*
* 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 (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*
- * Copyright (c) 2011, 2012, Whamcloud, Inc.
+ * Copyright (c) 2012, 2017, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
- * Lustre is a trademark of Sun Microsystems, Inc.
*
* lustre/ofd/lproc_ofd.c
+ *
+ * This file provides functions of procfs interface for OBD Filter Device (OFD).
+ *
+ * Author: Andreas Dilger <andreas.dilger@intel.com>
+ * Author: Mikhail Pershin <mike.pershin@intel.com>
+ * Author: Johann Lombardi <johann.lombardi@intel.com>
+ * Author: Fan Yong <fan.yong@intel.com>
*/
#define DEBUG_SUBSYSTEM S_CLASS
#include <obd.h>
#include <lprocfs_status.h>
#include <linux/seq_file.h>
+#include <lustre_lfsck.h>
+#include <uapi/linux/lustre/lustre_access_log.h>
+#include <obd_cksum.h>
#include "ofd_internal.h"
-#ifdef LPROCFS
+#ifdef CONFIG_PROC_FS
-static int lprocfs_ofd_rd_groups(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+/**
+ * Show number of FID allocation sequences.
+ *
+ * \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 seqs_allocated_show(struct kobject *kobj, struct attribute *attr,
+ char *buf)
{
- struct obd_device *obd = (struct obd_device *)data;
+ struct obd_device *obd = container_of(kobj, struct obd_device,
+ obd_kset.kobj);
struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- *eof = 1;
- return snprintf(page, count, "%u\n", ofd->ofd_max_group);
+ return sprintf(buf, "%u\n", ofd->ofd_seq_count);
}
+LUSTRE_RO_ATTR(seqs_allocated);
-static int lprocfs_ofd_rd_tot_dirty(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+/**
+ * Show total number of grants for precreate.
+ *
+ * \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 grant_precreate_show(struct kobject *kobj,
+ struct attribute *attr,
+ char *buf)
{
- struct obd_device *obd = (struct obd_device *)data;
- struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
+ struct obd_device *obd = container_of(kobj, struct obd_device,
+ obd_kset.kobj);
- LASSERT(obd != NULL);
- *eof = 1;
- return snprintf(page, count, LPU64"\n", ofd->ofd_tot_dirty);
+ return sprintf(buf, "%ld\n",
+ obd->obd_self_export->exp_target_data.ted_grant);
}
+LUSTRE_RO_ATTR(grant_precreate);
-static int lprocfs_ofd_rd_tot_granted(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+/**
+ * Show number of precreates allowed in a single transaction.
+ *
+ * \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 precreate_batch_show(struct kobject *kobj,
+ struct attribute *attr,
+ char *buf)
{
- struct obd_device *obd = (struct obd_device *)data;
+ struct obd_device *obd = container_of(kobj, struct obd_device,
+ obd_kset.kobj);
struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- LASSERT(obd != NULL);
- *eof = 1;
- return snprintf(page, count, LPU64"\n", ofd->ofd_tot_granted);
+ return sprintf(buf, "%d\n", ofd->ofd_precreate_batch);
}
-static int lprocfs_ofd_rd_tot_pending(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+/**
+ * Change number of precreates allowed in a single transaction.
+ *
+ * \param[in] file proc file
+ * \param[in] buffer string which represents maximum number
+ * \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 precreate_batch_store(struct kobject *kobj,
+ struct attribute *attr,
+ const char *buffer, size_t count)
{
- struct obd_device *obd = (struct obd_device *)data;
+ struct obd_device *obd = container_of(kobj, struct obd_device,
+ obd_kset.kobj);
struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
+ unsigned int val;
+ int rc;
- LASSERT(obd != NULL);
- *eof = 1;
- return snprintf(page, count, LPU64"\n", ofd->ofd_tot_pending);
-}
+ rc = kstrtouint(buffer, 0, &val);
+ if (rc)
+ return rc;
-static int lprocfs_ofd_rd_grant_precreate(char *page, char **start, off_t off,
- int count, int *eof, void *data)
-{
- struct obd_device *obd = (struct obd_device *)data;
+ if (val < 1 || val > 65536)
+ return -EINVAL;
- LASSERT(obd != NULL);
- *eof = 1;
- return snprintf(page, count, "%ld\n",
- obd->obd_self_export->exp_filter_data.fed_grant);
+ spin_lock(&ofd->ofd_batch_lock);
+ ofd->ofd_precreate_batch = val;
+ spin_unlock(&ofd->ofd_batch_lock);
+ return count;
}
+LUSTRE_RW_ATTR(precreate_batch);
-static int lprocfs_ofd_rd_grant_ratio(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+/**
+ * Show number of seconds to delay atime
+ *
+ * \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 atime_diff_show(struct kobject *kobj, struct attribute *attr,
+ char *buf)
{
- struct obd_device *obd = (struct obd_device *)data;
+ struct obd_device *obd = container_of(kobj, struct obd_device,
+ obd_kset.kobj);
struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- LASSERT(obd != NULL);
- *eof = 1;
- return snprintf(page, count, "%d%%\n",
- (int) ofd_grant_reserved(ofd, 100));
+ return scnprintf(buf, PAGE_SIZE, "%lld\n", ofd->ofd_atime_diff);
}
-static int lprocfs_ofd_wr_grant_ratio(struct file *file, const char *buffer,
- unsigned long count, void *data)
+/**
+ * Change number of seconds to delay atime
+ *
+ * \param[in] file proc file
+ * \param[in] buffer string which represents maximum number
+ * \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 atime_diff_store(struct kobject *kobj, struct attribute *attr,
+ const char *buffer, size_t count)
{
- struct obd_device *obd = (struct obd_device *)data;
- struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- int val;
- int rc;
+ struct obd_device *obd = container_of(kobj, struct obd_device,
+ obd_kset.kobj);
+ struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
+ unsigned int val;
+ int rc;
- rc = lprocfs_write_helper(buffer, count, &val);
+ rc = kstrtouint(buffer, 0, &val);
if (rc)
return rc;
- if (val > 100 || val < 0)
+ if (val > 86400)
return -EINVAL;
- if (val == 0)
- CWARN("%s: disabling grant error margin\n", obd->obd_name);
- if (val > 50)
- CWARN("%s: setting grant error margin >50%%, be warned that "
- "a huge part of the free space is now reserved for "
- "grants\n", obd->obd_name);
-
- cfs_spin_lock(&ofd->ofd_grant_lock);
- ofd->ofd_grant_ratio = ofd_grant_ratio_conv(val);
- cfs_spin_unlock(&ofd->ofd_grant_lock);
+ ofd->ofd_atime_diff = val;
return count;
}
+LUSTRE_RW_ATTR(atime_diff);
-static int lprocfs_ofd_rd_last_id(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+/**
+ * Show the last used ID for each FID sequence used by OFD.
+ *
+ * \param[in] m seq_file handle
+ * \param[in] data unused for single entry
+ *
+ * \retval 0 on success
+ * \retval negative value on error
+ */
+static int ofd_last_id_seq_show(struct seq_file *m, void *data)
{
- struct obd_device *obd = data;
- struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- int retval = 0, rc, i;
+ struct obd_device *obd = m->private;
+ struct ofd_device *ofd;
+ struct ofd_seq *oseq = NULL;
if (obd == NULL)
return 0;
- for (i = FID_SEQ_OST_MDT0; i <= ofd->ofd_max_group; i++) {
- rc = snprintf(page, count, LPU64"\n", ofd_last_id(ofd, i));
- if (rc < 0) {
- retval = rc;
- break;
- }
- page += rc;
- count -= rc;
- retval += rc;
+ ofd = ofd_dev(obd->obd_lu_dev);
+
+ read_lock(&ofd->ofd_seq_list_lock);
+ list_for_each_entry(oseq, &ofd->ofd_seq_list, os_list) {
+ __u64 seq;
+
+ seq = ostid_seq(&oseq->os_oi) == 0 ?
+ fid_idif_seq(ostid_id(&oseq->os_oi),
+ ofd->ofd_lut.lut_lsd.lsd_osd_index) & ~0xFFFF :
+ ostid_seq(&oseq->os_oi);
+ seq_printf(m, DOSTID"\n", seq, ostid_id(&oseq->os_oi));
}
- return retval;
+ read_unlock(&ofd->ofd_seq_list_lock);
+ return 0;
}
-int lprocfs_ofd_rd_fmd_max_num(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+LPROC_SEQ_FOPS_RO(ofd_last_id);
+
+/**
+ * Show if the OFD is in degraded mode.
+ *
+ * Degraded means OFD has a failed drive or is undergoing RAID rebuild.
+ * The MDS will try to avoid using this OST for new object allocations
+ * to reduce the impact to global IO performance when clients writing to
+ * this OST are slowed down. It also reduces the contention on the OST
+ * RAID device, allowing it to rebuild more quickly.
+ *
+ * \retval count of bytes written
+ */
+static ssize_t degraded_show(struct kobject *kobj, struct attribute *attr,
+ char *buf)
{
- struct obd_device *obd = data;
- struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- int rc;
+ struct obd_device *obd = container_of(kobj, struct obd_device,
+ obd_kset.kobj);
+ struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- rc = snprintf(page, count, "%u\n", ofd->ofd_fmd_max_num);
- return rc;
+ return sprintf(buf, "%u\n", ofd->ofd_raid_degraded);
}
-int lprocfs_ofd_wr_fmd_max_num(struct file *file, const char *buffer,
- unsigned long count, void *data)
+/**
+ * Set OFD to degraded mode.
+ *
+ * This is used to interface to userspace administrative tools for
+ * the underlying RAID storage, so that they can mark an OST
+ * as having degraded performance.
+ *
+ * \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 degraded_store(struct kobject *kobj, struct attribute *attr,
+ const char *buffer, size_t count)
{
- struct obd_device *obd = data;
- struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- int val;
- int rc;
+ struct obd_device *obd = container_of(kobj, struct obd_device,
+ obd_kset.kobj);
+ struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
+ bool val;
+ int rc;
- rc = lprocfs_write_helper(buffer, count, &val);
+ rc = kstrtobool(buffer, &val);
if (rc)
return rc;
- if (val > 65536 || val < 1)
- return -EINVAL;
-
- ofd->ofd_fmd_max_num = val;
+ spin_lock(&ofd->ofd_flags_lock);
+ ofd->ofd_raid_degraded = val;
+ spin_unlock(&ofd->ofd_flags_lock);
return count;
}
+LUSTRE_RW_ATTR(degraded);
-int lprocfs_ofd_rd_fmd_max_age(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+/**
+ * Show if the OFD is in no precreate mode.
+ *
+ * This means OFD has been adminstratively disabled at the OST to prevent
+ * the MDS from creating any new files on the OST, though existing files
+ * can still be read, written, and unlinked.
+ *
+ * \retval number of bytes written
+ */
+static ssize_t no_create_show(struct kobject *kobj, struct attribute *attr,
+ char *buf)
{
- struct obd_device *obd = data;
- struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- int rc;
+ struct obd_device *obd = container_of(kobj, struct obd_device,
+ obd_kset.kobj);
+ struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- rc = snprintf(page, count, "%ld\n", ofd->ofd_fmd_max_age / CFS_HZ);
- return rc;
+ return scnprintf(buf, PAGE_SIZE, "%u\n", ofd->ofd_lut.lut_no_create);
}
-int lprocfs_ofd_wr_fmd_max_age(struct file *file, const char *buffer,
- unsigned long count, void *data)
+/**
+ * Set OFD to no create mode.
+ *
+ * This is used to interface to userspace administrative tools to
+ * disable new object creation on the OST.
+ *
+ * \param[in] count \a buffer length
+ *
+ * \retval \a count on success
+ * \retval negative number on error
+ */
+static ssize_t no_create_store(struct kobject *kobj, struct attribute *attr,
+ const char *buffer, size_t count)
{
- struct obd_device *obd = data;
- struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- int val;
- int rc;
+ struct obd_device *obd = container_of(kobj, struct obd_device,
+ obd_kset.kobj);
+ struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
+ bool val;
+ int rc;
- rc = lprocfs_write_helper(buffer, count, &val);
+ rc = kstrtobool(buffer, &val);
if (rc)
return rc;
- if (val > 65536 || val < 1)
- return -EINVAL;
+ spin_lock(&ofd->ofd_flags_lock);
+ ofd->ofd_lut.lut_no_create = val;
+ spin_unlock(&ofd->ofd_flags_lock);
- ofd->ofd_fmd_max_age = val * CFS_HZ;
return count;
}
+LUSTRE_RW_ATTR(no_create);
+
+#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 20, 53, 0)
+/* compatibility entry for a few releases */
+#define no_precreate_show no_create_show
+#define no_precreate_store no_create_store
+LUSTRE_RW_ATTR(no_precreate);
+#endif
-static int lprocfs_ofd_rd_capa(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+/**
+ * Show OFD filesystem type.
+ *
+ * \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 fstype_show(struct kobject *kobj, struct attribute *attr,
+ char *buf)
{
- struct obd_device *obd = data;
- int rc;
+ struct obd_device *obd = container_of(kobj, struct obd_device,
+ obd_kset.kobj);
+ struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
+ struct lu_device *d;
- rc = snprintf(page, count, "capability on: %s\n",
- obd->u.filter.fo_fl_oss_capa ? "oss" : "");
- return rc;
+ LASSERT(ofd->ofd_osd);
+ d = &ofd->ofd_osd->dd_lu_dev;
+ LASSERT(d->ld_type);
+ return sprintf(buf, "%s\n", d->ld_type->ldt_name);
}
+LUSTRE_RO_ATTR(fstype);
-static int lprocfs_ofd_wr_capa(struct file *file, const char *buffer,
- unsigned long count, void *data)
+/**
+ * Show journal handling mode: synchronous or asynchronous.
+ *
+ * When running in asynchronous mode the journal transactions are not
+ * committed to disk before the RPC is replied back to the client.
+ * This will typically improve client performance when only a small number
+ * of clients are writing, since the client(s) can have more write RPCs
+ * in flight. However, it also means that the client has to handle recovery
+ * on bulk RPCs, and will have to keep more dirty pages in cache before they
+ * are committed on the OST.
+ *
+ * \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_journal_show(struct kobject *kobj, struct attribute *attr,
+ char *buf)
{
- struct obd_device *obd = data;
- int val, rc;
+ struct obd_device *obd = container_of(kobj, struct obd_device,
+ obd_kset.kobj);
+ struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- rc = lprocfs_write_helper(buffer, count, &val);
+ return sprintf(buf, "%u\n", ofd->ofd_sync_journal);
+}
+
+/**
+ * Set journal mode to synchronous or asynchronous.
+ *
+ * \param[in] file proc file
+ * \param[in] buffer string which represents mode
+ * 1: synchronous mode
+ * 0: asynchronous mode
+ * \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 sync_journal_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 ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
+ bool val;
+ int rc;
+
+ rc = kstrtobool(buffer, &val);
if (rc)
return rc;
- if (val & ~0x1) {
- CERROR("invalid capability mode, only 0/1 are accepted.\n"
- " 1: enable oss fid capability\n"
- " 0: disable oss fid capability\n");
- return -EINVAL;
- }
+ spin_lock(&ofd->ofd_flags_lock);
+ ofd->ofd_sync_journal = val;
+ ofd_slc_set(ofd);
+ spin_unlock(&ofd->ofd_flags_lock);
- obd->u.filter.fo_fl_oss_capa = val;
- LCONSOLE_INFO("OSS %s %s fid capability.\n", obd->obd_name,
- val ? "enabled" : "disabled");
return count;
}
+LUSTRE_RW_ATTR(sync_journal);
-static int lprocfs_ofd_rd_capa_count(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+static int ofd_brw_size_seq_show(struct seq_file *m, void *data)
{
- return snprintf(page, count, "%d %d\n",
- capa_count[CAPA_SITE_CLIENT],
- capa_count[CAPA_SITE_SERVER]);
+ struct obd_device *obd = m->private;
+ struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
+
+ seq_printf(m, "%u\n", ofd->ofd_brw_size / ONE_MB_BRW_SIZE);
+ return 0;
}
-int lprocfs_ofd_rd_degraded(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+static ssize_t
+ofd_brw_size_seq_write(struct file *file, const char __user *buffer,
+ size_t count, loff_t *off)
{
- struct obd_device *obd = data;
+ struct seq_file *m = file->private_data;
+ struct obd_device *obd = m->private;
struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
+ char kernbuf[22] = "";
+ u64 val;
+ int rc;
- return snprintf(page, count, "%u\n", ofd->ofd_raid_degraded);
-}
+ if (count >= sizeof(kernbuf))
+ return -EINVAL;
-int lprocfs_ofd_wr_degraded(struct file *file, const char *buffer,
- unsigned long count, void *data)
-{
- struct obd_device *obd = data;
- struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- int val, rc;
+ if (copy_from_user(kernbuf, buffer, count))
+ return -EFAULT;
+ kernbuf[count] = 0;
- rc = lprocfs_write_helper(buffer, count, &val);
- if (rc)
+ rc = sysfs_memparse(kernbuf, count, &val, "MiB");
+ if (rc < 0)
return rc;
- cfs_spin_lock(&ofd->ofd_flags_lock);
- ofd->ofd_raid_degraded = !!val;
- cfs_spin_unlock(&ofd->ofd_flags_lock);
+ if (val == 0)
+ return -EINVAL;
+
+ if (val > DT_MAX_BRW_SIZE ||
+ val < (1 << ofd->ofd_lut.lut_tgd.tgd_blockbits))
+ return -ERANGE;
+
+ spin_lock(&ofd->ofd_flags_lock);
+ ofd->ofd_brw_size = val;
+ spin_unlock(&ofd->ofd_flags_lock);
return count;
}
+LPROC_SEQ_FOPS(ofd_brw_size);
-int lprocfs_ofd_rd_fstype(char *page, char **start, off_t off, int count,
- int *eof, void *data)
+/*
+ * ofd_checksum_type(server) proc handling
+ */
+DECLARE_CKSUM_NAME;
+
+static int ofd_checksum_type_seq_show(struct seq_file *m, void *data)
{
- struct obd_device *obd = data;
- struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- struct lu_device *d;
+ struct obd_device *obd = m->private;
+ struct lu_target *lut;
+ enum cksum_types pref;
+ int i;
- LASSERT(ofd->ofd_osd);
- d = &ofd->ofd_osd->dd_lu_dev;
- LASSERT(d->ld_type);
- return snprintf(page, count, "%s\n", d->ld_type->ldt_name);
+ if (!obd)
+ return 0;
+
+ lut = obd2obt(obd)->obt_lut;
+ /* select fastest checksum type on the server */
+ pref = obd_cksum_type_select(obd->obd_name,
+ lut->lut_cksum_types_supported,
+ lut->lut_dt_conf.ddp_t10_cksum_type);
+
+ 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(ofd_checksum_type);
+
+
+#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 16, 53, 0)
+static ssize_t sync_on_lock_cancel_show(struct kobject *kobj,
+ struct attribute *attr, char *buf)
+{
+ return sync_lock_cancel_show(kobj, attr, buf);
}
-int lprocfs_ofd_rd_syncjournal(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+static ssize_t sync_on_lock_cancel_store(struct kobject *kobj,
+ struct attribute *attr,
+ const char *buffer, size_t count)
{
- struct obd_device *obd = data;
- struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- int rc;
+ static bool sync_on_lock_cancel_warned;
- rc = snprintf(page, count, "%u\n", ofd->ofd_syncjournal);
- return rc;
+ if (!sync_on_lock_cancel_warned) {
+ sync_on_lock_cancel_warned = true;
+ pr_info("ofd: 'obdfilter.*.sync_on_lock_cancel' is deprecated, use 'obdfilter.*.sync_lock_cancel' instead\n");
+ }
+ return sync_lock_cancel_store(kobj, attr, buffer, count);
}
+LUSTRE_RW_ATTR(sync_on_lock_cancel);
+#endif
-int lprocfs_ofd_wr_syncjournal(struct file *file, const char *buffer,
- unsigned long count, void *data)
+/**
+ * Show the limit of soft sync RPCs.
+ *
+ * This value defines how many IO RPCs with OBD_BRW_SOFT_SYNC flag
+ * are allowed before sync update will be triggered.
+ *
+ * \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 soft_sync_limit_show(struct kobject *kobj,
+ struct attribute *attr, char *buf)
{
- struct obd_device *obd = data;
- struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- int val;
- int rc;
+ struct obd_device *obd = container_of(kobj, struct obd_device,
+ obd_kset.kobj);
+ struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- rc = lprocfs_write_helper(buffer, count, &val);
- if (rc)
- return rc;
+ return sprintf(buf, "%u\n", ofd->ofd_soft_sync_limit);
+}
- if (val < 0)
- return -EINVAL;
+/**
+ * Change the limit of soft sync RPCs.
+ *
+ * Define how many IO RPCs with OBD_BRW_SOFT_SYNC flag
+ * allowed before sync update will be done.
+ *
+ * This limit is global across all exports.
+ *
+ * \param[in] file proc file
+ * \param[in] buffer string which represents limit
+ * \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 soft_sync_limit_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 ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
+ unsigned int val;
+ int rc;
- cfs_spin_lock(&ofd->ofd_flags_lock);
- ofd->ofd_syncjournal = !!val;
- ofd_slc_set(ofd);
- cfs_spin_unlock(&ofd->ofd_flags_lock);
+ rc = kstrtouint(buffer, 0, &val);
+ if (rc < 0)
+ return rc;
+ ofd->ofd_soft_sync_limit = val;
return count;
}
+LUSTRE_RW_ATTR(soft_sync_limit);
-static char *sync_on_cancel_states[] = {"never",
- "blocking",
- "always" };
-
-int lprocfs_ofd_rd_sync_lock_cancel(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+/**
+ * Show the LFSCK speed limit.
+ *
+ * The maximum number of items scanned per second.
+ *
+ * \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 lfsck_speed_limit_show(struct kobject *kobj,
+ struct attribute *attr, char *buf)
{
- struct obd_device *obd = data;
- struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- int rc;
+ struct obd_device *obd = container_of(kobj, struct obd_device,
+ obd_kset.kobj);
+ struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- rc = snprintf(page, count, "%s\n",
- sync_on_cancel_states[ofd->ofd_sync_lock_cancel]);
- return rc;
+ return lfsck_get_speed(buf, ofd->ofd_osd);
}
-int lprocfs_ofd_wr_sync_lock_cancel(struct file *file, const char *buffer,
- unsigned long count, void *data)
+/**
+ * Change the LFSCK speed limit.
+ *
+ * Limit number of items that may be scanned per second.
+ *
+ * \param[in] file proc file
+ * \param[in] buffer string which represents limit
+ * \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 lfsck_speed_limit_store(struct kobject *kobj,
+ struct attribute *attr,
+ const char *buffer, size_t count)
{
- struct obd_device *obd = data;
- struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- int val = -1;
- int i;
+ struct obd_device *obd = container_of(kobj, struct obd_device,
+ obd_kset.kobj);
+ struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
+ unsigned int val;
+ int rc;
- for (i = 0 ; i < NUM_SYNC_ON_CANCEL_STATES; i++) {
- if (memcmp(buffer, sync_on_cancel_states[i],
- strlen(sync_on_cancel_states[i])) == 0) {
- val = i;
- break;
- }
- }
- if (val == -1) {
- int rc;
+ rc = kstrtouint(buffer, 0, &val);
+ if (rc != 0)
+ return rc;
- rc = lprocfs_write_helper(buffer, count, &val);
- if (rc)
- return rc;
- }
+ rc = lfsck_set_speed(ofd->ofd_osd, val);
- if (val < 0 || val > 2)
- return -EINVAL;
+ return rc != 0 ? rc : count;
+}
+LUSTRE_RW_ATTR(lfsck_speed_limit);
- cfs_spin_lock(&ofd->ofd_flags_lock);
- ofd->ofd_sync_lock_cancel = val;
- cfs_spin_unlock(&ofd->ofd_flags_lock);
- return count;
+/**
+ * Show LFSCK layout verification stats from the most recent LFSCK run.
+ *
+ * \param[in] m seq_file handle
+ * \param[in] data unused for single entry
+ *
+ * \retval 0 on success
+ * \retval negative value on error
+ */
+static int ofd_lfsck_layout_seq_show(struct seq_file *m, void *data)
+{
+ struct obd_device *obd = m->private;
+ struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
+
+ return lfsck_dump(m, ofd->ofd_osd, LFSCK_TYPE_LAYOUT);
}
-int lprocfs_ofd_rd_grant_compat_disable(char *page, char **start, off_t off,
- int count, int *eof, void *data)
+LPROC_SEQ_FOPS_RO(ofd_lfsck_layout);
+
+/**
+ * Show if LFSCK performed parent FID verification.
+ *
+ * \param[in] m seq_file handle
+ * \param[in] data unused for single entry
+ *
+ * \retval 0 on success
+ * \retval negative value on error
+ */
+static int ofd_lfsck_verify_pfid_seq_show(struct seq_file *m, void *data)
{
- struct obd_device *obd = data;
- struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- int rc;
+ struct obd_device *obd = m->private;
+ struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- rc = snprintf(page, count, "%u\n", ofd->ofd_grant_compat_disable);
- return rc;
+ seq_printf(m, "switch: %s\ndetected: %llu\nrepaired: %llu\n",
+ ofd->ofd_lfsck_verify_pfid ? "on" : "off",
+ ofd->ofd_inconsistency_self_detected,
+ ofd->ofd_inconsistency_self_repaired);
+ return 0;
}
-int lprocfs_ofd_wr_grant_compat_disable(struct file *file, const char *buffer,
- unsigned long count, void *data)
+/**
+ * Set the LFSCK behavior to verify parent FID correctness.
+ *
+ * If flag ofd_lfsck_verify_pfid is set then LFSCK does parent FID
+ * verification during read/write operations.
+ *
+ * \param[in] file proc file
+ * \param[in] buffer string which represents behavior
+ * 1: verify parent FID
+ * 0: don't verify parent FID
+ * \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
+ofd_lfsck_verify_pfid_seq_write(struct file *file, const char __user *buffer,
+ size_t count, loff_t *off)
{
- struct obd_device *obd = data;
- struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- int val;
- int rc;
+ struct seq_file *m = file->private_data;
+ struct obd_device *obd = m->private;
+ struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
+ bool val;
+ int rc;
- rc = lprocfs_write_helper(buffer, count, &val);
+ rc = kstrtobool_from_user(buffer, count, &val);
if (rc)
return rc;
- if (val < 0)
- return -EINVAL;
-
- cfs_spin_lock(&ofd->ofd_flags_lock);
- ofd->ofd_grant_compat_disable = !!val;
- cfs_spin_unlock(&ofd->ofd_flags_lock);
+ ofd->ofd_lfsck_verify_pfid = val;
+ if (!ofd->ofd_lfsck_verify_pfid) {
+ ofd->ofd_inconsistency_self_detected = 0;
+ ofd->ofd_inconsistency_self_repaired = 0;
+ }
return count;
}
-static struct lprocfs_vars lprocfs_ofd_obd_vars[] = {
- { "uuid", lprocfs_rd_uuid, 0, 0 },
- { "blocksize", lprocfs_rd_blksize, 0, 0 },
- { "kbytestotal", lprocfs_rd_kbytestotal, 0, 0 },
- { "kbytesfree", lprocfs_rd_kbytesfree, 0, 0 },
- { "kbytesavail", lprocfs_rd_kbytesavail, 0, 0 },
- { "filestotal", lprocfs_rd_filestotal, 0, 0 },
- { "filesfree", lprocfs_rd_filesfree, 0, 0 },
- { "filegroups", lprocfs_ofd_rd_groups, 0, 0 },
- { "fstype", lprocfs_ofd_rd_fstype, 0, 0 },
- { "last_id", lprocfs_ofd_rd_last_id, 0, 0 },
- { "tot_dirty", lprocfs_ofd_rd_tot_dirty, 0, 0 },
- { "tot_pending", lprocfs_ofd_rd_tot_pending, 0, 0 },
- { "tot_granted", lprocfs_ofd_rd_tot_granted, 0, 0 },
- { "grant_precreate", lprocfs_ofd_rd_grant_precreate, 0, 0 },
- { "grant_ratio", lprocfs_ofd_rd_grant_ratio,
- lprocfs_ofd_wr_grant_ratio, 0, 0 },
- { "recovery_status", lprocfs_obd_rd_recovery_status, 0, 0 },
- { "recovery_time_soft", lprocfs_obd_rd_recovery_time_soft,
- lprocfs_obd_wr_recovery_time_soft, 0},
- { "recovery_time_hard", lprocfs_obd_rd_recovery_time_hard,
- lprocfs_obd_wr_recovery_time_hard, 0},
- { "evict_client", 0, lprocfs_wr_evict_client, 0,
- &lprocfs_evict_client_fops},
- { "num_exports", lprocfs_rd_num_exports, 0, 0 },
- { "degraded", lprocfs_ofd_rd_degraded,
- lprocfs_ofd_wr_degraded, 0},
- { "sync_journal", lprocfs_ofd_rd_syncjournal,
- lprocfs_ofd_wr_syncjournal, 0 },
- { "sync_on_lock_cancel", lprocfs_ofd_rd_sync_lock_cancel,
- lprocfs_ofd_wr_sync_lock_cancel, 0 },
- { "instance", lprocfs_target_rd_instance, 0 },
- { "ir_factor", lprocfs_obd_rd_ir_factor,
- lprocfs_obd_wr_ir_factor, 0},
- { "grant_compat_disable", lprocfs_ofd_rd_grant_compat_disable,
- lprocfs_ofd_wr_grant_compat_disable, 0 },
- { "client_cache_count", lprocfs_ofd_rd_fmd_max_num,
- lprocfs_ofd_wr_fmd_max_num, 0 },
- { "client_cache_seconds", lprocfs_ofd_rd_fmd_max_age,
- lprocfs_ofd_wr_fmd_max_age, 0 },
- { "capa", lprocfs_ofd_rd_capa,
- lprocfs_ofd_wr_capa, 0 },
- { "capa_count", lprocfs_ofd_rd_capa_count, 0, 0 },
- { 0 }
-};
+LPROC_SEQ_FOPS(ofd_lfsck_verify_pfid);
-static struct lprocfs_vars lprocfs_ofd_module_vars[] = {
- { "num_refs", lprocfs_rd_numrefs, 0, 0 },
- { 0 }
-};
+static ssize_t access_log_mask_show(struct kobject *kobj,
+ struct attribute *attr, char *buf)
+{
+ struct obd_device *obd = container_of(kobj, struct obd_device,
+ obd_kset.kobj);
+ struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
-#define pct(a,b) (b ? a * 100 / b : 0)
+ return scnprintf(buf, PAGE_SIZE, "%s%s%s\n",
+ (ofd->ofd_access_log_mask == 0) ? "0" : "",
+ (ofd->ofd_access_log_mask & OFD_ACCESS_READ) ? "r" : "",
+ (ofd->ofd_access_log_mask & OFD_ACCESS_WRITE) ? "w" : "");
+}
-static void display_brw_stats(struct seq_file *seq, char *name, char *units,
- struct obd_histogram *read,
- struct obd_histogram *write, int log2)
+static ssize_t access_log_mask_store(struct kobject *kobj,
+ struct attribute *attr,
+ const char *buffer, size_t count)
{
- unsigned long read_tot, write_tot, r, w, read_cum = 0, write_cum = 0;
- int i;
+ struct obd_device *obd = container_of(kobj, struct obd_device,
+ obd_kset.kobj);
+ struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
+ unsigned int mask = 0;
+ size_t i;
- seq_printf(seq, "\n%26s read | write\n", " ");
- seq_printf(seq, "%-22s %-5s %% cum %% | %-5s %% cum %%\n",
- name, units, units);
+ for (i = 0; i < count; i++) {
+ switch (tolower(buffer[i])) {
+ case '0':
+ break;
+ case 'r':
+ mask |= OFD_ACCESS_READ;
+ break;
+ case 'w':
+ mask |= OFD_ACCESS_WRITE;
+ break;
+ default:
+ return -EINVAL;
+ }
+ }
- read_tot = lprocfs_oh_sum(read);
- write_tot = lprocfs_oh_sum(write);
- for (i = 0; i < OBD_HIST_MAX; i++) {
- r = read->oh_buckets[i];
- w = write->oh_buckets[i];
- read_cum += r;
- write_cum += w;
- if (read_cum == 0 && write_cum == 0)
- continue;
+ ofd->ofd_access_log_mask = mask;
- if (!log2)
- seq_printf(seq, "%u", i);
- else if (i < 10)
- seq_printf(seq, "%u", 1 << i);
- else if (i < 20)
- seq_printf(seq, "%uK", 1 << (i - 10));
- else
- seq_printf(seq, "%uM", 1 << (i - 20));
+ return count;
+}
+LUSTRE_RW_ATTR(access_log_mask);
- seq_printf(seq, ":\t\t%10lu %3lu %3lu | %4lu %3lu %3lu\n",
- r, pct(r, read_tot), pct(read_cum, read_tot),
- w, pct(w, write_tot), pct(write_cum, write_tot));
+static ssize_t access_log_size_show(struct kobject *kobj,
+ struct attribute *attr, char *buf)
+{
+ struct obd_device *obd = container_of(kobj, struct obd_device,
+ obd_kset.kobj);
+ struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
- if (read_cum == read_tot && write_cum == write_tot)
- break;
- }
+ return scnprintf(buf, PAGE_SIZE, "%u\n", ofd->ofd_access_log_size);
}
-static void brw_stats_show(struct seq_file *seq, struct brw_stats *brw_stats)
+static ssize_t access_log_size_store(struct kobject *kobj,
+ struct attribute *attr,
+ const char *buffer, size_t count)
{
- struct timeval now;
- char title[24];
-
- /* this sampling races with updates */
- cfs_gettimeofday(&now);
- seq_printf(seq, "snapshot_time: %lu.%lu (secs.usecs)\n",
- now.tv_sec, now.tv_usec);
+ struct obd_device *obd = container_of(kobj, struct obd_device,
+ obd_kset.kobj);
+ struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev);
+ struct ofd_access_log *oal;
+ unsigned int size;
+ ssize_t rc;
- display_brw_stats(seq, "pages per bulk r/w", "rpcs",
- &brw_stats->hist[BRW_R_PAGES],
- &brw_stats->hist[BRW_W_PAGES], 1);
+ rc = kstrtouint(buffer, 0, &size);
+ if (rc < 0)
+ return rc;
- display_brw_stats(seq, "discontiguous pages", "rpcs",
- &brw_stats->hist[BRW_R_DISCONT_PAGES],
- &brw_stats->hist[BRW_W_DISCONT_PAGES], 0);
+ if (!ofd_access_log_size_is_valid(size))
+ return -EINVAL;
- display_brw_stats(seq, "discontiguous blocks", "rpcs",
- &brw_stats->hist[BRW_R_DISCONT_BLOCKS],
- &brw_stats->hist[BRW_W_DISCONT_BLOCKS], 0);
+ /* The size of the ofd_access_log cannot be changed after it
+ * has been created.
+ */
+ if (ofd->ofd_access_log_size == size)
+ return count;
+
+ oal = ofd_access_log_create(obd->obd_name, size);
+ if (IS_ERR(oal))
+ return PTR_ERR(oal);
+
+ spin_lock(&ofd->ofd_flags_lock);
+ if (ofd->ofd_access_log != NULL) {
+ rc = -EBUSY;
+ } else {
+ ofd->ofd_access_log = oal;
+ ofd->ofd_access_log_size = size;
+ oal = NULL;
+ rc = count;
+ }
+ spin_unlock(&ofd->ofd_flags_lock);
- display_brw_stats(seq, "disk fragmented I/Os", "ios",
- &brw_stats->hist[BRW_R_DIO_FRAGS],
- &brw_stats->hist[BRW_W_DIO_FRAGS], 0);
+ ofd_access_log_delete(oal);
- display_brw_stats(seq, "disk I/Os in flight", "ios",
- &brw_stats->hist[BRW_R_RPC_HIST],
- &brw_stats->hist[BRW_W_RPC_HIST], 0);
+ return rc;
+}
+LUSTRE_RW_ATTR(access_log_size);
- sprintf(title, "I/O time (1/%ds)", CFS_HZ);
- display_brw_stats(seq, title, "ios",
- &brw_stats->hist[BRW_R_IO_TIME],
- &brw_stats->hist[BRW_W_IO_TIME], 1);
+static int ofd_site_stats_seq_show(struct seq_file *m, void *data)
+{
+ struct obd_device *obd = m->private;
- display_brw_stats(seq, "disk I/O size", "ios",
- &brw_stats->hist[BRW_R_DISK_IOSIZE],
- &brw_stats->hist[BRW_W_DISK_IOSIZE], 1);
+ return lu_site_stats_seq_print(obd->obd_lu_dev->ld_site, m);
}
-#undef pct
+LPROC_SEQ_FOPS_RO(ofd_site_stats);
-static int ofd_brw_stats_seq_show(struct seq_file *seq, void *v)
+/**
+ * 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 *dev = seq->private;
- struct filter_obd *ofd = &dev->u.filter;
+ struct obd_device *obd = container_of(kobj, struct obd_device,
+ obd_kset.kobj);
+ struct lu_target *lut = obd2obt(obd)->obt_lut;
- brw_stats_show(seq, &ofd->fo_filter_stats);
-
- return 0;
+ return scnprintf(buf, PAGE_SIZE, "%u\n", lut->lut_cksum_t10pi_enforce);
}
-static ssize_t ofd_brw_stats_seq_write(struct file *file, const char *buf,
- size_t len, loff_t *off)
+/**
+ * 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 seq_file *seq = file->private_data;
- struct obd_device *dev = seq->private;
- struct filter_obd *ofd = &dev->u.filter;
- int i;
+ struct obd_device *obd = container_of(kobj, struct obd_device,
+ obd_kset.kobj);
+ struct lu_target *lut = obd2obt(obd)->obt_lut;
+ bool enforce;
+ int rc;
- for (i = 0; i < BRW_LAST; i++)
- lprocfs_oh_clear(&ofd->fo_filter_stats.hist[i]);
+ rc = kstrtobool(buffer, &enforce);
+ if (rc)
+ return rc;
- return len;
+ 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);
+
+LPROC_SEQ_FOPS_RO_TYPE(ofd, recovery_status);
+LUSTRE_RW_ATTR(recovery_time_hard);
+LUSTRE_RW_ATTR(recovery_time_soft);
+LUSTRE_RW_ATTR(ir_factor);
+
+LPROC_SEQ_FOPS_WR_ONLY(ofd, evict_client);
+LPROC_SEQ_FOPS_RW_TYPE(ofd, checksum_dump);
+LUSTRE_RW_ATTR(job_cleanup_interval);
+
+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);
+LUSTRE_RW_ATTR(grant_check_threshold);
+LUSTRE_RO_ATTR(eviction_count);
+
+struct lprocfs_vars lprocfs_ofd_obd_vars[] = {
+ { .name = "last_id",
+ .fops = &ofd_last_id_fops },
+ { .name = "recovery_status",
+ .fops = &ofd_recovery_status_fops },
+ { .name = "evict_client",
+ .fops = &ofd_evict_client_fops },
+ { .name = "brw_size",
+ .fops = &ofd_brw_size_fops },
+ { .name = "checksum_dump",
+ .fops = &ofd_checksum_dump_fops },
+ { .name = "lfsck_layout",
+ .fops = &ofd_lfsck_layout_fops },
+ { .name = "lfsck_verify_pfid",
+ .fops = &ofd_lfsck_verify_pfid_fops },
+ { .name = "site_stats",
+ .fops = &ofd_site_stats_fops },
+ { .name = "checksum_type",
+ .fops = &ofd_checksum_type_fops },
+ { NULL }
+};
-LPROC_SEQ_FOPS(ofd_brw_stats);
+LDEBUGFS_SEQ_FOPS_RO_TYPE(ofd, recovery_stale_clients);
-int lproc_ofd_attach_seqstat(struct obd_device *dev)
-{
- return lprocfs_obd_seq_create(dev, "brw_stats", 0444,
- &ofd_brw_stats_fops, dev);
-}
+struct ldebugfs_vars ldebugfs_ofd_obd_vars[] = {
+ { .name = "recovery_stale_clients",
+ .fops = &ofd_recovery_stale_clients_fops},
+ { NULL }
+};
+
+LDEBUGFS_SEQ_FOPS_RO_TYPE(ofd, srpc_serverctx);
-void lprocfs_ofd_init_vars(struct lprocfs_static_vars *lvars)
+static struct ldebugfs_vars ldebugfs_ofd_gss_vars[] = {
+ { .name = "srpc_serverctx",
+ .fops = &ofd_srpc_serverctx_fops },
+ { NULL }
+};
+
+/**
+ * Initialize OFD statistics counters
+ *
+ * param[in] stats statistics counters
+ */
+void ofd_stats_counter_init(struct lprocfs_stats *stats, unsigned int offset,
+ enum lprocfs_counter_config cntr_umask)
{
- lvars->module_vars = lprocfs_ofd_module_vars;
- lvars->obd_vars = lprocfs_ofd_obd_vars;
+ LASSERT(stats && stats->ls_num >= LPROC_OFD_STATS_LAST);
+
+ lprocfs_counter_init(stats, LPROC_OFD_STATS_READ_BYTES,
+ LPROCFS_TYPE_BYTES_FULL_HISTOGRAM & (~cntr_umask),
+ "read_bytes");
+ lprocfs_counter_init(stats, LPROC_OFD_STATS_WRITE_BYTES,
+ LPROCFS_TYPE_BYTES_FULL_HISTOGRAM & (~cntr_umask),
+ "write_bytes");
+ lprocfs_counter_init(stats, LPROC_OFD_STATS_READ,
+ LPROCFS_TYPE_LATENCY & (~cntr_umask), "read");
+ lprocfs_counter_init(stats, LPROC_OFD_STATS_WRITE,
+ LPROCFS_TYPE_LATENCY & (~cntr_umask), "write");
+ lprocfs_counter_init(stats, LPROC_OFD_STATS_GETATTR,
+ LPROCFS_TYPE_LATENCY & (~cntr_umask), "getattr");
+ lprocfs_counter_init(stats, LPROC_OFD_STATS_SETATTR,
+ LPROCFS_TYPE_LATENCY & (~cntr_umask), "setattr");
+ lprocfs_counter_init(stats, LPROC_OFD_STATS_PUNCH,
+ LPROCFS_TYPE_LATENCY & (~cntr_umask), "punch");
+ lprocfs_counter_init(stats, LPROC_OFD_STATS_SYNC,
+ LPROCFS_TYPE_LATENCY & (~cntr_umask), "sync");
+ lprocfs_counter_init(stats, LPROC_OFD_STATS_DESTROY,
+ LPROCFS_TYPE_LATENCY & (~cntr_umask), "destroy");
+ lprocfs_counter_init(stats, LPROC_OFD_STATS_CREATE,
+ LPROCFS_TYPE_LATENCY & (~cntr_umask), "create");
+ lprocfs_counter_init(stats, LPROC_OFD_STATS_STATFS,
+ LPROCFS_TYPE_LATENCY & (~cntr_umask), "statfs");
+ lprocfs_counter_init(stats, LPROC_OFD_STATS_GET_INFO,
+ LPROCFS_TYPE_LATENCY & (~cntr_umask), "get_info");
+ lprocfs_counter_init(stats, LPROC_OFD_STATS_SET_INFO,
+ LPROCFS_TYPE_LATENCY & (~cntr_umask), "set_info");
+ lprocfs_counter_init(stats, LPROC_OFD_STATS_QUOTACTL,
+ LPROCFS_TYPE_LATENCY & (~cntr_umask), "quotactl");
+ lprocfs_counter_init(stats, LPROC_OFD_STATS_PREALLOC,
+ LPROCFS_TYPE_LATENCY & (~cntr_umask), "prealloc");
}
-static int ofd_per_nid_stats_seq_show(struct seq_file *seq, void *v)
+LPROC_SEQ_FOPS(lprocfs_nid_stats_clear);
+
+LUSTRE_OBD_UINT_PARAM_ATTR(at_min);
+LUSTRE_OBD_UINT_PARAM_ATTR(at_max);
+LUSTRE_OBD_UINT_PARAM_ATTR(at_history);
+
+static struct attribute *ofd_attrs[] = {
+ &lustre_attr_access_log_mask.attr,
+ &lustre_attr_access_log_size.attr,
+ &lustre_attr_atime_diff.attr,
+ &lustre_attr_checksum_t10pi_enforce.attr,
+ &lustre_attr_degraded.attr,
+ &lustre_attr_eviction_count.attr,
+ &lustre_attr_fstype.attr,
+ &lustre_attr_grant_check_threshold.attr,
+ &lustre_attr_grant_compat_disable.attr,
+ &lustre_attr_grant_precreate.attr,
+ &lustre_attr_instance.attr,
+ &lustre_attr_ir_factor.attr,
+ &lustre_attr_job_cleanup_interval.attr,
+ &lustre_attr_lfsck_speed_limit.attr,
+ &lustre_attr_no_create.attr,
+#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 20, 53, 0)
+ &lustre_attr_no_precreate.attr,
+#endif
+ &lustre_attr_num_exports.attr,
+ &lustre_attr_precreate_batch.attr,
+ &lustre_attr_recovery_time_hard.attr,
+ &lustre_attr_recovery_time_soft.attr,
+ &lustre_attr_seqs_allocated.attr,
+ &lustre_attr_tot_dirty.attr,
+ &lustre_attr_tot_granted.attr,
+ &lustre_attr_tot_pending.attr,
+ &lustre_attr_soft_sync_limit.attr,
+ &lustre_attr_sync_journal.attr,
+#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 16, 53, 0)
+ &lustre_attr_sync_on_lock_cancel.attr,
+#endif
+ &lustre_attr_at_min.attr,
+ &lustre_attr_at_max.attr,
+ &lustre_attr_at_history.attr,
+ NULL,
+};
+
+KOBJ_ATTRIBUTE_GROUPS(ofd); /* creates ofd_groups from ofd_attrs */
+
+/**
+ * Initialize all needed procfs entries for OFD device.
+ *
+ * \param[in] ofd OFD device
+ *
+ * \retval 0 if successful
+ * \retval negative value on error
+ */
+int ofd_tunables_init(struct ofd_device *ofd)
{
- nid_stat_t *stat = seq->private;
+ struct obd_device *obd = ofd_obd(ofd);
+ struct proc_dir_entry *entry;
+ int rc = 0;
+
+ ENTRY;
+ /* lprocfs must be setup before the ofd so state can be safely added
+ * to /proc incrementally as the ofd is setup
+ */
+ obd->obd_ktype.default_groups = KOBJ_ATTR_GROUPS(ofd);
+ obd->obd_vars = lprocfs_ofd_obd_vars;
+ rc = lprocfs_obd_setup(obd, false);
+ if (rc) {
+ CERROR("%s: lprocfs_obd_setup failed: %d.\n",
+ obd->obd_name, rc);
+ RETURN(rc);
+ }
+ ldebugfs_add_vars(obd->obd_debugfs_entry, ldebugfs_ofd_obd_vars, obd);
- if (stat->nid_brw_stats)
- brw_stats_show(seq, stat->nid_brw_stats);
+ rc = tgt_tunables_init(&ofd->ofd_lut);
+ if (rc) {
+ CERROR("%s: tgt_tunables_init failed: rc = %d\n",
+ obd->obd_name, rc);
+ GOTO(obd_cleanup, rc);
+ }
- return 0;
-}
+ rc = lprocfs_alloc_obd_stats(obd, LPROC_OFD_STATS_LAST);
+ if (rc) {
+ CERROR("%s: lprocfs_alloc_obd_stats failed: %d.\n",
+ obd->obd_name, rc);
+ GOTO(tgt_cleanup, rc);
+ }
-static ssize_t ofd_per_nid_stats_seq_write(struct file *file, const char *buf,
- size_t len, loff_t *off)
-{
- struct seq_file *seq = file->private_data;
- nid_stat_t *stat = seq->private;
- int i;
+ obd->obd_debugfs_gss_dir = debugfs_create_dir("gss",
+ obd->obd_debugfs_entry);
+ if (obd->obd_debugfs_gss_dir)
+ ldebugfs_add_vars(obd->obd_debugfs_gss_dir,
+ ldebugfs_ofd_gss_vars, obd);
+
+ entry = lprocfs_register("exports", obd->obd_proc_entry, NULL, NULL);
+ if (IS_ERR(entry)) {
+ rc = PTR_ERR(entry);
+ CERROR("%s: error %d setting up lprocfs for %s\n",
+ obd->obd_name, rc, "exports");
+ GOTO(obd_free_stats, rc);
+ }
+ obd->obd_proc_exports_entry = entry;
+
+ entry = lprocfs_add_simple(obd->obd_proc_exports_entry, "clear",
+ obd, &lprocfs_nid_stats_clear_fops);
+ if (IS_ERR(entry)) {
+ rc = PTR_ERR(entry);
+ CERROR("%s: add proc entry 'clear' failed: %d.\n",
+ obd->obd_name, rc);
+ GOTO(obd_free_stats, rc);
+ }
- if (stat->nid_brw_stats)
- for (i = 0; i < BRW_LAST; i++)
- lprocfs_oh_clear(&stat->nid_brw_stats->hist[i]);
+ ofd_stats_counter_init(obd->obd_stats, 0, LPROCFS_CNTR_HISTOGRAM);
- return len;
-}
+ rc = lprocfs_job_stats_init(obd, LPROC_OFD_STATS_LAST,
+ ofd_stats_counter_init);
+ if (rc)
+ GOTO(obd_free_stats, rc);
-LPROC_SEQ_FOPS(ofd_per_nid_stats);
-#endif /* LPROCFS */
+ RETURN(0);
+
+obd_free_stats:
+ lprocfs_free_obd_stats(obd);
+tgt_cleanup:
+ tgt_tunables_fini(&ofd->ofd_lut);
+obd_cleanup:
+ lprocfs_obd_cleanup(obd);
+
+ return rc;
+}
+#endif /* CONFIG_PROC_FS */