X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Fosc%2Flproc_osc.c;h=e765e48b73c9bf0749e7c1cce19ad787fba4d604;hb=1d8e1b31d547bb2fdaf2649df68256ede1f363af;hp=89db417bea5d54edee2827e3341986a486ce0d83;hpb=70e80ade90af09300396706b8910e196a7928520;p=fs%2Flustre-release.git diff --git a/lustre/osc/lproc_osc.c b/lustre/osc/lproc_osc.c index 89db417..e765e48 100644 --- a/lustre/osc/lproc_osc.c +++ b/lustre/osc/lproc_osc.c @@ -1,6 +1,4 @@ -/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*- - * vim:expandtab:shiftwidth=8:tabstop=8: - * +/* * GPL HEADER START * * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. @@ -16,8 +14,8 @@ * in the LICENSE file that accompanied this code). * * You should have received a copy of the GNU General Public License - * version 2 along with this program; If not, see [sun.com URL with a - * copy of GPLv2]. + * 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 @@ -26,8 +24,10 @@ * GPL HEADER END */ /* - * Copyright 2008 Sun Microsystems, Inc. All rights reserved + * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. + * + * Copyright (c) 2011, 2013, Intel Corporation. */ /* * This file is part of Lustre, http://www.lustre.org/ @@ -61,59 +61,19 @@ static int osc_wr_active(struct file *file, const char *buffer, { struct obd_device *dev = data; int val, rc; - + rc = lprocfs_write_helper(buffer, count, &val); if (rc) return rc; if (val < 0 || val > 1) return -ERANGE; - LPROCFS_CLIMP_CHECK(dev); /* opposite senses */ - if (dev->u.cli.cl_import->imp_deactive == val) + if (dev->u.cli.cl_import->imp_deactive == val) rc = ptlrpc_set_import_active(dev->u.cli.cl_import, val); else CDEBUG(D_CONFIG, "activate %d: ignoring repeat request\n", val); - - LPROCFS_CLIMP_EXIT(dev); - return count; -} - -static int osc_rd_max_pages_per_rpc(char *page, char **start, off_t off, - int count, int *eof, void *data) -{ - struct obd_device *dev = data; - struct client_obd *cli = &dev->u.cli; - int rc; - - client_obd_list_lock(&cli->cl_loi_list_lock); - rc = snprintf(page, count, "%d\n", cli->cl_max_pages_per_rpc); - client_obd_list_unlock(&cli->cl_loi_list_lock); - return rc; -} - -static int osc_wr_max_pages_per_rpc(struct file *file, const char *buffer, - unsigned long count, void *data) -{ - struct obd_device *dev = data; - struct client_obd *cli = &dev->u.cli; - struct obd_connect_data *ocd = &cli->cl_import->imp_connect_data; - int val, rc; - rc = lprocfs_write_helper(buffer, count, &val); - if (rc) - return rc; - - LPROCFS_CLIMP_CHECK(dev); - if (val < 1 || val > ocd->ocd_brw_size >> CFS_PAGE_SHIFT) { - LPROCFS_CLIMP_EXIT(dev); - return -ERANGE; - } - client_obd_list_lock(&cli->cl_loi_list_lock); - cli->cl_max_pages_per_rpc = val; - client_obd_list_unlock(&cli->cl_loi_list_lock); - - LPROCFS_CLIMP_EXIT(dev); return count; } @@ -152,7 +112,7 @@ static int osc_wr_max_rpcs_in_flight(struct file *file, const char *buffer, client_obd_list_lock(&cli->cl_loi_list_lock); cli->cl_max_rpcs_in_flight = val; client_obd_list_unlock(&cli->cl_loi_list_lock); - + LPROCFS_CLIMP_EXIT(dev); return count; } @@ -174,28 +134,70 @@ static int osc_rd_max_dirty_mb(char *page, char **start, off_t off, int count, } static int osc_wr_max_dirty_mb(struct file *file, const char *buffer, - unsigned long count, void *data) + unsigned long count, void *data) { - struct obd_device *dev = data; - struct client_obd *cli = &dev->u.cli; - int pages_number, mult, rc; + struct obd_device *dev = data; + struct client_obd *cli = &dev->u.cli; + int pages_number, mult, rc; + + mult = 1 << (20 - PAGE_CACHE_SHIFT); + rc = lprocfs_write_frac_helper(buffer, count, &pages_number, mult); + if (rc) + return rc; + + if (pages_number <= 0 || + pages_number > OSC_MAX_DIRTY_MB_MAX << (20 - PAGE_CACHE_SHIFT) || + pages_number > totalram_pages / 4) /* 1/4 of RAM */ + return -ERANGE; + + client_obd_list_lock(&cli->cl_loi_list_lock); + cli->cl_dirty_max = (obd_count)(pages_number << PAGE_CACHE_SHIFT); + osc_wake_cache_waiters(cli); + client_obd_list_unlock(&cli->cl_loi_list_lock); + + return count; +} - mult = 1 << (20 - CFS_PAGE_SHIFT); - rc = lprocfs_write_frac_helper(buffer, count, &pages_number, mult); - if (rc) - return rc; +static int osc_rd_cached_mb(char *page, char **start, off_t off, int count, + int *eof, void *data) +{ + struct obd_device *dev = data; + struct client_obd *cli = &dev->u.cli; + int shift = 20 - PAGE_CACHE_SHIFT; + int rc; + + rc = snprintf(page, count, + "used_mb: %d\n" + "busy_cnt: %d\n", + (cfs_atomic_read(&cli->cl_lru_in_list) + + cfs_atomic_read(&cli->cl_lru_busy)) >> shift, + cfs_atomic_read(&cli->cl_lru_busy)); + + return rc; +} - if (pages_number < 0 || - pages_number > OSC_MAX_DIRTY_MB_MAX << (20 - CFS_PAGE_SHIFT) || - pages_number > num_physpages / 4) /* 1/4 of RAM */ - return -ERANGE; +/* shrink the number of caching pages to a specific number */ +static int osc_wr_cached_mb(struct file *file, const char *buffer, + unsigned long count, void *data) +{ + struct obd_device *dev = data; + struct client_obd *cli = &dev->u.cli; + int pages_number, mult, rc; - client_obd_list_lock(&cli->cl_loi_list_lock); - cli->cl_dirty_max = (obd_count)(pages_number << CFS_PAGE_SHIFT); - osc_wake_cache_waiters(cli); - client_obd_list_unlock(&cli->cl_loi_list_lock); + mult = 1 << (20 - PAGE_CACHE_SHIFT); + buffer = lprocfs_find_named_value(buffer, "used_mb:", &count); + rc = lprocfs_write_frac_helper(buffer, count, &pages_number, mult); + if (rc) + return rc; - return count; + if (pages_number < 0) + return -ERANGE; + + rc = cfs_atomic_read(&cli->cl_lru_in_list) - pages_number; + if (rc > 0) + (void)osc_lru_shrink(cli, rc, true); + + return count; } static int osc_rd_cur_dirty_bytes(char *page, char **start, off_t off, @@ -224,101 +226,64 @@ static int osc_rd_cur_grant_bytes(char *page, char **start, off_t off, return rc; } -static int osc_rd_create_count(char *page, char **start, off_t off, int count, - int *eof, void *data) +static int osc_wr_cur_grant_bytes(struct file *file, const char *buffer, + unsigned long count, void *data) { struct obd_device *obd = data; + struct client_obd *cli = &obd->u.cli; + int rc; + __u64 val; if (obd == NULL) return 0; - return snprintf(page, count, "%d\n", - obd->u.cli.cl_oscc.oscc_grow_count); -} - -/** - * Set OSC creator's osc_creator::oscc_grow_count - * - * \param file proc file - * \param buffer buffer containing the value - * \param count buffer size - * \param data obd device - * - * \retval \a count - */ -static int osc_wr_create_count(struct file *file, const char *buffer, - unsigned long count, void *data) -{ - struct obd_device *obd = data; - int val, rc, i; + rc = lprocfs_write_u64_helper(buffer, count, &val); + if (rc) + return rc; - if (obd == NULL) + /* this is only for shrinking grant */ + client_obd_list_lock(&cli->cl_loi_list_lock); + if (val >= cli->cl_avail_grant) { + client_obd_list_unlock(&cli->cl_loi_list_lock); return 0; + } + client_obd_list_unlock(&cli->cl_loi_list_lock); - rc = lprocfs_write_helper(buffer, count, &val); + LPROCFS_CLIMP_CHECK(obd); + if (cli->cl_import->imp_state == LUSTRE_IMP_FULL) + rc = osc_shrink_grant_to_target(cli, val); + LPROCFS_CLIMP_EXIT(obd); if (rc) return rc; + return count; +} - /* The MDT ALWAYS needs to limit the precreate count to - * OST_MAX_PRECREATE, and the constant cannot be changed - * because it is a value shared between the OSC and OST - * that is the maximum possible number of objects that will - * ever be handled by MDT->OST recovery processing. - * - * If the OST ever gets a request to delete more orphans, - * this implies that something has gone badly on the MDT - * and the OST will refuse to delete so much data from the - * filesystem as a safety measure. */ - if (val < OST_MIN_PRECREATE || val > OST_MAX_PRECREATE) - return -ERANGE; - if (val > obd->u.cli.cl_oscc.oscc_max_grow_count) - return -ERANGE; - - for (i = 1; (i << 1) <= val; i <<= 1) - ; - obd->u.cli.cl_oscc.oscc_grow_count = i; +static int osc_rd_cur_lost_grant_bytes(char *page, char **start, off_t off, + int count, int *eof, void *data) +{ + struct obd_device *dev = data; + struct client_obd *cli = &dev->u.cli; + int rc; - return count; + client_obd_list_lock(&cli->cl_loi_list_lock); + rc = snprintf(page, count, "%lu\n", cli->cl_lost_grant); + client_obd_list_unlock(&cli->cl_loi_list_lock); + return rc; } -/** - * Read OSC creator's osc_creator::oscc_max_grow_count - * - * \param page buffer to hold the returning string - * \param start - * \param off - * \param count - * \param eof - * proc read function parameters, please refer to kernel - * code fs/proc/generic.c proc_file_read() - * \param data obd device - * - * \retval number of characters printed. - */ -static int osc_rd_max_create_count(char *page, char **start, off_t off, - int count, int *eof, void *data) +static int osc_rd_grant_shrink_interval(char *page, char **start, off_t off, + int count, int *eof, void *data) { struct obd_device *obd = data; if (obd == NULL) return 0; - return snprintf(page, count, "%d\n", - obd->u.cli.cl_oscc.oscc_max_grow_count); + obd->u.cli.cl_grant_shrink_interval); } -/** - * Set OSC creator's osc_creator::oscc_max_grow_count - * - * \param file proc file - * \param buffer buffer containing the value - * \param count buffer size - * \param data obd device - * - * \retval \a count - */ -static int osc_wr_max_create_count(struct file *file, const char *buffer, - unsigned long count, void *data) +static int osc_wr_grant_shrink_interval(struct file *file, const char *buffer, + unsigned long count, void *data) { struct obd_device *obd = data; int val, rc; @@ -330,43 +295,14 @@ static int osc_wr_max_create_count(struct file *file, const char *buffer, if (rc) return rc; - if (val < 0) - return -ERANGE; - if (val > OST_MAX_PRECREATE) + if (val <= 0) return -ERANGE; - if (obd->u.cli.cl_oscc.oscc_grow_count > val) - obd->u.cli.cl_oscc.oscc_grow_count = val; - - obd->u.cli.cl_oscc.oscc_max_grow_count = val; + obd->u.cli.cl_grant_shrink_interval = val; return count; } -static int osc_rd_prealloc_next_id(char *page, char **start, off_t off, - int count, int *eof, void *data) -{ - struct obd_device *obd = data; - - if (obd == NULL) - return 0; - - return snprintf(page, count, LPU64"\n", - obd->u.cli.cl_oscc.oscc_next_id); -} - -static int osc_rd_prealloc_last_id(char *page, char **start, off_t off, - int count, int *eof, void *data) -{ - struct obd_device *obd = data; - - if (obd == NULL) - return 0; - - return snprintf(page, count, LPU64"\n", - obd->u.cli.cl_oscc.oscc_last_id); -} - static int osc_rd_checksum(char *page, char **start, off_t off, int count, int *eof, void *data) { @@ -435,7 +371,7 @@ static int osc_wd_checksum_type(struct file *file, const char *buffer, if (count > sizeof(kernbuf) - 1) return -EINVAL; - if (copy_from_user(kernbuf, buffer, count)) + if (copy_from_user(kernbuf, buffer, count)) return -EFAULT; if (count > 0 && kernbuf[count - 1] == '\n') kernbuf[count - 1] = '\0'; @@ -458,7 +394,8 @@ static int osc_rd_resend_count(char *page, char **start, off_t off, int count, { struct obd_device *obd = data; - return snprintf(page, count, "%u\n", atomic_read(&obd->u.cli.cl_resends)); + return snprintf(page, count, "%u\n", + cfs_atomic_read(&obd->u.cli.cl_resends)); } static int osc_wr_resend_count(struct file *file, const char *buffer, @@ -474,14 +411,110 @@ static int osc_wr_resend_count(struct file *file, const char *buffer, if (val < 0) return -EINVAL; - atomic_set(&obd->u.cli.cl_resends, val); + cfs_atomic_set(&obd->u.cli.cl_resends, val); return count; } +static int osc_rd_contention_seconds(char *page, char **start, off_t off, + int count, int *eof, void *data) +{ + struct obd_device *obd = data; + struct osc_device *od = obd2osc_dev(obd); + + return snprintf(page, count, "%u\n", od->od_contention_time); +} + +static int osc_wr_contention_seconds(struct file *file, const char *buffer, + unsigned long count, void *data) +{ + struct obd_device *obd = data; + struct osc_device *od = obd2osc_dev(obd); + + return lprocfs_write_helper(buffer, count, &od->od_contention_time) ?: + count; +} + +static int osc_rd_lockless_truncate(char *page, char **start, off_t off, + int count, int *eof, void *data) +{ + struct obd_device *obd = data; + struct osc_device *od = obd2osc_dev(obd); + + return snprintf(page, count, "%u\n", od->od_lockless_truncate); +} + +static int osc_wr_lockless_truncate(struct file *file, const char *buffer, + unsigned long count, void *data) +{ + struct obd_device *obd = data; + struct osc_device *od = obd2osc_dev(obd); + + return lprocfs_write_helper(buffer, count, &od->od_lockless_truncate) ?: + count; +} + +static int osc_rd_destroys_in_flight(char *page, char **start, off_t off, + int count, int *eof, void *data) +{ + struct obd_device *obd = data; + return snprintf(page, count, "%u\n", + cfs_atomic_read(&obd->u.cli.cl_destroy_in_flight)); +} + +static int lprocfs_osc_wr_max_pages_per_rpc(struct file *file, + const char *buffer, unsigned long count, void *data) +{ + struct obd_device *dev = data; + struct client_obd *cli = &dev->u.cli; + struct obd_connect_data *ocd = &cli->cl_import->imp_connect_data; + int chunk_mask, rc; + __u64 val; + + rc = lprocfs_write_u64_helper(buffer, count, &val); + if (rc) + return rc; + + /* if the max_pages is specified in bytes, convert to pages */ + if (val >= ONE_MB_BRW_SIZE) + val >>= PAGE_CACHE_SHIFT; + + LPROCFS_CLIMP_CHECK(dev); + + chunk_mask = ~((1 << (cli->cl_chunkbits - PAGE_CACHE_SHIFT)) - 1); + /* max_pages_per_rpc must be chunk aligned */ + val = (val + ~chunk_mask) & chunk_mask; + if (val == 0 || val > ocd->ocd_brw_size >> PAGE_CACHE_SHIFT) { + LPROCFS_CLIMP_EXIT(dev); + return -ERANGE; + } + client_obd_list_lock(&cli->cl_loi_list_lock); + cli->cl_max_pages_per_rpc = val; + client_obd_list_unlock(&cli->cl_loi_list_lock); + + LPROCFS_CLIMP_EXIT(dev); + return count; +} + +static int osc_rd_unstable_stats(char *page, char **start, off_t off, + int count, int *eof, void *data) +{ + struct obd_device *dev = data; + struct client_obd *cli = &dev->u.cli; + int pages, mb; + + pages = cfs_atomic_read(&cli->cl_unstable_count); + mb = (pages * PAGE_CACHE_SIZE) >> 20; + + return snprintf(page, count, + "unstable_pages: %8d\n" + "unstable_mb: %8d\n", + pages, mb); +} + static struct lprocfs_vars lprocfs_osc_obd_vars[] = { { "uuid", lprocfs_rd_uuid, 0, 0 }, - { "ping", 0, lprocfs_wr_ping, 0 }, + { "ping", 0, lprocfs_wr_ping, 0, 0, 0222 }, { "connect_flags", lprocfs_rd_connect_flags, 0, 0 }, { "blocksize", lprocfs_rd_blksize, 0, 0 }, { "kbytestotal", lprocfs_rd_kbytestotal, 0, 0 }, @@ -492,24 +525,35 @@ static struct lprocfs_vars lprocfs_osc_obd_vars[] = { //{ "filegroups", lprocfs_rd_filegroups, 0, 0 }, { "ost_server_uuid", lprocfs_rd_server_uuid, 0, 0 }, { "ost_conn_uuid", lprocfs_rd_conn_uuid, 0, 0 }, - { "active", osc_rd_active, + { "active", osc_rd_active, osc_wr_active, 0 }, - { "max_pages_per_rpc", osc_rd_max_pages_per_rpc, - osc_wr_max_pages_per_rpc, 0 }, + { "max_pages_per_rpc", lprocfs_obd_rd_max_pages_per_rpc, + lprocfs_osc_wr_max_pages_per_rpc, 0 }, { "max_rpcs_in_flight", osc_rd_max_rpcs_in_flight, osc_wr_max_rpcs_in_flight, 0 }, + { "destroys_in_flight", osc_rd_destroys_in_flight, 0, 0 }, { "max_dirty_mb", osc_rd_max_dirty_mb, osc_wr_max_dirty_mb, 0 }, + { "osc_cached_mb", osc_rd_cached_mb, osc_wr_cached_mb, 0 }, { "cur_dirty_bytes", osc_rd_cur_dirty_bytes, 0, 0 }, - { "cur_grant_bytes", osc_rd_cur_grant_bytes, 0, 0 }, - { "create_count", osc_rd_create_count, osc_wr_create_count, 0 }, - { "max_create_count", osc_rd_max_create_count, - osc_wr_max_create_count, 0}, - { "prealloc_next_id", osc_rd_prealloc_next_id, 0, 0 }, - { "prealloc_last_id", osc_rd_prealloc_last_id, 0, 0 }, + { "cur_grant_bytes", osc_rd_cur_grant_bytes, + osc_wr_cur_grant_bytes, 0 }, + { "cur_lost_grant_bytes", osc_rd_cur_lost_grant_bytes, 0, 0}, + { "grant_shrink_interval", osc_rd_grant_shrink_interval, + osc_wr_grant_shrink_interval, 0 }, { "checksums", osc_rd_checksum, osc_wr_checksum, 0 }, { "checksum_type", osc_rd_checksum_type, osc_wd_checksum_type, 0 }, { "resend_count", osc_rd_resend_count, osc_wr_resend_count, 0}, { "timeouts", lprocfs_rd_timeouts, 0, 0 }, + { "contention_seconds", osc_rd_contention_seconds, + osc_wr_contention_seconds, 0 }, + { "lockless_truncate", osc_rd_lockless_truncate, + osc_wr_lockless_truncate, 0 }, + { "import", lprocfs_rd_import, lprocfs_wr_import, 0 }, + { "state", lprocfs_rd_state, 0, 0 }, + { "pinger_recov", lprocfs_rd_pinger_recov, + lprocfs_wr_pinger_recov, 0, 0 }, + { "unstable_stats", osc_rd_unstable_stats, 0, 0}, + { 0 } }; @@ -526,11 +570,11 @@ static int osc_rpc_stats_seq_show(struct seq_file *seq, void *v) struct obd_device *dev = seq->private; struct client_obd *cli = &dev->u.cli; unsigned long read_tot = 0, write_tot = 0, read_cum, write_cum; - int i; + int i; - do_gettimeofday(&now); + do_gettimeofday(&now); - client_obd_list_lock(&cli->cl_loi_list_lock); + client_obd_list_lock(&cli->cl_loi_list_lock); seq_printf(seq, "snapshot_time: %lu.%lu (secs.usecs)\n", now.tv_sec, now.tv_usec); @@ -539,9 +583,9 @@ static int osc_rpc_stats_seq_show(struct seq_file *seq, void *v) seq_printf(seq, "write RPCs in flight: %d\n", cli->cl_w_in_flight); seq_printf(seq, "pending write pages: %d\n", - cli->cl_pending_w_pages); + cfs_atomic_read(&cli->cl_pending_w_pages)); seq_printf(seq, "pending read pages: %d\n", - cli->cl_pending_r_pages); + cfs_atomic_read(&cli->cl_pending_r_pages)); seq_printf(seq, "\n\t\t\tread\t\t\twrite\n"); seq_printf(seq, "pages per rpc rpcs %% cum %% |"); @@ -605,7 +649,7 @@ static int osc_rpc_stats_seq_show(struct seq_file *seq, void *v) write_cum += w; seq_printf(seq, "%d:\t\t%10lu %3lu %3lu | %10lu %3lu %3lu\n", (i == 0) ? 0 : 1 << (i - 1), - r, pct(r, read_tot), pct(read_cum, read_tot), + r, pct(r, read_tot), pct(read_cum, read_tot), w, pct(w, write_tot), pct(write_cum, write_tot)); if (read_cum == read_tot && write_cum == write_tot) break; @@ -636,10 +680,49 @@ static ssize_t osc_rpc_stats_seq_write(struct file *file, const char *buf, LPROC_SEQ_FOPS(osc_rpc_stats); +static int osc_stats_seq_show(struct seq_file *seq, void *v) +{ + struct timeval now; + struct obd_device *dev = seq->private; + struct osc_stats *stats = &obd2osc_dev(dev)->od_stats; + + do_gettimeofday(&now); + + seq_printf(seq, "snapshot_time: %lu.%lu (secs.usecs)\n", + now.tv_sec, now.tv_usec); + seq_printf(seq, "lockless_write_bytes\t\t"LPU64"\n", + stats->os_lockless_writes); + seq_printf(seq, "lockless_read_bytes\t\t"LPU64"\n", + stats->os_lockless_reads); + seq_printf(seq, "lockless_truncate\t\t"LPU64"\n", + stats->os_lockless_truncates); + return 0; +} + +static ssize_t osc_stats_seq_write(struct file *file, const char *buf, + size_t len, loff_t *off) +{ + struct seq_file *seq = file->private_data; + struct obd_device *dev = seq->private; + struct osc_stats *stats = &obd2osc_dev(dev)->od_stats; + + memset(stats, 0, sizeof(*stats)); + return len; +} + +LPROC_SEQ_FOPS(osc_stats); + int lproc_osc_attach_seqstat(struct obd_device *dev) { - return lprocfs_obd_seq_create(dev, "rpc_stats", 0444, - &osc_rpc_stats_fops, dev); + int rc; + + rc = lprocfs_seq_create(dev->obd_proc_entry, "osc_stats", 0644, + &osc_stats_fops, dev); + if (rc == 0) + rc = lprocfs_obd_seq_create(dev, "rpc_stats", 0644, + &osc_rpc_stats_fops, dev); + + return rc; } void lprocfs_osc_init_vars(struct lprocfs_static_vars *lvars)