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_filestotal(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);
119 rc = snprintf(page, count, LPU64"\n", osfs.os_files);
124 static int ll_rd_filesfree(char *page, char **start, off_t off, int count,
125 int *eof, void *data)
127 struct super_block *sb = (struct super_block *)data;
128 struct obd_statfs osfs;
132 rc = ll_statfs_internal(sb, &osfs, jiffies - HZ);
135 rc = snprintf(page, count, LPU64"\n", osfs.os_ffree);
142 static int ll_rd_path(char *page, char **start, off_t off, int count, int *eof,
149 static int ll_rd_fstype(char *page, char **start, off_t off, int count,
150 int *eof, void *data)
152 struct super_block *sb = (struct super_block*)data;
156 return snprintf(page, count, "%s\n", sb->s_type->name);
159 static int ll_rd_sb_uuid(char *page, char **start, off_t off, int count,
160 int *eof, void *data)
162 struct super_block *sb = (struct super_block *)data;
166 return snprintf(page, count, "%s\n", ll_s2sbi(sb)->ll_sb_uuid.uuid);
169 static int ll_rd_read_ahead(char *page, char **start, off_t off, int count,
170 int *eof, void *data)
172 struct super_block *sb = (struct super_block*)data;
173 struct ll_sb_info *sbi = ll_s2sbi(sb);
178 val = (sbi->ll_flags & LL_SBI_READAHEAD) ? 1 : 0;
179 rc = snprintf(page, count, "%d\n", val);
183 static int ll_wr_read_ahead(struct file *file, const char *buffer,
184 unsigned long count, void *data)
186 struct super_block *sb = (struct super_block*)data;
187 struct ll_sb_info *sbi = ll_s2sbi(sb);
191 if (1 != sscanf(buffer, "%d", &readahead))
195 sbi->ll_flags |= LL_SBI_READAHEAD;
197 sbi->ll_flags &= ~LL_SBI_READAHEAD;
202 static struct lprocfs_vars lprocfs_obd_vars[] = {
203 { "uuid", ll_rd_sb_uuid, 0, 0 },
204 //{ "mntpt_path", ll_rd_path, 0, 0 },
205 { "fstype", ll_rd_fstype, 0, 0 },
206 { "blocksize", ll_rd_blksize, 0, 0 },
207 { "kbytestotal", ll_rd_kbytestotal, 0, 0 },
208 { "kbytesfree", ll_rd_kbytesfree, 0, 0 },
209 { "filestotal", ll_rd_filestotal, 0, 0 },
210 { "filesfree", ll_rd_filesfree, 0, 0 },
211 //{ "filegroups", lprocfs_rd_filegroups, 0, 0 },
212 { "read_ahead", ll_rd_read_ahead, ll_wr_read_ahead, 0 },
216 #define MAX_STRING_SIZE 128
218 struct llite_file_opcode {
222 } llite_opcode_table[LPROC_LL_FILE_OPCODES] = {
224 { LPROC_LL_DIRTY_HITS, LPROCFS_TYPE_REGS, "dirty_pages_hits" },
225 { LPROC_LL_DIRTY_MISSES, LPROCFS_TYPE_REGS, "dirty_pages_misses" },
226 { LPROC_LL_WB_WRITEPAGE, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
227 "writeback_from_writepage" },
228 { LPROC_LL_WB_PRESSURE, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
229 "writeback_from_pressure" },
230 { LPROC_LL_WB_OK, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
231 "writeback_ok_pages" },
232 { LPROC_LL_WB_FAIL, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
233 "writeback_failed_pages" },
234 { LPROC_LL_READ_BYTES, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_BYTES,
236 { LPROC_LL_WRITE_BYTES, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_BYTES,
238 { LPROC_LL_BRW_READ, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
240 { LPROC_LL_BRW_WRITE, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
243 { LPROC_LL_IOCTL, LPROCFS_TYPE_REGS, "ioctl" },
244 { LPROC_LL_OPEN, LPROCFS_TYPE_REGS, "open" },
245 { LPROC_LL_RELEASE, LPROCFS_TYPE_REGS, "close" },
246 { LPROC_LL_MAP, LPROCFS_TYPE_REGS, "mmap" },
247 { LPROC_LL_LLSEEK, LPROCFS_TYPE_REGS, "seek" },
248 { LPROC_LL_FSYNC, LPROCFS_TYPE_REGS, "fsync" },
249 /* inode operation */
250 { LPROC_LL_SETATTR, LPROCFS_TYPE_REGS, "setattr" },
251 { LPROC_LL_TRUNC, LPROCFS_TYPE_REGS, "punch" },
252 #if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0))
253 { LPROC_LL_GETATTR, LPROCFS_TYPE_REGS, "getattr" },
255 { LPROC_LL_REVALIDATE, LPROCFS_TYPE_REGS, "getattr" },
257 /* special inode operation */
258 { LPROC_LL_STAFS, LPROCFS_TYPE_REGS, "statfs" },
259 { LPROC_LL_ALLOC_INODE, LPROCFS_TYPE_REGS, "alloc_inode" },
260 { LPROC_LL_DIRECT_READ, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
262 { LPROC_LL_DIRECT_WRITE, LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
267 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
268 struct super_block *sb, char *osc, char *mdc)
270 struct lprocfs_vars lvars[2];
271 struct ll_sb_info *sbi = ll_s2sbi(sb);
272 struct obd_device *obd;
273 char name[MAX_STRING_SIZE + 1];
275 struct lprocfs_stats *svc_stats = NULL;
276 struct proc_dir_entry *entry;
279 memset(lvars, 0, sizeof(lvars));
281 name[MAX_STRING_SIZE] = '\0';
282 lvars[0].name = name;
284 LASSERT(sbi != NULL);
285 LASSERT(mdc != NULL);
286 LASSERT(osc != NULL);
289 snprintf(name, MAX_STRING_SIZE, "fs%llu", mnt_instance);
292 sbi->ll_proc_root = lprocfs_register(name, parent, NULL, NULL);
293 if (IS_ERR(sbi->ll_proc_root)) {
294 err = PTR_ERR(sbi->ll_proc_root);
295 sbi->ll_proc_root = NULL;
299 entry = create_proc_entry("dump_page_cache", 0444, sbi->ll_proc_root);
301 GOTO(out, err = -ENOMEM);
302 entry->proc_fops = &llite_dump_pgcache_fops;
305 svc_stats = lprocfs_alloc_stats(LPROC_LL_FILE_OPCODES);
306 if (svc_stats == NULL) {
310 /* do counter init */
311 for (id = 0; id < LPROC_LL_FILE_OPCODES; id++) {
312 __u32 type = llite_opcode_table[id].type;
314 if (type & LPROCFS_TYPE_REGS)
317 if (type & LPROCFS_TYPE_BYTES)
320 if (type & LPROCFS_TYPE_PAGES)
324 lprocfs_counter_init(svc_stats, llite_opcode_table[id].opcode,
325 (type & LPROCFS_CNTR_AVGMINMAX),
326 llite_opcode_table[id].opname, ptr);
328 err = lprocfs_register_stats(sbi->ll_proc_root, "stats", svc_stats);
332 sbi->ll_stats = svc_stats;
333 /* need place to keep svc_stats */
335 /* Static configuration info */
336 err = lprocfs_add_vars(sbi->ll_proc_root, lprocfs_obd_vars, sb);
341 obd = class_name2obd(mdc);
343 LASSERT(obd != NULL);
344 LASSERT(obd->obd_type != NULL);
345 LASSERT(obd->obd_type->typ_name != NULL);
347 snprintf(name, MAX_STRING_SIZE, "%s/common_name",
348 obd->obd_type->typ_name);
349 lvars[0].read_fptr = lprocfs_rd_name;
350 err = lprocfs_add_vars(sbi->ll_proc_root, lvars, obd);
354 snprintf(name, MAX_STRING_SIZE, "%s/uuid", obd->obd_type->typ_name);
355 lvars[0].read_fptr = lprocfs_rd_uuid;
356 err = lprocfs_add_vars(sbi->ll_proc_root, lvars, obd);
361 obd = class_name2obd(osc);
363 LASSERT(obd != NULL);
364 LASSERT(obd->obd_type != NULL);
365 LASSERT(obd->obd_type->typ_name != NULL);
367 snprintf(name, MAX_STRING_SIZE, "%s/common_name",
368 obd->obd_type->typ_name);
369 lvars[0].read_fptr = lprocfs_rd_name;
370 err = lprocfs_add_vars(sbi->ll_proc_root, lvars, obd);
374 snprintf(name, MAX_STRING_SIZE, "%s/uuid", obd->obd_type->typ_name);
375 lvars[0].read_fptr = lprocfs_rd_uuid;
376 err = lprocfs_add_vars(sbi->ll_proc_root, lvars, obd);
380 lprocfs_free_stats(svc_stats);
381 if (sbi->ll_proc_root)
382 lprocfs_remove(sbi->ll_proc_root);
387 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi)
389 if (sbi->ll_proc_root) {
390 struct proc_dir_entry *file_stats =
391 lprocfs_srch(sbi->ll_proc_root, "stats");
394 lprocfs_free_stats(sbi->ll_stats);
395 lprocfs_remove(file_stats);
399 #undef MAX_STRING_SIZE
401 static struct ll_async_page *llite_pglist_next_llap(struct ll_sb_info *sbi,
402 struct list_head *list)
404 struct ll_async_page *llap;
405 struct list_head *pos;
407 list_for_each(pos, list) {
408 if (pos == &sbi->ll_pglist)
410 llap = list_entry(pos, struct ll_async_page, llap_proc_item);
411 if (llap->llap_page == NULL)
419 #define seq_page_flag(seq, page, flag, has_flags) do { \
420 if (test_bit(PG_##flag, &(page)->flags)) { \
424 seq_putc(seq, '|'); \
425 seq_puts(seq, #flag); \
429 static int llite_dump_pgcache_seq_show(struct seq_file *seq, void *v)
431 struct ll_async_page *llap, *dummy_llap = seq->private;
432 struct ll_sb_info *sbi = dummy_llap->llap_cookie;
434 /* 2.4 doesn't seem to have SEQ_START_TOKEN, so we implement
435 * it in our own state */
436 if (dummy_llap->llap_magic == 0) {
437 seq_printf(seq, "generation | llap .cookie | page ");
438 seq_printf(seq, "inode .index [ page flags ]\n");
442 spin_lock(&sbi->ll_pglist_lock);
444 llap = llite_pglist_next_llap(sbi, &dummy_llap->llap_proc_item);
447 struct page *page = llap->llap_page;
449 seq_printf(seq, "%lu | %p %p | %p %p %lu [",
451 llap, llap->llap_cookie,
452 page, page->mapping->host, page->index);
453 seq_page_flag(seq, page, locked, has_flags);
454 seq_page_flag(seq, page, error, has_flags);
455 seq_page_flag(seq, page, referenced, has_flags);
456 seq_page_flag(seq, page, uptodate, has_flags);
457 seq_page_flag(seq, page, dirty, has_flags);
458 seq_page_flag(seq, page, highmem, has_flags);
460 seq_puts(seq, "-]\n");
462 seq_puts(seq, "]\n");
465 spin_unlock(&sbi->ll_pglist_lock);
470 static void *llite_dump_pgcache_seq_start(struct seq_file *seq, loff_t *pos)
472 struct ll_async_page *llap = seq->private;
474 if (llap->llap_magic == 2)
480 static void *llite_dump_pgcache_seq_next(struct seq_file *seq, void *v,
483 struct ll_async_page *llap, *dummy_llap = seq->private;
484 struct ll_sb_info *sbi = dummy_llap->llap_cookie;
486 /* bail if we just displayed the banner */
487 if (dummy_llap->llap_magic == 0) {
488 dummy_llap->llap_magic = 1;
492 /* we've just displayed the llap that is after us in the list.
493 * we advance to a position beyond it, returning null if there
494 * isn't another llap in the list beyond that new position. */
495 spin_lock(&sbi->ll_pglist_lock);
496 llap = llite_pglist_next_llap(sbi, &dummy_llap->llap_proc_item);
497 list_del_init(&dummy_llap->llap_proc_item);
499 list_add(&dummy_llap->llap_proc_item, &llap->llap_proc_item);
500 llap = llite_pglist_next_llap(sbi, &dummy_llap->llap_proc_item);
502 spin_unlock(&sbi->ll_pglist_lock);
506 dummy_llap->llap_magic = 2;
512 static void llite_dump_pgcache_seq_stop(struct seq_file *seq, void *v)
516 struct seq_operations llite_dump_pgcache_seq_sops = {
517 .start = llite_dump_pgcache_seq_start,
518 .stop = llite_dump_pgcache_seq_stop,
519 .next = llite_dump_pgcache_seq_next,
520 .show = llite_dump_pgcache_seq_show,
523 /* we're displaying llaps in a list_head list. we don't want to hold a lock
524 * while we walk the entire list, and we don't want to have to seek into
525 * the right position in the list as an app advances with many syscalls. we
526 * allocate a dummy llap and hang it off file->private. its position in
527 * the list records where the app is currently displaying. this way our
528 * seq .start and .stop don't actually do anything. .next returns null
529 * when the dummy hits the end of the list which eventually leads to .release
530 * where we tear down. this kind of displaying is super-racey, so we put
531 * a generation counter on the list so the output shows when the list
532 * changes between reads.
534 static int llite_dump_pgcache_seq_open(struct inode *inode, struct file *file)
536 struct proc_dir_entry *dp = PDE(inode);
537 struct ll_async_page *llap;
538 struct seq_file *seq;
539 struct ll_sb_info *sbi = dp->data;
542 llap = kmalloc(sizeof(*llap), GFP_KERNEL);
545 llap->llap_page = NULL;
546 llap->llap_cookie = sbi;
547 llap->llap_magic = 0;
549 rc = seq_open(file, &llite_dump_pgcache_seq_sops);
554 seq = file->private_data;
557 spin_lock(&sbi->ll_pglist_lock);
558 list_add(&llap->llap_proc_item, &sbi->ll_pglist);
559 spin_unlock(&sbi->ll_pglist_lock);
564 static int llite_dump_pgcache_seq_release(struct inode *inode,
567 struct seq_file *seq = file->private_data;
568 struct ll_async_page *llap = seq->private;
569 struct ll_sb_info *sbi = llap->llap_cookie;
571 spin_lock(&sbi->ll_pglist_lock);
572 if (!list_empty(&llap->llap_proc_item))
573 list_del_init(&llap->llap_proc_item);
574 spin_unlock(&sbi->ll_pglist_lock);
577 return seq_release(inode, file);
580 struct file_operations llite_dump_pgcache_fops = {
581 .open = llite_dump_pgcache_seq_open,
583 .release = llite_dump_pgcache_seq_release,