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>
29 #include "llite_internal.h"
31 /* /proc/lustre/llite mount point registration */
32 struct proc_dir_entry *proc_lustre_fs_root;
33 struct file_operations llite_dump_pgcache_fops;
36 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
37 struct super_block *sb, char *osc, char *mdc)
41 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi){}
44 long long mnt_instance;
46 static int ll_rd_blksize(char *page, char **start, off_t off, int count,
49 struct super_block *sb = (struct super_block *)data;
50 struct obd_statfs osfs;
54 rc = ll_statfs_internal(sb, &osfs, jiffies - HZ);
57 rc = snprintf(page, count, "%u\n", osfs.os_bsize);
63 static int ll_rd_kbytestotal(char *page, char **start, off_t off, int count,
66 struct super_block *sb = (struct super_block *)data;
67 struct obd_statfs osfs;
71 rc = ll_statfs_internal(sb, &osfs, jiffies - HZ);
73 __u32 blk_size = osfs.os_bsize >> 10;
74 __u64 result = osfs.os_blocks;
76 while (blk_size >>= 1)
80 rc = snprintf(page, count, LPU64"\n", result);
86 static int ll_rd_kbytesfree(char *page, char **start, off_t off, int count,
89 struct super_block *sb = (struct super_block *)data;
90 struct obd_statfs osfs;
94 rc = ll_statfs_internal(sb, &osfs, jiffies - HZ);
96 __u32 blk_size = osfs.os_bsize >> 10;
97 __u64 result = osfs.os_bfree;
99 while (blk_size >>= 1)
103 rc = snprintf(page, count, LPU64"\n", result);
108 static int ll_rd_kbytesavail(char *page, char **start, off_t off, int count,
109 int *eof, void *data)
111 struct super_block *sb = (struct super_block *)data;
112 struct obd_statfs osfs;
116 rc = ll_statfs_internal(sb, &osfs, jiffies - HZ);
118 __u32 blk_size = osfs.os_bsize >> 10;
119 __u64 result = osfs.os_bavail;
121 while (blk_size >>= 1)
125 rc = snprintf(page, count, LPU64"\n", result);
130 static int ll_rd_filestotal(char *page, char **start, off_t off, int count,
131 int *eof, void *data)
133 struct super_block *sb = (struct super_block *)data;
134 struct obd_statfs osfs;
138 rc = ll_statfs_internal(sb, &osfs, jiffies - HZ);
141 rc = snprintf(page, count, LPU64"\n", osfs.os_files);
146 static int ll_rd_filesfree(char *page, char **start, off_t off, int count,
147 int *eof, void *data)
149 struct super_block *sb = (struct super_block *)data;
150 struct obd_statfs osfs;
154 rc = ll_statfs_internal(sb, &osfs, jiffies - HZ);
157 rc = snprintf(page, count, LPU64"\n", osfs.os_ffree);
164 static int ll_rd_path(char *page, char **start, off_t off, int count, int *eof,
171 static int ll_rd_fstype(char *page, char **start, off_t off, int count,
172 int *eof, void *data)
174 struct super_block *sb = (struct super_block*)data;
178 return snprintf(page, count, "%s\n", sb->s_type->name);
181 static int ll_rd_sb_uuid(char *page, char **start, off_t off, int count,
182 int *eof, void *data)
184 struct super_block *sb = (struct super_block *)data;
188 return snprintf(page, count, "%s\n", ll_s2sbi(sb)->ll_sb_uuid.uuid);
191 static int ll_rd_read_ahead(char *page, char **start, off_t off, int count,
192 int *eof, void *data)
194 struct super_block *sb = (struct super_block*)data;
195 struct ll_sb_info *sbi = ll_s2sbi(sb);
200 val = (sbi->ll_flags & LL_SBI_READAHEAD) ? 1 : 0;
201 rc = snprintf(page, count, "%d\n", val);
205 static int ll_wr_read_ahead(struct file *file, const char *buffer,
206 unsigned long count, void *data)
208 struct super_block *sb = (struct super_block*)data;
209 struct ll_sb_info *sbi = ll_s2sbi(sb);
213 if (1 != sscanf(buffer, "%d", &readahead))
217 sbi->ll_flags |= LL_SBI_READAHEAD;
219 sbi->ll_flags &= ~LL_SBI_READAHEAD;
224 static struct lprocfs_vars lprocfs_obd_vars[] = {
225 { "uuid", ll_rd_sb_uuid, 0, 0 },
226 //{ "mntpt_path", ll_rd_path, 0, 0 },
227 { "fstype", ll_rd_fstype, 0, 0 },
228 { "blocksize", ll_rd_blksize, 0, 0 },
229 { "kbytestotal", ll_rd_kbytestotal, 0, 0 },
230 { "kbytesfree", ll_rd_kbytesfree, 0, 0 },
231 { "kbytesavail", ll_rd_kbytesavail, 0, 0 },
232 { "filestotal", ll_rd_filestotal, 0, 0 },
233 { "filesfree", ll_rd_filesfree, 0, 0 },
234 //{ "filegroups", lprocfs_rd_filegroups, 0, 0 },
235 { "read_ahead", ll_rd_read_ahead, ll_wr_read_ahead, 0 },
239 #define MAX_STRING_SIZE 128
241 struct llite_file_opcode {
245 } llite_opcode_table[LPROC_LL_FILE_OPCODES] = {
247 { LPROC_LL_DIRTY_HITS, LPROCFS_TYPE_REGS, "dirty_pages_hits" },
248 { LPROC_LL_DIRTY_MISSES, LPROCFS_TYPE_REGS, "dirty_pages_misses" },
249 { LPROC_LL_WB_WRITEPAGE, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
250 "writeback_from_writepage" },
251 { LPROC_LL_WB_PRESSURE, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
252 "writeback_from_pressure" },
253 { LPROC_LL_WB_OK, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
254 "writeback_ok_pages" },
255 { LPROC_LL_WB_FAIL, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
256 "writeback_failed_pages" },
257 { LPROC_LL_READ_BYTES, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_BYTES,
259 { LPROC_LL_WRITE_BYTES, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_BYTES,
261 { LPROC_LL_BRW_READ, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
263 { LPROC_LL_BRW_WRITE, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
266 { LPROC_LL_IOCTL, LPROCFS_TYPE_REGS, "ioctl" },
267 { LPROC_LL_OPEN, LPROCFS_TYPE_REGS, "open" },
268 { LPROC_LL_RELEASE, LPROCFS_TYPE_REGS, "close" },
269 { LPROC_LL_MAP, LPROCFS_TYPE_REGS, "mmap" },
270 { LPROC_LL_LLSEEK, LPROCFS_TYPE_REGS, "seek" },
271 { LPROC_LL_FSYNC, LPROCFS_TYPE_REGS, "fsync" },
272 /* inode operation */
273 { LPROC_LL_SETATTR, LPROCFS_TYPE_REGS, "setattr" },
274 { LPROC_LL_TRUNC, LPROCFS_TYPE_REGS, "punch" },
275 #if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0))
276 { LPROC_LL_GETATTR, LPROCFS_TYPE_REGS, "getattr" },
278 { LPROC_LL_REVALIDATE, LPROCFS_TYPE_REGS, "getattr" },
280 /* special inode operation */
281 { LPROC_LL_STAFS, LPROCFS_TYPE_REGS, "statfs" },
282 { LPROC_LL_ALLOC_INODE, LPROCFS_TYPE_REGS, "alloc_inode" },
283 { LPROC_LL_DIRECT_READ, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
285 { LPROC_LL_DIRECT_WRITE, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
290 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
291 struct super_block *sb, char *osc, char *mdc)
293 struct lprocfs_vars lvars[2];
294 struct ll_sb_info *sbi = ll_s2sbi(sb);
295 struct obd_device *obd;
296 char name[MAX_STRING_SIZE + 1];
298 struct lprocfs_stats *svc_stats = NULL;
299 struct proc_dir_entry *entry;
302 memset(lvars, 0, sizeof(lvars));
304 name[MAX_STRING_SIZE] = '\0';
305 lvars[0].name = name;
307 LASSERT(sbi != NULL);
308 LASSERT(mdc != NULL);
309 LASSERT(osc != NULL);
312 snprintf(name, MAX_STRING_SIZE, "fs%llu", mnt_instance);
315 sbi->ll_proc_root = lprocfs_register(name, parent, NULL, NULL);
316 if (IS_ERR(sbi->ll_proc_root)) {
317 err = PTR_ERR(sbi->ll_proc_root);
318 sbi->ll_proc_root = NULL;
322 entry = create_proc_entry("dump_page_cache", 0444, sbi->ll_proc_root);
324 GOTO(out, err = -ENOMEM);
325 entry->proc_fops = &llite_dump_pgcache_fops;
328 svc_stats = lprocfs_alloc_stats(LPROC_LL_FILE_OPCODES);
329 if (svc_stats == NULL) {
333 /* do counter init */
334 for (id = 0; id < LPROC_LL_FILE_OPCODES; id++) {
335 __u32 type = llite_opcode_table[id].type;
337 if (type & LPROCFS_TYPE_REGS)
340 if (type & LPROCFS_TYPE_BYTES)
343 if (type & LPROCFS_TYPE_PAGES)
347 lprocfs_counter_init(svc_stats, llite_opcode_table[id].opcode,
348 (type & LPROCFS_CNTR_AVGMINMAX),
349 llite_opcode_table[id].opname, ptr);
351 err = lprocfs_register_stats(sbi->ll_proc_root, "stats", svc_stats);
355 sbi->ll_stats = svc_stats;
356 /* need place to keep svc_stats */
358 /* Static configuration info */
359 err = lprocfs_add_vars(sbi->ll_proc_root, lprocfs_obd_vars, sb);
364 obd = class_name2obd(mdc);
366 LASSERT(obd != NULL);
367 LASSERT(obd->obd_type != NULL);
368 LASSERT(obd->obd_type->typ_name != NULL);
370 snprintf(name, MAX_STRING_SIZE, "%s/common_name",
371 obd->obd_type->typ_name);
372 lvars[0].read_fptr = lprocfs_rd_name;
373 err = lprocfs_add_vars(sbi->ll_proc_root, lvars, obd);
377 snprintf(name, MAX_STRING_SIZE, "%s/uuid", obd->obd_type->typ_name);
378 lvars[0].read_fptr = lprocfs_rd_uuid;
379 err = lprocfs_add_vars(sbi->ll_proc_root, lvars, obd);
384 obd = class_name2obd(osc);
386 LASSERT(obd != NULL);
387 LASSERT(obd->obd_type != NULL);
388 LASSERT(obd->obd_type->typ_name != NULL);
390 snprintf(name, MAX_STRING_SIZE, "%s/common_name",
391 obd->obd_type->typ_name);
392 lvars[0].read_fptr = lprocfs_rd_name;
393 err = lprocfs_add_vars(sbi->ll_proc_root, lvars, obd);
397 snprintf(name, MAX_STRING_SIZE, "%s/uuid", obd->obd_type->typ_name);
398 lvars[0].read_fptr = lprocfs_rd_uuid;
399 err = lprocfs_add_vars(sbi->ll_proc_root, lvars, obd);
403 lprocfs_free_stats(svc_stats);
404 if (sbi->ll_proc_root)
405 lprocfs_remove(sbi->ll_proc_root);
410 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi)
412 if (sbi->ll_proc_root) {
413 struct proc_dir_entry *file_stats =
414 lprocfs_srch(sbi->ll_proc_root, "stats");
417 lprocfs_free_stats(sbi->ll_stats);
418 lprocfs_remove(file_stats);
422 #undef MAX_STRING_SIZE
424 static struct ll_async_page *llite_pglist_next_llap(struct ll_sb_info *sbi,
425 struct list_head *list)
427 struct ll_async_page *llap;
428 struct list_head *pos;
430 list_for_each(pos, list) {
431 if (pos == &sbi->ll_pglist)
433 llap = list_entry(pos, struct ll_async_page, llap_proc_item);
434 if (llap->llap_page == NULL)
442 #define seq_page_flag(seq, page, flag, has_flags) do { \
443 if (test_bit(PG_##flag, &(page)->flags)) { \
447 seq_putc(seq, '|'); \
448 seq_puts(seq, #flag); \
452 static int llite_dump_pgcache_seq_show(struct seq_file *seq, void *v)
454 struct ll_async_page *llap, *dummy_llap = seq->private;
455 struct ll_sb_info *sbi = dummy_llap->llap_cookie;
457 /* 2.4 doesn't seem to have SEQ_START_TOKEN, so we implement
458 * it in our own state */
459 if (dummy_llap->llap_magic == 0) {
460 seq_printf(seq, "generation | llap .cookie | page ");
461 seq_printf(seq, "inode .index [ page flags ]\n");
465 spin_lock(&sbi->ll_pglist_lock);
467 llap = llite_pglist_next_llap(sbi, &dummy_llap->llap_proc_item);
470 struct page *page = llap->llap_page;
472 seq_printf(seq, "%lu | %p %p | %p %p %lu [",
474 llap, llap->llap_cookie,
475 page, page->mapping->host, page->index);
476 seq_page_flag(seq, page, locked, has_flags);
477 seq_page_flag(seq, page, error, has_flags);
478 seq_page_flag(seq, page, referenced, has_flags);
479 seq_page_flag(seq, page, uptodate, has_flags);
480 seq_page_flag(seq, page, dirty, has_flags);
481 seq_page_flag(seq, page, highmem, has_flags);
483 seq_puts(seq, "-]\n");
485 seq_puts(seq, "]\n");
488 spin_unlock(&sbi->ll_pglist_lock);
493 static void *llite_dump_pgcache_seq_start(struct seq_file *seq, loff_t *pos)
495 struct ll_async_page *llap = seq->private;
497 if (llap->llap_magic == 2)
503 static void *llite_dump_pgcache_seq_next(struct seq_file *seq, void *v,
506 struct ll_async_page *llap, *dummy_llap = seq->private;
507 struct ll_sb_info *sbi = dummy_llap->llap_cookie;
509 /* bail if we just displayed the banner */
510 if (dummy_llap->llap_magic == 0) {
511 dummy_llap->llap_magic = 1;
515 /* we've just displayed the llap that is after us in the list.
516 * we advance to a position beyond it, returning null if there
517 * isn't another llap in the list beyond that new position. */
518 spin_lock(&sbi->ll_pglist_lock);
519 llap = llite_pglist_next_llap(sbi, &dummy_llap->llap_proc_item);
520 list_del_init(&dummy_llap->llap_proc_item);
522 list_add(&dummy_llap->llap_proc_item, &llap->llap_proc_item);
523 llap = llite_pglist_next_llap(sbi, &dummy_llap->llap_proc_item);
525 spin_unlock(&sbi->ll_pglist_lock);
529 dummy_llap->llap_magic = 2;
535 static void llite_dump_pgcache_seq_stop(struct seq_file *seq, void *v)
539 struct seq_operations llite_dump_pgcache_seq_sops = {
540 .start = llite_dump_pgcache_seq_start,
541 .stop = llite_dump_pgcache_seq_stop,
542 .next = llite_dump_pgcache_seq_next,
543 .show = llite_dump_pgcache_seq_show,
546 /* we're displaying llaps in a list_head list. we don't want to hold a lock
547 * while we walk the entire list, and we don't want to have to seek into
548 * the right position in the list as an app advances with many syscalls. we
549 * allocate a dummy llap and hang it off file->private. its position in
550 * the list records where the app is currently displaying. this way our
551 * seq .start and .stop don't actually do anything. .next returns null
552 * when the dummy hits the end of the list which eventually leads to .release
553 * where we tear down. this kind of displaying is super-racey, so we put
554 * a generation counter on the list so the output shows when the list
555 * changes between reads.
557 static int llite_dump_pgcache_seq_open(struct inode *inode, struct file *file)
559 struct proc_dir_entry *dp = PDE(inode);
560 struct ll_async_page *llap;
561 struct seq_file *seq;
562 struct ll_sb_info *sbi = dp->data;
565 llap = kmalloc(sizeof(*llap), GFP_KERNEL);
568 llap->llap_page = NULL;
569 llap->llap_cookie = sbi;
570 llap->llap_magic = 0;
572 rc = seq_open(file, &llite_dump_pgcache_seq_sops);
577 seq = file->private_data;
580 spin_lock(&sbi->ll_pglist_lock);
581 list_add(&llap->llap_proc_item, &sbi->ll_pglist);
582 spin_unlock(&sbi->ll_pglist_lock);
587 static int llite_dump_pgcache_seq_release(struct inode *inode,
590 struct seq_file *seq = file->private_data;
591 struct ll_async_page *llap = seq->private;
592 struct ll_sb_info *sbi = llap->llap_cookie;
594 spin_lock(&sbi->ll_pglist_lock);
595 if (!list_empty(&llap->llap_proc_item))
596 list_del_init(&llap->llap_proc_item);
597 spin_unlock(&sbi->ll_pglist_lock);
600 return seq_release(inode, file);
603 struct file_operations llite_dump_pgcache_fops = {
604 .open = llite_dump_pgcache_seq_open,
606 .release = llite_dump_pgcache_seq_release,