X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Fobdclass%2Flprocfs_status.c;h=c3f0249d76c66952e816b52d06f3bd6d986ab6c4;hb=c7a42267f5740d587f726e610695f1c72818a21e;hp=01ac77bf5c783581ace303aae55b8e0dc77f36de;hpb=98a2ab6915c16fd8745437bfed71889588880e12;p=fs%2Flustre-release.git diff --git a/lustre/obdclass/lprocfs_status.c b/lustre/obdclass/lprocfs_status.c index 01ac77b..c3f0249 100644 --- a/lustre/obdclass/lprocfs_status.c +++ b/lustre/obdclass/lprocfs_status.c @@ -1,26 +1,41 @@ /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*- * vim:expandtab:shiftwidth=8:tabstop=8: * - * Copyright (C) 2002, 2003 Cluster File Systems, Inc. - * Author: Hariharan Thantry + * GPL HEADER START * - * This file is part of the Lustre file system, http://www.lustre.org - * Lustre is a trademark of Cluster File Systems, Inc. + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * - * You may have signed or agreed to another license before downloading - * this software. If so, you are bound by the terms and conditions - * of that agreement, and the following does not apply to you. See the - * LICENSE file included with this distribution for more information. + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. * - * If you did not agree to a different license, then this copy of Lustre - * is open source software; you can redistribute it and/or modify it - * under the terms of version 2 of the GNU General Public License as - * published by the Free Software Foundation. + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). * - * In either case, Lustre is distributed in the hope that it will be - * useful, but WITHOUT ANY WARRANTY; without even the implied warranty - * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - * license text for more details. + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf + * + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright 2008 Sun Microsystems, Inc. All rights reserved + * Use is subject to license terms. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + * + * lustre/obdclass/lprocfs_status.c + * + * Author: Hariharan Thantry */ #ifndef EXPORT_SYMTAB @@ -35,6 +50,8 @@ #include #include #include +#include +#include #if defined(LPROCFS) @@ -58,14 +75,18 @@ struct proc_dir_entry *lprocfs_srch(struct proc_dir_entry *head, if (head == NULL) return NULL; + LPROCFS_ENTRY(); temp = head->subdir; while (temp != NULL) { - if (strcmp(temp->name, name) == 0) + if (strcmp(temp->name, name) == 0) { + LPROCFS_EXIT(); return temp; + } temp = temp->next; } + LPROCFS_EXIT(); return NULL; } @@ -92,28 +113,50 @@ static int lprocfs_obd_snprintf(char **page, int end, int *len, return n; } -int lprocfs_add_simple(struct proc_dir_entry *root, char *name, - read_proc_t *read_proc, write_proc_t *write_proc, - void *data) +cfs_proc_dir_entry_t *lprocfs_add_simple(struct proc_dir_entry *root, + char *name, + read_proc_t *read_proc, + write_proc_t *write_proc, + void *data, + struct file_operations *fops) { - struct proc_dir_entry *proc; + cfs_proc_dir_entry_t *proc; mode_t mode = 0; - + if (root == NULL || name == NULL) - return -EINVAL; + return ERR_PTR(-EINVAL); if (read_proc) mode = 0444; if (write_proc) mode |= 0200; + if (fops) + mode = 0644; proc = create_proc_entry(name, mode, root); if (!proc) { CERROR("LprocFS: No memory to create /proc entry %s", name); - return -ENOMEM; + return ERR_PTR(-ENOMEM); } proc->read_proc = read_proc; proc->write_proc = write_proc; proc->data = data; - return 0; + if (fops) + proc->proc_fops = fops; + return proc; +} + +struct proc_dir_entry *lprocfs_add_symlink(const char *name, + struct proc_dir_entry *parent, const char *dest) +{ + struct proc_dir_entry *entry; + + if (parent == NULL || dest == NULL) + return NULL; + + entry = proc_symlink(name, parent, dest); + if (entry == NULL) + CERROR("LprocFS: Could not create symbolic link from %s to %s", + name, dest); + return entry; } static ssize_t lprocfs_fops_read(struct file *f, char __user *buf, @@ -133,8 +176,8 @@ static ssize_t lprocfs_fops_read(struct file *f, char __user *buf, LPROCFS_ENTRY(); OBD_FAIL_TIMEOUT(OBD_FAIL_LPROC_REMOVE, 10); if (!dp->deleted && dp->read_proc) - rc = dp->read_proc(page, &start, *ppos, CFS_PAGE_SIZE, - &eof, dp->data); + rc = dp->read_proc(page, &start, *ppos, CFS_PAGE_SIZE, + &eof, dp->data); LPROCFS_EXIT(); if (rc <= 0) goto out; @@ -165,10 +208,11 @@ out: return rc; } -static ssize_t lprocfs_fops_write(struct file *f, const char __user *buf, size_t size, loff_t *ppos) +static ssize_t lprocfs_fops_write(struct file *f, const char __user *buf, + size_t size, loff_t *ppos) { struct proc_dir_entry *dp = PDE(f->f_dentry->d_inode); - int rc = 0; + int rc = -EIO; LPROCFS_ENTRY(); if (!dp->deleted && dp->write_proc) @@ -213,6 +257,17 @@ struct file_operations lprocfs_evict_client_fops = { }; EXPORT_SYMBOL(lprocfs_evict_client_fops); +/** + * Add /proc entries. + * + * \param root [in] The parent proc entry on which new entry will be added. + * \param list [in] Array of proc entries to be added. + * \param data [in] The argument to be passed when entries read/write routines + * are called through /proc file. + * + * \retval 0 on success + * < 0 on error + */ int lprocfs_add_vars(struct proc_dir_entry *root, struct lprocfs_vars *list, void *data) { @@ -252,10 +307,14 @@ int lprocfs_add_vars(struct proc_dir_entry *root, struct lprocfs_vars *list, proc_mkdir(cur, cur_root)); } else if (proc == NULL) { mode_t mode = 0; - if (list->read_fptr) - mode = 0444; - if (list->write_fptr) - mode |= 0200; + if (list->proc_mode != 0000) { + mode = list->proc_mode; + } else { + if (list->read_fptr) + mode = 0444; + if (list->write_fptr) + mode |= 0200; + } proc = create_proc_entry(cur, mode, cur_root); } } @@ -288,13 +347,14 @@ void lprocfs_remove(struct proc_dir_entry **rooth) struct proc_dir_entry *rm_entry; struct proc_dir_entry *parent; - if (!root) + if (!root) return; *rooth = NULL; parent = root->parent; LASSERT(parent != NULL); - + LPROCFS_WRITE_ENTRY(); /* search vs remove race */ + while (1) { while (temp->subdir != NULL) temp = temp->subdir; @@ -308,15 +368,20 @@ void lprocfs_remove(struct proc_dir_entry **rooth) "0x%p %s/%s len %d\n", rm_entry, temp->name, rm_entry->name, (int)strlen(rm_entry->name)); - /* Now, the rm_entry->deleted flags is protected + /* Now, the rm_entry->deleted flags is protected * by _lprocfs_lock. */ - down_write(&_lprocfs_lock); rm_entry->data = NULL; - remove_proc_entry(rm_entry->name, rm_entry->parent); - up_write(&_lprocfs_lock); + remove_proc_entry(rm_entry->name, temp); if (temp == parent) break; } + LPROCFS_WRITE_EXIT(); +} + +void lprocfs_remove_proc_entry(const char *name, struct proc_dir_entry *parent) +{ + LASSERT(parent != NULL); + remove_proc_entry(name, parent); } struct proc_dir_entry *lprocfs_register(const char *name, @@ -347,7 +412,7 @@ struct proc_dir_entry *lprocfs_register(const char *name, int lprocfs_rd_uint(char *page, char **start, off_t off, int count, int *eof, void *data) { - unsigned int *temp = (unsigned int *)data; + unsigned int *temp = data; return snprintf(page, count, "%u\n", *temp); } @@ -381,7 +446,7 @@ int lprocfs_rd_u64(char *page, char **start, off_t off, int lprocfs_rd_atomic(char *page, char **start, off_t off, int count, int *eof, void *data) { - atomic_t *atom = (atomic_t *)data; + atomic_t *atom = data; LASSERT(atom != NULL); *eof = 1; return snprintf(page, count, "%d\n", atomic_read(atom)); @@ -393,14 +458,14 @@ int lprocfs_wr_atomic(struct file *file, const char *buffer, atomic_t *atm = data; int val = 0; int rc; - + rc = lprocfs_write_helper(buffer, count, &val); if (rc < 0) return rc; if (val <= 0) return -ERANGE; - + atomic_set(atm, val); return count; } @@ -408,7 +473,7 @@ int lprocfs_wr_atomic(struct file *file, const char *buffer, int lprocfs_rd_uuid(char *page, char **start, off_t off, int count, int *eof, void *data) { - struct obd_device *obd = (struct obd_device*)data; + struct obd_device *obd = data; LASSERT(obd != NULL); *eof = 1; @@ -416,9 +481,9 @@ int lprocfs_rd_uuid(char *page, char **start, off_t off, int count, } int lprocfs_rd_name(char *page, char **start, off_t off, int count, - int *eof, void* data) + int *eof, void *data) { - struct obd_device *dev = (struct obd_device *)data; + struct obd_device *dev = data; LASSERT(dev != NULL); LASSERT(dev->obd_name != NULL); @@ -429,7 +494,7 @@ int lprocfs_rd_name(char *page, char **start, off_t off, int count, int lprocfs_rd_fstype(char *page, char **start, off_t off, int count, int *eof, void *data) { - struct obd_device *obd = (struct obd_device *)data; + struct obd_device *obd = data; LASSERT(obd != NULL); LASSERT(obd->obd_fsops != NULL); @@ -441,7 +506,8 @@ int lprocfs_rd_blksize(char *page, char **start, off_t off, int count, int *eof, void *data) { struct obd_statfs osfs; - int rc = obd_statfs(data, &osfs, cfs_time_current_64() - HZ); + int rc = obd_statfs(data, &osfs, cfs_time_current_64() - HZ, + OBD_STATFS_NODELAY); if (!rc) { *eof = 1; rc = snprintf(page, count, "%u\n", osfs.os_bsize); @@ -453,7 +519,8 @@ int lprocfs_rd_kbytestotal(char *page, char **start, off_t off, int count, int *eof, void *data) { struct obd_statfs osfs; - int rc = obd_statfs(data, &osfs, cfs_time_current_64() - HZ); + int rc = obd_statfs(data, &osfs, cfs_time_current_64() - HZ, + OBD_STATFS_NODELAY); if (!rc) { __u32 blk_size = osfs.os_bsize >> 10; __u64 result = osfs.os_blocks; @@ -471,7 +538,8 @@ int lprocfs_rd_kbytesfree(char *page, char **start, off_t off, int count, int *eof, void *data) { struct obd_statfs osfs; - int rc = obd_statfs(data, &osfs, cfs_time_current_64() - HZ); + int rc = obd_statfs(data, &osfs, cfs_time_current_64() - HZ, + OBD_STATFS_NODELAY); if (!rc) { __u32 blk_size = osfs.os_bsize >> 10; __u64 result = osfs.os_bfree; @@ -489,7 +557,8 @@ int lprocfs_rd_kbytesavail(char *page, char **start, off_t off, int count, int *eof, void *data) { struct obd_statfs osfs; - int rc = obd_statfs(data, &osfs, cfs_time_current_64() - HZ); + int rc = obd_statfs(data, &osfs, cfs_time_current_64() - HZ, + OBD_STATFS_NODELAY); if (!rc) { __u32 blk_size = osfs.os_bsize >> 10; __u64 result = osfs.os_bavail; @@ -507,7 +576,8 @@ int lprocfs_rd_filestotal(char *page, char **start, off_t off, int count, int *eof, void *data) { struct obd_statfs osfs; - int rc = obd_statfs(data, &osfs, cfs_time_current_64() - HZ); + int rc = obd_statfs(data, &osfs, cfs_time_current_64() - HZ, + OBD_STATFS_NODELAY); if (!rc) { *eof = 1; rc = snprintf(page, count, LPU64"\n", osfs.os_files); @@ -520,7 +590,8 @@ int lprocfs_rd_filesfree(char *page, char **start, off_t off, int count, int *eof, void *data) { struct obd_statfs osfs; - int rc = obd_statfs(data, &osfs, cfs_time_current_64() - HZ); + int rc = obd_statfs(data, &osfs, cfs_time_current_64() - HZ, + OBD_STATFS_NODELAY); if (!rc) { *eof = 1; rc = snprintf(page, count, LPU64"\n", osfs.os_ffree); @@ -531,7 +602,7 @@ int lprocfs_rd_filesfree(char *page, char **start, off_t off, int count, int lprocfs_rd_server_uuid(char *page, char **start, off_t off, int count, int *eof, void *data) { - struct obd_device *obd = (struct obd_device *)data; + struct obd_device *obd = data; struct obd_import *imp; char *imp_state_name = NULL; int rc = 0; @@ -552,7 +623,7 @@ int lprocfs_rd_server_uuid(char *page, char **start, off_t off, int count, int lprocfs_rd_conn_uuid(char *page, char **start, off_t off, int count, int *eof, void *data) { - struct obd_device *obd = (struct obd_device*)data; + struct obd_device *obd = data; struct ptlrpc_connection *conn; int rc = 0; @@ -573,6 +644,77 @@ int lprocfs_rd_conn_uuid(char *page, char **start, off_t off, int count, return rc; } +/** add up per-cpu counters */ +void lprocfs_stats_collect(struct lprocfs_stats *stats, int idx, + struct lprocfs_counter *cnt) +{ + unsigned int num_cpu; + struct lprocfs_counter t; + struct lprocfs_counter *percpu_cntr; + int centry, i; + + memset(cnt, 0, sizeof(*cnt)); + + if (stats == NULL) { + /* set count to 1 to avoid divide-by-zero errs in callers */ + cnt->lc_count = 1; + return; + } + + cnt->lc_min = LC_MIN_INIT; + + if (stats->ls_flags & LPROCFS_STATS_FLAG_NOPERCPU) + num_cpu = 1; + else + num_cpu = num_possible_cpus(); + + for (i = 0; i < num_cpu; i++) { + percpu_cntr = &(stats->ls_percpu[i])->lp_cntr[idx]; + + do { + centry = atomic_read(&percpu_cntr->lc_cntl.la_entry); + t.lc_count = percpu_cntr->lc_count; + t.lc_sum = percpu_cntr->lc_sum; + t.lc_min = percpu_cntr->lc_min; + t.lc_max = percpu_cntr->lc_max; + t.lc_sumsquare = percpu_cntr->lc_sumsquare; + } while (centry != atomic_read(&percpu_cntr->lc_cntl.la_entry) && + centry != atomic_read(&percpu_cntr->lc_cntl.la_exit)); + cnt->lc_count += t.lc_count; + cnt->lc_sum += t.lc_sum; + if (t.lc_min < cnt->lc_min) + cnt->lc_min = t.lc_min; + if (t.lc_max > cnt->lc_max) + cnt->lc_max = t.lc_max; + cnt->lc_sumsquare += t.lc_sumsquare; + } + + cnt->lc_units = stats->ls_percpu[0]->lp_cntr[idx].lc_units; +} + +/** + * Append a space separated list of current set flags to str. + */ +#define flag2str(flag) \ + if (imp->imp_##flag && max - len > 0) \ + len += snprintf(str + len, max - len, "%s" #flag, len ? ", " : ""); +static int obd_import_flags2str(struct obd_import *imp, char *str, int max) +{ + int len = 0; + + if (imp->imp_obd->obd_no_recov) + len += snprintf(str, max - len, "no_recov"); + + flag2str(invalid); + flag2str(deactive); + flag2str(replayable); + flag2str(pingable); + flag2str(recon_bk); + flag2str(last_recon); + return len; +} +#undef flags2str + static const char *obd_connect_names[] = { "read_only", "lov_index", @@ -597,34 +739,280 @@ static const char *obd_connect_names[] = { "mds_capability", "oss_capability", "early_lock_cancel", - "size_on_mds", + "som", "adaptive_timeouts", "lru_resize", "mds_mds_connection", "real_conn", + "change_qunit_size", + "alt_checksum_algorithm", + "fid_is_enabled", + "version_recovery", + "pools", + "grant_shrink", + "skip_orphan", NULL }; -int lprocfs_rd_connect_flags(char *page, char **start, off_t off, - int count, int *eof, void *data) +static int obd_connect_flags2str(char *page, int count, __u64 flags, char *sep) { - struct obd_device *obd = data; - __u64 mask = 1, flags; + __u64 mask = 1; int i, ret = 0; - LPROCFS_CLIMP_CHECK(obd); - flags = obd->u.cli.cl_import->imp_connect_data.ocd_connect_flags; - ret = snprintf(page, count, "flags="LPX64"\n", flags); for (i = 0; obd_connect_names[i] != NULL; i++, mask <<= 1) { if (flags & mask) - ret += snprintf(page + ret, count - ret, "%s\n", - obd_connect_names[i]); + ret += snprintf(page + ret, count - ret, "%s%s", + ret ? sep : "", obd_connect_names[i]); } if (flags & ~(mask - 1)) ret += snprintf(page + ret, count - ret, - "unknown flags "LPX64"\n", flags & ~(mask - 1)); + "%sunknown flags "LPX64, + ret ? sep : "", flags & ~(mask - 1)); + return ret; +} + +int lprocfs_rd_import(char *page, char **start, off_t off, int count, + int *eof, void *data) +{ + struct lprocfs_counter ret; + struct obd_device *obd = (struct obd_device *)data; + struct obd_import *imp; + int i, j, k, rw = 0; + + LASSERT(obd != NULL); + LPROCFS_CLIMP_CHECK(obd); + imp = obd->u.cli.cl_import; + *eof = 1; + + i = snprintf(page, count, + "import:\n" + " name: %s\n" + " target: %s\n" + " current_connection: %s\n" + " state: %s\n" + " connect_flags: [", + obd->obd_name, + obd2cli_tgt(obd), + imp->imp_connection->c_remote_uuid.uuid, + ptlrpc_import_state_name(imp->imp_state)); + i += obd_connect_flags2str(page + i, count - i, + imp->imp_connect_data.ocd_connect_flags, + ", "); + i += snprintf(page + i, count - i, + "]\n" + " import_flags: ["); + i += obd_import_flags2str(imp, page + i, count - i); + + i += snprintf(page + i, count - i, + "]\n" + " connection:\n" + " connection_attempts: %u\n" + " generation: %u\n" + " in-progress_invalidations: %u\n", + imp->imp_conn_cnt, + imp->imp_generation, + atomic_read(&imp->imp_inval_count)); + + lprocfs_stats_collect(obd->obd_svc_stats, PTLRPC_REQWAIT_CNTR, &ret); + do_div(ret.lc_sum, ret.lc_count); + i += snprintf(page + i, count - i, + " rpcs:\n" + " inflight: %u\n" + " unregistering: %u\n" + " timeouts: %u\n" + " avg_waittime: "LPU64" %s\n", + atomic_read(&imp->imp_inflight), + atomic_read(&imp->imp_unregistering), + atomic_read(&imp->imp_timeouts), + ret.lc_sum, ret.lc_units); + + k = 0; + for(j = 0; j < IMP_AT_MAX_PORTALS; j++) { + if (imp->imp_at.iat_portal[j] == 0) + break; + k = max_t(unsigned int, k, + at_get(&imp->imp_at.iat_service_estimate[j])); + } + i += snprintf(page + i, count - i, + " service_estimates:\n" + " services: %u sec\n" + " network: %u sec\n", + k, + at_get(&imp->imp_at.iat_net_latency)); + + i += snprintf(page + i, count - i, + " transactions:\n" + " last_replay: "LPU64"\n" + " peer_committed: "LPU64"\n" + " last_checked: "LPU64"\n", + imp->imp_last_replay_transno, + imp->imp_peer_committed_transno, + imp->imp_last_transno_checked); + + /* avg data rates */ + for (rw = 0; rw <= 1; rw++) { + lprocfs_stats_collect(obd->obd_svc_stats, + PTLRPC_LAST_CNTR + BRW_READ_BYTES + rw, + &ret); + if (ret.lc_sum > 0) { + do_div(ret.lc_sum, ret.lc_count); + i += snprintf(page + i, count - i, + " %s_data_averages:\n" + " bytes_per_rpc: "LPU64"\n", + rw ? "write" : "read", + ret.lc_sum); + } + k = (int)ret.lc_sum; + j = opcode_offset(OST_READ + rw) + EXTRA_MAX_OPCODES; + lprocfs_stats_collect(obd->obd_svc_stats, j, &ret); + if (ret.lc_sum > 0) { + do_div(ret.lc_sum, ret.lc_count); + i += snprintf(page + i, count - i, + " %s_per_rpc: "LPU64"\n", + ret.lc_units, ret.lc_sum); + j = (int)ret.lc_sum; + if (j > 0) + i += snprintf(page + i, count - i, + " MB_per_sec: %u.%.02u\n", + k / j, (100 * k / j) % 100); + } + } LPROCFS_CLIMP_EXIT(obd); + return i; +} + +int lprocfs_rd_state(char *page, char **start, off_t off, int count, + int *eof, void *data) +{ + struct obd_device *obd = (struct obd_device *)data; + struct obd_import *imp; + int i, j, k; + + LASSERT(obd != NULL); + LPROCFS_CLIMP_CHECK(obd); + imp = obd->u.cli.cl_import; + *eof = 1; + + i = snprintf(page, count, "current_state: %s\n", + ptlrpc_import_state_name(imp->imp_state)); + i += snprintf(page + i, count - i, + "state_history:\n"); + k = imp->imp_state_hist_idx; + for (j = 0; j < IMP_STATE_HIST_LEN; j++) { + struct import_state_hist *ish = + &imp->imp_state_hist[(k + j) % IMP_STATE_HIST_LEN]; + if (ish->ish_state == 0) + continue; + i += snprintf(page + i, count - i, " - ["CFS_TIME_T", %s]\n", + ish->ish_time, + ptlrpc_import_state_name(ish->ish_state)); + } + + LPROCFS_CLIMP_EXIT(obd); + return i; +} + +int lprocfs_at_hist_helper(char *page, int count, int rc, + struct adaptive_timeout *at) +{ + int i; + for (i = 0; i < AT_BINS; i++) + rc += snprintf(page + rc, count - rc, "%3u ", at->at_hist[i]); + rc += snprintf(page + rc, count - rc, "\n"); + return rc; +} + +int lprocfs_rd_quota_resend_count(char *page, char **start, off_t off, + int count, int *eof, void *data) +{ + struct obd_device *obd = data; + + return snprintf(page, count, "%d\n", + atomic_read(&obd->u.cli.cl_quota_resends)); +} + +int lprocfs_wr_quota_resend_count(struct file *file, const char *buffer, + unsigned long count, void *data) +{ + struct obd_device *obd = data; + int val, rc; + + rc = lprocfs_write_helper(buffer, count, &val); + if (rc) + return rc; + + atomic_set(&obd->u.cli.cl_quota_resends, val); + + return count; +} + +/* See also ptlrpc_lprocfs_rd_timeouts */ +int lprocfs_rd_timeouts(char *page, char **start, off_t off, int count, + int *eof, void *data) +{ + struct obd_device *obd = (struct obd_device *)data; + struct obd_import *imp; + unsigned int cur, worst; + time_t now, worstt; + struct dhms ts; + int i, rc = 0; + + LASSERT(obd != NULL); + LPROCFS_CLIMP_CHECK(obd); + imp = obd->u.cli.cl_import; + *eof = 1; + + now = cfs_time_current_sec(); + + /* Some network health info for kicks */ + s2dhms(&ts, now - imp->imp_last_reply_time); + rc += snprintf(page + rc, count - rc, + "%-10s : %ld, "DHMS_FMT" ago\n", + "last reply", imp->imp_last_reply_time, DHMS_VARS(&ts)); + + cur = at_get(&imp->imp_at.iat_net_latency); + worst = imp->imp_at.iat_net_latency.at_worst_ever; + worstt = imp->imp_at.iat_net_latency.at_worst_time; + s2dhms(&ts, now - worstt); + rc += snprintf(page + rc, count - rc, + "%-10s : cur %3u worst %3u (at %ld, "DHMS_FMT" ago) ", + "network", cur, worst, worstt, DHMS_VARS(&ts)); + rc = lprocfs_at_hist_helper(page, count, rc, + &imp->imp_at.iat_net_latency); + + for(i = 0; i < IMP_AT_MAX_PORTALS; i++) { + if (imp->imp_at.iat_portal[i] == 0) + break; + cur = at_get(&imp->imp_at.iat_service_estimate[i]); + worst = imp->imp_at.iat_service_estimate[i].at_worst_ever; + worstt = imp->imp_at.iat_service_estimate[i].at_worst_time; + s2dhms(&ts, now - worstt); + rc += snprintf(page + rc, count - rc, + "portal %-2d : cur %3u worst %3u (at %ld, " + DHMS_FMT" ago) ", imp->imp_at.iat_portal[i], + cur, worst, worstt, DHMS_VARS(&ts)); + rc = lprocfs_at_hist_helper(page, count, rc, + &imp->imp_at.iat_service_estimate[i]); + } + + LPROCFS_CLIMP_EXIT(obd); + return rc; +} + +int lprocfs_rd_connect_flags(char *page, char **start, off_t off, + int count, int *eof, void *data) +{ + struct obd_device *obd = data; + __u64 flags; + int ret = 0; + + LPROCFS_CLIMP_CHECK(obd); + flags = obd->u.cli.cl_import->imp_connect_data.ocd_connect_flags; + ret = snprintf(page, count, "flags="LPX64"\n", flags); + ret += obd_connect_flags2str(page + ret, count - ret, flags, "\n"); + ret += snprintf(page + ret, count - ret, "\n"); + LPROCFS_CLIMP_EXIT(obd); return ret; } EXPORT_SYMBOL(lprocfs_rd_connect_flags); @@ -632,7 +1020,7 @@ EXPORT_SYMBOL(lprocfs_rd_connect_flags); int lprocfs_rd_num_exports(char *page, char **start, off_t off, int count, int *eof, void *data) { - struct obd_device *obd = (struct obd_device*)data; + struct obd_device *obd = data; LASSERT(obd != NULL); *eof = 1; @@ -670,24 +1058,72 @@ int lprocfs_obd_setup(struct obd_device *obd, struct lprocfs_vars *list) int lprocfs_obd_cleanup(struct obd_device *obd) { - if (!obd) + if (!obd) return -EINVAL; - if (obd->obd_proc_exports) { + if (obd->obd_proc_exports_entry) { /* Should be no exports left */ - LASSERT(obd->obd_proc_exports->subdir == NULL); - lprocfs_remove(&obd->obd_proc_exports); + LASSERT(obd->obd_proc_exports_entry->subdir == NULL); + lprocfs_remove(&obd->obd_proc_exports_entry); + obd->obd_proc_exports_entry = NULL; + } + if (obd->obd_proc_entry) { + lprocfs_remove(&obd->obd_proc_entry); + obd->obd_proc_entry = NULL; } - lprocfs_remove(&obd->obd_proc_entry); return 0; } +static void lprocfs_free_client_stats(struct nid_stat *client_stat) +{ + CDEBUG(D_CONFIG, "stat %p - data %p/%p/%p\n", client_stat, + client_stat->nid_proc, client_stat->nid_stats, + client_stat->nid_brw_stats); + + LASSERTF(atomic_read(&client_stat->nid_exp_ref_count) == 0, + "count %d\n", atomic_read(&client_stat->nid_exp_ref_count)); + + hlist_del_init(&client_stat->nid_hash); + + if (client_stat->nid_proc) + lprocfs_remove(&client_stat->nid_proc); + + if (client_stat->nid_stats) + lprocfs_free_stats(&client_stat->nid_stats); + + if (client_stat->nid_brw_stats) + OBD_FREE_PTR(client_stat->nid_brw_stats); + + if (client_stat->nid_ldlm_stats) + lprocfs_free_stats(&client_stat->nid_ldlm_stats); + + OBD_FREE_PTR(client_stat); + return; + +} + +void lprocfs_free_per_client_stats(struct obd_device *obd) +{ + struct nid_stat *stat; + ENTRY; + + /* we need extra list - because hash_exit called to early */ + /* not need locking because all clients is died */ + while(!list_empty(&obd->obd_nid_stats)) { + stat = list_entry(obd->obd_nid_stats.next, + struct nid_stat, nid_list); + list_del_init(&stat->nid_list); + lprocfs_free_client_stats(stat); + } + + EXIT; +} + struct lprocfs_stats *lprocfs_alloc_stats(unsigned int num, enum lprocfs_stats_flags flags) { struct lprocfs_stats *stats; - struct lprocfs_percpu *percpu; unsigned int percpusize; - unsigned int i; + unsigned int i, j; unsigned int num_cpu; if (num == 0) @@ -710,12 +1146,20 @@ struct lprocfs_stats *lprocfs_alloc_stats(unsigned int num, stats->ls_flags = 0; } - percpusize = offsetof(typeof(*percpu), lp_cntr[num]); + percpusize = offsetof(struct lprocfs_percpu, lp_cntr[num]); if (num_cpu > 1) percpusize = L1_CACHE_ALIGN(percpusize); - stats->ls_percpu_size = num_cpu * percpusize; - OBD_ALLOC(stats->ls_percpu[0], stats->ls_percpu_size); + for (i = 0; i < num_cpu; i++) { + OBD_ALLOC(stats->ls_percpu[i], percpusize); + if (stats->ls_percpu[i] == NULL) { + for (j = 0; j < i; j++) { + OBD_FREE(stats->ls_percpu[j], percpusize); + stats->ls_percpu[j] = NULL; + } + break; + } + } if (stats->ls_percpu[0] == NULL) { OBD_FREE(stats, offsetof(typeof(*stats), ls_percpu[num_cpu])); @@ -723,10 +1167,6 @@ struct lprocfs_stats *lprocfs_alloc_stats(unsigned int num, } stats->ls_num = num; - for (i = 1; i < num_cpu; i++) - stats->ls_percpu[i] = (void *)(stats->ls_percpu[i - 1]) + - percpusize; - return stats; } @@ -734,6 +1174,8 @@ void lprocfs_free_stats(struct lprocfs_stats **statsh) { struct lprocfs_stats *stats = *statsh; unsigned int num_cpu; + unsigned int percpusize; + unsigned int i; if (stats == NULL || stats->ls_num == 0) return; @@ -744,7 +1186,11 @@ void lprocfs_free_stats(struct lprocfs_stats **statsh) else num_cpu = num_possible_cpus(); - OBD_FREE(stats->ls_percpu[0], stats->ls_percpu_size); + percpusize = offsetof(struct lprocfs_percpu, lp_cntr[stats->ls_num]); + if (num_cpu > 1) + percpusize = L1_CACHE_ALIGN(percpusize); + for (i = 0; i < num_cpu; i++) + OBD_FREE(stats->ls_percpu[i], percpusize); OBD_FREE(stats, offsetof(typeof(*stats), ls_percpu[num_cpu])); } @@ -757,12 +1203,12 @@ void lprocfs_clear_stats(struct lprocfs_stats *stats) num_cpu = lprocfs_stats_lock(stats, LPROCFS_GET_NUM_CPU); for (i = 0; i < num_cpu; i++) { - for (j = 0; j < stats->ls_num; j++) { + for (j = 0; j < stats->ls_num; j++) { percpu_cntr = &(stats->ls_percpu[i])->lp_cntr[j]; atomic_inc(&percpu_cntr->lc_cntl.la_entry); percpu_cntr->lc_count = 0; percpu_cntr->lc_sum = 0; - percpu_cntr->lc_min = ~(__u64)0; + percpu_cntr->lc_min = LC_MIN_INIT; percpu_cntr->lc_max = 0; percpu_cntr->lc_sumsquare = 0; atomic_inc(&percpu_cntr->lc_cntl.la_exit); @@ -807,10 +1253,9 @@ static void *lprocfs_stats_seq_next(struct seq_file *p, void *v, loff_t *pos) static int lprocfs_stats_seq_show(struct seq_file *p, void *v) { struct lprocfs_stats *stats = p->private; - struct lprocfs_counter *cntr = v; - struct lprocfs_counter t, ret = { .lc_min = ~(__u64)0 }; - int i, idx, rc; - unsigned int num_cpu; + struct lprocfs_counter *cntr = v; + struct lprocfs_counter ret; + int idx, rc = 0; if (cntr == &(stats->ls_percpu[0])->lp_cntr[0]) { struct timeval now; @@ -822,46 +1267,24 @@ static int lprocfs_stats_seq_show(struct seq_file *p, void *v) } idx = cntr - &(stats->ls_percpu[0])->lp_cntr[0]; - if (stats->ls_flags & LPROCFS_STATS_FLAG_NOPERCPU) - num_cpu = 1; - else - num_cpu = num_possible_cpus(); - - for (i = 0; i < num_cpu; i++) { - struct lprocfs_counter *percpu_cntr = - &(stats->ls_percpu[i])->lp_cntr[idx]; - int centry; - - do { - centry = atomic_read(&percpu_cntr->lc_cntl.la_entry); - t.lc_count = percpu_cntr->lc_count; - t.lc_sum = percpu_cntr->lc_sum; - t.lc_min = percpu_cntr->lc_min; - t.lc_max = percpu_cntr->lc_max; - t.lc_sumsquare = percpu_cntr->lc_sumsquare; - } while (centry != atomic_read(&percpu_cntr->lc_cntl.la_entry) && - centry != atomic_read(&percpu_cntr->lc_cntl.la_exit)); - ret.lc_count += t.lc_count; - ret.lc_sum += t.lc_sum; - if (t.lc_min < ret.lc_min) - ret.lc_min = t.lc_min; - if (t.lc_max > ret.lc_max) - ret.lc_max = t.lc_max; - ret.lc_sumsquare += t.lc_sumsquare; - } + lprocfs_stats_collect(stats, idx, &ret); + + if (ret.lc_count == 0) + goto out; - rc = seq_printf(p, "%-25s "LPU64" samples [%s]", cntr->lc_name, + rc = seq_printf(p, "%-25s "LPD64" samples [%s]", cntr->lc_name, ret.lc_count, cntr->lc_units); + if (rc < 0) goto out; if ((cntr->lc_config & LPROCFS_CNTR_AVGMINMAX) && (ret.lc_count > 0)) { - rc = seq_printf(p, " "LPU64" "LPU64" "LPU64, + rc = seq_printf(p, " "LPD64" "LPD64" "LPD64, ret.lc_min, ret.lc_max, ret.lc_sum); if (rc < 0) goto out; if (cntr->lc_config & LPROCFS_CNTR_STDDEV) - rc = seq_printf(p, " "LPU64, ret.lc_sumsquare); + rc = seq_printf(p, " "LPD64, ret.lc_sumsquare); if (rc < 0) goto out; } @@ -913,7 +1336,7 @@ int lprocfs_register_stats(struct proc_dir_entry *root, const char *name, if (entry == NULL) return -ENOMEM; entry->proc_fops = &lprocfs_stats_seq_fops; - entry->data = (void *)stats; + entry->data = stats; return 0; } @@ -933,7 +1356,7 @@ void lprocfs_counter_init(struct lprocfs_stats *stats, int index, c->lc_config = conf; c->lc_count = 0; c->lc_sum = 0; - c->lc_min = ~(__u64)0; + c->lc_min = LC_MIN_INIT; c->lc_max = 0; c->lc_name = name; c->lc_units = units; @@ -979,19 +1402,13 @@ void lprocfs_init_ops_stats(int num_private_stats, struct lprocfs_stats *stats) LPROCFS_OBD_OP_INIT(num_private_stats, stats, preallocate); LPROCFS_OBD_OP_INIT(num_private_stats, stats, precreate); LPROCFS_OBD_OP_INIT(num_private_stats, stats, create); + LPROCFS_OBD_OP_INIT(num_private_stats, stats, create_async); LPROCFS_OBD_OP_INIT(num_private_stats, stats, destroy); LPROCFS_OBD_OP_INIT(num_private_stats, stats, setattr); LPROCFS_OBD_OP_INIT(num_private_stats, stats, setattr_async); LPROCFS_OBD_OP_INIT(num_private_stats, stats, getattr); LPROCFS_OBD_OP_INIT(num_private_stats, stats, getattr_async); LPROCFS_OBD_OP_INIT(num_private_stats, stats, brw); - LPROCFS_OBD_OP_INIT(num_private_stats, stats, brw_async); - LPROCFS_OBD_OP_INIT(num_private_stats, stats, prep_async_page); - LPROCFS_OBD_OP_INIT(num_private_stats, stats, queue_async_io); - LPROCFS_OBD_OP_INIT(num_private_stats, stats, queue_group_io); - LPROCFS_OBD_OP_INIT(num_private_stats, stats, trigger_group_io); - LPROCFS_OBD_OP_INIT(num_private_stats, stats, set_async_flags); - LPROCFS_OBD_OP_INIT(num_private_stats, stats, teardown_async_page); LPROCFS_OBD_OP_INIT(num_private_stats, stats, merge_lvb); LPROCFS_OBD_OP_INIT(num_private_stats, stats, adjust_kms); LPROCFS_OBD_OP_INIT(num_private_stats, stats, punch); @@ -1002,11 +1419,9 @@ void lprocfs_init_ops_stats(int num_private_stats, struct lprocfs_stats *stats) LPROCFS_OBD_OP_INIT(num_private_stats, stats, preprw); LPROCFS_OBD_OP_INIT(num_private_stats, stats, commitrw); LPROCFS_OBD_OP_INIT(num_private_stats, stats, enqueue); - LPROCFS_OBD_OP_INIT(num_private_stats, stats, match); LPROCFS_OBD_OP_INIT(num_private_stats, stats, change_cbdata); LPROCFS_OBD_OP_INIT(num_private_stats, stats, cancel); LPROCFS_OBD_OP_INIT(num_private_stats, stats, cancel_unused); - LPROCFS_OBD_OP_INIT(num_private_stats, stats, join_lru); LPROCFS_OBD_OP_INIT(num_private_stats, stats, init_export); LPROCFS_OBD_OP_INIT(num_private_stats, stats, destroy_export); LPROCFS_OBD_OP_INIT(num_private_stats, stats, extent_calc); @@ -1018,9 +1433,17 @@ void lprocfs_init_ops_stats(int num_private_stats, struct lprocfs_stats *stats) LPROCFS_OBD_OP_INIT(num_private_stats, stats, import_event); LPROCFS_OBD_OP_INIT(num_private_stats, stats, notify); LPROCFS_OBD_OP_INIT(num_private_stats, stats, health_check); + LPROCFS_OBD_OP_INIT(num_private_stats, stats, get_uuid); LPROCFS_OBD_OP_INIT(num_private_stats, stats, quotacheck); LPROCFS_OBD_OP_INIT(num_private_stats, stats, quotactl); + LPROCFS_OBD_OP_INIT(num_private_stats, stats, quota_adjust_qunit); LPROCFS_OBD_OP_INIT(num_private_stats, stats, ping); + LPROCFS_OBD_OP_INIT(num_private_stats, stats, pool_new); + LPROCFS_OBD_OP_INIT(num_private_stats, stats, pool_rem); + LPROCFS_OBD_OP_INIT(num_private_stats, stats, pool_add); + LPROCFS_OBD_OP_INIT(num_private_stats, stats, pool_del); + LPROCFS_OBD_OP_INIT(num_private_stats, stats, getref); + LPROCFS_OBD_OP_INIT(num_private_stats, stats, putref); } int lprocfs_alloc_obd_stats(struct obd_device *obd, unsigned num_private_stats) @@ -1063,7 +1486,7 @@ int lprocfs_alloc_obd_stats(struct obd_device *obd, unsigned num_private_stats) void lprocfs_free_obd_stats(struct obd_device *obd) { - if (obd->obd_stats) + if (obd->obd_stats) lprocfs_free_stats(&obd->obd_stats); } @@ -1085,7 +1508,7 @@ int lprocfs_alloc_md_stats(struct obd_device *obd, LASSERT(obd->obd_proc_entry != NULL); LASSERT(obd->md_cntr_base == 0); - num_stats = 1 + MD_COUNTER_OFFSET(get_remote_perm) + + num_stats = 1 + MD_COUNTER_OFFSET(revalidate_lock) + num_private_stats; stats = lprocfs_alloc_stats(num_stats, 0); if (stats == NULL) @@ -1118,7 +1541,10 @@ int lprocfs_alloc_md_stats(struct obd_device *obd, LPROCFS_MD_OP_INIT(num_private_stats, stats, lock_match); LPROCFS_MD_OP_INIT(num_private_stats, stats, cancel_unused); LPROCFS_MD_OP_INIT(num_private_stats, stats, renew_capa); + LPROCFS_MD_OP_INIT(num_private_stats, stats, unpack_capa); LPROCFS_MD_OP_INIT(num_private_stats, stats, get_remote_perm); + LPROCFS_MD_OP_INIT(num_private_stats, stats, intent_getattr_async); + LPROCFS_MD_OP_INIT(num_private_stats, stats, revalidate_lock); for (i = num_private_stats; i < num_stats; i++) { if (stats->ls_percpu[0]->lp_cntr[i].lc_name == NULL) { @@ -1128,7 +1554,7 @@ int lprocfs_alloc_md_stats(struct obd_device *obd, LBUG(); } } - rc = lprocfs_register_stats(obd->obd_proc_entry, "stats", stats); + rc = lprocfs_register_stats(obd->obd_proc_entry, "md_stats", stats); if (rc < 0) { lprocfs_free_stats(&stats); } else { @@ -1144,103 +1570,295 @@ void lprocfs_free_md_stats(struct obd_device *obd) if (stats != NULL) { obd->md_stats = NULL; + obd->md_cntr_base = 0; lprocfs_free_stats(&stats); } } +void lprocfs_init_ldlm_stats(struct lprocfs_stats *ldlm_stats) +{ + lprocfs_counter_init(ldlm_stats, + LDLM_ENQUEUE - LDLM_FIRST_OPC, + 0, "ldlm_enqueue", "reqs"); + lprocfs_counter_init(ldlm_stats, + LDLM_CONVERT - LDLM_FIRST_OPC, + 0, "ldlm_convert", "reqs"); + lprocfs_counter_init(ldlm_stats, + LDLM_CANCEL - LDLM_FIRST_OPC, + 0, "ldlm_cancel", "reqs"); + lprocfs_counter_init(ldlm_stats, + LDLM_BL_CALLBACK - LDLM_FIRST_OPC, + 0, "ldlm_bl_callback", "reqs"); + lprocfs_counter_init(ldlm_stats, + LDLM_CP_CALLBACK - LDLM_FIRST_OPC, + 0, "ldlm_cp_callback", "reqs"); + lprocfs_counter_init(ldlm_stats, + LDLM_GL_CALLBACK - LDLM_FIRST_OPC, + 0, "ldlm_gl_callback", "reqs"); +} + int lprocfs_exp_rd_nid(char *page, char **start, off_t off, int count, int *eof, void *data) { - struct obd_export *exp = (struct obd_export*)data; + struct obd_export *exp = data; LASSERT(exp != NULL); *eof = 1; return snprintf(page, count, "%s\n", obd_export_nid2str(exp)); } +struct exp_uuid_cb_data { + char *page; + int count; + int *eof; + int *len; +}; + +static void +lprocfs_exp_rd_cb_data_init(struct exp_uuid_cb_data *cb_data, char *page, + int count, int *eof, int *len) +{ + cb_data->page = page; + cb_data->count = count; + cb_data->eof = eof; + cb_data->len = len; +} + +void lprocfs_exp_print_uuid(void *obj, void *cb_data) +{ + struct obd_export *exp = (struct obd_export *)obj; + struct exp_uuid_cb_data *data = (struct exp_uuid_cb_data *)cb_data; + + if (exp->exp_nid_stats) + *data->len += snprintf((data->page + *data->len), + data->count, "%s\n", + obd_uuid2str(&exp->exp_client_uuid)); +} + int lprocfs_exp_rd_uuid(char *page, char **start, off_t off, int count, - int *eof, void *data) + int *eof, void *data) { - struct obd_export *exp = (struct obd_export*)data; - LASSERT(exp != NULL); + struct nid_stat *stats = (struct nid_stat *)data; + struct exp_uuid_cb_data cb_data; + struct obd_device *obd = stats->nid_obd; + int len = 0; + *eof = 1; - return snprintf(page, count, "%s\n", - obd_uuid2str(&exp->exp_client_uuid)); + page[0] = '\0'; + lprocfs_exp_rd_cb_data_init(&cb_data, page, count, eof, &len); + lustre_hash_for_each_key(obd->obd_nid_hash, &stats->nid, + lprocfs_exp_print_uuid, &cb_data); + return (*cb_data.len); } - -int lprocfs_exp_setup(struct obd_export *exp) + +void lprocfs_exp_print_hash(void *obj, void *cb_data) { - char name[sizeof (exp->exp_client_uuid.uuid) + 3]; - int i = 1, rc; + struct exp_uuid_cb_data *data = cb_data; + struct obd_export *exp = obj; + lustre_hash_t *lh; + + lh = exp->exp_lock_hash; + if (lh) { + if (!*data->len) + *data->len += lustre_hash_debug_header(data->page, + data->count); + + *data->len += lustre_hash_debug_str(lh, data->page + *data->len, + data->count); + } +} + +int lprocfs_exp_rd_hash(char *page, char **start, off_t off, int count, + int *eof, void *data) +{ + struct nid_stat *stats = (struct nid_stat *)data; + struct exp_uuid_cb_data cb_data; + struct obd_device *obd = stats->nid_obd; + int len = 0; + + *eof = 1; + page[0] = '\0'; + lprocfs_exp_rd_cb_data_init(&cb_data, page, count, eof, &len); + + lustre_hash_for_each_key(obd->obd_nid_hash, &stats->nid, + lprocfs_exp_print_hash, &cb_data); + return (*cb_data.len); +} + +int lprocfs_nid_stats_clear_read(char *page, char **start, off_t off, + int count, int *eof, void *data) +{ + *eof = 1; + return snprintf(page, count, "%s\n", + "Write into this file to clear all nid stats and " + "stale nid entries"); +} +EXPORT_SYMBOL(lprocfs_nid_stats_clear_read); + +void lprocfs_nid_stats_clear_write_cb(void *obj, void *data) +{ + struct nid_stat *stat = obj; + int i; ENTRY; - if (!exp || !exp->exp_obd || !exp->exp_obd->obd_proc_exports) + /* object has only hash + iterate_all references. + * add/delete blocked by hash bucket lock */ + CDEBUG(D_INFO,"refcnt %d\n", atomic_read(&stat->nid_exp_ref_count)); + if (atomic_read(&stat->nid_exp_ref_count) == 2) { + hlist_del_init(&stat->nid_hash); + nidstat_putref(stat); + spin_lock(&stat->nid_obd->obd_nid_lock); + list_move(&stat->nid_list, data); + spin_unlock(&stat->nid_obd->obd_nid_lock); + EXIT; + return; + } + /* we has reference to object - only clear data*/ + if (stat->nid_stats) + lprocfs_clear_stats(stat->nid_stats); + + if (stat->nid_brw_stats) { + for (i = 0; i < BRW_LAST; i++) + lprocfs_oh_clear(&stat->nid_brw_stats->hist[i]); + } + EXIT; + return; +} + +int lprocfs_nid_stats_clear_write(struct file *file, const char *buffer, + unsigned long count, void *data) +{ + struct obd_device *obd = (struct obd_device *)data; + struct nid_stat *client_stat; + CFS_LIST_HEAD(free_list); + + lustre_hash_for_each(obd->obd_nid_stats_hash, + lprocfs_nid_stats_clear_write_cb, &free_list); + + while (!list_empty(&free_list)) { + client_stat = list_entry(free_list.next, struct nid_stat, + nid_list); + list_del_init(&client_stat->nid_list); + lprocfs_free_client_stats(client_stat); + } + + return count; +} +EXPORT_SYMBOL(lprocfs_nid_stats_clear_write); + +int lprocfs_exp_setup(struct obd_export *exp, lnet_nid_t *nid, int *newnid) +{ + struct nid_stat *new_stat, *old_stat; + struct obd_device *obd = NULL; + cfs_proc_dir_entry_t *entry; + int rc = 0; + ENTRY; + + *newnid = 0; + + if (!exp || !exp->exp_obd || !exp->exp_obd->obd_proc_exports_entry || + !exp->exp_obd->obd_nid_stats_hash) RETURN(-EINVAL); - mutex_down(&exp->exp_obd->obd_proc_exp_sem); - sprintf(name, "%s", (char *)exp->exp_client_uuid.uuid); - while (lprocfs_srch(exp->exp_obd->obd_proc_exports, name)) { - /* We might add a new export before deleting the old one during - an eviction (recovery-small 19a). Suckage. We - could block, or come up with a new name, or just give up. */ - if (++i > 9) - GOTO(out, rc = -EEXIST); - sprintf(name, "%s:%d", (char *)exp->exp_client_uuid.uuid, i); + + /* not test against zero because eric say: + * You may only test nid against another nid, or LNET_NID_ANY. + * Anything else is nonsense.*/ + if (!nid || *nid == LNET_NID_ANY) + RETURN(0); + + obd = exp->exp_obd; + + CDEBUG(D_CONFIG, "using hash %p\n", obd->obd_nid_stats_hash); + + OBD_ALLOC_PTR(new_stat); + if (new_stat == NULL) + RETURN(-ENOMEM); + + new_stat->nid = *nid; + new_stat->nid_obd = exp->exp_obd; + atomic_set(&new_stat->nid_exp_ref_count, 0); + + old_stat = lustre_hash_findadd_unique(obd->obd_nid_stats_hash, + nid, &new_stat->nid_hash); + CDEBUG(D_INFO, "Found stats %p for nid %s - ref %d\n", + old_stat, libcfs_nid2str(*nid), + atomic_read(&new_stat->nid_exp_ref_count)); + + /* Return -EALREADY here so that we know that the /proc + * entry already has been created */ + if (old_stat != new_stat) { + spin_lock(&obd->obd_nid_lock); + if (exp->exp_nid_stats != old_stat) { + if (exp->exp_nid_stats) + nidstat_putref(exp->exp_nid_stats); + exp->exp_nid_stats = old_stat; + } else { + /* lustre_hash_findadd_unique() has added + * old_stat's refcount */ + nidstat_putref(old_stat); + } + + spin_unlock(&obd->obd_nid_lock); + + GOTO(destroy_new, rc = -EALREADY); + } + /* not found - create */ + new_stat->nid_proc = lprocfs_register(libcfs_nid2str(*nid), + obd->obd_proc_exports_entry, + NULL, NULL); + if (new_stat->nid_proc == NULL) { + CERROR("Error making export directory for nid %s\n", + libcfs_nid2str(*nid)); + GOTO(destroy_new_ns, rc = -ENOMEM); } - /* Create a proc entry for this export */ - exp->exp_proc = proc_mkdir(name, exp->exp_obd->obd_proc_exports); - if (!exp->exp_proc) { - CERROR("Error making export directory for %s\n", name); - GOTO(out, rc = -ENOMEM); + entry = lprocfs_add_simple(new_stat->nid_proc, "uuid", + lprocfs_exp_rd_uuid, NULL, new_stat, NULL); + if (IS_ERR(entry)) { + CWARN("Error adding the NID stats file\n"); + rc = PTR_ERR(entry); + GOTO(destroy_new_ns, rc); } - /* Always add nid and uuid */ - rc = lprocfs_add_simple(exp->exp_proc, "nid", - lprocfs_exp_rd_nid, NULL, exp); - if (rc) - GOTO(out, rc); - rc = lprocfs_add_simple(exp->exp_proc, "uuid", - lprocfs_exp_rd_uuid, NULL, exp); - if (rc) - GOTO(out, rc); - /* Always add ldlm stats */ - exp->exp_ldlm_stats = lprocfs_alloc_stats(LDLM_LAST_OPC - - LDLM_FIRST_OPC, 0); - if (exp->exp_ldlm_stats == NULL) { - lprocfs_remove(&exp->exp_proc); - GOTO(out, rc = -ENOMEM); + entry = lprocfs_add_simple(new_stat->nid_proc, "hash", + lprocfs_exp_rd_hash, NULL, new_stat, NULL); + if (IS_ERR(entry)) { + CWARN("Error adding the hash file\n"); + rc = PTR_ERR(entry); + GOTO(destroy_new_ns, rc); } - lprocfs_counter_init(exp->exp_ldlm_stats, - LDLM_ENQUEUE - LDLM_FIRST_OPC, - 0, "ldlm_enqueue", "reqs"); - lprocfs_counter_init(exp->exp_ldlm_stats, - LDLM_CONVERT - LDLM_FIRST_OPC, - 0, "ldlm_convert", "reqs"); - lprocfs_counter_init(exp->exp_ldlm_stats, - LDLM_CANCEL - LDLM_FIRST_OPC, - 0, "ldlm_cancel", "reqs"); - lprocfs_counter_init(exp->exp_ldlm_stats, - LDLM_BL_CALLBACK - LDLM_FIRST_OPC, - 0, "ldlm_bl_callback", "reqs"); - lprocfs_counter_init(exp->exp_ldlm_stats, - LDLM_CP_CALLBACK - LDLM_FIRST_OPC, - 0, "ldlm_cp_callback", "reqs"); - lprocfs_counter_init(exp->exp_ldlm_stats, - LDLM_GL_CALLBACK - LDLM_FIRST_OPC, - 0, "ldlm_gl_callback", "reqs"); - lprocfs_register_stats(exp->exp_proc, "ldlm_stats", - exp->exp_ldlm_stats); -out: - mutex_up(&exp->exp_obd->obd_proc_exp_sem); + if (exp->exp_nid_stats) + nidstat_putref(exp->exp_nid_stats); + nidstat_getref(new_stat); + exp->exp_nid_stats = new_stat; + *newnid = 1; + /* protect competitive add to list, not need locking on destroy */ + spin_lock(&obd->obd_nid_lock); + list_add(&new_stat->nid_list, &obd->obd_nid_stats); + spin_unlock(&obd->obd_nid_lock); + + RETURN(rc); + +destroy_new_ns: + if (new_stat->nid_proc != NULL) + lprocfs_remove(&new_stat->nid_proc); + lustre_hash_del(obd->obd_nid_stats_hash, nid, &new_stat->nid_hash); + +destroy_new: + OBD_FREE_PTR(new_stat); RETURN(rc); } int lprocfs_exp_cleanup(struct obd_export *exp) { - mutex_down(&exp->exp_obd->obd_proc_exp_sem); - lprocfs_remove(&exp->exp_proc); - lprocfs_free_stats(&exp->exp_ops_stats); - lprocfs_free_stats(&exp->exp_ldlm_stats); - mutex_up(&exp->exp_obd->obd_proc_exp_sem); + struct nid_stat *stat = exp->exp_nid_stats; + + if(!stat || !exp->exp_obd) + RETURN(0); + + nidstat_putref(exp->exp_nid_stats); + exp->exp_nid_stats = NULL; + lprocfs_free_md_stats(exp->exp_obd); + return 0; } @@ -1292,7 +1910,8 @@ int lprocfs_write_frac_helper(const char *buffer, unsigned long count, return 0; } -int lprocfs_read_frac_helper(char *buffer, unsigned long count, long val, int mult) +int lprocfs_read_frac_helper(char *buffer, unsigned long count, long val, + int mult) { long decimal_val, frac_val; int prtn; @@ -1310,24 +1929,26 @@ int lprocfs_read_frac_helper(char *buffer, unsigned long count, long val, int mu temp_frac = frac_val * 10; buffer[prtn++] = '.'; - while (frac_bits < 2 && (temp_frac / mult) < 1 ) { /*only reserved 2bits fraction*/ + while (frac_bits < 2 && (temp_frac / mult) < 1 ) { + /* only reserved 2 bits fraction */ buffer[prtn++] ='0'; temp_frac *= 10; frac_bits++; } /* - Need to think these cases : - 1. #echo x.00 > /proc/xxx output result : x - 2. #echo x.0x > /proc/xxx output result : x.0x - 3. #echo x.x0 > /proc/xxx output result : x.x - 4. #echo x.xx > /proc/xxx output result : x.xx - Only reserved 2bits fraction. + * Need to think these cases : + * 1. #echo x.00 > /proc/xxx output result : x + * 2. #echo x.0x > /proc/xxx output result : x.0x + * 3. #echo x.x0 > /proc/xxx output result : x.x + * 4. #echo x.xx > /proc/xxx output result : x.xx + * Only reserved 2 bits fraction. */ for (i = 0; i < (5 - prtn); i++) temp_mult *= 10; frac_bits = min((int)count - prtn, 3 - frac_bits); - prtn += snprintf(buffer + prtn, frac_bits, "%ld", frac_val * temp_mult / mult); + prtn += snprintf(buffer + prtn, frac_bits, "%ld", + frac_val * temp_mult / mult); prtn--; while(buffer[prtn] < '1' || buffer[prtn] > '9') { @@ -1355,7 +1976,7 @@ int lprocfs_write_frac_u64_helper(const char *buffer, unsigned long count, __u64 whole, frac = 0, units; unsigned frac_d = 1; - if (count > (sizeof(kernbuf) - 1) ) + if (count > (sizeof(kernbuf) - 1)) return -EINVAL; if (copy_from_user(kernbuf, buffer, count)) @@ -1400,7 +2021,7 @@ int lprocfs_write_frac_u64_helper(const char *buffer, unsigned long count, units <<= 10; } /* Specified units override the multiplier */ - if (units) + if (units) mult = mult < 0 ? -units : units; frac *= mult; @@ -1409,8 +2030,7 @@ int lprocfs_write_frac_u64_helper(const char *buffer, unsigned long count, return 0; } -int lprocfs_seq_create(cfs_proc_dir_entry_t *parent, - char *name, mode_t mode, +int lprocfs_seq_create(cfs_proc_dir_entry_t *parent, char *name, mode_t mode, struct file_operations *seq_fops, void *data) { struct proc_dir_entry *entry; @@ -1431,7 +2051,7 @@ __inline__ int lprocfs_obd_seq_create(struct obd_device *dev, char *name, struct file_operations *seq_fops, void *data) { - return (lprocfs_seq_create(dev->obd_proc_entry, name, + return (lprocfs_seq_create(dev->obd_proc_entry, name, mode, seq_fops, data)); } EXPORT_SYMBOL(lprocfs_obd_seq_create); @@ -1477,6 +2097,24 @@ void lprocfs_oh_clear(struct obd_histogram *oh) } EXPORT_SYMBOL(lprocfs_oh_clear); +int lprocfs_obd_rd_hash(char *page, char **start, off_t off, + int count, int *eof, void *data) +{ + struct obd_device *obd = data; + int c = 0; + + if (obd == NULL) + return 0; + + c += lustre_hash_debug_header(page, count); + c += lustre_hash_debug_str(obd->obd_uuid_hash, page + c, count - c); + c += lustre_hash_debug_str(obd->obd_nid_hash, page + c, count - c); + c += lustre_hash_debug_str(obd->obd_nid_stats_hash, page+c, count-c); + + return c; +} +EXPORT_SYMBOL(lprocfs_obd_rd_hash); + int lprocfs_obd_rd_recovery_status(char *page, char **start, off_t off, int count, int *eof, void *data) { @@ -1503,9 +2141,10 @@ int lprocfs_obd_rd_recovery_status(char *page, char **start, off_t off, if (lprocfs_obd_snprintf(&page, size, &len, "status: ") <= 0) goto out; - if (obd->obd_max_recoverable_clients == 0) { - lprocfs_obd_snprintf(&page, size, &len, "INACTIVE\n"); + if (lprocfs_obd_snprintf(&page, size, &len, "INACTIVE\n") <= 0) + goto out; + goto fclose; } @@ -1513,64 +2152,73 @@ int lprocfs_obd_rd_recovery_status(char *page, char **start, off_t off, if (obd->obd_recovering == 0) { if (lprocfs_obd_snprintf(&page, size, &len, "COMPLETE\n") <= 0) goto out; - - if (lprocfs_obd_snprintf(&page, size, &len, "recovery_start: %lu\n", - obd->obd_recovery_start) <= 0) + if (lprocfs_obd_snprintf(&page, size, &len, + "recovery_start: %lu\n", + obd->obd_recovery_start) <= 0) goto out; - - if (lprocfs_obd_snprintf(&page, size, &len, "recovery_end: %lu\n", - obd->obd_recovery_end) <= 0) + if (lprocfs_obd_snprintf(&page, size, &len, + "recovery_duration: %lu\n", + obd->obd_recovery_end - + obd->obd_recovery_start) <= 0) goto out; - - /* Number of clients have have completed recovery */ - if (lprocfs_obd_snprintf(&page, size, &len, "recovered_clients: %d\n", - obd->obd_max_recoverable_clients - obd->obd_recoverable_clients) <= 0) + /* Number of clients that have completed recovery */ + if (lprocfs_obd_snprintf(&page, size, &len, + "completed_clients: %d/%d\n", + obd->obd_max_recoverable_clients - + obd->obd_stale_clients, + obd->obd_max_recoverable_clients) <= 0) goto out; - - if (lprocfs_obd_snprintf(&page, size, &len, "unrecovered_clients: %d\n", - obd->obd_recoverable_clients) <= 0) + if (lprocfs_obd_snprintf(&page, size, &len, + "replayed_requests: %d\n", + obd->obd_replayed_requests) <= 0) goto out; - - if (lprocfs_obd_snprintf(&page, size, &len, "last_transno: "LPD64"\n", - obd->obd_next_recovery_transno - 1) <= 0) + if (lprocfs_obd_snprintf(&page, size, &len, + "last_transno: "LPD64"\n", + obd->obd_next_recovery_transno - 1)<=0) + goto out; + if (lprocfs_obd_snprintf(&page, size, &len, "VBR: %s\n", + obd->obd_version_recov ? "ON" : "OFF")<=0) goto out; - - lprocfs_obd_snprintf(&page, size, &len, "replayed_requests: %d\n", obd->obd_replayed_requests); goto fclose; } if (lprocfs_obd_snprintf(&page, size, &len, "RECOVERING\n") <= 0) goto out; - if (lprocfs_obd_snprintf(&page, size, &len, "recovery_start: %lu\n", - obd->obd_recovery_start) <= 0) + obd->obd_recovery_start) <= 0) goto out; - - if (lprocfs_obd_snprintf(&page, size, &len, "time remaining: %lu\n", - CURRENT_SECONDS >= obd->obd_recovery_end ? 0 : - obd->obd_recovery_end - CURRENT_SECONDS) <= 0) - goto out; - - if(lprocfs_obd_snprintf(&page, size, &len, "connected_clients: %d/%d\n", - obd->obd_connected_clients, - obd->obd_max_recoverable_clients) <= 0) + if (lprocfs_obd_snprintf(&page, size, &len, "time_remaining: %lu\n", + cfs_time_current_sec() >= obd->obd_recovery_end ? 0 : + obd->obd_recovery_end - cfs_time_current_sec()) <= 0) goto out; - - /* Number of clients have have completed recovery */ - if (lprocfs_obd_snprintf(&page, size, &len, "completed_clients: %d/%d\n", - obd->obd_max_recoverable_clients - obd->obd_recoverable_clients, + if (lprocfs_obd_snprintf(&page, size, &len,"connected_clients: %d/%d\n", + obd->obd_connected_clients, obd->obd_max_recoverable_clients) <= 0) goto out; - - if (lprocfs_obd_snprintf(&page, size, &len, "replayed_requests: %d/??\n", + /* Number of clients that have completed recovery */ + if (lprocfs_obd_snprintf(&page, size, &len,"req_replay_clients: %d\n", + atomic_read(&obd->obd_req_replay_clients))<= 0) + goto out; + if (lprocfs_obd_snprintf(&page, size, &len,"lock_repay_clients: %d\n", + atomic_read(&obd->obd_lock_replay_clients))<=0) + goto out; + if (lprocfs_obd_snprintf(&page, size, &len,"completed_clients: %d\n", + obd->obd_connected_clients - + atomic_read(&obd->obd_lock_replay_clients))<=0) + goto out; + if (lprocfs_obd_snprintf(&page, size, &len,"evicted_clients: %d\n", + obd->obd_stale_clients) <= 0) + goto out; + if (lprocfs_obd_snprintf(&page, size, &len,"replayed_requests: %d\n", obd->obd_replayed_requests) <= 0) goto out; - if (lprocfs_obd_snprintf(&page, size, &len, "queued_requests: %d\n", obd->obd_requests_queued_for_recovery) <= 0) goto out; - lprocfs_obd_snprintf(&page, size, &len, "next_transno: "LPD64"\n", obd->obd_next_recovery_transno); + if (lprocfs_obd_snprintf(&page, size, &len, "next_transno: "LPD64"\n", + obd->obd_next_recovery_transno) <= 0) + goto out; fclose: *eof = 1; @@ -1579,18 +2227,51 @@ out: } EXPORT_SYMBOL(lprocfs_obd_rd_recovery_status); +int lprocfs_obd_rd_recovery_maxtime(char *page, char **start, off_t off, + int count, int *eof, void *data) +{ + struct obd_device *obd = data; + LASSERT(obd != NULL); + + return snprintf(page, count, "%lu\n", obd->obd_recovery_max_time); +} +EXPORT_SYMBOL(lprocfs_obd_rd_recovery_maxtime); + +int lprocfs_obd_wr_recovery_maxtime(struct file *file, const char *buffer, + unsigned long count, void *data) +{ + struct obd_device *obd = data; + int val, rc; + LASSERT(obd != NULL); + + rc = lprocfs_write_helper(buffer, count, &val); + if (rc) + return rc; + + obd->obd_recovery_max_time = val; + return count; +} +EXPORT_SYMBOL(lprocfs_obd_wr_recovery_maxtime); + + EXPORT_SYMBOL(lprocfs_register); EXPORT_SYMBOL(lprocfs_srch); EXPORT_SYMBOL(lprocfs_remove); +EXPORT_SYMBOL(lprocfs_remove_proc_entry); EXPORT_SYMBOL(lprocfs_add_vars); EXPORT_SYMBOL(lprocfs_obd_setup); EXPORT_SYMBOL(lprocfs_obd_cleanup); +EXPORT_SYMBOL(lprocfs_add_simple); +EXPORT_SYMBOL(lprocfs_add_symlink); +EXPORT_SYMBOL(lprocfs_free_per_client_stats); EXPORT_SYMBOL(lprocfs_alloc_stats); EXPORT_SYMBOL(lprocfs_free_stats); EXPORT_SYMBOL(lprocfs_clear_stats); EXPORT_SYMBOL(lprocfs_register_stats); EXPORT_SYMBOL(lprocfs_init_ops_stats); +EXPORT_SYMBOL(lprocfs_init_ldlm_stats); EXPORT_SYMBOL(lprocfs_alloc_obd_stats); +EXPORT_SYMBOL(lprocfs_alloc_md_stats); EXPORT_SYMBOL(lprocfs_free_obd_stats); EXPORT_SYMBOL(lprocfs_exp_setup); EXPORT_SYMBOL(lprocfs_exp_cleanup); @@ -1607,17 +2288,23 @@ EXPORT_SYMBOL(lprocfs_rd_server_uuid); EXPORT_SYMBOL(lprocfs_rd_conn_uuid); EXPORT_SYMBOL(lprocfs_rd_num_exports); EXPORT_SYMBOL(lprocfs_rd_numrefs); - +EXPORT_SYMBOL(lprocfs_at_hist_helper); +EXPORT_SYMBOL(lprocfs_rd_import); +EXPORT_SYMBOL(lprocfs_rd_state); +EXPORT_SYMBOL(lprocfs_rd_timeouts); EXPORT_SYMBOL(lprocfs_rd_blksize); EXPORT_SYMBOL(lprocfs_rd_kbytestotal); EXPORT_SYMBOL(lprocfs_rd_kbytesfree); EXPORT_SYMBOL(lprocfs_rd_kbytesavail); EXPORT_SYMBOL(lprocfs_rd_filestotal); EXPORT_SYMBOL(lprocfs_rd_filesfree); +EXPORT_SYMBOL(lprocfs_rd_quota_resend_count); +EXPORT_SYMBOL(lprocfs_wr_quota_resend_count); EXPORT_SYMBOL(lprocfs_write_helper); EXPORT_SYMBOL(lprocfs_write_frac_helper); EXPORT_SYMBOL(lprocfs_read_frac_helper); EXPORT_SYMBOL(lprocfs_write_u64_helper); EXPORT_SYMBOL(lprocfs_write_frac_u64_helper); +EXPORT_SYMBOL(lprocfs_stats_collect); #endif /* LPROCFS*/