X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Fmdt%2Fmdt_lproc.c;h=c6445fbf3ff9c250fc8f1626ea04e52c91daae1e;hp=e3da0d9f1ee20230082b9f07f372aacecf803217;hb=2c787065441ee60c6c163dc77851d0964f81a89c;hpb=b18cecc6a17e9b30064ed9b24927d1905729d89b diff --git a/lustre/mdt/mdt_lproc.c b/lustre/mdt/mdt_lproc.c index e3da0d9..c6445fb 100644 --- a/lustre/mdt/mdt_lproc.c +++ b/lustre/mdt/mdt_lproc.c @@ -1,6 +1,4 @@ -/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*- - * vim:expandtab:shiftwidth=8:tabstop=8: - * +/* * GPL HEADER START * * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. @@ -17,11 +15,7 @@ * * 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. + * http://www.gnu.org/licenses/gpl-2.0.html * * GPL HEADER END */ @@ -29,12 +23,10 @@ * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. * - * Copyright (c) 2011 Whamcloud, Inc. - * + * Copyright (c) 2011, 2017, Intel Corporation. */ /* * This file is part of Lustre, http://www.lustre.org/ - * Lustre is a trademark of Sun Microsystems, Inc. * * lustre/mdt/mdt_lproc.c * @@ -42,18 +34,15 @@ * Author: Fan Yong */ -#ifndef EXPORT_SYMTAB -# define EXPORT_SYMTAB -#endif #define DEBUG_SUBSYSTEM S_MDS #include #include #include - +#include /* LUSTRE_VERSION_CODE */ -#include +#include /* * struct OBD_{ALLOC,FREE}*() * MDT_FAIL_CHECK @@ -65,22 +54,16 @@ #include #include #include -#include #include -#include #include "mdt_internal.h" -#include - -enum { - LPROC_MDT_NR -}; -static const char *mdt_proc_names[LPROC_MDT_NR] = { -}; +#include /** * The rename stats output would be YAML formats, like * rename_stats: - * - snapshot_time: 1234567890.123456 + * - snapshot_time: 1234567890.123456789 + * - start_time: 1234567880.987654321 + * - elapsed_time: 9.135802468 * - same_dir: * 4kB: { samples: 1230, pct: 33, cum_pct: 45 } * 8kB: { samples: 1242, pct: 33, cum_pct: 78 } @@ -95,968 +78,1549 @@ static const char *mdt_proc_names[LPROC_MDT_NR] = { * 16kB: { samples: 12, pct: 3, cum_pct: 81 } **/ -#define pct(a, b) (b ? a * 100 / b : 0) - static void display_rename_stats(struct seq_file *seq, char *name, - struct obd_histogram *hist) + struct obd_histogram *rs_hist) { - unsigned long tot, t, cum = 0; - int i; + unsigned long tot, t, cum = 0; + int i; + + tot = lprocfs_oh_sum(rs_hist); + if (tot > 0) + seq_printf(seq, "- %s\n", name); + + for (i = 0; i < OBD_HIST_MAX; i++) { + t = rs_hist->oh_buckets[i]; + cum += t; + if (cum == 0) + continue; + + if (i < 10) + seq_printf(seq, "%6s%d%s", " ", 1 << i, "bytes:"); + else if (i < 20) + seq_printf(seq, "%6s%d%s", " ", 1 << (i - 10), "KB:"); + else + seq_printf(seq, "%6s%d%s", " ", 1 << (i - 20), "MB:"); + + seq_printf(seq, " { sample: %3lu, pct: %3u, cum_pct: %3u }\n", + t, pct(t, tot), pct(cum, tot)); + + if (cum == tot) + break; + } +} - tot = lprocfs_oh_sum(hist); - if (tot > 0) - seq_printf(seq, "- %-15s\n", name); - /* dir size start from 4K, start i from 10(2^10) here */ - for (i = 0; i < OBD_HIST_MAX; i++) { - t = hist->oh_buckets[i]; - cum += t; - if (cum == 0) - continue; +static void rename_stats_show(struct seq_file *seq, + struct rename_stats *rename_stats) +{ + /* this sampling races with updates */ + seq_puts(seq, "rename_stats:\n- "); + lprocfs_stats_header(seq, ktime_get(), rename_stats->rs_init, 15, ":", + false); + + display_rename_stats(seq, "same_dir", + &rename_stats->rs_hist[RENAME_SAMEDIR_SIZE]); + display_rename_stats(seq, "crossdir_src", + &rename_stats->rs_hist[RENAME_CROSSDIR_SRC_SIZE]); + display_rename_stats(seq, "crossdir_tgt", + &rename_stats->rs_hist[RENAME_CROSSDIR_TGT_SIZE]); +} - if (i < 10) - seq_printf(seq, "%6s%d%s", " ", 1<< i, "bytes:"); - else if (i < 20) - seq_printf(seq, "%6s%d%s", " ", 1<<(i-10), "KB:"); - else - seq_printf(seq, "%6s%d%s", " ", 1<<(i-20), "MB:"); +static int mdt_rename_stats_seq_show(struct seq_file *seq, void *v) +{ + struct mdt_device *mdt = seq->private; - seq_printf(seq, " { sample: %3lu, pct: %3lu, cum_pct: %3lu }\n", - t, pct(t, tot), pct(cum, tot)); + rename_stats_show(seq, &mdt->mdt_rename_stats); - if (cum == tot) - break; - } + return 0; } -static void rename_stats_show(struct seq_file *seq, - struct rename_stats *rename_stats) +static ssize_t +mdt_rename_stats_seq_write(struct file *file, const char __user *buf, + size_t len, loff_t *off) { - struct timeval now; + struct seq_file *seq = file->private_data; + struct mdt_device *mdt = seq->private; + int i; - /* this sampling races with updates */ - do_gettimeofday(&now); - seq_printf(seq, "rename_stats:\n"); - seq_printf(seq, "- %-15s %lu.%lu\n", "snapshot_time:", - now.tv_sec, now.tv_usec); + for (i = 0; i < RENAME_LAST; i++) + lprocfs_oh_clear(&mdt->mdt_rename_stats.rs_hist[i]); + mdt->mdt_rename_stats.rs_init = ktime_get(); - display_rename_stats(seq, "same_dir", - &rename_stats->hist[RENAME_SAMEDIR_SIZE]); - display_rename_stats(seq, "crossdir_src", - &rename_stats->hist[RENAME_CROSSDIR_SRC_SIZE]); - display_rename_stats(seq, "crossdir_tgt", - &rename_stats->hist[RENAME_CROSSDIR_TGT_SIZE]); + return len; } +LPROC_SEQ_FOPS(mdt_rename_stats); + +static int lproc_mdt_attach_rename_seqstat(struct mdt_device *mdt) +{ + int i; -#undef pct + for (i = 0; i < RENAME_LAST; i++) + spin_lock_init(&mdt->mdt_rename_stats.rs_hist[i].oh_lock); -static int mdt_rename_stats_seq_show(struct seq_file *seq, void *v) + return lprocfs_obd_seq_create(mdt2obd_dev(mdt), "rename_stats", 0644, + &mdt_rename_stats_fops, mdt); +} + +void mdt_rename_counter_tally(struct mdt_thread_info *info, + struct mdt_device *mdt, + struct ptlrpc_request *req, + struct mdt_object *src, + struct mdt_object *tgt, long count) { - struct mdt_device *mdt = seq->private; + struct md_attr *ma = &info->mti_attr; + struct rename_stats *rstats = &mdt->mdt_rename_stats; + int rc; + + ma->ma_need = MA_INODE; + ma->ma_valid = 0; + rc = mo_attr_get(info->mti_env, mdt_object_child(src), ma); + if (rc) { + CERROR("%s: "DFID" attr_get, rc = %d\n", + mdt_obd_name(mdt), PFID(mdt_object_fid(src)), rc); + return; + } + + if (src == tgt) { + mdt_counter_incr(req, LPROC_MDT_SAMEDIR_RENAME, count); + lprocfs_oh_tally_log2(&rstats->rs_hist[RENAME_SAMEDIR_SIZE], + (unsigned int)ma->ma_attr.la_size); + return; + } + + mdt_counter_incr(req, LPROC_MDT_CROSSDIR_RENAME, count); + lprocfs_oh_tally_log2(&rstats->rs_hist[RENAME_CROSSDIR_SRC_SIZE], + (unsigned int)ma->ma_attr.la_size); + + ma->ma_need = MA_INODE; + ma->ma_valid = 0; + rc = mo_attr_get(info->mti_env, mdt_object_child(tgt), ma); + if (rc) { + CERROR("%s: "DFID" attr_get, rc = %d\n", + mdt_obd_name(mdt), PFID(mdt_object_fid(tgt)), rc); + return; + } + + lprocfs_oh_tally_log2(&rstats->rs_hist[RENAME_CROSSDIR_TGT_SIZE], + (unsigned int)ma->ma_attr.la_size); +} - rename_stats_show(seq, &mdt->mdt_rename_stats); +static ssize_t identity_expire_show(struct kobject *kobj, + struct attribute *attr, char *buf) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - return 0; + return scnprintf(buf, PAGE_SIZE, "%lld\n", + mdt->mdt_identity_cache->uc_entry_expire); } -static ssize_t mdt_rename_stats_seq_write(struct file *file, const char *buf, - size_t len, loff_t *off) +static ssize_t identity_expire_store(struct kobject *kobj, + struct attribute *attr, + const char *buffer, size_t count) { - struct seq_file *seq = file->private_data; - struct mdt_device *mdt = seq->private; - int i; + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + time64_t val; + int rc; + + rc = kstrtoll(buffer, 10, &val); + if (rc) + return rc; - for (i = 0; i < RENAME_LAST; i++) - lprocfs_oh_clear(&mdt->mdt_rename_stats.hist[i]); + if (val < 0) + return -ERANGE; - return len; + mdt->mdt_identity_cache->uc_entry_expire = val; + + return count; } +LUSTRE_RW_ATTR(identity_expire); -LPROC_SEQ_FOPS(mdt_rename_stats); +static ssize_t identity_acquire_expire_show(struct kobject *kobj, + struct attribute *attr, char *buf) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); -static int lproc_mdt_attach_rename_seqstat(struct mdt_device *mdt) + return scnprintf(buf, PAGE_SIZE, "%lld\n", + mdt->mdt_identity_cache->uc_acquire_expire); +} + +static ssize_t identity_acquire_expire_store(struct kobject *kobj, + struct attribute *attr, + const char *buffer, size_t count) { - struct lu_device *ld = &mdt->mdt_md_dev.md_lu_dev; - struct obd_device *obd = ld->ld_obd; - int i; + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + time64_t val; + int rc; - for (i = 0; i < RENAME_LAST; i++) - spin_lock_init(&mdt->mdt_rename_stats.hist[i].oh_lock); + rc = kstrtoll(buffer, 0, &val); + if (rc) + return rc; - return lprocfs_obd_seq_create(obd, "rename_stats", 0444, - &mdt_rename_stats_fops, mdt); + if (val < 0 || val > INT_MAX) + return -ERANGE; + + mdt->mdt_identity_cache->uc_acquire_expire = val; + + return count; } +LUSTRE_RW_ATTR(identity_acquire_expire); -void mdt_rename_counter_tally(struct mdt_thread_info *info, - struct mdt_device *mdt, - struct obd_export *exp, - struct mdt_object *src, - struct mdt_object *tgt) -{ - struct md_attr *ma = &info->mti_attr; - struct rename_stats *rstats = &mdt->mdt_rename_stats; - int rc; - - ma->ma_need = MA_INODE; - ma->ma_valid = 0; - rc = mo_attr_get(info->mti_env, mdt_object_child(src), ma); - if (rc) { - CERROR("%s: "DFID" attr_get, rc = %d\n", - exp->exp_obd->obd_name, PFID(mdt_object_fid(src)), rc); - return; - } - - if (src == tgt) { - mdt_counter_incr(exp, LPROC_MDT_SAMEDIR_RENAME); - lprocfs_oh_tally_log2(&rstats->hist[RENAME_SAMEDIR_SIZE], - (unsigned int)ma->ma_attr.la_size); - return; - } - - mdt_counter_incr(exp, LPROC_MDT_CROSSDIR_RENAME); - lprocfs_oh_tally_log2(&rstats->hist[RENAME_CROSSDIR_SRC_SIZE], - (unsigned int)ma->ma_attr.la_size); - - ma->ma_need = MA_INODE; - ma->ma_valid = 0; - rc = mo_attr_get(info->mti_env, mdt_object_child(tgt), ma); - if (rc) { - CERROR("%s: "DFID" attr_get, rc = %d\n", - exp->exp_obd->obd_name, PFID(mdt_object_fid(tgt)), rc); - return; - } +static ssize_t identity_upcall_show(struct kobject *kobj, + struct attribute *attr, char *buf) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + struct upcall_cache *hash = mdt->mdt_identity_cache; + int rc; + + down_read(&hash->uc_upcall_rwsem); + rc = scnprintf(buf, PAGE_SIZE, "%s\n", hash->uc_upcall); + up_read(&hash->uc_upcall_rwsem); + return rc; +} + +static ssize_t identity_upcall_store(struct kobject *kobj, + struct attribute *attr, + const char *buffer, size_t count) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + struct upcall_cache *hash = mdt->mdt_identity_cache; + + if (count >= UC_CACHE_UPCALL_MAXPATH) { + CERROR("%s: identity upcall too long\n", mdt_obd_name(mdt)); + return -EINVAL; + } + + /* Remove any extraneous bits from the upcall (e.g. linefeeds) */ + down_write(&hash->uc_upcall_rwsem); + sscanf(buffer, "%s", hash->uc_upcall); + up_write(&hash->uc_upcall_rwsem); + + if (strcmp(hash->uc_name, mdt_obd_name(mdt)) != 0) + CWARN("%s: write to upcall name %s\n", + mdt_obd_name(mdt), hash->uc_upcall); + + if (strcmp(hash->uc_upcall, "NONE") == 0 && mdt->mdt_opts.mo_acl) + CWARN("%s: disable \"identity_upcall\" with ACL enabled maybe " + "cause unexpected \"EACCESS\"\n", mdt_obd_name(mdt)); + + CDEBUG(D_CONFIG, "%s: identity upcall set to %s\n", mdt_obd_name(mdt), + hash->uc_upcall); + RETURN(count); +} +LUSTRE_RW_ATTR(identity_upcall); + +static ssize_t identity_flush_store(struct kobject *kobj, + struct attribute *attr, + const char *buffer, size_t count) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + int uid; + int rc; + + rc = kstrtoint(buffer, 0, &uid); + if (rc) + return rc; + + mdt_flush_identity(mdt->mdt_identity_cache, uid); + return count; +} +LUSTRE_WO_ATTR(identity_flush); + +static ssize_t +lprocfs_identity_info_seq_write(struct file *file, const char __user *buffer, + size_t count, void *data) +{ + struct seq_file *m = file->private_data; + struct obd_device *obd = m->private; + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + struct identity_downcall_data *param; + int size = sizeof(*param), rc, checked = 0; - lprocfs_oh_tally_log2(&rstats->hist[RENAME_CROSSDIR_TGT_SIZE], - (unsigned int)ma->ma_attr.la_size); -} +again: + if (count < size) { + CERROR("%s: invalid data count = %lu, size = %d\n", + mdt_obd_name(mdt), (unsigned long) count, size); + return -EINVAL; + } + + OBD_ALLOC(param, size); + if (param == NULL) + return -ENOMEM; + + if (copy_from_user(param, buffer, size)) { + CERROR("%s: bad identity data\n", mdt_obd_name(mdt)); + GOTO(out, rc = -EFAULT); + } + + if (checked == 0) { + checked = 1; + if (param->idd_magic != IDENTITY_DOWNCALL_MAGIC) { + CERROR("%s: MDS identity downcall bad params\n", + mdt_obd_name(mdt)); + GOTO(out, rc = -EINVAL); + } + + if (param->idd_nperms > N_PERMS_MAX) { + CERROR("%s: perm count %d more than maximum %d\n", + mdt_obd_name(mdt), param->idd_nperms, + N_PERMS_MAX); + GOTO(out, rc = -EINVAL); + } + + if (param->idd_ngroups > NGROUPS_MAX) { + CERROR("%s: group count %d more than maximum %d\n", + mdt_obd_name(mdt), param->idd_ngroups, + NGROUPS_MAX); + GOTO(out, rc = -EINVAL); + } + + if (param->idd_ngroups) { + rc = param->idd_ngroups; /* save idd_ngroups */ + OBD_FREE(param, size); + size = offsetof(struct identity_downcall_data, + idd_groups[rc]); + goto again; + } + } + + rc = upcall_cache_downcall(mdt->mdt_identity_cache, param->idd_err, + param->idd_uid, param); -int mdt_procfs_init(struct mdt_device *mdt, const char *name) -{ - struct lu_device *ld = &mdt->mdt_md_dev.md_lu_dev; - struct obd_device *obd = ld->ld_obd; - struct lprocfs_static_vars lvars; - int rc; - ENTRY; +out: + if (param != NULL) + OBD_FREE(param, size); - LASSERT(name != NULL); + return rc ? rc : count; +} +LPROC_SEQ_FOPS_WR_ONLY(mdt, identity_info); - lprocfs_mdt_init_vars(&lvars); - rc = lprocfs_obd_setup(obd, lvars.obd_vars); - if (rc) { - CERROR("Can't init lprocfs, rc %d\n", rc); - return rc; - } - ptlrpc_lprocfs_register_obd(obd); +static int mdt_site_stats_seq_show(struct seq_file *m, void *data) +{ + struct obd_device *obd = m->private; + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - mdt->mdt_proc_entry = obd->obd_proc_entry; - LASSERT(mdt->mdt_proc_entry != NULL); + return lu_site_stats_seq_print(mdt_lu_site(mdt), m); +} +LPROC_SEQ_FOPS_RO(mdt_site_stats); - rc = lu_time_init(&mdt->mdt_stats, mdt->mdt_proc_entry, - mdt_proc_names, ARRAY_SIZE(mdt_proc_names)); - if (rc == 0) - rc = lu_time_named_init(&ld->ld_site->ls_time_stats, - "site_time", mdt->mdt_proc_entry, - lu_time_names, - ARRAY_SIZE(lu_time_names)); - if (rc) - return rc; +#define BUFLEN (UUID_MAX + 4) - obd->obd_proc_exports_entry = proc_mkdir("exports", - obd->obd_proc_entry); - if (obd->obd_proc_exports_entry) - lprocfs_add_simple(obd->obd_proc_exports_entry, - "clear", lprocfs_nid_stats_clear_read, - lprocfs_nid_stats_clear_write, obd, NULL); - rc = lprocfs_alloc_md_stats(obd, LPROC_MDT_LAST); - if (rc == 0) - mdt_stats_counter_init(obd->md_stats); +static ssize_t +lprocfs_mds_evict_client_seq_write(struct file *file, const char __user *buf, + size_t count, loff_t *off) +{ + struct seq_file *m = file->private_data; + struct obd_device *obd = m->private; + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + char *kbuf; + char *tmpbuf; + int rc = 0; + + OBD_ALLOC(kbuf, BUFLEN); + if (kbuf == NULL) + return -ENOMEM; + + /* + * OBD_ALLOC() will zero kbuf, but we only copy BUFLEN - 1 + * bytes into kbuf, to ensure that the string is NUL-terminated. + * UUID_MAX should include a trailing NUL already. + */ + if (copy_from_user(kbuf, buf, min_t(unsigned long, BUFLEN - 1, count))) + GOTO(out, rc = -EFAULT); + tmpbuf = skip_spaces(kbuf); + tmpbuf = strsep(&tmpbuf, " \t\n\f\v\r"); + + if (strncmp(tmpbuf, "nid:", 4) != 0) { + count = lprocfs_evict_client_seq_write(file, buf, count, off); + goto out; + } + + if (mdt->mdt_opts.mo_evict_tgt_nids) { + rc = obd_set_info_async(NULL, mdt->mdt_child_exp, + sizeof(KEY_EVICT_BY_NID), + KEY_EVICT_BY_NID, + strlen(tmpbuf + 4) + 1, + tmpbuf + 4, NULL); + if (rc) + CERROR("Failed to evict nid %s from OSTs: rc %d\n", + tmpbuf + 4, rc); + } + + /* See the comments in function lprocfs_wr_evict_client() + * in ptlrpc/lproc_ptlrpc.c for details. - jay */ + class_incref(obd, __func__, current); + obd_export_evict_by_nid(obd, tmpbuf + 4); + class_decref(obd, __func__, current); - rc = lproc_mdt_attach_rename_seqstat(mdt); - if (rc) - CERROR("%s: MDT can not create rename stats rc = %d\n", - obd->obd_name, rc); - RETURN(rc); +out: + OBD_FREE(kbuf, BUFLEN); + return rc < 0 ? rc : count; } -int mdt_procfs_fini(struct mdt_device *mdt) +#undef BUFLEN + +static ssize_t evict_tgt_nids_show(struct kobject *kobj, + struct attribute *attr, char *buf) { - struct lu_device *ld = &mdt->mdt_md_dev.md_lu_dev; - struct obd_device *obd = ld->ld_obd; + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - if (obd->obd_proc_exports_entry) { - lprocfs_remove_proc_entry("clear", obd->obd_proc_exports_entry); - obd->obd_proc_exports_entry = NULL; - } - lprocfs_free_per_client_stats(obd); - lprocfs_obd_cleanup(obd); - ptlrpc_lprocfs_unregister_obd(obd); - if (mdt->mdt_proc_entry) { - lu_time_fini(&ld->ld_site->ls_time_stats); - lu_time_fini(&mdt->mdt_stats); - mdt->mdt_proc_entry = NULL; - } - lprocfs_free_md_stats(obd); - lprocfs_free_obd_stats(obd); + return scnprintf(buf, PAGE_SIZE, "%u\n", + mdt->mdt_opts.mo_evict_tgt_nids); +} - RETURN(0); +static ssize_t evict_tgt_nids_store(struct kobject *kobj, + struct attribute *attr, const char *buffer, + size_t count) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + bool val; + int rc; + + rc = kstrtobool(buffer, &val); + if (rc) + return rc; + + mdt->mdt_opts.mo_evict_tgt_nids = val; + return count; } +LUSTRE_RW_ATTR(evict_tgt_nids); -void mdt_time_start(const struct mdt_thread_info *info) +static ssize_t commit_on_sharing_show(struct kobject *kobj, + struct attribute *attr, char *buf) { - lu_lprocfs_time_start(info->mti_env); + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + + return scnprintf(buf, PAGE_SIZE, "%u\n", mdt_cos_is_enabled(mdt)); } -void mdt_time_end(const struct mdt_thread_info *info, int idx) +static ssize_t commit_on_sharing_store(struct kobject *kobj, + struct attribute *attr, + const char *buffer, size_t count) { - lu_lprocfs_time_end(info->mti_env, info->mti_mdt->mdt_stats, idx); + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + bool val; + int rc; + + rc = kstrtobool(buffer, &val); + if (rc) + return rc; + + mdt_enable_cos(mdt, val); + return count; } +LUSTRE_RW_ATTR(commit_on_sharing); -static int lprocfs_rd_identity_expire(char *page, char **start, off_t off, - int count, int *eof, void *data) +static ssize_t local_recovery_show(struct kobject *kobj, + struct attribute *attr, char *buf) { - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); - *eof = 1; - return snprintf(page, count, "%u\n", - mdt->mdt_identity_cache->uc_entry_expire); + return scnprintf(buf, PAGE_SIZE, "%u\n", + obd->u.obt.obt_lut->lut_local_recovery); } -static int lprocfs_wr_identity_expire(struct file *file, const char *buffer, - unsigned long count, void *data) +static ssize_t local_recovery_store(struct kobject *kobj, + struct attribute *attr, + const char *buffer, size_t count) { - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - int rc, val; + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + bool val; + int rc; - rc = lprocfs_write_helper(buffer, count, &val); - if (rc) - return rc; + rc = kstrtobool(buffer, &val); + if (rc) + return rc; - mdt->mdt_identity_cache->uc_entry_expire = val; - return count; + obd->u.obt.obt_lut->lut_local_recovery = !!val; + return count; } +LUSTRE_RW_ATTR(local_recovery); -static int lprocfs_rd_identity_acquire_expire(char *page, char **start, - off_t off, int count, int *eof, - void *data) +static int mdt_root_squash_seq_show(struct seq_file *m, void *data) { - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + struct obd_device *obd = m->private; + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + struct root_squash_info *squash = &mdt->mdt_squash; - *eof = 1; - return snprintf(page, count, "%u\n", - mdt->mdt_identity_cache->uc_acquire_expire); + seq_printf(m, "%u:%u\n", squash->rsi_uid, + squash->rsi_gid); + return 0; } -static int lprocfs_wr_identity_acquire_expire(struct file *file, - const char *buffer, - unsigned long count, - void *data) +static ssize_t +mdt_root_squash_seq_write(struct file *file, const char __user *buffer, + size_t count, loff_t *off) { - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - int rc, val; + struct seq_file *m = file->private_data; + struct obd_device *obd = m->private; + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + struct root_squash_info *squash = &mdt->mdt_squash; - rc = lprocfs_write_helper(buffer, count, &val); - if (rc) - return rc; + return lprocfs_wr_root_squash(buffer, count, squash, + mdt_obd_name(mdt)); +} +LPROC_SEQ_FOPS(mdt_root_squash); - mdt->mdt_identity_cache->uc_acquire_expire = val; - return count; +static int mdt_nosquash_nids_seq_show(struct seq_file *m, void *data) +{ + struct obd_device *obd = m->private; + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + struct root_squash_info *squash = &mdt->mdt_squash; + int len = 0; + + spin_lock(&squash->rsi_lock); + if (!list_empty(&squash->rsi_nosquash_nids)) { + len = cfs_print_nidlist(m->buf + m->count, m->size - m->count, + &squash->rsi_nosquash_nids); + m->count += len; + seq_putc(m, '\n'); + } else + seq_puts(m, "NONE\n"); + spin_unlock(&squash->rsi_lock); + + return 0; } -static int lprocfs_rd_identity_upcall(char *page, char **start, off_t off, - int count, int *eof, void *data) +static ssize_t +mdt_nosquash_nids_seq_write(struct file *file, const char __user *buffer, + size_t count, loff_t *off) { - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - struct upcall_cache *hash = mdt->mdt_identity_cache; - int len; + struct seq_file *m = file->private_data; + struct obd_device *obd = m->private; + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + struct root_squash_info *squash = &mdt->mdt_squash; - *eof = 1; - cfs_read_lock(&hash->uc_upcall_rwlock); - len = snprintf(page, count, "%s\n", hash->uc_upcall); - cfs_read_unlock(&hash->uc_upcall_rwlock); - return len; + return lprocfs_wr_nosquash_nids(buffer, count, squash, + mdt_obd_name(mdt)); } +LPROC_SEQ_FOPS(mdt_nosquash_nids); -static int lprocfs_wr_identity_upcall(struct file *file, const char *buffer, - unsigned long count, void *data) +static ssize_t enable_remote_dir_show(struct kobject *kobj, + struct attribute *attr, char *buf) { - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - struct upcall_cache *hash = mdt->mdt_identity_cache; - int rc; - char *kernbuf; + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + + return scnprintf(buf, PAGE_SIZE, "%u\n", mdt->mdt_enable_remote_dir); +} - if (count >= UC_CACHE_UPCALL_MAXPATH) { - CERROR("%s: identity upcall too long\n", obd->obd_name); - return -EINVAL; - } - OBD_ALLOC(kernbuf, count + 1); - if (kernbuf == NULL) - GOTO(failed, rc = -ENOMEM); - if (cfs_copy_from_user(kernbuf, buffer, count)) - GOTO(failed, rc = -EFAULT); +static ssize_t enable_remote_dir_store(struct kobject *kobj, + struct attribute *attr, + const char *buffer, size_t count) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + bool val; + int rc; + + rc = kstrtobool(buffer, &val); + if (rc) + return rc; + + mdt->mdt_enable_remote_dir = val; + return count; +} +LUSTRE_RW_ATTR(enable_remote_dir); - /* Remove any extraneous bits from the upcall (e.g. linefeeds) */ - cfs_write_lock(&hash->uc_upcall_rwlock); - sscanf(kernbuf, "%s", hash->uc_upcall); - cfs_write_unlock(&hash->uc_upcall_rwlock); +static ssize_t enable_remote_dir_gid_show(struct kobject *kobj, + struct attribute *attr, char *buf) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - if (strcmp(hash->uc_name, obd->obd_name) != 0) - CWARN("%s: write to upcall name %s\n", - obd->obd_name, hash->uc_upcall); + return scnprintf(buf, PAGE_SIZE, "%d\n", + (int)mdt->mdt_enable_remote_dir_gid); +} - if (strcmp(hash->uc_upcall, "NONE") == 0 && mdt->mdt_opts.mo_acl) - CWARN("%s: disable \"identity_upcall\" with ACL enabled maybe " - "cause unexpected \"EACCESS\"\n", obd->obd_name); +static ssize_t enable_remote_dir_gid_store(struct kobject *kobj, + struct attribute *attr, + const char *buffer, size_t count) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + int val; + int rc; + + rc = kstrtoint(buffer, 0, &val); + if (rc) + return rc; + + mdt->mdt_enable_remote_dir_gid = val; + return count; +} +LUSTRE_RW_ATTR(enable_remote_dir_gid); - CWARN("%s: identity upcall set to %s\n", obd->obd_name, hash->uc_upcall); - OBD_FREE(kernbuf, count + 1); - RETURN(count); +static ssize_t enable_chprojid_gid_show(struct kobject *kobj, + struct attribute *attr, char *buf) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - failed: - if (kernbuf) - OBD_FREE(kernbuf, count + 1); - RETURN(rc); + return scnprintf(buf, PAGE_SIZE, "%d\n", + (int)mdt->mdt_enable_chprojid_gid); } -static int lprocfs_wr_identity_flush(struct file *file, const char *buffer, - unsigned long count, void *data) +static ssize_t enable_chprojid_gid_store(struct kobject *kobj, + struct attribute *attr, + const char *buffer, size_t count) { - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - int rc, uid; + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + int val; + int rc; + + rc = kstrtoint(buffer, 0, &val); + if (rc) + return rc; + + mdt->mdt_enable_chprojid_gid = val; + return count; +} +LUSTRE_RW_ATTR(enable_chprojid_gid); - rc = lprocfs_write_helper(buffer, count, &uid); - if (rc) - return rc; +static ssize_t enable_striped_dir_show(struct kobject *kobj, + struct attribute *attr, char *buf) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - mdt_flush_identity(mdt->mdt_identity_cache, uid); - return count; + return scnprintf(buf, PAGE_SIZE, "%u\n", mdt->mdt_enable_striped_dir); } -static int lprocfs_wr_identity_info(struct file *file, const char *buffer, - unsigned long count, void *data) +static ssize_t enable_striped_dir_store(struct kobject *kobj, + struct attribute *attr, + const char *buffer, size_t count) { - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - struct identity_downcall_data *param; - int size = sizeof(*param), rc, checked = 0; + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + bool val; + int rc; + + rc = kstrtobool(buffer, &val); + if (rc) + return rc; + + mdt->mdt_enable_striped_dir = val; + return count; +} +LUSTRE_RW_ATTR(enable_striped_dir); -again: - if (count < size) { - CERROR("%s: invalid data count = %lu, size = %d\n", - obd->obd_name, count, size); - return -EINVAL; - } - - OBD_ALLOC(param, size); - if (param == NULL) - return -ENOMEM; - - if (cfs_copy_from_user(param, buffer, size)) { - CERROR("%s: bad identity data\n", obd->obd_name); - GOTO(out, rc = -EFAULT); - } - - if (checked == 0) { - checked = 1; - if (param->idd_magic != IDENTITY_DOWNCALL_MAGIC) { - CERROR("%s: MDS identity downcall bad params\n", - obd->obd_name); - GOTO(out, rc = -EINVAL); - } - - if (param->idd_nperms > N_PERMS_MAX) { - CERROR("%s: perm count %d more than maximum %d\n", - obd->obd_name, param->idd_nperms, N_PERMS_MAX); - GOTO(out, rc = -EINVAL); - } - - if (param->idd_ngroups > NGROUPS_MAX) { - CERROR("%s: group count %d more than maximum %d\n", - obd->obd_name, param->idd_ngroups, NGROUPS_MAX); - GOTO(out, rc = -EINVAL); - } - - if (param->idd_ngroups) { - rc = param->idd_ngroups; /* save idd_ngroups */ - OBD_FREE(param, size); - size = offsetof(struct identity_downcall_data, - idd_groups[rc]); - goto again; - } - } - - rc = upcall_cache_downcall(mdt->mdt_identity_cache, param->idd_err, - param->idd_uid, param); +static ssize_t enable_dir_migration_show(struct kobject *kobj, + struct attribute *attr, char *buf) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); -out: - if (param != NULL) - OBD_FREE(param, size); + return scnprintf(buf, PAGE_SIZE, "%u\n", mdt->mdt_enable_dir_migration); +} - return rc ? rc : count; +static ssize_t enable_dir_migration_store(struct kobject *kobj, + struct attribute *attr, + const char *buffer, size_t count) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + bool val; + int rc; + + rc = kstrtobool(buffer, &val); + if (rc) + return rc; + + mdt->mdt_enable_dir_migration = val; + return count; } +LUSTRE_RW_ATTR(enable_dir_migration); -/* for debug only */ -static int lprocfs_rd_capa(char *page, char **start, off_t off, - int count, int *eof, void *data) +static ssize_t enable_dir_restripe_show(struct kobject *kobj, + struct attribute *attr, char *buf) { - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - return snprintf(page, count, "capability on: %s %s\n", - mdt->mdt_opts.mo_oss_capa ? "oss" : "", - mdt->mdt_opts.mo_mds_capa ? "mds" : ""); + return scnprintf(buf, PAGE_SIZE, "%u\n", mdt->mdt_enable_dir_restripe); } -static int lprocfs_wr_capa(struct file *file, const char *buffer, - unsigned long count, void *data) +static ssize_t enable_dir_restripe_store(struct kobject *kobj, + struct attribute *attr, + const char *buffer, size_t count) { - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - int val, rc; - - rc = lprocfs_write_helper(buffer, count, &val); - if (rc) - return rc; + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + bool val; + int rc; + + rc = kstrtobool(buffer, &val); + if (rc) + return rc; + + mdt->mdt_enable_dir_restripe = val; + return count; +} +LUSTRE_RW_ATTR(enable_dir_restripe); - if (val < 0 || val > 3) { - CERROR("invalid capability mode, only 0/2/3 is accepted.\n" - " 0: disable fid capability\n" - " 2: enable MDS fid capability\n" - " 3: enable both MDS and OSS fid capability\n"); - return -EINVAL; - } +static ssize_t enable_dir_auto_split_show(struct kobject *kobj, + struct attribute *attr, char *buf) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - /* OSS fid capability needs enable both MDS and OSS fid capability on - * MDS */ - if (val == 1) { - CERROR("can't enable OSS fid capability only, you should use " - "'3' to enable both MDS and OSS fid capability.\n"); - return -EINVAL; - } + return scnprintf(buf, PAGE_SIZE, "%u\n", + mdt->mdt_enable_dir_auto_split); +} - mdt->mdt_opts.mo_oss_capa = (val & 0x1); - mdt->mdt_opts.mo_mds_capa = !!(val & 0x2); - mdt->mdt_capa_conf = 1; - LCONSOLE_INFO("MDS %s %s MDS fid capability.\n", - obd->obd_name, - mdt->mdt_opts.mo_mds_capa ? "enabled" : "disabled"); - LCONSOLE_INFO("MDS %s %s OSS fid capability.\n", - obd->obd_name, - mdt->mdt_opts.mo_oss_capa ? "enabled" : "disabled"); - return count; +static ssize_t enable_dir_auto_split_store(struct kobject *kobj, + struct attribute *attr, + const char *buffer, size_t count) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + bool val; + int rc; + + rc = kstrtobool(buffer, &val); + if (rc) + return rc; + + mdt->mdt_enable_dir_auto_split = val; + return count; } +LUSTRE_RW_ATTR(enable_dir_auto_split); -static int lprocfs_rd_capa_count(char *page, char **start, off_t off, - int count, int *eof, void *data) +/** + * Show MDT async commit count. + * + * @m seq_file handle + * @data unused for single entry + * + * Return: 0 on success + * negative value on error + */ +static ssize_t async_commit_count_show(struct kobject *kobj, + struct attribute *attr, char *buf) { - return snprintf(page, count, "%d %d\n", - capa_count[CAPA_SITE_CLIENT], - capa_count[CAPA_SITE_SERVER]); + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + + return scnprintf(buf, PAGE_SIZE, "%d\n", + atomic_read(&mdt->mdt_async_commit_count)); } -static int lprocfs_rd_site_stats(char *page, char **start, off_t off, - int count, int *eof, void *data) +static ssize_t async_commit_count_store(struct kobject *kobj, + struct attribute *attr, + const char *buffer, size_t count) { - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + int val; + int rc; + + rc = kstrtoint(buffer, 10, &val); + if (rc) + return rc; + + atomic_set(&mdt->mdt_async_commit_count, val); - return lu_site_stats_print(mdt_lu_site(mdt), page, count); + return count; } +LUSTRE_RW_ATTR(async_commit_count); -static int lprocfs_rd_capa_timeout(char *page, char **start, off_t off, - int count, int *eof, void *data) +/** + * Show MDT sync count. + * + * \param[in] m seq_file handle + * \param[in] data unused for single entry + * + * \retval 0 on success + * \retval negative value on error + */ +static ssize_t sync_count_show(struct kobject *kobj, struct attribute *attr, + char *buf) { - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct lu_target *tgt = obd->u.obt.obt_lut; - return snprintf(page, count, "%lu\n", mdt->mdt_capa_timeout); + return scnprintf(buf, PAGE_SIZE, "%d\n", + atomic_read(&tgt->lut_sync_count)); } -static int lprocfs_wr_capa_timeout(struct file *file, const char *buffer, - unsigned long count, void *data) +static ssize_t sync_count_store(struct kobject *kobj, struct attribute *attr, + const char *buffer, size_t count) { - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - int val, rc; + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct lu_target *tgt = obd->u.obt.obt_lut; + int val; + int rc; + + rc = kstrtoint(buffer, 0, &val); + if (rc) + return rc; - rc = lprocfs_write_helper(buffer, count, &val); - if (rc) - return rc; + atomic_set(&tgt->lut_sync_count, val); - mdt->mdt_capa_timeout = (unsigned long)val; - mdt->mdt_capa_conf = 1; - return count; + return count; } +LUSTRE_RW_ATTR(sync_count); + +static const char *dom_open_lock_modes[NUM_DOM_LOCK_ON_OPEN_MODES] = { + [NO_DOM_LOCK_ON_OPEN] = "never", + [TRYLOCK_DOM_ON_OPEN] = "trylock", + [ALWAYS_DOM_LOCK_ON_OPEN] = "always", +}; + +/* This must be longer than the longest string above */ +#define DOM_LOCK_MODES_MAXLEN 16 -static int lprocfs_rd_ck_timeout(char *page, char **start, off_t off, int count, - int *eof, void *data) +/** + * Show MDT policy for data prefetch on open for DoM files.. + * + * \param[in] m seq_file handle + * \param[in] data unused + * + * \retval 0 on success + * \retval negative value on error + */ +static ssize_t dom_lock_show(struct kobject *kobj, struct attribute *attr, + char *buf) { - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - return snprintf(page, count, "%lu\n", mdt->mdt_ck_timeout); + return scnprintf(buf, PAGE_SIZE, "%s\n", + dom_open_lock_modes[mdt->mdt_opts.mo_dom_lock]); } -static int lprocfs_wr_ck_timeout(struct file *file, const char *buffer, - unsigned long count, void *data) +/** + * Change MDT policy for data prefetch on open for DoM files. + * + * This variable defines how DOM lock is taken at open enqueue. + * There are three possible modes: + * 1) never - never take DoM lock on open. DoM lock will be taken as separate + * IO lock with own enqueue. + * 2) trylock - DoM lock will be taken only if non-blocked. + * 3) always - DoM lock will be taken always even if it is blocking lock. + * + * If dom_read_open is enabled too then DoM lock is taken in PR mode and + * is paired with LAYOUT lock when possible. + * + * \param[in] file proc file + * \param[in] buffer string which represents policy + * \param[in] count \a buffer length + * \param[in] off unused for single entry + * + * \retval \a count on success + * \retval negative number on error + */ +static ssize_t dom_lock_store(struct kobject *kobj, struct attribute *attr, + const char *buffer, size_t count) { - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - int val, rc; + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + int val = -1; + int i, rc; + + if (count == 0 || count >= DOM_LOCK_MODES_MAXLEN) + return -EINVAL; + + for (i = 0 ; i < NUM_DOM_LOCK_ON_OPEN_MODES; i++) { + /* buffer might have '\n' but using strlen() avoids it */ + if (strncmp(buffer, dom_open_lock_modes[i], + strlen(dom_open_lock_modes[i])) == 0) { + val = i; + break; + } + } + + /* Legacy numeric codes */ + if (val == -1) { + rc = kstrtoint(buffer, 0, &val); + if (rc) + return rc; + } + + if (val == ALWAYS_DOM_LOCK_ON_OPEN) + val = TRYLOCK_DOM_ON_OPEN; + + if (val < 0 || val >= NUM_DOM_LOCK_ON_OPEN_MODES) + return -EINVAL; + + mdt->mdt_opts.mo_dom_lock = val; + return count; +} +LUSTRE_RW_ATTR(dom_lock); - rc = lprocfs_write_helper(buffer, count, &val); - if (rc) - return rc; +/** + * Show MDT policy for data prefetch on open for DoM files.. + * + * \param[in] m seq_file handle + * \param[in] data unused + * + * \retval 0 on success + * \retval negative value on error + */ +static ssize_t dom_read_open_show(struct kobject *kobj, + struct attribute *attr, char *buf) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - mdt->mdt_ck_timeout = (unsigned long)val; - mdt->mdt_capa_conf = 1; - return count; + return scnprintf(buf, PAGE_SIZE, "%u\n", + !!mdt->mdt_opts.mo_dom_read_open); } -#define BUFLEN (UUID_MAX + 4) +/** + * Modify MDT policy for data prefetch on open for DoM files. + * + * If enabled then Data-on-MDT file data may be read during open and + * returned back in reply. It works only with mo_dom_lock enabled. + * + * \param[in] file proc file + * \param[in] buffer string which represents policy + * \param[in] count \a buffer length + * \param[in] off unused for single entry + * + * \retval \a count on success + * \retval negative number on error + */ +static ssize_t dom_read_open_store(struct kobject *kobj, + struct attribute *attr, const char *buffer, + size_t count) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + bool val; + int rc; + + rc = kstrtobool(buffer, &val); + if (rc) + return rc; + + mdt->mdt_opts.mo_dom_read_open = !!val; + return count; +} +LUSTRE_RW_ATTR(dom_read_open); -static int lprocfs_mdt_wr_evict_client(struct file *file, const char *buffer, - unsigned long count, void *data) +static ssize_t migrate_hsm_allowed_show(struct kobject *kobj, + struct attribute *attr, char *buf) { - char *kbuf; - char *tmpbuf; + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - OBD_ALLOC(kbuf, BUFLEN); - if (kbuf == NULL) - return -ENOMEM; + return scnprintf(buf, PAGE_SIZE, "%u\n", + mdt->mdt_opts.mo_migrate_hsm_allowed); +} - /* - * OBD_ALLOC() will zero kbuf, but we only copy BUFLEN - 1 - * bytes into kbuf, to ensure that the string is NUL-terminated. - * UUID_MAX should include a trailing NUL already. - */ - if (cfs_copy_from_user(kbuf, buffer, - min_t(unsigned long, BUFLEN - 1, count))) { - count = -EFAULT; - goto out; - } - tmpbuf = cfs_firststr(kbuf, min_t(unsigned long, BUFLEN - 1, count)); +static ssize_t migrate_hsm_allowed_store(struct kobject *kobj, + struct attribute *attr, + const char *buffer, size_t count) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + bool val; + int rc; + + rc = kstrtobool(buffer, &val); + if (rc) + return rc; + + mdt->mdt_opts.mo_migrate_hsm_allowed = val; + return count; +} +LUSTRE_RW_ATTR(migrate_hsm_allowed); - if (strncmp(tmpbuf, "nid:", 4) != 0) { - count = lprocfs_wr_evict_client(file, buffer, count, data); - goto out; - } +static ssize_t readonly_show(struct kobject *kobj, struct attribute *attr, + char *buf) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - CERROR("NOT implement evict client by nid %s\n", tmpbuf); + return scnprintf(buf, PAGE_SIZE, "%u\n", mdt->mdt_readonly); +} -out: - OBD_FREE(kbuf, BUFLEN); - return count; +static ssize_t readonly_store(struct kobject *kobj, struct attribute *attr, + const char *buffer, size_t count) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + bool val; + int rc; + + rc = kstrtobool(buffer, &val); + if (rc) + return rc; + + mdt->mdt_readonly = val; + return count; } +LUSTRE_RW_ATTR(readonly); -#undef BUFLEN +static ssize_t enable_remote_rename_show(struct kobject *kobj, + struct attribute *attr, + char *buf) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + + return scnprintf(buf, PAGE_SIZE, "%u\n", + mdt->mdt_enable_remote_rename); +} + +static ssize_t enable_remote_rename_store(struct kobject *kobj, + struct attribute *attr, + const char *buffer, size_t count) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + bool val; + int rc; + + rc = kstrtobool(buffer, &val); + if (rc) + return rc; + + mdt->mdt_enable_remote_rename = val; + return count; +} +LUSTRE_RW_ATTR(enable_remote_rename); -static int lprocfs_rd_sec_level(char *page, char **start, off_t off, - int count, int *eof, void *data) +static ssize_t dir_split_count_show(struct kobject *kobj, + struct attribute *attr, + char *buf) { - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - return snprintf(page, count, "%d\n", mdt->mdt_sec_level); + return scnprintf(buf, PAGE_SIZE, "%llu\n", + mdt->mdt_restriper.mdr_dir_split_count); } -static int lprocfs_wr_sec_level(struct file *file, const char *buffer, - unsigned long count, void *data) +static ssize_t dir_split_count_store(struct kobject *kobj, + struct attribute *attr, + const char *buffer, size_t count) { - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - int val, rc; + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + s64 val; + int rc; - rc = lprocfs_write_helper(buffer, count, &val); - if (rc) - return rc; + rc = sysfs_memparse(buffer, count, &val, "B"); + if (rc < 0) + return rc; - if (val > LUSTRE_SEC_ALL || val < LUSTRE_SEC_NONE) - return -EINVAL; + if (val < 0) + return -ERANGE; - if (val == LUSTRE_SEC_SPECIFY) { - CWARN("security level %d will be supported in future.\n", - LUSTRE_SEC_SPECIFY); - return -EINVAL; - } + mdt->mdt_restriper.mdr_dir_split_count = val; - mdt->mdt_sec_level = val; - return count; + return count; } +LUSTRE_RW_ATTR(dir_split_count); -static int lprocfs_rd_cos(char *page, char **start, off_t off, - int count, int *eof, void *data) +static ssize_t dir_split_delta_show(struct kobject *kobj, + struct attribute *attr, + char *buf) { - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - return snprintf(page, count, "%u\n", mdt_cos_is_enabled(mdt)); + return scnprintf(buf, PAGE_SIZE, "%u\n", + mdt->mdt_restriper.mdr_dir_split_delta); } -static int lprocfs_wr_cos(struct file *file, const char *buffer, - unsigned long count, void *data) +static ssize_t dir_split_delta_store(struct kobject *kobj, + struct attribute *attr, + const char *buffer, size_t count) { - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - int val, rc; + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + u32 val; + int rc; + + rc = kstrtouint(buffer, 0, &val); + if (rc) + return rc; + + mdt->mdt_restriper.mdr_dir_split_delta = val; - rc = lprocfs_write_helper(buffer, count, &val); - if (rc) - return rc; - mdt_enable_cos(mdt, val); - return count; + return count; } +LUSTRE_RW_ATTR(dir_split_delta); -static int lprocfs_rd_root_squash(char *page, char **start, off_t off, - int count, int *eof, void *data) +static ssize_t dir_restripe_nsonly_show(struct kobject *kobj, + struct attribute *attr, char *buf) { - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - return snprintf(page, count, "%u:%u\n", mdt->mdt_squash_uid, - mdt->mdt_squash_gid); + return scnprintf(buf, PAGE_SIZE, "%u\n", mdt->mdt_dir_restripe_nsonly); } -static int safe_strtoul(const char *str, char **endp, unsigned long *res) +static ssize_t dir_restripe_nsonly_store(struct kobject *kobj, + struct attribute *attr, + const char *buffer, size_t count) { - char n[24]; + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + bool val; + int rc; + + rc = kstrtobool(buffer, &val); + if (rc) + return rc; + + mdt->mdt_dir_restripe_nsonly = val; + return count; +} +LUSTRE_RW_ATTR(dir_restripe_nsonly); - *res = simple_strtoul(str, endp, 0); - if (str == *endp) - return 1; +static ssize_t enable_remote_subdir_mount_show(struct kobject *kobj, + struct attribute *attr, + char *buf) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - sprintf(n, "%lu", *res); - if (strncmp(n, str, *endp - str)) - /* overflow */ - return 1; - return 0; + return scnprintf(buf, PAGE_SIZE, "%u\n", + mdt->mdt_enable_remote_subdir_mount); } -static int lprocfs_wr_root_squash(struct file *file, const char *buffer, - unsigned long count, void *data) +static ssize_t enable_remote_subdir_mount_store(struct kobject *kobj, + struct attribute *attr, + const char *buffer, + size_t count) { - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - int rc; - char kernbuf[50], *tmp, *end, *errmsg; - unsigned long uid, gid; - int nouid, nogid; - ENTRY; - - if (count >= sizeof(kernbuf)) { - errmsg = "string too long"; - GOTO(failed, rc = -EINVAL); - } - if (cfs_copy_from_user(kernbuf, buffer, count)) { - errmsg = "bad address"; - GOTO(failed, rc = -EFAULT); - } - kernbuf[count] = '\0'; - - nouid = nogid = 0; - if (safe_strtoul(buffer, &tmp, &uid)) { - uid = mdt->mdt_squash_uid; - nouid = 1; - } - - /* skip ':' */ - if (*tmp == ':') { - tmp++; - if (safe_strtoul(tmp, &end, &gid)) { - gid = mdt->mdt_squash_gid; - nogid = 1; - } - } else { - gid = mdt->mdt_squash_gid; - nogid = 1; - } - - mdt->mdt_squash_uid = uid; - mdt->mdt_squash_gid = gid; - - if (nouid && nogid) { - errmsg = "needs uid:gid format"; - GOTO(failed, rc = -EINVAL); - } - - LCONSOLE_INFO("%s: root_squash is set to %u:%u\n", - obd->obd_name, - mdt->mdt_squash_uid, mdt->mdt_squash_gid); - RETURN(count); - - failed: - CWARN("%s: failed to set root_squash to \"%s\", %s: rc %d\n", - obd->obd_name, buffer, errmsg, rc); - RETURN(rc); -} - -static int lprocfs_rd_nosquash_nids(char *page, char **start, off_t off, - int count, int *eof, void *data) -{ - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - - if (mdt->mdt_nosquash_str) - return snprintf(page, count, "%s\n", mdt->mdt_nosquash_str); - return snprintf(page, count, "NONE\n"); -} - -static int lprocfs_wr_nosquash_nids(struct file *file, const char *buffer, - unsigned long count, void *data) -{ - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - int rc; - char *kernbuf, *errmsg; - cfs_list_t tmp; - ENTRY; - - OBD_ALLOC(kernbuf, count + 1); - if (kernbuf == NULL) { - errmsg = "no memory"; - GOTO(failed, rc = -ENOMEM); - } - if (cfs_copy_from_user(kernbuf, buffer, count)) { - errmsg = "bad address"; - GOTO(failed, rc = -EFAULT); - } - kernbuf[count] = '\0'; - - if (!strcmp(kernbuf, "NONE") || !strcmp(kernbuf, "clear")) { - /* empty string is special case */ - cfs_down_write(&mdt->mdt_squash_sem); - if (!cfs_list_empty(&mdt->mdt_nosquash_nids)) { - cfs_free_nidlist(&mdt->mdt_nosquash_nids); - OBD_FREE(mdt->mdt_nosquash_str, - mdt->mdt_nosquash_strlen); - mdt->mdt_nosquash_str = NULL; - mdt->mdt_nosquash_strlen = 0; - } - cfs_up_write(&mdt->mdt_squash_sem); - LCONSOLE_INFO("%s: nosquash_nids is cleared\n", - obd->obd_name); - OBD_FREE(kernbuf, count + 1); - RETURN(count); - } - - CFS_INIT_LIST_HEAD(&tmp); - if (cfs_parse_nidlist(kernbuf, count, &tmp) <= 0) { - errmsg = "can't parse"; - GOTO(failed, rc = -EINVAL); - } - - cfs_down_write(&mdt->mdt_squash_sem); - if (!cfs_list_empty(&mdt->mdt_nosquash_nids)) { - cfs_free_nidlist(&mdt->mdt_nosquash_nids); - OBD_FREE(mdt->mdt_nosquash_str, mdt->mdt_nosquash_strlen); - } - mdt->mdt_nosquash_str = kernbuf; - mdt->mdt_nosquash_strlen = count + 1; - cfs_list_splice(&tmp, &mdt->mdt_nosquash_nids); - - LCONSOLE_INFO("%s: nosquash_nids is set to %s\n", - obd->obd_name, kernbuf); - cfs_up_write(&mdt->mdt_squash_sem); - RETURN(count); - - failed: - CWARN("%s: failed to set nosquash_nids to \"%s\", %s: rc %d\n", - obd->obd_name, kernbuf, errmsg, rc); - if (kernbuf) - OBD_FREE(kernbuf, count + 1); - RETURN(rc); -} - -static int lprocfs_rd_mdt_som(char *page, char **start, off_t off, - int count, int *eof, void *data) -{ - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - - return snprintf(page, count, "%sabled\n", - mdt->mdt_som_conf ? "en" : "dis"); -} - -static int lprocfs_wr_mdt_som(struct file *file, const char *buffer, - unsigned long count, void *data) -{ - struct obd_export *exp; - struct obd_device *obd = data; - struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); - char kernbuf[16]; - unsigned long val = 0; - - if (count > (sizeof(kernbuf) - 1)) - return -EINVAL; - - if (cfs_copy_from_user(kernbuf, buffer, count)) - return -EFAULT; - - kernbuf[count] = '\0'; - - if (!strcmp(kernbuf, "enabled")) - val = 1; - else if (strcmp(kernbuf, "disabled")) - return -EINVAL; - - if (mdt->mdt_som_conf == val) - return count; - - if (!obd->obd_process_conf) { - CERROR("Temporary SOM change is not supported, use lctl " - "conf_param for permanent setting\n"); - return count; - } - - /* 1 stands for self export. */ - cfs_list_for_each_entry(exp, &obd->obd_exports, exp_obd_chain) { - if (exp == obd->obd_self_export) - continue; - if (exp->exp_connect_flags & OBD_CONNECT_MDS_MDS) - continue; - /* Some clients are already connected, skip the change */ - LCONSOLE_INFO("%s is already connected, SOM will be %s on " - "the next mount\n", exp->exp_client_uuid.uuid, - val ? "enabled" : "disabled"); - return count; - } - - mdt->mdt_som_conf = val; - LCONSOLE_INFO("Enabling SOM\n"); - - return count; -} - -/* Temporary; for testing purposes only */ -static int lprocfs_mdt_wr_mdc(struct file *file, const char *buffer, - unsigned long count, void *data) -{ - struct obd_device *obd = data; - struct obd_export *exp = NULL; - struct obd_uuid *uuid; - char *kbuf; - char *tmpbuf; - - OBD_ALLOC(kbuf, UUID_MAX); - if (kbuf == NULL) - return -ENOMEM; - - /* - * OBD_ALLOC() will zero kbuf, but we only copy UUID_MAX - 1 - * bytes into kbuf, to ensure that the string is NUL-terminated. - * UUID_MAX should include a trailing NUL already. - */ - if (cfs_copy_from_user(kbuf, buffer, - min_t(unsigned long, UUID_MAX - 1, count))) { - count = -EFAULT; - goto out; - } - tmpbuf = cfs_firststr(kbuf, min_t(unsigned long, UUID_MAX - 1, count)); - - OBD_ALLOC(uuid, UUID_MAX); - if (uuid == NULL) { - count = -ENOMEM; - goto out; - } - - obd_str2uuid(uuid, tmpbuf); - exp = cfs_hash_lookup(obd->obd_uuid_hash, uuid); - if (exp == NULL) { - CERROR("%s: no export %s found\n", - obd->obd_name, obd_uuid2str(uuid)); - } else { - mdt_hsm_copytool_send(exp); - class_export_put(exp); - } - - OBD_FREE(uuid, UUID_MAX); -out: - OBD_FREE(kbuf, UUID_MAX); - return count; + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev); + bool val; + int rc; + + rc = kstrtobool(buffer, &val); + if (rc) + return rc; + + mdt->mdt_enable_remote_subdir_mount = val; + return count; } +LUSTRE_RW_ATTR(enable_remote_subdir_mount); -static struct lprocfs_vars lprocfs_mdt_obd_vars[] = { - { "uuid", lprocfs_rd_uuid, 0, 0 }, - { "recovery_status", lprocfs_obd_rd_recovery_status, 0, 0 }, - { "num_exports", lprocfs_rd_num_exports, 0, 0 }, - { "identity_expire", lprocfs_rd_identity_expire, - lprocfs_wr_identity_expire, 0 }, - { "identity_acquire_expire", lprocfs_rd_identity_acquire_expire, - lprocfs_wr_identity_acquire_expire, 0 }, - { "identity_upcall", lprocfs_rd_identity_upcall, - lprocfs_wr_identity_upcall, 0 }, - { "identity_flush", 0, lprocfs_wr_identity_flush, 0 }, - { "identity_info", 0, lprocfs_wr_identity_info, 0 }, - { "capa", lprocfs_rd_capa, - lprocfs_wr_capa, 0 }, - { "capa_timeout", lprocfs_rd_capa_timeout, - lprocfs_wr_capa_timeout, 0 }, - { "capa_key_timeout", lprocfs_rd_ck_timeout, - lprocfs_wr_ck_timeout, 0 }, - { "capa_count", lprocfs_rd_capa_count, 0, 0 }, - { "site_stats", lprocfs_rd_site_stats, 0, 0 }, - { "evict_client", 0, lprocfs_mdt_wr_evict_client, 0 }, - { "hash_stats", lprocfs_obd_rd_hash, 0, 0 }, - { "sec_level", lprocfs_rd_sec_level, - lprocfs_wr_sec_level, 0 }, - { "commit_on_sharing", lprocfs_rd_cos, lprocfs_wr_cos, 0 }, - { "root_squash", lprocfs_rd_root_squash, - lprocfs_wr_root_squash, 0 }, - { "nosquash_nids", lprocfs_rd_nosquash_nids, - lprocfs_wr_nosquash_nids, 0 }, - { "som", lprocfs_rd_mdt_som, - lprocfs_wr_mdt_som, 0 }, - { "mdccomm", 0, lprocfs_mdt_wr_mdc, 0 }, - { "instance", lprocfs_target_rd_instance, 0 }, - { "ir_factor", lprocfs_obd_rd_ir_factor, - lprocfs_obd_wr_ir_factor, 0 }, - { 0 } +/** + * Show if the OFD enforces T10PI checksum. + * + * \param[in] m seq_file handle + * \param[in] data unused for single entry + * + * \retval 0 on success + * \retval negative value on error + */ +static ssize_t checksum_t10pi_enforce_show(struct kobject *kobj, + struct attribute *attr, + char *buf) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct lu_target *lut = obd->u.obt.obt_lut; + + return scnprintf(buf, PAGE_SIZE, "%u\n", lut->lut_cksum_t10pi_enforce); +} + +/** + * Force specific T10PI checksum modes to be enabled + * + * If T10PI *is* supported in hardware, allow only the supported T10PI type + * to be used. If T10PI is *not* supported by the OSD, setting the enforce + * parameter forces all T10PI types to be enabled (even if slower) for + * testing. + * + * The final determination of which algorithm to be used depends whether + * the client supports T10PI or not, and is handled at client connect time. + * + * \param[in] file proc file + * \param[in] buffer string which represents mode + * 1: set T10PI checksums enforced + * 0: unset T10PI checksums enforced + * \param[in] count \a buffer length + * \param[in] off unused for single entry + * + * \retval \a count on success + * \retval negative number on error + */ +static ssize_t checksum_t10pi_enforce_store(struct kobject *kobj, + struct attribute *attr, + const char *buffer, size_t count) +{ + struct obd_device *obd = container_of(kobj, struct obd_device, + obd_kset.kobj); + struct lu_target *lut = obd->u.obt.obt_lut; + bool enforce; + int rc; + + rc = kstrtobool(buffer, &enforce); + if (rc) + return rc; + + spin_lock(&lut->lut_flags_lock); + lut->lut_cksum_t10pi_enforce = enforce; + spin_unlock(&lut->lut_flags_lock); + return count; +} +LUSTRE_RW_ATTR(checksum_t10pi_enforce); + +/* + * mdt_checksum_type(server) proc handling + */ +DECLARE_CKSUM_NAME; + +static int mdt_checksum_type_seq_show(struct seq_file *m, void *data) +{ + struct obd_device *obd = m->private; + struct lu_target *lut; + enum cksum_types pref; + int i; + + if (!obd) + return 0; + + lut = obd->u.obt.obt_lut; + + /* select fastest checksum type on the server */ + pref = obd_cksum_type_select(obd->obd_name, + lut->lut_cksum_types_supported, 0); + + for (i = 0; i < ARRAY_SIZE(cksum_name); i++) { + if ((BIT(i) & lut->lut_cksum_types_supported) == 0) + continue; + + if (pref == BIT(i)) + seq_printf(m, "[%s] ", cksum_name[i]); + else + seq_printf(m, "%s ", cksum_name[i]); + } + seq_puts(m, "\n"); + + return 0; +} + +LPROC_SEQ_FOPS_RO(mdt_checksum_type); + +LPROC_SEQ_FOPS_RO_TYPE(mdt, hash); +LPROC_SEQ_FOPS_WR_ONLY(mdt, mds_evict_client); +LPROC_SEQ_FOPS_RW_TYPE(mdt, checksum_dump); +LUSTRE_RW_ATTR(job_cleanup_interval); +LPROC_SEQ_FOPS_RW_TYPE(mdt, nid_stats_clear); +LUSTRE_RW_ATTR(hsm_control); + +LPROC_SEQ_FOPS_RO_TYPE(mdt, recovery_status); +LUSTRE_RW_ATTR(recovery_time_hard); +LUSTRE_RW_ATTR(recovery_time_soft); +LUSTRE_RW_ATTR(ir_factor); + +LUSTRE_RO_ATTR(tot_dirty); +LUSTRE_RO_ATTR(tot_granted); +LUSTRE_RO_ATTR(tot_pending); +LUSTRE_RW_ATTR(grant_compat_disable); +LUSTRE_RO_ATTR(instance); + +LUSTRE_RO_ATTR(num_exports); + +static struct attribute *mdt_attrs[] = { + &lustre_attr_tot_dirty.attr, + &lustre_attr_tot_granted.attr, + &lustre_attr_tot_pending.attr, + &lustre_attr_grant_compat_disable.attr, + &lustre_attr_instance.attr, + &lustre_attr_recovery_time_hard.attr, + &lustre_attr_recovery_time_soft.attr, + &lustre_attr_ir_factor.attr, + &lustre_attr_num_exports.attr, + &lustre_attr_identity_expire.attr, + &lustre_attr_identity_acquire_expire.attr, + &lustre_attr_identity_upcall.attr, + &lustre_attr_identity_flush.attr, + &lustre_attr_evict_tgt_nids.attr, + &lustre_attr_enable_remote_dir.attr, + &lustre_attr_enable_remote_dir_gid.attr, + &lustre_attr_enable_chprojid_gid.attr, + &lustre_attr_enable_striped_dir.attr, + &lustre_attr_enable_dir_migration.attr, + &lustre_attr_enable_dir_restripe.attr, + &lustre_attr_enable_dir_auto_split.attr, + &lustre_attr_enable_remote_rename.attr, + &lustre_attr_commit_on_sharing.attr, + &lustre_attr_local_recovery.attr, + &lustre_attr_async_commit_count.attr, + &lustre_attr_sync_count.attr, + &lustre_attr_dom_lock.attr, + &lustre_attr_dom_read_open.attr, + &lustre_attr_migrate_hsm_allowed.attr, + &lustre_attr_hsm_control.attr, + &lustre_attr_job_cleanup_interval.attr, + &lustre_attr_readonly.attr, + &lustre_attr_dir_split_count.attr, + &lustre_attr_dir_split_delta.attr, + &lustre_attr_dir_restripe_nsonly.attr, + &lustre_attr_checksum_t10pi_enforce.attr, + &lustre_attr_enable_remote_subdir_mount.attr, + NULL, }; -static struct lprocfs_vars lprocfs_mdt_module_vars[] = { - { "num_refs", lprocfs_rd_numrefs, 0, 0 }, - { 0 } +static struct lprocfs_vars lprocfs_mdt_obd_vars[] = { + { .name = "recovery_status", + .fops = &mdt_recovery_status_fops }, + { .name = "identity_info", + .fops = &mdt_identity_info_fops }, + { .name = "site_stats", + .fops = &mdt_site_stats_fops }, + { .name = "evict_client", + .fops = &mdt_mds_evict_client_fops }, + { .name = "checksum_dump", + .fops = &mdt_checksum_dump_fops }, + { .name = "hash_stats", + .fops = &mdt_hash_fops }, + { .name = "root_squash", + .fops = &mdt_root_squash_fops }, + { .name = "nosquash_nids", + .fops = &mdt_nosquash_nids_fops }, + { .name = "checksum_type", + .fops = &mdt_checksum_type_fops }, + { NULL } }; -void lprocfs_mdt_init_vars(struct lprocfs_static_vars *lvars) +static int +lprocfs_mdt_print_open_files(struct obd_export *exp, void *v) +{ + struct seq_file *seq = v; + + if (exp->exp_lock_hash != NULL) { + struct mdt_export_data *med = &exp->exp_mdt_data; + struct mdt_file_data *mfd; + + spin_lock(&med->med_open_lock); + list_for_each_entry(mfd, &med->med_open_head, mfd_list) { + seq_printf(seq, DFID"\n", + PFID(mdt_object_fid(mfd->mfd_object))); + } + spin_unlock(&med->med_open_lock); + } + + return 0; +} + +static int lprocfs_mdt_open_files_seq_show(struct seq_file *seq, void *v) +{ + struct nid_stat *stats = seq->private; + + return obd_nid_export_for_each(stats->nid_obd, stats->nid, + lprocfs_mdt_print_open_files, seq); +} + +int lprocfs_mdt_open_files_seq_open(struct inode *inode, struct file *file) +{ + struct seq_file *seq; + int rc; + + rc = single_open(file, &lprocfs_mdt_open_files_seq_show, NULL); + if (rc != 0) + return rc; + + seq = file->private_data; + seq->private = PDE_DATA(inode); + + return 0; +} + +void mdt_counter_incr(struct ptlrpc_request *req, int opcode, long amount) { - lvars->module_vars = lprocfs_mdt_module_vars; - lvars->obd_vars = lprocfs_mdt_obd_vars; + struct obd_export *exp = req->rq_export; + + if (exp->exp_obd && exp->exp_obd->obd_md_stats) + lprocfs_counter_add(exp->exp_obd->obd_md_stats, + opcode + LPROC_MD_LAST_OPC, amount); + if (exp->exp_nid_stats && exp->exp_nid_stats->nid_stats != NULL) + lprocfs_counter_add(exp->exp_nid_stats->nid_stats, opcode, + amount); + if (exp->exp_obd && exp->exp_obd->u.obt.obt_jobstats.ojs_hash && + (exp_connect_flags(exp) & OBD_CONNECT_JOBSTATS)) + lprocfs_job_stats_log(exp->exp_obd, + lustre_msg_get_jobid(req->rq_reqmsg), + opcode, amount); } -void mdt_counter_incr(struct obd_export *exp, int opcode) +static const char * const mdt_stats[] = { + [LPROC_MDT_OPEN] = "open", + [LPROC_MDT_CLOSE] = "close", + [LPROC_MDT_MKNOD] = "mknod", + [LPROC_MDT_LINK] = "link", + [LPROC_MDT_UNLINK] = "unlink", + [LPROC_MDT_MKDIR] = "mkdir", + [LPROC_MDT_RMDIR] = "rmdir", + [LPROC_MDT_RENAME] = "rename", + [LPROC_MDT_GETATTR] = "getattr", + [LPROC_MDT_SETATTR] = "setattr", + [LPROC_MDT_GETXATTR] = "getxattr", + [LPROC_MDT_SETXATTR] = "setxattr", + [LPROC_MDT_STATFS] = "statfs", + [LPROC_MDT_SYNC] = "sync", + [LPROC_MDT_SAMEDIR_RENAME] = "samedir_rename", + [LPROC_MDT_CROSSDIR_RENAME] = "crossdir_rename", + [LPROC_MDT_IO_READ] = "read_bytes", + [LPROC_MDT_IO_WRITE] = "write_bytes", + [LPROC_MDT_IO_PUNCH] = "punch", + [LPROC_MDT_MIGRATE] = "migrate", + [LPROC_MDT_FALLOCATE] = "fallocate", +}; + +void mdt_stats_counter_init(struct lprocfs_stats *stats, unsigned int offset) { - if (exp->exp_obd && exp->exp_obd->md_stats) - lprocfs_counter_incr(exp->exp_obd->md_stats, opcode); - if (exp->exp_nid_stats && exp->exp_nid_stats->nid_stats != NULL) - lprocfs_counter_incr(exp->exp_nid_stats->nid_stats, opcode); + int array_size = ARRAY_SIZE(mdt_stats); + int oidx; /* obd_md_stats index */ + int midx; /* mdt_stats index */ + + LASSERT(stats && stats->ls_num >= offset + array_size); + + for (midx = 0; midx < array_size; midx++) { + oidx = midx + offset; + if (midx == LPROC_MDT_IO_READ || midx == LPROC_MDT_IO_WRITE) + lprocfs_counter_init(stats, oidx, + LPROCFS_TYPE_BYTES_FULL, + mdt_stats[midx], "bytes"); + else + lprocfs_counter_init(stats, oidx, + LPROCFS_TYPE_LATENCY, + mdt_stats[midx], "usecs"); + } +} +int mdt_tunables_init(struct mdt_device *mdt, const char *name) +{ + struct obd_device *obd = mdt2obd_dev(mdt); + int rc; + + ENTRY; + LASSERT(name != NULL); + + obd->obd_ktype.default_attrs = mdt_attrs; + obd->obd_vars = lprocfs_mdt_obd_vars; + rc = lprocfs_obd_setup(obd, true); + if (rc) { + CERROR("%s: cannot create proc entries: rc = %d\n", + mdt_obd_name(mdt), rc); + return rc; + } + + rc = tgt_tunables_init(&mdt->mdt_lut); + if (rc) { + CERROR("%s: failed to init target tunables: rc = %d\n", + mdt_obd_name(mdt), rc); + return rc; + } + + rc = hsm_cdt_tunables_init(mdt); + if (rc) { + CERROR("%s: cannot create hsm proc entries: rc = %d\n", + mdt_obd_name(mdt), rc); + return rc; + } + + obd->obd_proc_exports_entry = proc_mkdir("exports", + obd->obd_proc_entry); + if (obd->obd_proc_exports_entry) + lprocfs_add_simple(obd->obd_proc_exports_entry, "clear", + obd, &mdt_nid_stats_clear_fops); + + rc = lprocfs_alloc_md_stats(obd, ARRAY_SIZE(mdt_stats)); + if (rc) + return rc; + + /* add additional MDT md_stats after the default ones */ + mdt_stats_counter_init(obd->obd_md_stats, LPROC_MD_LAST_OPC); + rc = lprocfs_job_stats_init(obd, ARRAY_SIZE(mdt_stats), + mdt_stats_counter_init); + + rc = lproc_mdt_attach_rename_seqstat(mdt); + if (rc) + CERROR("%s: MDT can not create rename stats rc = %d\n", + mdt_obd_name(mdt), rc); + + RETURN(rc); } -void mdt_stats_counter_init(struct lprocfs_stats *stats) +void mdt_tunables_fini(struct mdt_device *mdt) { - lprocfs_counter_init(stats, LPROC_MDT_OPEN, 0, "open", "reqs"); - lprocfs_counter_init(stats, LPROC_MDT_CLOSE, 0, "close", "reqs"); - lprocfs_counter_init(stats, LPROC_MDT_MKNOD, 0, "mknod", "reqs"); - lprocfs_counter_init(stats, LPROC_MDT_LINK, 0, "link", "reqs"); - lprocfs_counter_init(stats, LPROC_MDT_UNLINK, 0, "unlink", "reqs"); - lprocfs_counter_init(stats, LPROC_MDT_MKDIR, 0, "mkdir", "reqs"); - lprocfs_counter_init(stats, LPROC_MDT_RMDIR, 0, "rmdir", "reqs"); - lprocfs_counter_init(stats, LPROC_MDT_RENAME, 0, "rename", "reqs"); - lprocfs_counter_init(stats, LPROC_MDT_GETATTR, 0, "getattr", "reqs"); - lprocfs_counter_init(stats, LPROC_MDT_SETATTR, 0, "setattr", "reqs"); - lprocfs_counter_init(stats, LPROC_MDT_GETXATTR, 0, "getxattr", "reqs"); - lprocfs_counter_init(stats, LPROC_MDT_SETXATTR, 0, "setxattr", "reqs"); - lprocfs_counter_init(stats, LPROC_MDT_STATFS, 0, "statfs", "reqs"); - lprocfs_counter_init(stats, LPROC_MDT_SYNC, 0, "sync", "reqs"); - lprocfs_counter_init(stats, LPROC_MDT_SAMEDIR_RENAME, 0, - "samedir_rename", "reqs"); - lprocfs_counter_init(stats, LPROC_MDT_CROSSDIR_RENAME, 0, - "crossdir_rename", "reqs"); + struct obd_device *obd = mdt2obd_dev(mdt); + + if (obd->obd_proc_exports_entry != NULL) { + lprocfs_remove_proc_entry("clear", obd->obd_proc_exports_entry); + obd->obd_proc_exports_entry = NULL; + } + + lprocfs_free_per_client_stats(obd); + /* hsm_cdt_tunables is disabled earlier than this to avoid + * coordinator restart. + */ + hsm_cdt_tunables_fini(mdt); + tgt_tunables_fini(&mdt->mdt_lut); + lprocfs_obd_cleanup(obd); + lprocfs_free_md_stats(obd); + lprocfs_free_obd_stats(obd); + lprocfs_job_stats_fini(obd); }