_fs_root;
34 struct file_operations llite_dump_pgcache_fops;
37 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
38 struct super_block *sb, char *osc, char *mdc)
42 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi){}
45 long long mnt_instance;
47 static int ll_rd_blksize(char *page, char **start, off_t off, int count,
50 struct super_block *sb = (struct super_block *)data;
51 struct obd_statfs osfs;
55 rc = ll_statfs_internal(sb, &osfs, jiffies - HZ);
58 rc = snprintf(page, count, "%u\n", osfs.os_bsize);
64 static int ll_rd_kbytestotal(char *page, char **start, off_t off, int count,
67 struct super_block *sb = (struct super_block *)data;
68 struct obd_statfs osfs;
72 rc = ll_statfs_internal(sb, &osfs, jiffies - HZ);
74 __u32 blk_size = osfs.os_bsize >> 10;
75 __u64 result = osfs.os_blocks;
77 while (blk_size >>= 1)
81 rc = snprintf(page, count, LPU64"\n", result);
87 static int ll_rd_kbytesfree(char *page, char **start, off_t off, int count,
90 struct super_block *sb = (struct super_block *)data;
91 struct obd_statfs osfs;
95 rc = ll_statfs_internal(sb, &osfs, jiffies - HZ);
97 __u32 blk_size = osfs.os_bsize >> 10;
98 __u64 result = osfs.os_bfree;
100 while (blk_size >>= 1)
104 rc = snprintf(page, count, LPU64"\n", result);
109 static int ll_rd_kbytesavail(char *page, char **start, off_t off, int count,
110 int *eof, void *data)
112 struct super_block *sb = (struct super_block *)data;
113 struct obd_statfs osfs;
117 rc = ll_statfs_internal(sb, &osfs, jiffies - HZ);
119 __u32 blk_size = osfs.os_bsize >> 10;
120 __u64 result = osfs.os_bavail;
122 while (blk_size >>= 1)
126 rc = snprintf(page, count, LPU64"\n", result);
131 static int ll_rd_filestotal(char *page, char **start, off_t off, int count,
132 int *eof, void *data)
134 struct super_block *sb = (struct super_block *)data;
135 struct obd_statfs osfs;
139 rc = ll_statfs_internal(sb, &osfs, jiffies - HZ);
142 rc = snprintf(page, count, LPU64"\n", osfs.os_files);
147 static int ll_rd_filesfree(char *page, char **start, off_t off, int count,
148 int *eof, void *data)
150 struct super_block *sb = (struct super_block *)data;
151 struct obd_statfs osfs;
155 rc = ll_statfs_internal(sb, &osfs, jiffies - HZ);
158 rc = snprintf(page, count, LPU64"\n", osfs.os_ffree);
165 static int ll_rd_path(char *page, char **start, off_t off, int count, int *eof,
172 static int ll_rd_fstype(char *page, char **start, off_t off, int count,
173 int *eof, void *data)
175 struct super_block *sb = (struct super_block*)data;
179 return snprintf(page, count, "%s\n", sb->s_type->name);
182 static int ll_rd_sb_uuid(char *page, char **start, off_t off, int count,
183 int *eof, void *data)
185 struct super_block *sb = (struct super_block *)data;
189 return snprintf(page, count, "%s\n", ll_s2sbi(sb)->ll_sb_uuid.uuid);
192 static int ll_rd_read_ahead(char *page, char **start, off_t off, int count,
193 int *eof, void *data)
195 struct super_block *sb = (struct super_block*)data;
196 struct ll_sb_info *sbi = ll_s2sbi(sb);
201 val = (sbi->ll_flags & LL_SBI_READAHEAD) ? 1 : 0;
202 rc = snprintf(page, count, "%d\n", val);
206 static int ll_wr_read_ahead(struct file *file, const char *buffer,
207 unsigned long count, void *data)
209 struct super_block *sb = (struct super_block*)data;
210 struct ll_sb_info *sbi = ll_s2sbi(sb);
214 if (1 != sscanf(buffer, "%d", &readahead))
218 sbi->ll_flags |= LL_SBI_READAHEAD;
220 sbi->ll_flags &= ~LL_SBI_READAHEAD;
225 static struct lprocfs_vars lprocfs_obd_vars[] = {
226 { "uuid", ll_rd_sb_uuid, 0, 0 },
227 //{ "mntpt_path", ll_rd_path, 0, 0 },
228 { "fstype", ll_rd_fstype, 0, 0 },
229 { "blocksize", ll_rd_blksize, 0, 0 },
230 { "kbytestotal", ll_rd_kbytestotal, 0, 0 },
231 { "kbytesfree", ll_rd_kbytesfree, 0, 0 },
232 { "kbytesavail", ll_rd_kbytesavail, 0, 0 },
233 { "filestotal", ll_rd_filestotal, 0, 0 },
234 { "filesfree", ll_rd_filesfree, 0, 0 },
235 //{ "filegroups", lprocfs_rd_filegroups, 0, 0 },
236 { "read_ahead", ll_rd_read_ahead, ll_wr_read_ahead, 0 },
240 #define MAX_STRING_SIZE 128
242 struct llite_file_opcode {
246 } llite_opcode_table[LPROC_LL_FILE_OPCODES] = {
248 { LPROC_LL_DIRTY_HITS, LPROCFS_TYPE_REGS, "dirty_pages_hits" },
249 { LPROC_LL_DIRTY_MISSES, LPROCFS_TYPE_REGS, "dirty_pages_misses" },
250 { LPROC_LL_WB_WRITEPAGE, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
251 "writeback_from_writepage" },
252 { LPROC_LL_WB_PRESSURE, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
253 "writeback_from_pressure" },
254 { LPROC_LL_WB_OK, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
255 "writeback_ok_pages" },
256 { LPROC_LL_WB_FAIL, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
257 "writeback_failed_pages" },
258 { LPROC_LL_READ_BYTES, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_BYTES,
260 { LPROC_LL_WRITE_BYTES, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_BYTES,
262 { LPROC_LL_BRW_READ, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
264 { LPROC_LL_BRW_WRITE, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
267 { LPROC_LL_IOCTL, LPROCFS_TYPE_REGS, "ioctl" },
268 { LPROC_LL_OPEN, LPROCFS_TYPE_REGS, "open" },
269 { LPROC_LL_RELEASE, LPROCFS_TYPE_REGS, "close" },
270 { LPROC_LL_MAP, LPROCFS_TYPE_REGS, "mmap" },
271 { LPROC_LL_LLSEEK, LPROCFS_TYPE_REGS, "seek" },
272 { LPROC_LL_FSYNC, LPROCFS_TYPE_REGS, "fsync" },
273 /* inode operation */
274 { LPROC_LL_SETATTR, LPROCFS_TYPE_REGS, "setattr" },
275 { LPROC_LL_TRUNC, LPROCFS_TYPE_REGS, "punch" },
276 #if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0))
277 { LPROC_LL_GETATTR, LPROCFS_TYPE_REGS, "getattr" },
279 { LPROC_LL_REVALIDATE, LPROCFS_TYPE_REGS, "getattr" },
281 /* special inode operation */
282 { LPROC_LL_STAFS, LPROCFS_TYPE_REGS, "statfs" },
283 { LPROC_LL_ALLOC_INODE, LPROCFS_TYPE_REGS, "alloc_inode" },
284 { LPROC_LL_DIRECT_READ, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
286 { LPROC_LL_DIRECT_WRITE, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
291 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
292 struct super_block *sb, char *osc, char *mdc)
294 struct lprocfs_vars lvars[2];
295 struct ll_sb_info *sbi = ll_s2sbi(sb);
296 struct obd_device *obd;
297 char name[MAX_STRING_SIZE + 1];
299 struct lprocfs_stats *svc_stats = NULL;
300 struct proc_dir_entry *entry;
303 memset(lvars, 0, sizeof(lvars));
305 name[MAX_STRING_SIZE] = '\0';
306 lvars[0].name = name;
308 LASSERT(sbi != NULL);
309 LASSERT(mdc != NULL);
310 LASSERT(osc != NULL);
313 snprintf(name, MAX_STRING_SIZE, "fs%llu", mnt_instance);
316 sbi->ll_proc_root = lprocfs_register(name, parent, NULL, NULL);
317 if (IS_ERR(sbi->ll_proc_root)) {
318 err = PTR_ERR(sbi->ll_proc_root);
319 sbi->ll_proc_root = NULL;
323 entry = create_proc_entry("dump_page_cache", 0444, sbi->ll_proc_root);
325 GOTO(out, err = -ENOMEM);
326 entry->proc_fops = &llite_dump_pgcache_fops;
329 svc_stats = lprocfs_alloc_stats(LPROC_LL_FILE_OPCODES);
330 if (svc_stats == NULL) {
334 /* do counter init */
335 for (id = 0; id < LPROC_LL_FILE_OPCODES; id++) {
336 __u32 type = llite_opcode_table[id].type;
338 if (type & LPROCFS_TYPE_REGS)
341 if (type & LPROCFS_TYPE_BYTES)
344 if (type & LPROCFS_TYPE_PAGES)
348 lprocfs_counter_init(svc_stats, llite_opcode_table[id].opcode,
349 (type & LPROCFS_CNTR_AVGMINMAX),
350 llite_opcode_table[id].opname, ptr);
352 err = lprocfs_register_stats(sbi->ll_proc_root, "stats", svc_stats);
356 sbi->ll_stats = svc_stats;
357 /* need place to keep svc_stats */
359 /* Static configuration info */
360 err = lprocfs_add_vars(sbi->ll_proc_root, lprocfs_obd_vars, sb);
365 obd = class_name2obd(mdc);
367 LASSERT(obd != NULL);
368 LASSERT(obd->obd_type != NULL);
369 LASSERT(obd->obd_type->typ_name != NULL);
371 snprintf(name, MAX_STRING_SIZE, "%s/common_name",
372 obd->obd_type->typ_name);
373 lvars[0].read_fptr = lprocfs_rd_name;
374 err = lprocfs_add_vars(sbi->ll_proc_root, lvars, obd);
378 snprintf(name, MAX_STRING_SIZE, "%s/uuid", obd->obd_type->typ_name);
379 lvars[0].read_fptr = lprocfs_rd_uuid;
380 err = lprocfs_add_vars(sbi->ll_proc_root, lvars, obd);
385 obd = class_name2obd(osc);
387 LASSERT(obd != NULL);
388 LASSERT(obd->obd_type != NULL);
389 LASSERT(obd->obd_type->typ_name != NULL);
391 snprintf(name, MAX_STRING_SIZE, "%s/common_name",
392 obd->obd_type->typ_name);
393 lvars[0].read_fptr = lprocfs_rd_name;
394 err = lprocfs_add_vars(sbi->ll_proc_root, lvars, obd);
398 snprintf(name, MAX_STRING_SIZE, "%s/uuid", obd->obd_type->typ_name);
399 lvars[0].read_fptr = lprocfs_rd_uuid;
400 err = lprocfs_add_vars(sbi->ll_proc_root, lvars, obd);
404 lprocfs_free_stats(svc_stats);
405 if (sbi->ll_proc_root)
406 lprocfs_remove(sbi->ll_proc_root);
411 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi)
413 if (sbi->ll_proc_root) {
414 struct proc_dir_entry *file_stats =
415 lprocfs_srch(sbi->ll_proc_root, "stats");
418 lprocfs_free_stats(sbi->ll_stats);
419 lprocfs_remove(file_stats);
423 #undef MAX_STRING_SIZE
425 static struct ll_async_page *llite_pglist_next_llap(struct ll_sb_info *sbi,
426 struct list_head *list)
428 struct ll_async_page *llap;
429 struct list_head *pos;
431 list_for_each(pos, list) {
432 if (pos == &sbi->ll_pglist)
434 llap = list_entry(pos, struct ll_async_page, llap_proc_item);
435 if (llap->llap_page == NULL)
443 #define seq_page_flag(seq, page, flag, has_flags) do { \
444 if (test_bit(PG_##flag, &(page)->flags)) { \
448 seq_putc(seq, '|'); \
449 seq_puts(seq, #flag); \
453 static int llite_dump_pgcache_seq_show(struct seq_file *seq, void *v)
455 struct ll_async_page *llap, *dummy_llap = seq->private;
456 struct ll_sb_info *sbi = dummy_llap->llap_cookie;
458 /* 2.4 doesn't seem to have SEQ_START_TOKEN, so we implement
459 * it in our own state */
460 if (dummy_llap->llap_magic == 0) {
461 seq_printf(seq, "generation | llap .cookie | page ");
462 seq_printf(seq, "inode .index [ page flags ]\n");
466 spin_lock(&sbi->ll_pglist_lock);
468 llap = llite_pglist_next_llap(sbi, &dummy_llap->llap_proc_item);
471 struct page *page = llap->llap_page;
473 seq_printf(seq, "%lu | %p %p | %p %p %lu [",
475 llap, llap->llap_cookie,
476 page, page->mapping->host, page->index);
477 seq_page_flag(seq, page, locked, has_flags);
478 seq_page_flag(seq, page, error, has_flags);
479 seq_page_flag(seq, page, referenced, has_flags);
480 seq_page_flag(seq, page, uptodate, has_flags);
481 seq_page_flag(seq, page, dirty, has_flags);
482 seq_page_flag(seq, page, highmem, has_flags);
484 seq_puts(seq, "-]\n");
486 seq_puts(seq, "]\n");
489 spin_unlock(&sbi->ll_pglist_lock);
494 static void *llite_dump_pgcache_seq_start(struct seq_file *seq, loff_t *pos)
496 struct ll_async_page *llap = seq->private;
498 if (llap->llap_magic == 2)
504 static void *llite_dump_pgcache_seq_next(struct seq_file *seq, void *v,
507 struct ll_async_page *llap, *dummy_llap = seq->private;
508 struct ll_sb_info *sbi = dummy_llap->llap_cookie;
510 /* bail if we just displayed the banner */
511 if (dummy_llap->llap_magic == 0) {
512 dummy_llap->llap_magic = 1;
516 /* we've just displayed the llap that is after us in the list.
517 * we advance to a position beyond it, returning null if there
518 * isn't another llap in the list beyond that new position. */
519 spin_lock(&sbi->ll_pglist_lock);
520 llap = llite_pglist_next_llap(sbi, &dummy_llap->llap_proc_item);
521 list_del_init(&dummy_llap->llap_proc_item);
523 list_add(&dummy_llap->llap_proc_item, &llap->llap_proc_item);
524 llap = llite_pglist_next_llap(sbi, &dummy_llap->llap_proc_item);
526 spin_unlock(&sbi->ll_pglist_lock);
530 dummy_llap->llap_magic = 2;
536 static void llite_dump_pgcache_seq_stop(struct seq_file *seq, void *v)
540 struct seq_operations llite_dump_pgcache_seq_sops = {
541 .start = llite_dump_pgcache_seq_start,
542 .stop = llite_dump_pgcache_seq_stop,
543 .next = llite_dump_pgcache_seq_next,
544 .show = llite_dump_pgcache_seq_show,
547 /* we're displaying llaps in a list_head list. we don't want to hold a lock
548 * while we walk the entire list, and we don't want to have to seek into
549 * the right position in the list as an app advances with many syscalls. we
550 * allocate a dummy llap and hang it off file->private. its position in
551 * the list records where the app is currently displaying. this way our
552 * seq .start and .stop don't actually do anything. .next returns null
553 * when the dummy hits the end of the list which eventually leads to .release
554 * where we tear down. this kind of displaying is super-racey, so we put
555 * a generation counter on the list so the output shows when the list
556 * changes between reads.
558 static int llite_dump_pgcache_seq_open(struct inode *inode, struct file *file)
560 struct proc_dir_entry *dp = PDE(inode);
561 struct ll_async_page *llap;
562 struct seq_file *seq;
563 struct ll_sb_info *sbi = dp->data;
566 OBD_ALLOC_GFP(llap, sizeof(*llap), GFP_KERNEL);
569 llap->llap_page = NULL;
570 llap->llap_cookie = sbi;
571 llap->llap_magic = 0;
573 rc = seq_open(file, &llite_dump_pgcache_seq_sops);
575 OBD_FREE(llap, sizeof(*llap));
578 seq = file->private_data;
581 spin_lock(&sbi->ll_pglist_lock);
582 list_add(&llap->llap_proc_item, &sbi->ll_pglist);
583 spin_unlock(&sbi->ll_pglist_lock);
588 static int llite_dump_pgcache_seq_release(struct inode *inode,
591 struct seq_file *seq = file->private_data;
592 struct ll_async_page *llap = seq->private;
593 struct ll_sb_info *sbi = llap->llap_cookie;
595 spin_lock(&sbi->ll_pglist_lock);
596 if (!list_empty(&llap->llap_proc_item))
597 list_del_init(&llap->llap_proc_item);
598 spin_unlock(&sbi->ll_pglist_lock);
599 OBD_FREE(llap, sizeof(*llap));
601 return seq_release(inode, file);
604 struct file_operations llite_dump_pgcache_fops = {
605 .owner = THIS_MODULE,
606 .open = llite_dump_pgcache_seq_open,
608 .release = llite_dump_pgcache_seq_release,