-/* -*- 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.
*
* 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
*/
/*
- * Copyright 2008 Sun Microsystems, Inc. All rights reserved
+ * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
+ *
+ * Copyright (c) 2011, 2017, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
* Author: Fan Yong <fanyong@clusterfs.com>
*/
-#ifndef EXPORT_SYMTAB
-# define EXPORT_SYMTAB
-#endif
#define DEBUG_SUBSYSTEM S_MDS
#include <linux/version.h>
#include <asm/statfs.h>
#include <linux/module.h>
-
+#include <uapi/linux/lnet/nidstr.h>
/* LUSTRE_VERSION_CODE */
-#include <lustre_ver.h>
+#include <uapi/linux/lustre/lustre_ver.h>
/*
* struct OBD_{ALLOC,FREE}*()
* MDT_FAIL_CHECK
#include <obd.h>
#include <obd_class.h>
#include <lustre_mds.h>
-#include <lustre_mdt.h>
#include <lprocfs_status.h>
-#include <lu_time.h>
#include "mdt_internal.h"
-#include <lnet/lib-lnet.h>
-static const char *mdt_proc_names[LPROC_MDT_NR] = {
-};
-
-int mdt_procfs_init(struct mdt_device *mdt, const char *name)
+/**
+ * The rename stats output would be YAML formats, like
+ * rename_stats:
+ * - snapshot_time: 1234567890.123456
+ * - same_dir:
+ * 4kB: { samples: 1230, pct: 33, cum_pct: 45 }
+ * 8kB: { samples: 1242, pct: 33, cum_pct: 78 }
+ * 16kB: { samples: 132, pct: 3, cum_pct: 81 }
+ * - crossdir_src:
+ * 4kB: { samples: 123, pct: 33, cum_pct: 45 }
+ * 8kB: { samples: 124, pct: 33, cum_pct: 78 }
+ * 16kB: { samples: 12, pct: 3, cum_pct: 81 }
+ * - crossdir_tgt:
+ * 4kB: { samples: 123, pct: 33, cum_pct: 45 }
+ * 8kB: { samples: 124, pct: 33, cum_pct: 78 }
+ * 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 lu_device *ld = &mdt->mdt_md_dev.md_lu_dev;
- int result;
- ENTRY;
+ unsigned long tot, t, cum = 0;
+ int i;
+
+ 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;
- LASSERT(name != NULL);
- mdt->mdt_proc_entry = ld->ld_obd->obd_proc_entry;
- LASSERT(mdt->mdt_proc_entry != NULL);
+ 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:");
- result = lu_time_init(&mdt->mdt_stats, mdt->mdt_proc_entry,
- mdt_proc_names, ARRAY_SIZE(mdt_proc_names));
- if (result == 0)
- result = lu_time_named_init(&ld->ld_site->ls_time_stats,
- "site_time", mdt->mdt_proc_entry,
- lu_time_names,
- ARRAY_SIZE(lu_time_names));
- RETURN(result);
-}
+ seq_printf(seq, " { sample: %3lu, pct: %3lu, cum_pct: %3lu }\n",
+ t, pct(t, tot), pct(cum, tot));
-int mdt_procfs_fini(struct mdt_device *mdt)
-{
- struct lu_device *ld = &mdt->mdt_md_dev.md_lu_dev;
- 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;
+ if (cum == tot)
+ break;
}
- RETURN(0);
}
-void mdt_time_start(const struct mdt_thread_info *info)
+static void rename_stats_show(struct seq_file *seq,
+ struct rename_stats *rename_stats)
{
- lu_lprocfs_time_start(info->mti_env);
+ struct timespec64 now;
+
+ /* this sampling races with updates */
+ ktime_get_real_ts64(&now);
+ seq_printf(seq, "rename_stats:\n");
+ seq_printf(seq, "- %-15s %llu.%9lu\n", "snapshot_time:",
+ (s64)now.tv_sec, now.tv_nsec);
+
+ 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]);
}
-void mdt_time_end(const struct mdt_thread_info *info, int idx)
-{
- lu_lprocfs_time_end(info->mti_env, info->mti_mdt->mdt_stats, idx);
-}
+#undef pct
-static int lprocfs_rd_identity_expire(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+static int mdt_rename_stats_seq_show(struct seq_file *seq, void *v)
{
- struct obd_device *obd = data;
- struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev);
+ struct mdt_device *mdt = seq->private;
- *eof = 1;
- return snprintf(page, count, "%lu\n",
- mdt->mdt_identity_cache->uc_entry_expire / HZ);
+ rename_stats_show(seq, &mdt->mdt_rename_stats);
+
+ return 0;
}
-static int lprocfs_wr_identity_expire(struct file *file, const char *buffer,
- unsigned long count, void *data)
+static ssize_t
+mdt_rename_stats_seq_write(struct file *file, const char __user *buf,
+ size_t len, loff_t *off)
{
- struct obd_device *obd = data;
- struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev);
- int rc, val;
+ struct seq_file *seq = file->private_data;
+ struct mdt_device *mdt = seq->private;
+ int i;
- rc = lprocfs_write_helper(buffer, count, &val);
- if (rc)
- return rc;
+ for (i = 0; i < RENAME_LAST; i++)
+ lprocfs_oh_clear(&mdt->mdt_rename_stats.hist[i]);
- mdt->mdt_identity_cache->uc_entry_expire = val * HZ;
- return count;
+ return len;
}
+LPROC_SEQ_FOPS(mdt_rename_stats);
-static int lprocfs_rd_identity_acquire_expire(char *page, char **start,
- off_t off, int count, int *eof,
- void *data)
+static int lproc_mdt_attach_rename_seqstat(struct mdt_device *mdt)
{
- struct obd_device *obd = data;
- struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev);
+ int i;
+
+ for (i = 0; i < RENAME_LAST; i++)
+ spin_lock_init(&mdt->mdt_rename_stats.hist[i].oh_lock);
- *eof = 1;
- return snprintf(page, count, "%lu\n",
- mdt->mdt_identity_cache->uc_acquire_expire / HZ);
+ return lprocfs_obd_seq_create(mdt2obd_dev(mdt), "rename_stats", 0644,
+ &mdt_rename_stats_fops, mdt);
}
-static int lprocfs_wr_identity_acquire_expire(struct file *file,
- const char *buffer,
- unsigned long count,
- void *data)
+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)
{
- struct obd_device *obd = data;
- struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev);
- int rc, val;
+ struct md_attr *ma = &info->mti_attr;
+ struct rename_stats *rstats = &mdt->mdt_rename_stats;
+ int rc;
- rc = lprocfs_write_helper(buffer, count, &val);
- if (rc)
- return 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;
+ }
- mdt->mdt_identity_cache->uc_acquire_expire = val * HZ;
- return count;
-}
+ if (src == tgt) {
+ mdt_counter_incr(req, LPROC_MDT_SAMEDIR_RENAME);
+ lprocfs_oh_tally_log2(&rstats->hist[RENAME_SAMEDIR_SIZE],
+ (unsigned int)ma->ma_attr.la_size);
+ return;
+ }
-static int lprocfs_rd_identity_upcall(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);
- struct upcall_cache *hash = mdt->mdt_identity_cache;
- int len;
+ mdt_counter_incr(req, 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",
+ mdt_obd_name(mdt), PFID(mdt_object_fid(tgt)), rc);
+ return;
+ }
- *eof = 1;
- read_lock(&hash->uc_upcall_rwlock);
- len = snprintf(page, count, "%s\n", hash->uc_upcall);
- read_unlock(&hash->uc_upcall_rwlock);
- return len;
+ lprocfs_oh_tally_log2(&rstats->hist[RENAME_CROSSDIR_TGT_SIZE],
+ (unsigned int)ma->ma_attr.la_size);
}
-static int lprocfs_wr_identity_upcall(struct file *file, const char *buffer,
- unsigned long count, void *data)
+static int mdt_identity_expire_seq_show(struct seq_file *m, void *data)
{
- struct obd_device *obd = data;
- struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev);
- struct upcall_cache *hash = mdt->mdt_identity_cache;
- char kernbuf[UC_CACHE_UPCALL_MAXPATH] = { '\0' };
+ struct obd_device *obd = m->private;
+ struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev);
- if (count >= UC_CACHE_UPCALL_MAXPATH) {
- CERROR("%s: identity upcall too long\n", obd->obd_name);
- return -EINVAL;
- }
+ seq_printf(m, "%lld\n", mdt->mdt_identity_cache->uc_entry_expire);
+ return 0;
+}
- if (copy_from_user(kernbuf, buffer, min_t(unsigned long, count,
- UC_CACHE_UPCALL_MAXPATH - 1)))
- return -EFAULT;
+static ssize_t
+mdt_identity_expire_seq_write(struct file *file, const char __user *buffer,
+ 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);
+ time64_t val;
+ int rc;
- /* Remove any extraneous bits from the upcall (e.g. linefeeds) */
- write_lock(&hash->uc_upcall_rwlock);
- sscanf(kernbuf, "%s", hash->uc_upcall);
- write_unlock(&hash->uc_upcall_rwlock);
+ rc = kstrtoll_from_user(buffer, count, 0, &val);
+ if (rc)
+ return rc;
- if (strcmp(hash->uc_name, obd->obd_name) != 0)
- CWARN("%s: write to upcall name %s\n",
- obd->obd_name, hash->uc_upcall);
+ if (val < 0)
+ return -ERANGE;
- 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);
+ mdt->mdt_identity_cache->uc_entry_expire = val;
- CWARN("%s: identity upcall set to %s\n", obd->obd_name, hash->uc_upcall);
- return count;
+ return count;
}
+LPROC_SEQ_FOPS(mdt_identity_expire);
-static int lprocfs_wr_identity_flush(struct file *file, const char *buffer,
- unsigned long count, void *data)
+static int mdt_identity_acquire_expire_seq_show(struct seq_file *m, void *data)
{
- struct obd_device *obd = data;
- struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev);
- int rc, uid;
-
- rc = lprocfs_write_helper(buffer, count, &uid);
- if (rc)
- return rc;
+ struct obd_device *obd = m->private;
+ struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev);
- mdt_flush_identity(mdt->mdt_identity_cache, uid);
- return count;
+ seq_printf(m, "%lld\n", mdt->mdt_identity_cache->uc_acquire_expire);
+ return 0;
}
-static int lprocfs_wr_identity_info(struct file *file, const char *buffer,
- unsigned long count, void *data)
+static ssize_t
+mdt_identity_acquire_expire_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 identity_downcall_data sparam, *param = &sparam;
- int size = 0, rc = 0;
+ struct seq_file *m = file->private_data;
+ struct obd_device *obd = m->private;
+ struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev);
+ time64_t val;
+ int rc;
- if (count < sizeof(*param)) {
- CERROR("%s: invalid data size %lu\n", obd->obd_name, count);
- return count;
- }
+ rc = kstrtoll_from_user(buffer, count, 0, &val);
+ if (rc)
+ return rc;
- if (copy_from_user(&sparam, buffer, sizeof(sparam))) {
- CERROR("%s: bad identity data\n", obd->obd_name);
- GOTO(out, rc = -EFAULT);
- }
+ if (val < 0 || val > INT_MAX)
+ return -ERANGE;
- if (sparam.idd_magic != IDENTITY_DOWNCALL_MAGIC) {
- CERROR("%s: MDS identity downcall bad params\n", obd->obd_name);
- GOTO(out, rc = -EINVAL);
- }
+ mdt->mdt_identity_cache->uc_acquire_expire = val;
- if (sparam.idd_nperms > N_PERMS_MAX) {
- CERROR("%s: perm count %d more than maximum %d\n",
- obd->obd_name, sparam.idd_nperms, N_PERMS_MAX);
- GOTO(out, rc = -EINVAL);
- }
+ return count;
+}
+LPROC_SEQ_FOPS(mdt_identity_acquire_expire);
- if (sparam.idd_ngroups > NGROUPS_MAX) {
- CERROR("%s: group count %d more than maximum %d\n",
- obd->obd_name, sparam.idd_ngroups, NGROUPS_MAX);
- GOTO(out, rc = -EINVAL);
- }
+static int mdt_identity_upcall_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 upcall_cache *hash = mdt->mdt_identity_cache;
+
+ down_read(&hash->uc_upcall_rwsem);
+ seq_printf(m, "%s\n", hash->uc_upcall);
+ up_read(&hash->uc_upcall_rwsem);
+ return 0;
+}
- if (sparam.idd_ngroups) {
- size = offsetof(struct identity_downcall_data,
- idd_groups[sparam.idd_ngroups]);
- OBD_ALLOC(param, size);
- if (!param) {
- CERROR("%s: fail to alloc %d bytes for uid %u"
- " with %d groups\n", obd->obd_name, size,
- sparam.idd_uid, sparam.idd_ngroups);
- param = &sparam;
- param->idd_ngroups = 0;
- } else if (copy_from_user(param, buffer, size)) {
- CERROR("%s: uid %u bad supplementary group data\n",
- obd->obd_name, sparam.idd_uid);
- OBD_FREE(param, size);
- param = &sparam;
- param->idd_ngroups = 0;
- }
- }
+static ssize_t
+mdt_identity_upcall_seq_write(struct file *file, const char __user *buffer,
+ 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);
+ struct upcall_cache *hash = mdt->mdt_identity_cache;
+ int rc;
+ char *kernbuf;
+
+ if (count >= UC_CACHE_UPCALL_MAXPATH) {
+ CERROR("%s: identity upcall too long\n", mdt_obd_name(mdt));
+ return -EINVAL;
+ }
+ OBD_ALLOC(kernbuf, count + 1);
+ if (kernbuf == NULL)
+ GOTO(failed, rc = -ENOMEM);
+ if (copy_from_user(kernbuf, buffer, count))
+ GOTO(failed, rc = -EFAULT);
+
+ /* Remove any extraneous bits from the upcall (e.g. linefeeds) */
+ down_write(&hash->uc_upcall_rwsem);
+ sscanf(kernbuf, "%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);
+ OBD_FREE(kernbuf, count + 1);
+ RETURN(count);
+
+ failed:
+ if (kernbuf)
+ OBD_FREE(kernbuf, count + 1);
+ RETURN(rc);
+}
+LPROC_SEQ_FOPS(mdt_identity_upcall);
+
+static ssize_t
+lprocfs_identity_flush_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);
+ int uid;
+ int rc;
+
+ rc = kstrtoint_from_user(buffer, count, 0, &uid);
+ if (rc)
+ return rc;
+
+ mdt_flush_identity(mdt->mdt_identity_cache, uid);
+ return count;
+}
+LPROC_SEQ_FOPS_WR_ONLY(mdt, identity_flush);
- rc = upcall_cache_downcall(mdt->mdt_identity_cache, param->idd_err,
- param->idd_uid, param);
+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;
+
+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);
out:
- if (param && (param != &sparam))
- OBD_FREE(param, size);
+ if (param != NULL)
+ OBD_FREE(param, size);
- return rc ?: count;
+ return rc ? rc : count;
}
+LPROC_SEQ_FOPS_WR_ONLY(mdt, identity_info);
-/* for debug only */
-static int lprocfs_rd_capa(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+static int mdt_site_stats_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);
- return snprintf(page, count, "capability on: %s %s\n",
- mdt->mdt_opts.mo_oss_capa ? "oss" : "",
- mdt->mdt_opts.mo_mds_capa ? "mds" : "");
+ return lu_site_stats_seq_print(mdt_lu_site(mdt), m);
}
+LPROC_SEQ_FOPS_RO(mdt_site_stats);
+
+#define BUFLEN (UUID_MAX + 4)
-static int lprocfs_wr_capa(struct file *file, const char *buffer,
- unsigned long count, void *data)
+static ssize_t
+lprocfs_mds_evict_client_seq_write(struct file *file, const char __user *buf,
+ size_t count, loff_t *off)
{
- struct obd_device *obd = data;
- struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev);
- int val, rc;
+ 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 = cfs_firststr(kbuf, min_t(unsigned long, BUFLEN - 1, count));
+
+ 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 = lprocfs_write_helper(buffer, count, &val);
- if (rc)
- return rc;
- 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;
- }
+out:
+ OBD_FREE(kbuf, BUFLEN);
+ return rc < 0 ? rc : count;
+}
- /* 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;
- }
+#undef BUFLEN
+
+static int mdt_evict_tgt_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);
- 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;
+ seq_printf(m, "%u\n", mdt->mdt_opts.mo_evict_tgt_nids);
+ return 0;
}
-static int lprocfs_rd_capa_count(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+static ssize_t
+mdt_evict_tgt_nids_seq_write(struct file *file, const char __user *buffer,
+ size_t count, loff_t *off)
{
- return snprintf(page, count, "%d %d\n",
- capa_count[CAPA_SITE_CLIENT],
- capa_count[CAPA_SITE_SERVER]);
+ struct seq_file *m = file->private_data;
+ struct obd_device *obd = m->private;
+ struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev);
+ bool val;
+ int rc;
+
+ rc = kstrtobool_from_user(buffer, count, &val);
+ if (rc)
+ return rc;
+
+ mdt->mdt_opts.mo_evict_tgt_nids = val;
+ return count;
}
+LPROC_SEQ_FOPS(mdt_evict_tgt_nids);
-static int lprocfs_rd_site_stats(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+static int mdt_cos_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);
- return lu_site_stats_print(mdt_lu_site(mdt), page, count);
+ seq_printf(m, "%u\n", mdt_cos_is_enabled(mdt));
+ return 0;
}
-static int lprocfs_rd_capa_timeout(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+static ssize_t
+mdt_cos_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);
-
- return snprintf(page, count, "%lu\n", mdt->mdt_capa_timeout);
+ struct seq_file *m = file->private_data;
+ struct obd_device *obd = m->private;
+ struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev);
+ bool val;
+ int rc;
+
+ rc = kstrtobool_from_user(buffer, count, &val);
+ if (rc)
+ return rc;
+
+ mdt_enable_cos(mdt, val);
+ return count;
}
+LPROC_SEQ_FOPS(mdt_cos);
-static int lprocfs_wr_capa_timeout(struct file *file, const char *buffer,
- unsigned long count, 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);
- int val, rc;
+ 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;
-
- mdt->mdt_capa_timeout = (unsigned long)val;
- mdt->mdt_capa_conf = 1;
- return count;
+ seq_printf(m, "%u:%u\n", squash->rsi_uid,
+ squash->rsi_gid);
+ return 0;
}
-static int lprocfs_rd_ck_timeout(char *page, char **start, off_t off, int count,
- int *eof, 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);
+ 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;
- return snprintf(page, count, "%lu\n", mdt->mdt_ck_timeout);
+ return lprocfs_wr_root_squash(buffer, count, squash,
+ mdt_obd_name(mdt));
}
+LPROC_SEQ_FOPS(mdt_root_squash);
-static int lprocfs_wr_ck_timeout(struct file *file, const char *buffer,
- unsigned long count, void *data)
+static int mdt_nosquash_nids_seq_show(struct seq_file *m, void *data)
{
- 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;
-
- mdt->mdt_ck_timeout = (unsigned long)val;
- mdt->mdt_capa_conf = 1;
- return count;
+ 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;
+
+ down_read(&squash->rsi_sem);
+ 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");
+ up_read(&squash->rsi_sem);
+
+ return 0;
}
-static int lprocfs_mdt_wr_evict_client(struct file *file, const char *buffer,
- unsigned long count, void *data)
+static ssize_t
+mdt_nosquash_nids_seq_write(struct file *file, const char __user *buffer,
+ size_t count, loff_t *off)
{
- char tmpbuf[sizeof(struct obd_uuid)];
-
- sscanf(buffer, "%40s", tmpbuf);
-
- if (strncmp(tmpbuf, "nid:", 4) != 0)
- return lprocfs_wr_evict_client(file, buffer, count, 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 root_squash_info *squash = &mdt->mdt_squash;
- CERROR("NOT implement evict client by nid %s\n", tmpbuf);
-
- return count;
+ return lprocfs_wr_nosquash_nids(buffer, count, squash,
+ mdt_obd_name(mdt));
}
+LPROC_SEQ_FOPS(mdt_nosquash_nids);
-static int lprocfs_rd_sec_level(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+static int mdt_enable_remote_dir_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);
- return snprintf(page, count, "%d\n", mdt->mdt_sec_level);
+ seq_printf(m, "%u\n", mdt->mdt_enable_remote_dir);
+ return 0;
}
-static int lprocfs_wr_sec_level(struct file *file, const char *buffer,
- unsigned long count, void *data)
+static ssize_t
+mdt_enable_remote_dir_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 val, rc;
-
- rc = lprocfs_write_helper(buffer, count, &val);
- if (rc)
- return rc;
+ struct seq_file *m = file->private_data;
+ struct obd_device *obd = m->private;
+ struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev);
+ bool val;
+ int rc;
+
+ rc = kstrtobool_from_user(buffer, count, &val);
+ if (rc)
+ return rc;
+
+ mdt->mdt_enable_remote_dir = val;
+ return count;
+}
+LPROC_SEQ_FOPS(mdt_enable_remote_dir);
- if (val > LUSTRE_SEC_ALL || val < LUSTRE_SEC_NONE)
- return -EINVAL;
+static int mdt_enable_remote_dir_gid_seq_show(struct seq_file *m, void *data)
+{
+ struct obd_device *obd = m->private;
+ struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev);
- if (val == LUSTRE_SEC_SPECIFY) {
- CWARN("security level %d will be supported in future.\n",
- LUSTRE_SEC_SPECIFY);
- return -EINVAL;
- }
+ seq_printf(m, "%d\n",
+ (int)mdt->mdt_enable_remote_dir_gid);
+ return 0;
+}
- mdt->mdt_sec_level = val;
- return count;
+static ssize_t
+mdt_enable_remote_dir_gid_seq_write(struct file *file,
+ const char __user *buffer,
+ 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);
+ int val;
+ int rc;
+
+ rc = kstrtoint_from_user(buffer, count, 0, &val);
+ if (rc)
+ return rc;
+
+ mdt->mdt_enable_remote_dir_gid = val;
+ return count;
}
+LPROC_SEQ_FOPS(mdt_enable_remote_dir_gid);
-static int lprocfs_rd_cos(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+/**
+ * Show MDT policy for handling dirty metadata under a lock being cancelled.
+ *
+ * \param[in] m seq_file handle
+ * \param[in] data unused for single entry
+ *
+ * \retval 0 on success
+ * \retval negative value on error
+ */
+static int mdt_slc_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 lu_target *tgt = obd->u.obt.obt_lut;
+ char *slc_states[] = {"never", "blocking", "always" };
- return snprintf(page, count, "%u\n", mdt_cos_is_enabled(mdt));
+ seq_printf(m, "%s\n", slc_states[tgt->lut_sync_lock_cancel]);
+ return 0;
}
+LPROC_SEQ_FOPS_RO(mdt_slc);
-static int lprocfs_wr_cos(struct file *file, const char *buffer,
- unsigned long count, void *data)
+/**
+ * Show MDT async commit count.
+ *
+ * \param[in] m seq_file handle
+ * \param[in] data unused for single entry
+ *
+ * \retval 0 on success
+ * \retval negative value on error
+ */
+static int mdt_async_commit_count_seq_show(struct seq_file *m, void *data)
{
- struct obd_device *obd = data;
- struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev);
- int val, rc;
+ struct obd_device *obd = m->private;
+ struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev);
- rc = lprocfs_write_helper(buffer, count, &val);
- if (rc)
- return rc;
- mdt_enable_cos(mdt, val);
- return count;
+ seq_printf(m, "%d\n", atomic_read(&mdt->mdt_async_commit_count));
+ return 0;
}
-static int lprocfs_rd_root_squash(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+static ssize_t
+mdt_async_commit_count_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);
- ENTRY;
+ struct seq_file *m = file->private_data;
+ struct obd_device *obd = m->private;
+ struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev);
+ int val;
+ int rc;
- return snprintf(page, count, "%u:%u\n", mdt->mdt_squash_uid,
- mdt->mdt_squash_gid);
+ rc = kstrtoint_from_user(buffer, count, 0, &val);
+ if (rc)
+ return rc;
+
+ atomic_set(&mdt->mdt_async_commit_count, val);
+
+ return count;
}
+LPROC_SEQ_FOPS(mdt_async_commit_count);
-static int safe_strtoul(const char *str, char **endp, unsigned long *res)
+/**
+ * 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 int mdt_sync_count_seq_show(struct seq_file *m, void *data)
{
- char n[24];
+ struct obd_device *obd = m->private;
+ struct lu_target *tgt = obd->u.obt.obt_lut;
- *res = simple_strtoul(str, endp, 0);
- if (str == *endp)
- return 1;
-
- sprintf(n, "%lu", *res);
- if (strncmp(n, str, *endp - str))
- /* overflow */
- return 1;
- return 0;
+ seq_printf(m, "%d\n", atomic_read(&tgt->lut_sync_count));
+ return 0;
}
-static int lprocfs_wr_root_squash(struct file *file, const char *buffer,
- unsigned long count, void *data)
+static ssize_t
+mdt_sync_count_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;
- 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 (copy_from_user(kernbuf, buffer, count)) {
- errmsg = "bad address";
- GOTO(failed, rc = -EFAULT);
- }
- kernbuf[count] = '\0';
+ struct seq_file *m = file->private_data;
+ struct obd_device *obd = m->private;
+ struct lu_target *tgt = obd->u.obt.obt_lut;
+ int val;
+ int rc;
- nouid = nogid = 0;
- if (safe_strtoul(buffer, &tmp, &uid)) {
- uid = mdt->mdt_squash_uid;
- nouid = 1;
- }
+ rc = kstrtoint_from_user(buffer, count, 0, &val);
+ if (rc)
+ return rc;
- /* 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;
- }
+ atomic_set(&tgt->lut_sync_count, val);
- mdt->mdt_squash_uid = uid;
- mdt->mdt_squash_gid = gid;
+ return count;
+}
+LPROC_SEQ_FOPS(mdt_sync_count);
- if (nouid && nogid) {
- errmsg = "needs uid:gid format";
- GOTO(failed, rc = -EINVAL);
- }
+static 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",
+};
- LCONSOLE_INFO("%s: root_squash is set to %u:%u\n",
- obd->obd_name,
- mdt->mdt_squash_uid, mdt->mdt_squash_gid);
- RETURN(count);
+/* This must be longer than the longest string above */
+#define DOM_LOCK_MODES_MAXLEN 16
- failed:
- CWARN("%s: failed to set root_squash to \"%s\", %s: rc %d\n",
- obd->obd_name, buffer, errmsg, 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 int mdt_dom_lock_seq_show(struct seq_file *m, void *data)
+{
+ struct obd_device *obd = m->private;
+ struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev);
+
+ seq_printf(m, "%s\n", dom_open_lock_modes[mdt->mdt_opts.mo_dom_lock]);
+ return 0;
}
-static int lprocfs_rd_nosquash_nids(char *page, char **start, off_t off,
- int count, int *eof, 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
+mdt_dom_lock_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);
-
- if (mdt->mdt_nosquash_str)
- return snprintf(page, count, "%s\n", mdt->mdt_nosquash_str);
- return snprintf(page, count, "NONE\n");
+ struct seq_file *m = file->private_data;
+ struct obd_device *obd = m->private;
+ struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev);
+ char kernbuf[DOM_LOCK_MODES_MAXLEN];
+ int val = -1;
+ int i, rc;
+
+ if (count == 0 || count >= sizeof(kernbuf))
+ return -EINVAL;
+
+ if (copy_from_user(kernbuf, buffer, count))
+ return -EFAULT;
+
+ kernbuf[count] = 0;
+ if (kernbuf[count - 1] == '\n')
+ kernbuf[count - 1] = 0;
+
+ for (i = 0 ; i < NUM_DOM_LOCK_ON_OPEN_MODES; i++) {
+ if (strcmp(kernbuf, dom_open_lock_modes[i]) == 0) {
+ val = i;
+ break;
+ }
+ }
+
+ /* Legacy numeric codes */
+ if (val == -1) {
+ rc = kstrtoint_from_user(buffer, count, 0, &val);
+ if (rc)
+ return rc;
+ }
+
+ if (val < 0 || val >= NUM_DOM_LOCK_ON_OPEN_MODES)
+ return -EINVAL;
+
+ mdt->mdt_opts.mo_dom_lock = val;
+ return count;
}
+LPROC_SEQ_FOPS(mdt_dom_lock);
-static int lprocfs_wr_nosquash_nids(struct file *file, const char *buffer,
- unsigned long count, 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 int mdt_dom_read_open_seq_show(struct seq_file *m, void *data)
{
- struct obd_device *obd = data;
- struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev);
- int rc;
- char *kernbuf, *errmsg;
- struct list_head tmp;
- ENTRY;
-
- OBD_ALLOC(kernbuf, count + 1);
- if (kernbuf == NULL) {
- errmsg = "no memory";
- GOTO(failed, rc = -ENOMEM);
- }
- if (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 */
- down_write(&mdt->mdt_squash_sem);
- if (!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;
- }
- up_write(&mdt->mdt_squash_sem);
- LCONSOLE_INFO("%s: nosquash_nids is cleared\n",
- obd->obd_name);
- OBD_FREE(kernbuf, count + 1);
- RETURN(count);
- }
+ struct obd_device *obd = m->private;
+ struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev);
- CFS_INIT_LIST_HEAD(&tmp);
- if (cfs_parse_nidlist(kernbuf, count, &tmp) <= 0) {
- errmsg = "can't parse";
- GOTO(failed, rc = -EINVAL);
- }
+ seq_printf(m, "%u\n", !!mdt->mdt_opts.mo_dom_read_open);
+ return 0;
+}
- down_write(&mdt->mdt_squash_sem);
- if (!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;
- list_splice(&tmp, &mdt->mdt_nosquash_nids);
+/**
+ * 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
+mdt_dom_read_open_seq_write(struct file *file, const char __user *buffer,
+ 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);
+ bool val;
+ int rc;
+
+ rc = kstrtobool_from_user(buffer, count, &val);
+ if (rc)
+ return rc;
+
+ mdt->mdt_opts.mo_dom_read_open = !!val;
+ return count;
+}
+LPROC_SEQ_FOPS(mdt_dom_read_open);
- LCONSOLE_INFO("%s: nosquash_nids is set to %s\n",
- obd->obd_name, kernbuf);
- up_write(&mdt->mdt_squash_sem);
- RETURN(count);
+static int mdt_migrate_hsm_allowed_seq_show(struct seq_file *m, void *data)
+{
+ struct obd_device *obd = m->private;
+ struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev);
- 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);
+ seq_printf(m, "%u\n", (mdt->mdt_opts.mo_migrate_hsm_allowed != 0));
+ return 0;
}
-static int lprocfs_rd_mdt_som(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+static ssize_t
+mdt_migrate_hsm_allowed_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 seq_file *m = file->private_data;
+ struct obd_device *obd = m->private;
+ struct mdt_device *mdt = mdt_dev(obd->obd_lu_dev);
+ bool val;
+ int rc;
+
+ rc = kstrtobool_from_user(buffer, count, &val);
+ if (rc)
+ return rc;
+
+ mdt->mdt_opts.mo_migrate_hsm_allowed = val;
+ return count;
+}
+LPROC_SEQ_FOPS(mdt_migrate_hsm_allowed);
+
+LPROC_SEQ_FOPS_RO_TYPE(mdt, recovery_status);
+LPROC_SEQ_FOPS_RO_TYPE(mdt, num_exports);
+LPROC_SEQ_FOPS_RO_TYPE(mdt, target_instance);
+LPROC_SEQ_FOPS_RO_TYPE(mdt, hash);
+LPROC_SEQ_FOPS_WR_ONLY(mdt, mds_evict_client);
+LPROC_SEQ_FOPS_RW_TYPE(mdt, job_interval);
+LPROC_SEQ_FOPS_RW_TYPE(mdt, ir_factor);
+LPROC_SEQ_FOPS_RW_TYPE(mdt, nid_stats_clear);
+LPROC_SEQ_FOPS(mdt_hsm_cdt_control);
+
+LPROC_SEQ_FOPS_RW_TYPE(mdt, recovery_time_hard);
+LPROC_SEQ_FOPS_RW_TYPE(mdt, recovery_time_soft);
+
+LPROC_SEQ_FOPS_RO(tgt_tot_dirty);
+LPROC_SEQ_FOPS_RO(tgt_tot_granted);
+LPROC_SEQ_FOPS_RO(tgt_tot_pending);
+LPROC_SEQ_FOPS(tgt_grant_compat_disable);
- return snprintf(page, count, "%sabled\n",
- mdt->mdt_som_conf ? "en" : "dis");
+static struct lprocfs_vars lprocfs_mdt_obd_vars[] = {
+ { .name = "tot_dirty",
+ .fops = &tgt_tot_dirty_fops },
+ { .name = "tot_pending",
+ .fops = &tgt_tot_pending_fops },
+ { .name = "tot_granted",
+ .fops = &tgt_tot_granted_fops },
+ { .name = "grant_compat_disable",
+ .fops = &tgt_grant_compat_disable_fops },
+ { .name = "recovery_status",
+ .fops = &mdt_recovery_status_fops },
+ { .name = "num_exports",
+ .fops = &mdt_num_exports_fops },
+ { .name = "identity_expire",
+ .fops = &mdt_identity_expire_fops },
+ { .name = "identity_acquire_expire",
+ .fops = &mdt_identity_acquire_expire_fops },
+ { .name = "identity_upcall",
+ .fops = &mdt_identity_upcall_fops },
+ { .name = "identity_flush",
+ .fops = &mdt_identity_flush_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 = "evict_tgt_nids",
+ .fops = &mdt_evict_tgt_nids_fops },
+ { .name = "hash_stats",
+ .fops = &mdt_hash_fops },
+ { .name = "commit_on_sharing",
+ .fops = &mdt_cos_fops },
+ { .name = "root_squash",
+ .fops = &mdt_root_squash_fops },
+ { .name = "nosquash_nids",
+ .fops = &mdt_nosquash_nids_fops },
+ { .name = "instance",
+ .fops = &mdt_target_instance_fops },
+ { .name = "ir_factor",
+ .fops = &mdt_ir_factor_fops },
+ { .name = "job_cleanup_interval",
+ .fops = &mdt_job_interval_fops },
+ { .name = "enable_remote_dir",
+ .fops = &mdt_enable_remote_dir_fops },
+ { .name = "enable_remote_dir_gid",
+ .fops = &mdt_enable_remote_dir_gid_fops },
+ { .name = "hsm_control",
+ .fops = &mdt_hsm_cdt_control_fops },
+ { .name = "recovery_time_hard",
+ .fops = &mdt_recovery_time_hard_fops },
+ { .name = "recovery_time_soft",
+ .fops = &mdt_recovery_time_soft_fops },
+ { .name = "sync_lock_cancel",
+ .fops = &mdt_slc_fops },
+ { .name = "async_commit_count",
+ .fops = &mdt_async_commit_count_fops },
+ { .name = "sync_count",
+ .fops = &mdt_sync_count_fops },
+ { .name = "dom_lock",
+ .fops = &mdt_dom_lock_fops },
+ { .name = "dom_read_open",
+ .fops = &mdt_dom_read_open_fops },
+ { .name = "migrate_hsm_allowed",
+ .fops = &mdt_migrate_hsm_allowed_fops },
+ { NULL }
+};
+
+static int
+lprocfs_mdt_print_open_files(struct cfs_hash *hs, struct cfs_hash_bd *bd,
+ struct hlist_node *hnode, void *v)
+{
+ struct obd_export *exp = cfs_hash_object(hs, hnode);
+ 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 mdt_quota_off(struct mdt_device *mdt)
+static int lprocfs_mdt_open_files_seq_show(struct seq_file *seq, void *v)
{
- struct md_device *next = mdt->mdt_child;
- const struct md_quota_operations *mqo = &next->md_ops->mdo_quota;
- struct lu_env env;
- int rc;
+ struct nid_stat *stats = seq->private;
+ struct obd_device *obd = stats->nid_obd;
+
+ cfs_hash_for_each_key(obd->obd_nid_hash, &stats->nid,
+ lprocfs_mdt_print_open_files, seq);
- lu_env_init(&env, LCT_MD_THREAD);
- rc = mqo->mqo_off(&env, next, UGQUOTA | IMMQUOTA);
- lu_env_fini(&env);
- return rc;
+ return 0;
}
-static int lprocfs_wr_mdt_som(struct file *file, const char *buffer,
- unsigned long count, void *data)
+int lprocfs_mdt_open_files_seq_open(struct inode *inode, struct file *file)
{
- 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;
- int rc;
+ struct seq_file *seq;
+ int rc;
- if (count > (sizeof(kernbuf) - 1))
- return -EINVAL;
+ rc = single_open(file, &lprocfs_mdt_open_files_seq_show, NULL);
+ if (rc != 0)
+ return rc;
- if (copy_from_user(kernbuf, buffer, count))
- return -EFAULT;
+ seq = file->private_data;
+ seq->private = PDE_DATA(inode);
- kernbuf[count] = '\0';
+ return 0;
+}
- if (!strcmp(kernbuf, "enabled"))
- val = 1;
- else if (strcmp(kernbuf, "disabled"))
- return -EINVAL;
+void mdt_counter_incr(struct ptlrpc_request *req, int opcode)
+{
+ struct obd_export *exp = req->rq_export;
+
+ if (exp->exp_obd && exp->exp_obd->obd_md_stats)
+ lprocfs_counter_incr(exp->exp_obd->obd_md_stats,
+ opcode + LPROC_MD_LAST_OPC);
+ if (exp->exp_nid_stats && exp->exp_nid_stats->nid_stats != NULL)
+ lprocfs_counter_incr(exp->exp_nid_stats->nid_stats, opcode);
+ 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, 1);
+}
- if (mdt->mdt_som_conf == val)
- return count;
+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",
+};
- if (!obd->obd_process_conf) {
- CERROR("Temporary SOM change is not supported, use lctl "
- "conf_param for permanent setting\n");
- return count;
- }
+void mdt_stats_counter_init(struct lprocfs_stats *stats)
+{
+ int idx;
- /* 1 stands for self export. */
- 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;
- }
+ LASSERT(stats && stats->ls_num >= ARRAY_SIZE(mdt_stats));
- if ((rc = mdt_quota_off(mdt))) {
- if (rc == -EALREADY)
- rc = 0;
- else
- return rc;
- }
+ for (idx = 0; idx < ARRAY_SIZE(mdt_stats); idx++) {
+ int flags = 0;
- mdt->mdt_som_conf = val;
- LCONSOLE_INFO("Enabling SOM\n");
+ if (idx == LPROC_MDT_IO_WRITE || idx == LPROC_MDT_IO_READ)
+ flags = LPROCFS_CNTR_AVGMINMAX;
- return count;
+ lprocfs_counter_init(stats, idx, flags, mdt_stats[idx], "reqs");
+ }
}
-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 },
- { 0 }
-};
-
-static struct lprocfs_vars lprocfs_mdt_module_vars[] = {
- { "num_refs", lprocfs_rd_numrefs, 0, 0 },
- { 0 }
-};
+int mdt_procfs_init(struct mdt_device *mdt, const char *name)
+{
+ struct obd_device *obd = mdt2obd_dev(mdt);
+ int rc;
+ int i;
+
+ ENTRY;
+ LASSERT(name != NULL);
+
+ 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 = hsm_cdt_procfs_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 */
+ for (i = 0; i < ARRAY_SIZE(mdt_stats); i++) {
+ int idx = i + LPROC_MD_LAST_OPC;
+ int flags = 0;
+
+ if (idx == LPROC_MDT_IO_WRITE || idx == LPROC_MDT_IO_READ)
+ flags = LPROCFS_CNTR_AVGMINMAX;
+
+ lprocfs_counter_init(obd->obd_md_stats, idx, flags,
+ mdt_stats[i], "reqs");
+ }
+
+ 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 lprocfs_mdt_init_vars(struct lprocfs_static_vars *lvars)
+void mdt_procfs_fini(struct mdt_device *mdt)
{
- lvars->module_vars = lprocfs_mdt_module_vars;
- lvars->obd_vars = lprocfs_mdt_obd_vars;
+ 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_procfs_fini(mdt);
+ lprocfs_obd_cleanup(obd);
+ lprocfs_free_md_stats(obd);
+ lprocfs_free_obd_stats(obd);
+ lprocfs_job_stats_fini(obd);
}
git://git.whamcloud.com / fs / lustre-release.git / blobdiff