X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Fofd%2Flproc_ofd.c;h=38093a50e8dce1c21ab94806ce082a40dd3e62b6;hp=435449785ef86825de1711861640d5d213cc6d3d;hb=13834f5aeef42d3c358574ac59475c0758dce300;hpb=9761d5c52aeef31a8c3112c3fb3ec9e24b37c800 diff --git a/lustre/ofd/lproc_ofd.c b/lustre/ofd/lproc_ofd.c index 4354497..38093a5 100644 --- a/lustre/ofd/lproc_ofd.c +++ b/lustre/ofd/lproc_ofd.c @@ -15,11 +15,7 @@ * * You should have received a copy of the GNU General Public License * version 2 along with this program; If not, see - * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf - * - * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, - * CA 95054 USA or visit www.sun.com if you need additional information or - * have any questions. + * http://www.gnu.org/licenses/gpl-2.0.html * * GPL HEADER END */ @@ -27,13 +23,20 @@ * Copyright (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. * - * Copyright (c) 2012, 2013, Intel Corporation. + * Copyright (c) 2012, 2014 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 + * Author: Mikhail Pershin + * Author: Johann Lombardi + * Author: Fan Yong */ #define DEBUG_SUBSYSTEM S_CLASS @@ -41,134 +44,160 @@ #include #include #include +#include #include "ofd_internal.h" -#ifdef LPROCFS +#ifdef CONFIG_PROC_FS -static int lprocfs_ofd_rd_seqs(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 int ofd_seqs_seq_show(struct seq_file *m, void *data) { - struct obd_device *obd = (struct obd_device *)data; + struct obd_device *obd = m->private; struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev); - *eof = 1; - return snprintf(page, count, "%u\n", ofd->ofd_seq_count); + seq_printf(m, "%u\n", ofd->ofd_seq_count); + return 0; } +LPROC_SEQ_FOPS_RO(ofd_seqs); -static int lprocfs_ofd_rd_tot_dirty(char *page, char **start, off_t off, - int count, int *eof, void *data) +/** + * Show estimate of total amount of dirty data on clients. + * + * \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_tot_dirty_seq_show(struct seq_file *m, void *data) { - struct obd_device *obd = (struct obd_device *)data; + struct obd_device *obd = m->private; struct ofd_device *ofd; LASSERT(obd != NULL); ofd = ofd_dev(obd->obd_lu_dev); - *eof = 1; - return snprintf(page, count, LPU64"\n", ofd->ofd_tot_dirty); + seq_printf(m, LPU64"\n", ofd->ofd_tot_dirty); + return 0; } +LPROC_SEQ_FOPS_RO(ofd_tot_dirty); -static int lprocfs_ofd_rd_tot_granted(char *page, char **start, off_t off, - int count, int *eof, void *data) +/** + * Show total amount of space granted to clients. + * + * \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_tot_granted_seq_show(struct seq_file *m, void *data) { - struct obd_device *obd = (struct obd_device *)data; + struct obd_device *obd = m->private; struct ofd_device *ofd; LASSERT(obd != NULL); ofd = ofd_dev(obd->obd_lu_dev); - *eof = 1; - return snprintf(page, count, LPU64"\n", ofd->ofd_tot_granted); + seq_printf(m, LPU64"\n", ofd->ofd_tot_granted); + return 0; } +LPROC_SEQ_FOPS_RO(ofd_tot_granted); -static int lprocfs_ofd_rd_tot_pending(char *page, char **start, off_t off, - int count, int *eof, void *data) +/** + * Show total amount of space used by IO in progress. + * + * \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_tot_pending_seq_show(struct seq_file *m, void *data) { - struct obd_device *obd = (struct obd_device *)data; + struct obd_device *obd = m->private; struct ofd_device *ofd; LASSERT(obd != NULL); ofd = ofd_dev(obd->obd_lu_dev); - *eof = 1; - return snprintf(page, count, LPU64"\n", ofd->ofd_tot_pending); + seq_printf(m, LPU64"\n", ofd->ofd_tot_pending); + return 0; } +LPROC_SEQ_FOPS_RO(ofd_tot_pending); -static int lprocfs_ofd_rd_grant_precreate(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 int ofd_grant_precreate_seq_show(struct seq_file *m, void *data) { - struct obd_device *obd = (struct obd_device *)data; + struct obd_device *obd = m->private; LASSERT(obd != NULL); - *eof = 1; - return snprintf(page, count, "%ld\n", - obd->obd_self_export->exp_filter_data.fed_grant); + seq_printf(m, "%ld\n", + obd->obd_self_export->exp_filter_data.fed_grant); + return 0; } +LPROC_SEQ_FOPS_RO(ofd_grant_precreate); -static int lprocfs_ofd_rd_grant_ratio(char *page, char **start, off_t off, - int count, int *eof, void *data) -{ - struct obd_device *obd = (struct obd_device *)data; - struct ofd_device *ofd; - - LASSERT(obd != NULL); - ofd = ofd_dev(obd->obd_lu_dev); - *eof = 1; - return snprintf(page, count, "%d%%\n", - (int) ofd_grant_reserved(ofd, 100)); -} - -static int lprocfs_ofd_wr_grant_ratio(struct file *file, const char *buffer, - unsigned long count, void *data) -{ - struct obd_device *obd = (struct obd_device *)data; - struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev); - int val; - int rc; - - rc = lprocfs_write_helper(buffer, count, &val); - if (rc) - return rc; - - if (val > 100 || val < 0) - 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); - - spin_lock(&ofd->ofd_grant_lock); - ofd->ofd_grant_ratio = ofd_grant_ratio_conv(val); - spin_unlock(&ofd->ofd_grant_lock); - return count; -} - -static int lprocfs_ofd_rd_precreate_batch(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 int ofd_precreate_batch_seq_show(struct seq_file *m, void *data) { - struct obd_device *obd = (struct obd_device *)data; + struct obd_device *obd = m->private; struct ofd_device *ofd; LASSERT(obd != NULL); ofd = ofd_dev(obd->obd_lu_dev); - *eof = 1; - return snprintf(page, count, "%d\n", ofd->ofd_precreate_batch); + seq_printf(m, "%d\n", ofd->ofd_precreate_batch); + return 0; } -static int lprocfs_ofd_wr_precreate_batch(struct file *file, const char *buffer, - unsigned long count, 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 +ofd_precreate_batch_seq_write(struct file *file, const char __user *buffer, + size_t count, loff_t *off) { - struct obd_device *obd = (struct obd_device *)data; + struct seq_file *m = file->private_data; + struct obd_device *obd = m->private; struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev); - int val; + __s64 val; int rc; - rc = lprocfs_write_helper(buffer, count, &val); + rc = lprocfs_str_to_s64(buffer, count, &val); if (rc) return rc; - if (val < 1) + if (val < 1 || val > INT_MAX) return -EINVAL; spin_lock(&ofd->ofd_batch_lock); @@ -176,14 +205,22 @@ static int lprocfs_ofd_wr_precreate_batch(struct file *file, const char *buffer, spin_unlock(&ofd->ofd_batch_lock); return count; } +LPROC_SEQ_FOPS(ofd_precreate_batch); -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 obd_device *obd = m->private; struct ofd_device *ofd; struct ofd_seq *oseq = NULL; - int retval = 0, rc; if (obd == NULL) return 0; @@ -191,47 +228,62 @@ static int lprocfs_ofd_rd_last_id(char *page, char **start, off_t off, ofd = ofd_dev(obd->obd_lu_dev); read_lock(&ofd->ofd_seq_list_lock); - cfs_list_for_each_entry(oseq, &ofd->ofd_seq_list, os_list) { + 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) : ostid_seq(&oseq->os_oi); - rc = snprintf(page, count, DOSTID"\n", seq, - ostid_id(&oseq->os_oi)); - if (rc < 0) { - retval = rc; - break; - } - page += rc; - count -= rc; - retval += rc; + seq_printf(m, DOSTID"\n", seq, ostid_id(&oseq->os_oi)); } read_unlock(&ofd->ofd_seq_list_lock); - return retval; + return 0; } +LPROC_SEQ_FOPS_RO(ofd_last_id); -int lprocfs_ofd_rd_fmd_max_num(char *page, char **start, off_t off, - int count, int *eof, void *data) +/** + * Show maximum number of Filter Modification Data (FMD) maintained 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_fmd_max_num_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_fmd_max_num); - return rc; + seq_printf(m, "%u\n", ofd->ofd_fmd_max_num); + return 0; } -int lprocfs_ofd_wr_fmd_max_num(struct file *file, const char *buffer, - unsigned long count, void *data) +/** + * Change number of FMDs maintained by OFD. + * + * This defines how large the list of FMDs can be. + * + * \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 +ofd_fmd_max_num_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); + __s64 val; + int rc; - rc = lprocfs_write_helper(buffer, count, &val); + rc = lprocfs_str_to_s64(buffer, count, &val); if (rc) return rc; @@ -241,188 +293,333 @@ int lprocfs_ofd_wr_fmd_max_num(struct file *file, const char *buffer, ofd->ofd_fmd_max_num = val; return count; } +LPROC_SEQ_FOPS(ofd_fmd_max_num); -int lprocfs_ofd_rd_fmd_max_age(char *page, char **start, off_t off, - int count, int *eof, void *data) +/** + * Show the maximum age of FMD data in seconds. + * + * Though it is shown in seconds, it is stored internally in units + * of jiffies for efficiency. + * + * \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_fmd_max_age_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, "%ld\n", ofd->ofd_fmd_max_age / HZ); - return rc; + seq_printf(m, "%ld\n", jiffies_to_msecs(ofd->ofd_fmd_max_age) / + MSEC_PER_SEC); + return 0; } -int lprocfs_ofd_wr_fmd_max_age(struct file *file, const char *buffer, - unsigned long count, void *data) +/** + * Set the maximum age of FMD data in seconds. + * + * This defines how long FMD data stays in the FMD list. + * It is stored internally in units of jiffies for efficiency. + * + * \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 +ofd_fmd_max_age_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); - int val; + __s64 val; int rc; - rc = lprocfs_write_helper(buffer, count, &val); + rc = lprocfs_str_to_s64(buffer, count, &val); if (rc) return rc; if (val > 65536 || val < 1) return -EINVAL; - ofd->ofd_fmd_max_age = val * HZ; + ofd->ofd_fmd_max_age = msecs_to_jiffies(val * MSEC_PER_SEC); return count; } +LPROC_SEQ_FOPS(ofd_fmd_max_age); -static int lprocfs_ofd_rd_capa(char *page, char **start, off_t off, - int count, int *eof, void *data) +/** + * 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. + * + * \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_degraded_seq_show(struct seq_file *m, void *data) { - struct obd_device *obd = data; - int rc; + struct obd_device *obd = m->private; + struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev); - rc = snprintf(page, count, "capability on: %s\n", - obd->u.filter.fo_fl_oss_capa ? "oss" : ""); - return rc; + seq_printf(m, "%u\n", ofd->ofd_raid_degraded); + return 0; } -static int lprocfs_ofd_wr_capa(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] file proc file + * \param[in] buffer string which represents mode + * 1: set degraded mode + * 0: unset degraded 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 +ofd_degraded_seq_write(struct file *file, const char __user *buffer, + size_t count, loff_t *off) { - struct obd_device *obd = data; - int val, rc; + struct seq_file *m = file->private_data; + struct obd_device *obd = m->private; + struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev); + int rc; + __s64 val; - rc = lprocfs_write_helper(buffer, count, &val); + rc = lprocfs_str_to_s64(buffer, count, &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; - } - - obd->u.filter.fo_fl_oss_capa = val; - LCONSOLE_INFO("OSS %s %s fid capability.\n", obd->obd_name, - val ? "enabled" : "disabled"); + spin_lock(&ofd->ofd_flags_lock); + ofd->ofd_raid_degraded = !!val; + spin_unlock(&ofd->ofd_flags_lock); return count; } +LPROC_SEQ_FOPS(ofd_degraded); -static int lprocfs_ofd_rd_capa_count(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 int ofd_fstype_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); + struct lu_device *d; + + LASSERT(ofd->ofd_osd); + d = &ofd->ofd_osd->dd_lu_dev; + LASSERT(d->ld_type); + seq_printf(m, "%s\n", d->ld_type->ldt_name); + return 0; } +LPROC_SEQ_FOPS_RO(ofd_fstype); -int lprocfs_ofd_rd_degraded(char *page, char **start, off_t off, - int count, int *eof, 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 int ofd_syncjournal_seq_show(struct seq_file *m, void *data) { - struct obd_device *obd = data; - struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev); + struct obd_device *obd = m->private; + struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev); - return snprintf(page, count, "%u\n", ofd->ofd_raid_degraded); + seq_printf(m, "%u\n", ofd->ofd_syncjournal); + return 0; } -int lprocfs_ofd_wr_degraded(struct file *file, const char *buffer, - unsigned long count, void *data) +/** + * 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 +ofd_syncjournal_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, rc; + struct seq_file *m = file->private_data; + struct obd_device *obd = m->private; + struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev); + __s64 val; + int rc; - rc = lprocfs_write_helper(buffer, count, &val); + rc = lprocfs_str_to_s64(buffer, count, &val); if (rc) return rc; + if (val < 0) + return -EINVAL; + spin_lock(&ofd->ofd_flags_lock); - ofd->ofd_raid_degraded = !!val; + ofd->ofd_syncjournal = !!val; + ofd_slc_set(ofd); spin_unlock(&ofd->ofd_flags_lock); return count; } +LPROC_SEQ_FOPS(ofd_syncjournal); -int lprocfs_ofd_rd_fstype(char *page, char **start, off_t off, int count, - int *eof, void *data) -{ - struct obd_device *obd = data; - struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev); - struct lu_device *d; +/* This must be longer than the longest string below */ +#define SYNC_STATES_MAXLEN 16 - 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); -} - -int lprocfs_ofd_rd_syncjournal(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) { - struct obd_device *obd = data; + struct obd_device *obd = m->private; struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev); - int rc; - rc = snprintf(page, count, "%u\n", ofd->ofd_syncjournal); - return rc; + seq_printf(m, "%u\n", ofd->ofd_brw_size / ONE_MB_BRW_SIZE); + return 0; } -int lprocfs_ofd_wr_syncjournal(struct file *file, const char *buffer, - unsigned long count, 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 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); + __s64 val; + int rc; - rc = lprocfs_write_helper(buffer, count, &val); + rc = lprocfs_str_to_s64(buffer, count, &val); if (rc) return rc; if (val < 0) return -EINVAL; + val = val * ONE_MB_BRW_SIZE; + if (val <= 0 || val > DT_MAX_BRW_SIZE || + val < (1 << ofd->ofd_blockbits)) + return -ERANGE; + spin_lock(&ofd->ofd_flags_lock); - ofd->ofd_syncjournal = !!val; - ofd_slc_set(ofd); + ofd->ofd_brw_size = val; spin_unlock(&ofd->ofd_flags_lock); return count; } +LPROC_SEQ_FOPS(ofd_brw_size); 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 OFD policy for handling dirty data 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 ofd_sync_lock_cancel_seq_show(struct seq_file *m, void *data) { - struct obd_device *obd = data; + struct obd_device *obd = m->private; struct lu_target *tgt = obd->u.obt.obt_lut; - int rc; - rc = snprintf(page, count, "%s\n", - sync_on_cancel_states[tgt->lut_sync_lock_cancel]); - return rc; + seq_printf(m, "%s\n", + sync_on_cancel_states[tgt->lut_sync_lock_cancel]); + return 0; } -int lprocfs_ofd_wr_sync_lock_cancel(struct file *file, const char *buffer, - unsigned long count, void *data) +/** + * Change OFD policy for handling dirty data under a lock being cancelled. + * + * This variable defines what action OFD takes upon lock cancel + * There are three possible modes: + * 1) never - never do sync upon lock cancel. This can lead to data + * inconsistencies if both the OST and client crash while writing a file + * that is also concurrently being read by another client. In these cases, + * this may allow the file data to "rewind" to an earlier state. + * 2) blocking - do sync only if there is blocking lock, e.g. if another + * client is trying to access this same object + * 3) always - do sync always + * + * \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 +ofd_sync_lock_cancel_seq_write(struct file *file, const char __user *buffer, + size_t count, loff_t *off) { - struct obd_device *obd = data; - struct lu_target *tgt = obd->u.obt.obt_lut; - int val = -1; - int i; + struct seq_file *m = file->private_data; + struct obd_device *obd = m->private; + struct lu_target *tgt = obd->u.obt.obt_lut; + char kernbuf[SYNC_STATES_MAXLEN]; + __s64 val = -1; + int i; + + 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_SYNC_ON_CANCEL_STATES; i++) { - if (memcmp(buffer, sync_on_cancel_states[i], - strlen(sync_on_cancel_states[i])) == 0) { + if (strcmp(kernbuf, sync_on_cancel_states[i]) == 0) { val = i; break; } } - if (val == -1) { - int rc; - rc = lprocfs_write_helper(buffer, count, &val); + /* Legacy numeric codes */ + if (val == -1) { + int rc = lprocfs_str_to_s64(buffer, count, &val); if (rc) return rc; } @@ -435,27 +632,60 @@ int lprocfs_ofd_wr_sync_lock_cancel(struct file *file, const char *buffer, spin_unlock(&tgt->lut_flags_lock); return count; } +LPROC_SEQ_FOPS(ofd_sync_lock_cancel); -int lprocfs_ofd_rd_grant_compat_disable(char *page, char **start, off_t off, - int count, int *eof, void *data) +/** + * Show if grants compatibility mode is disabled. + * + * When ofd_grant_compat_disable is set, we don't grant any space to clients + * not supporting OBD_CONNECT_GRANT_PARAM. Otherwise, space granted to such + * a client is inflated since it consumes PAGE_CACHE_SIZE of grant space per + * block, (i.e. typically 4kB units), but underlaying file system might have + * block size bigger than page size, e.g. ZFS. See LU-2049 for details. + * + * \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_grant_compat_disable_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, "%u\n", ofd->ofd_grant_compat_disable); + return 0; } -int lprocfs_ofd_wr_grant_compat_disable(struct file *file, const char *buffer, - unsigned long count, void *data) +/** + * Change grant compatibility mode. + * + * Setting ofd_grant_compat_disable prohibit any space granting to clients + * not supporting OBD_CONNECT_GRANT_PARAM. See details above. + * + * \param[in] file proc file + * \param[in] buffer string which represents mode + * 1: disable compatibility mode + * 0: enable compatibility 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 +ofd_grant_compat_disable_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); + __s64 val; + int rc; - rc = lprocfs_write_helper(buffer, count, &val); + rc = lprocfs_str_to_s64(buffer, count, &val); if (rc) return rc; @@ -468,89 +698,291 @@ int lprocfs_ofd_wr_grant_compat_disable(struct file *file, const char *buffer, return count; } +LPROC_SEQ_FOPS(ofd_grant_compat_disable); -int lprocfs_ofd_rd_soft_sync_limit(char *page, char **start, off_t off, - int count, int *eof, 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 int ofd_soft_sync_limit_seq_show(struct seq_file *m, void *data) { - struct obd_device *obd = data; + struct obd_device *obd = m->private; struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev); - return lprocfs_rd_uint(page, start, off, count, eof, - &ofd->ofd_soft_sync_limit); + return lprocfs_uint_seq_show(m, &ofd->ofd_soft_sync_limit); +} + +/** + * 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 +ofd_soft_sync_limit_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 ofd_device *ofd = ofd_dev(obd->obd_lu_dev); + + return lprocfs_uint_seq_write(file, buffer, count, + (loff_t *) &ofd->ofd_soft_sync_limit); } +LPROC_SEQ_FOPS(ofd_soft_sync_limit); -int lprocfs_ofd_wr_soft_sync_limit(struct file *file, const char *buffer, - unsigned long count, 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 int ofd_lfsck_speed_limit_seq_show(struct seq_file *m, void *data) { - struct obd_device *obd = data; + struct obd_device *obd = m->private; struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev); - return lprocfs_wr_uint(file, buffer, count, &ofd->ofd_soft_sync_limit); -} - -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 }, - { "seqs_allocated", lprocfs_ofd_rd_seqs, 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 }, - { "precreate_batch", lprocfs_ofd_rd_precreate_batch, - lprocfs_ofd_wr_precreate_batch, 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 }, - { "job_cleanup_interval", lprocfs_rd_job_interval, - lprocfs_wr_job_interval, 0}, - { "soft_sync_limit", lprocfs_ofd_rd_soft_sync_limit, - lprocfs_ofd_wr_soft_sync_limit, 0}, - { 0 } -}; + return lfsck_get_speed(m, ofd->ofd_osd); +} -static struct lprocfs_vars lprocfs_ofd_module_vars[] = { - { "num_refs", lprocfs_rd_numrefs, 0, 0 }, - { 0 } -}; +/** + * 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 +ofd_lfsck_speed_limit_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 ofd_device *ofd = ofd_dev(obd->obd_lu_dev); + __s64 val; + int rc; + + rc = lprocfs_str_to_s64(buffer, count, &val); + if (rc != 0) + return rc; + + if (val < 0) + return -ERANGE; + + rc = lfsck_set_speed(ofd->ofd_osd, val); + + return rc != 0 ? rc : count; +} +LPROC_SEQ_FOPS(ofd_lfsck_speed_limit); + +/** + * 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); +} +LPROC_SEQ_FOPS_RO(ofd_lfsck_layout); -void lprocfs_ofd_init_vars(struct lprocfs_static_vars *lvars) +/** + * 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) { - lvars->module_vars = lprocfs_ofd_module_vars; - lvars->obd_vars = lprocfs_ofd_obd_vars; + struct obd_device *obd = m->private; + struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev); + + seq_printf(m, "switch: %s\ndetected: "LPU64"\nrepaired: "LPU64"\n", + ofd->ofd_lfsck_verify_pfid ? "on" : "off", + ofd->ofd_inconsistency_self_detected, + ofd->ofd_inconsistency_self_repaired); + return 0; } +/** + * 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 seq_file *m = file->private_data; + struct obd_device *obd = m->private; + struct ofd_device *ofd = ofd_dev(obd->obd_lu_dev); + __s64 val; + int rc; + + rc = lprocfs_str_to_s64(buffer, count, &val); + if (rc != 0) + return rc; + + ofd->ofd_lfsck_verify_pfid = !!val; + + return count; +} +LPROC_SEQ_FOPS(ofd_lfsck_verify_pfid); + +static int ofd_site_stats_seq_show(struct seq_file *m, void *data) +{ + struct obd_device *obd = m->private; + + return lu_site_stats_seq_print(obd->obd_lu_dev->ld_site, m); +} +LPROC_SEQ_FOPS_RO(ofd_site_stats); + +LPROC_SEQ_FOPS_RO_TYPE(ofd, uuid); +LPROC_SEQ_FOPS_RO_TYPE(ofd, blksize); +LPROC_SEQ_FOPS_RO_TYPE(ofd, kbytestotal); +LPROC_SEQ_FOPS_RO_TYPE(ofd, kbytesfree); +LPROC_SEQ_FOPS_RO_TYPE(ofd, kbytesavail); +LPROC_SEQ_FOPS_RO_TYPE(ofd, filestotal); +LPROC_SEQ_FOPS_RO_TYPE(ofd, filesfree); + +LPROC_SEQ_FOPS_RO_TYPE(ofd, recovery_status); +LPROC_SEQ_FOPS_RW_TYPE(ofd, recovery_time_soft); +LPROC_SEQ_FOPS_RW_TYPE(ofd, recovery_time_hard); +LPROC_SEQ_FOPS_WO_TYPE(ofd, evict_client); +LPROC_SEQ_FOPS_RO_TYPE(ofd, num_exports); +LPROC_SEQ_FOPS_RO_TYPE(ofd, target_instance); +LPROC_SEQ_FOPS_RW_TYPE(ofd, ir_factor); +LPROC_SEQ_FOPS_RW_TYPE(ofd, job_interval); + +struct lprocfs_vars lprocfs_ofd_obd_vars[] = { + { .name = "uuid", + .fops = &ofd_uuid_fops }, + { .name = "blocksize", + .fops = &ofd_blksize_fops }, + { .name = "kbytestotal", + .fops = &ofd_kbytestotal_fops }, + { .name = "kbytesfree", + .fops = &ofd_kbytesfree_fops }, + { .name = "kbytesavail", + .fops = &ofd_kbytesavail_fops }, + { .name = "filestotal", + .fops = &ofd_filestotal_fops }, + { .name = "filesfree", + .fops = &ofd_filesfree_fops }, + { .name = "seqs_allocated", + .fops = &ofd_seqs_fops }, + { .name = "fstype", + .fops = &ofd_fstype_fops }, + { .name = "last_id", + .fops = &ofd_last_id_fops }, + { .name = "tot_dirty", + .fops = &ofd_tot_dirty_fops }, + { .name = "tot_pending", + .fops = &ofd_tot_pending_fops }, + { .name = "tot_granted", + .fops = &ofd_tot_granted_fops }, + { .name = "grant_precreate", + .fops = &ofd_grant_precreate_fops }, + { .name = "precreate_batch", + .fops = &ofd_precreate_batch_fops }, + { .name = "recovery_status", + .fops = &ofd_recovery_status_fops }, + { .name = "recovery_time_soft", + .fops = &ofd_recovery_time_soft_fops }, + { .name = "recovery_time_hard", + .fops = &ofd_recovery_time_hard_fops }, + { .name = "evict_client", + .fops = &ofd_evict_client_fops }, + { .name = "num_exports", + .fops = &ofd_num_exports_fops }, + { .name = "degraded", + .fops = &ofd_degraded_fops }, + { .name = "sync_journal", + .fops = &ofd_syncjournal_fops }, + { .name = "brw_size", + .fops = &ofd_brw_size_fops }, + { .name = "sync_on_lock_cancel", + .fops = &ofd_sync_lock_cancel_fops }, + { .name = "instance", + .fops = &ofd_target_instance_fops }, + { .name = "ir_factor", + .fops = &ofd_ir_factor_fops }, + { .name = "grant_compat_disable", + .fops = &ofd_grant_compat_disable_fops }, + { .name = "client_cache_count", + .fops = &ofd_fmd_max_num_fops }, + { .name = "client_cache_seconds", + .fops = &ofd_fmd_max_age_fops }, + { .name = "job_cleanup_interval", + .fops = &ofd_job_interval_fops }, + { .name = "soft_sync_limit", + .fops = &ofd_soft_sync_limit_fops }, + { .name = "lfsck_speed_limit", + .fops = &ofd_lfsck_speed_limit_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 }, + { NULL } +}; + +/** + * Initialize OFD statistics counters + * + * param[in] stats statistics counters + */ void ofd_stats_counter_init(struct lprocfs_stats *stats) { LASSERT(stats && stats->ls_num >= LPROC_OFD_STATS_LAST); @@ -581,4 +1013,4 @@ void ofd_stats_counter_init(struct lprocfs_stats *stats) 0, "quotactl", "reqs"); } -#endif /* LPROCFS */ +#endif /* CONFIG_PROC_FS */