X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Fobdclass%2Fllog.c;h=8a24d05dcfa3712d18b6d20359417f2d662218da;hp=932b3847b8013ddc72e271f3739bcb13bcac2f37;hb=d6bd5e9cc49b3bb9901ada503107e8b0eca44f7e;hpb=e2677595ab7fff054c5a728cb1b3e5080bd01f5e diff --git a/lustre/obdclass/llog.c b/lustre/obdclass/llog.c index 932b384..8a24d05 100644 --- a/lustre/obdclass/llog.c +++ b/lustre/obdclass/llog.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 */ @@ -27,7 +23,7 @@ * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. * - * Copyright (c) 2012, 2013, Intel Corporation. + * Copyright (c) 2012, 2017, Intel Corporation. */ /* * This file is part of Lustre, http://www.lustre.org/ @@ -47,16 +43,18 @@ #define DEBUG_SUBSYSTEM S_LOG - -#include +#include +#include +#include #include +#include +#include #include "llog_internal.h" - /* * Allocate a new log or catalog handle * Used inside llog_open(). */ -struct llog_handle *llog_alloc_handle(void) +static struct llog_handle *llog_alloc_handle(void) { struct llog_handle *loghandle; @@ -65,9 +63,10 @@ struct llog_handle *llog_alloc_handle(void) return NULL; init_rwsem(&loghandle->lgh_lock); - spin_lock_init(&loghandle->lgh_hdr_lock); + mutex_init(&loghandle->lgh_hdr_mutex); + init_rwsem(&loghandle->lgh_last_sem); INIT_LIST_HEAD(&loghandle->u.phd.phd_entry); - atomic_set(&loghandle->lgh_refcount, 1); + refcount_set(&loghandle->lgh_refcount, 1); return loghandle; } @@ -75,103 +74,273 @@ struct llog_handle *llog_alloc_handle(void) /* * Free llog handle and header data if exists. Used in llog_close() only */ -void llog_free_handle(struct llog_handle *loghandle) +static void llog_free_handle(struct llog_handle *loghandle) { LASSERT(loghandle != NULL); /* failed llog_init_handle */ - if (!loghandle->lgh_hdr) + if (loghandle->lgh_hdr == NULL) goto out; if (loghandle->lgh_hdr->llh_flags & LLOG_F_IS_PLAIN) LASSERT(list_empty(&loghandle->u.phd.phd_entry)); else if (loghandle->lgh_hdr->llh_flags & LLOG_F_IS_CAT) LASSERT(list_empty(&loghandle->u.chd.chd_head)); - LASSERT(sizeof(*(loghandle->lgh_hdr)) == LLOG_CHUNK_SIZE); - OBD_FREE(loghandle->lgh_hdr, LLOG_CHUNK_SIZE); + OBD_FREE_LARGE(loghandle->lgh_hdr, loghandle->lgh_hdr_size); out: OBD_FREE_PTR(loghandle); } -void llog_handle_get(struct llog_handle *loghandle) +struct llog_handle *llog_handle_get(struct llog_handle *loghandle) { - atomic_inc(&loghandle->lgh_refcount); + if (refcount_inc_not_zero(&loghandle->lgh_refcount)) + return loghandle; + return NULL; } -void llog_handle_put(struct llog_handle *loghandle) +int llog_handle_put(const struct lu_env *env, struct llog_handle *loghandle) { - LASSERT(atomic_read(&loghandle->lgh_refcount) > 0); - if (atomic_dec_and_test(&loghandle->lgh_refcount)) + int rc = 0; + + if (refcount_dec_and_test(&loghandle->lgh_refcount)) { + struct llog_operations *lop; + + rc = llog_handle2ops(loghandle, &lop); + if (!rc) { + if (lop->lop_close) + rc = lop->lop_close(env, loghandle); + else + rc = -EOPNOTSUPP; + } llog_free_handle(loghandle); + } + return rc; +} + +static int llog_declare_destroy(const struct lu_env *env, + struct llog_handle *handle, + struct thandle *th) +{ + struct llog_operations *lop; + int rc; + + ENTRY; + + rc = llog_handle2ops(handle, &lop); + if (rc) + RETURN(rc); + if (lop->lop_declare_destroy == NULL) + RETURN(-EOPNOTSUPP); + + rc = lop->lop_declare_destroy(env, handle, th); + + RETURN(rc); +} + +int llog_trans_destroy(const struct lu_env *env, struct llog_handle *handle, + struct thandle *th) +{ + struct llog_operations *lop; + int rc; + ENTRY; + + rc = llog_handle2ops(handle, &lop); + if (rc < 0) + RETURN(rc); + if (lop->lop_destroy == NULL) + RETURN(-EOPNOTSUPP); + + LASSERT(handle->lgh_obj != NULL); + if (!llog_exist(handle)) + RETURN(0); + + rc = lop->lop_destroy(env, handle, th); + + RETURN(rc); +} + +int llog_destroy(const struct lu_env *env, struct llog_handle *handle) +{ + struct llog_operations *lop; + struct dt_device *dt; + struct thandle *th; + int rc; + + ENTRY; + + rc = llog_handle2ops(handle, &lop); + if (rc < 0) + RETURN(rc); + if (lop->lop_destroy == NULL) + RETURN(-EOPNOTSUPP); + + if (handle->lgh_obj == NULL) { + /* if lgh_obj == NULL, then it is from client side destroy */ + rc = lop->lop_destroy(env, handle, NULL); + RETURN(rc); + } + + if (!llog_exist(handle)) + RETURN(0); + + dt = lu2dt_dev(handle->lgh_obj->do_lu.lo_dev); + + if (unlikely(unlikely(dt->dd_rdonly))) + RETURN(-EROFS); + + th = dt_trans_create(env, dt); + if (IS_ERR(th)) + RETURN(PTR_ERR(th)); + + rc = llog_declare_destroy(env, handle, th); + if (rc != 0) + GOTO(out_trans, rc); + + rc = dt_trans_start_local(env, dt, th); + if (rc < 0) + GOTO(out_trans, rc); + + rc = lop->lop_destroy(env, handle, th); + +out_trans: + dt_trans_stop(env, dt, th); + + RETURN(rc); } +EXPORT_SYMBOL(llog_destroy); /* returns negative on error; 0 if success; 1 if success & log destroyed */ -int llog_cancel_rec(const struct lu_env *env, struct llog_handle *loghandle, - int index) +int llog_cancel_arr_rec(const struct lu_env *env, struct llog_handle *loghandle, + int num, int *index) { - struct llog_log_hdr *llh = loghandle->lgh_hdr; - int rc = 0; - ENTRY; + struct llog_thread_info *lgi = llog_info(env); + struct dt_device *dt; + struct llog_log_hdr *llh; + struct thandle *th; + __u32 tmp_lgc_index; + int rc, i = 0; + int rc1; + bool subtract_count = false; - CDEBUG(D_RPCTRACE, "Canceling %d in log "DOSTID"\n", - index, POSTID(&loghandle->lgh_id.lgl_oi)); + ENTRY; - if (index == 0) { - CERROR("Can't cancel index 0 which is header\n"); - RETURN(-EINVAL); - } + LASSERT(loghandle != NULL); + LASSERT(loghandle->lgh_ctxt != NULL); + LASSERT(loghandle->lgh_obj != NULL); - spin_lock(&loghandle->lgh_hdr_lock); - if (!ext2_clear_bit(index, llh->llh_bitmap)) { - spin_unlock(&loghandle->lgh_hdr_lock); - CDEBUG(D_RPCTRACE, "Catalog index %u already clear?\n", index); - RETURN(-ENOENT); + llh = loghandle->lgh_hdr; + + CDEBUG(D_RPCTRACE, "Canceling %d records, first %d in log "DFID"\n", + num, index[0], PFID(&loghandle->lgh_id.lgl_oi.oi_fid)); + + dt = lu2dt_dev(loghandle->lgh_obj->do_lu.lo_dev); + + if (unlikely(unlikely(dt->dd_rdonly))) + RETURN(0); + + th = dt_trans_create(env, dt); + if (IS_ERR(th)) + RETURN(PTR_ERR(th)); + + rc = llog_declare_write_rec(env, loghandle, &llh->llh_hdr, 0, th); + if (rc < 0) + GOTO(out_trans, rc); + + if ((llh->llh_flags & LLOG_F_ZAP_WHEN_EMPTY)) { + rc = llog_declare_destroy(env, loghandle, th); + if (rc < 0) + GOTO(out_trans, rc); } - llh->llh_count--; + th->th_wait_submit = 1; + rc = dt_trans_start_local(env, dt, th); + if (rc < 0) + GOTO(out_trans, rc); + + down_write(&loghandle->lgh_lock); + /* clear bitmap */ + mutex_lock(&loghandle->lgh_hdr_mutex); + for (i = 0; i < num; ++i) { + if (index[i] == 0) { + CERROR("Can't cancel index 0 which is header\n"); + GOTO(out_unlock, rc = -EINVAL); + } + if (!ext2_clear_bit(index[i], LLOG_HDR_BITMAP(llh))) { + CDEBUG(D_RPCTRACE, "Catalog index %u already clear?\n", + index[i]); + GOTO(out_unlock, rc = -ENOENT); + } + } + loghandle->lgh_hdr->llh_count -= num; + subtract_count = true; + + /* Since llog_process_thread use lgi_cookie, it`s better to save them + * and restore after using + */ + tmp_lgc_index = lgi->lgi_cookie.lgc_index; + /* Pass this index to llog_osd_write_rec(), which will use the index + * to only update the necesary bitmap. */ + lgi->lgi_cookie.lgc_index = index[0]; + /* update header */ + rc = llog_write_rec(env, loghandle, &llh->llh_hdr, (num != 1 ? NULL : + &lgi->lgi_cookie), LLOG_HEADER_IDX, th); + lgi->lgi_cookie.lgc_index = tmp_lgc_index; + + if (rc != 0) + GOTO(out_unlock, rc); if ((llh->llh_flags & LLOG_F_ZAP_WHEN_EMPTY) && (llh->llh_count == 1) && - (loghandle->lgh_last_idx == (LLOG_BITMAP_BYTES * 8) - 1)) { - spin_unlock(&loghandle->lgh_hdr_lock); - rc = llog_destroy(env, loghandle); + ((loghandle->lgh_last_idx == LLOG_HDR_BITMAP_SIZE(llh) - 1) || + (loghandle->u.phd.phd_cat_handle != NULL && + loghandle->u.phd.phd_cat_handle->u.chd.chd_current_log != + loghandle))) { + /* never try to destroy it again */ + llh->llh_flags &= ~LLOG_F_ZAP_WHEN_EMPTY; + rc = llog_trans_destroy(env, loghandle, th); if (rc < 0) { - CERROR("%s: can't destroy empty llog #"DOSTID - "#%08x: rc = %d\n", + /* Sigh, can not destroy the final plain llog, but + * the bitmap has been clearly, so the record can not + * be accessed anymore, let's return 0 for now, and + * the orphan will be handled by LFSCK. */ + CERROR("%s: can't destroy empty llog "DFID": rc = %d\n", loghandle->lgh_ctxt->loc_obd->obd_name, - POSTID(&loghandle->lgh_id.lgl_oi), - loghandle->lgh_id.lgl_ogen, rc); - GOTO(out_err, rc); + PFID(&loghandle->lgh_id.lgl_oi.oi_fid), rc); + GOTO(out_unlock, rc = 0); } - RETURN(LLOG_DEL_PLAIN); + rc = LLOG_DEL_PLAIN; } - spin_unlock(&loghandle->lgh_hdr_lock); - rc = llog_write(env, loghandle, &llh->llh_hdr, LLOG_HEADER_IDX); +out_unlock: + mutex_unlock(&loghandle->lgh_hdr_mutex); + up_write(&loghandle->lgh_lock); +out_trans: + rc1 = dt_trans_stop(env, dt, th); + if (rc == 0) + rc = rc1; if (rc < 0) { - CERROR("%s: fail to write header for llog #"DOSTID - "#%08x: rc = %d\n", - loghandle->lgh_ctxt->loc_obd->obd_name, - POSTID(&loghandle->lgh_id.lgl_oi), - loghandle->lgh_id.lgl_ogen, rc); - GOTO(out_err, rc); + mutex_lock(&loghandle->lgh_hdr_mutex); + if (subtract_count) + loghandle->lgh_hdr->llh_count += num; + for (i = i - 1; i >= 0; i--) + ext2_set_bit(index[i], LLOG_HDR_BITMAP(llh)); + mutex_unlock(&loghandle->lgh_hdr_mutex); } - RETURN(0); -out_err: - spin_lock(&loghandle->lgh_hdr_lock); - ext2_set_bit(index, llh->llh_bitmap); - llh->llh_count++; - spin_unlock(&loghandle->lgh_hdr_lock); - return rc; + RETURN(rc); } -EXPORT_SYMBOL(llog_cancel_rec); -static int llog_read_header(const struct lu_env *env, - struct llog_handle *handle, - struct obd_uuid *uuid) +int llog_cancel_rec(const struct lu_env *env, struct llog_handle *loghandle, + int index) +{ + return llog_cancel_arr_rec(env, loghandle, 1, &index); +} + +int llog_read_header(const struct lu_env *env, struct llog_handle *handle, + const struct obd_uuid *uuid) { struct llog_operations *lop; int rc; + ENTRY; rc = llog_handle2ops(handle, &lop); if (rc) @@ -184,35 +353,51 @@ static int llog_read_header(const struct lu_env *env, if (rc == LLOG_EEMPTY) { struct llog_log_hdr *llh = handle->lgh_hdr; + /* lrh_len should be initialized in llog_init_handle */ handle->lgh_last_idx = 0; /* header is record with index 0 */ llh->llh_count = 1; /* for the header record */ llh->llh_hdr.lrh_type = LLOG_HDR_MAGIC; - llh->llh_hdr.lrh_len = llh->llh_tail.lrt_len = LLOG_CHUNK_SIZE; - llh->llh_hdr.lrh_index = llh->llh_tail.lrt_index = 0; - llh->llh_timestamp = cfs_time_current_sec(); + LASSERT(handle->lgh_ctxt->loc_chunk_size >= + LLOG_MIN_CHUNK_SIZE); + llh->llh_hdr.lrh_len = handle->lgh_ctxt->loc_chunk_size; + llh->llh_hdr.lrh_index = 0; + llh->llh_timestamp = ktime_get_real_seconds(); if (uuid) memcpy(&llh->llh_tgtuuid, uuid, sizeof(llh->llh_tgtuuid)); llh->llh_bitmap_offset = offsetof(typeof(*llh), llh_bitmap); - ext2_set_bit(0, llh->llh_bitmap); + /* Since update llog header might also call this function, + * let's reset the bitmap to 0 here */ + memset(LLOG_HDR_BITMAP(llh), 0, llh->llh_hdr.lrh_len - + llh->llh_bitmap_offset - + sizeof(llh->llh_tail)); + ext2_set_bit(0, LLOG_HDR_BITMAP(llh)); + LLOG_HDR_TAIL(llh)->lrt_len = llh->llh_hdr.lrh_len; + LLOG_HDR_TAIL(llh)->lrt_index = llh->llh_hdr.lrh_index; rc = 0; } - return rc; + RETURN(rc); } +EXPORT_SYMBOL(llog_read_header); int llog_init_handle(const struct lu_env *env, struct llog_handle *handle, int flags, struct obd_uuid *uuid) { struct llog_log_hdr *llh; + enum llog_flag fmt = flags & LLOG_F_EXT_MASK; int rc; - + int chunk_size = handle->lgh_ctxt->loc_chunk_size; ENTRY; + LASSERT(handle->lgh_hdr == NULL); - OBD_ALLOC_PTR(llh); + LASSERT(chunk_size >= LLOG_MIN_CHUNK_SIZE); + OBD_ALLOC_LARGE(llh, chunk_size); if (llh == NULL) RETURN(-ENOMEM); + handle->lgh_hdr = llh; + handle->lgh_hdr_size = chunk_size; /* first assign flags to use llog_client_ops */ llh->llh_flags = flags; rc = llog_read_header(env, handle, uuid); @@ -252,15 +437,17 @@ int llog_init_handle(const struct lu_env *env, struct llog_handle *handle, LASSERT(list_empty(&handle->u.chd.chd_head)); INIT_LIST_HEAD(&handle->u.chd.chd_head); llh->llh_size = sizeof(struct llog_logid_rec); + llh->llh_flags |= LLOG_F_IS_FIXSIZE; } else if (!(flags & LLOG_F_IS_PLAIN)) { CERROR("%s: unknown flags: %#x (expected %#x or %#x)\n", handle->lgh_ctxt->loc_obd->obd_name, flags, LLOG_F_IS_CAT, LLOG_F_IS_PLAIN); rc = -EINVAL; } + llh->llh_flags |= fmt; out: if (rc) { - OBD_FREE_PTR(llh); + OBD_FREE_LARGE(llh, chunk_size); handle->lgh_hdr = NULL; } RETURN(rc); @@ -273,146 +460,288 @@ static int llog_process_thread(void *arg) struct llog_handle *loghandle = lpi->lpi_loghandle; struct llog_log_hdr *llh = loghandle->lgh_hdr; struct llog_process_cat_data *cd = lpi->lpi_catdata; + struct llog_thread_info *lti; char *buf; - __u64 cur_offset = LLOG_CHUNK_SIZE; - __u64 last_offset; + size_t chunk_size; + __u64 cur_offset; int rc = 0, index = 1, last_index; int saved_index = 0; int last_called_index = 0; + bool repeated = false; ENTRY; - LASSERT(llh); + if (llh == NULL) + RETURN(-EINVAL); + + lti = lpi->lpi_env == NULL ? NULL : llog_info(lpi->lpi_env); + + cur_offset = chunk_size = llh->llh_hdr.lrh_len; + /* expect chunk_size to be power of two */ + LASSERT(is_power_of_2(chunk_size)); - OBD_ALLOC(buf, LLOG_CHUNK_SIZE); - if (!buf) { - lpi->lpi_rc = -ENOMEM; + OBD_ALLOC_LARGE(buf, chunk_size); + if (buf == NULL) { + lpi->lpi_rc = -ENOMEM; RETURN(0); - } + } - if (cd != NULL) { - last_called_index = cd->lpcd_first_idx; - index = cd->lpcd_first_idx + 1; - } - if (cd != NULL && cd->lpcd_last_idx) - last_index = cd->lpcd_last_idx; - else - last_index = LLOG_BITMAP_BYTES * 8 - 1; - - while (rc == 0) { - struct llog_rec_hdr *rec; - - /* skip records not set in bitmap */ - while (index <= last_index && - !ext2_test_bit(index, llh->llh_bitmap)) - ++index; - - LASSERT(index <= last_index + 1); - if (index == last_index + 1) - break; -repeat: - CDEBUG(D_OTHER, "index: %d last_index %d\n", - index, last_index); + if (cd != NULL) { + last_called_index = cd->lpcd_first_idx; + index = cd->lpcd_first_idx + 1; + } + if (cd != NULL && cd->lpcd_last_idx) + last_index = cd->lpcd_last_idx; + else + last_index = LLOG_HDR_BITMAP_SIZE(llh) - 1; + + while (rc == 0) { + struct llog_rec_hdr *rec; + off_t chunk_offset = 0; + unsigned int buf_offset = 0; + bool partial_chunk; + int lh_last_idx; + int synced_idx = 0; + + /* skip records not set in bitmap */ + while (index <= last_index && + !ext2_test_bit(index, LLOG_HDR_BITMAP(llh))) + ++index; + + /* There are no indices prior the last_index */ + if (index > last_index) + break; + + CDEBUG(D_OTHER, "index: %d last_index %d\n", index, + last_index); - /* get the buf with our target record; avoid old garbage */ - memset(buf, 0, LLOG_CHUNK_SIZE); - last_offset = cur_offset; +repeat: + /* get the buf with our target record; avoid old garbage */ + memset(buf, 0, chunk_size); + /* the record index for outdated chunk data */ + /* it is safe to process buffer until saved lgh_last_idx */ + lh_last_idx = LLOG_HDR_TAIL(llh)->lrt_index; rc = llog_next_block(lpi->lpi_env, loghandle, &saved_index, - index, &cur_offset, buf, LLOG_CHUNK_SIZE); - if (rc) - GOTO(out, rc); + index, &cur_offset, buf, chunk_size); + if (repeated && rc) + CDEBUG(D_OTHER, "cur_offset %llu, chunk_offset %llu," + " buf_offset %u, rc = %d\n", cur_offset, + (__u64)chunk_offset, buf_offset, rc); + /* we`ve tried to reread the chunk, but there is no + * new records */ + if (rc == -EIO && repeated && (chunk_offset + buf_offset) == + cur_offset) + GOTO(out, rc = 0); + if (rc != 0) + GOTO(out, rc); + + /* NB: after llog_next_block() call the cur_offset is the + * offset of the next block after read one. + * The absolute offset of the current chunk is calculated + * from cur_offset value and stored in chunk_offset variable. + */ + if ((cur_offset & (chunk_size - 1)) != 0) { + partial_chunk = true; + chunk_offset = cur_offset & ~(chunk_size - 1); + } else { + partial_chunk = false; + chunk_offset = cur_offset - chunk_size; + } - /* NB: when rec->lrh_len is accessed it is already swabbed - * since it is used at the "end" of the loop and the rec - * swabbing is done at the beginning of the loop. */ - for (rec = (struct llog_rec_hdr *)buf; - (char *)rec < buf + LLOG_CHUNK_SIZE; - rec = (struct llog_rec_hdr *)((char *)rec + rec->lrh_len)){ + /* NB: when rec->lrh_len is accessed it is already swabbed + * since it is used at the "end" of the loop and the rec + * swabbing is done at the beginning of the loop. */ + for (rec = (struct llog_rec_hdr *)(buf + buf_offset); + (char *)rec < buf + chunk_size; + rec = llog_rec_hdr_next(rec)) { - CDEBUG(D_OTHER, "processing rec 0x%p type %#x\n", - rec, rec->lrh_type); + CDEBUG(D_OTHER, "processing rec 0x%p type %#x\n", + rec, rec->lrh_type); - if (LLOG_REC_HDR_NEEDS_SWABBING(rec)) + if (LLOG_REC_HDR_NEEDS_SWABBING(rec)) lustre_swab_llog_rec(rec); - CDEBUG(D_OTHER, "after swabbing, type=%#x idx=%d\n", - rec->lrh_type, rec->lrh_index); + CDEBUG(D_OTHER, "after swabbing, type=%#x idx=%d\n", + rec->lrh_type, rec->lrh_index); - if (rec->lrh_index == 0) { - /* probably another rec just got added? */ - if (index <= loghandle->lgh_last_idx) - GOTO(repeat, rc = 0); - GOTO(out, rc = 0); /* no more records */ + if (index == (synced_idx + 1) && + synced_idx == LLOG_HDR_TAIL(llh)->lrt_index) + GOTO(out, rc = 0); + + if (OBD_FAIL_PRECHECK(OBD_FAIL_LLOG_PROCESS_TIMEOUT) && + cfs_fail_val == (unsigned int) + (loghandle->lgh_id.lgl_oi.oi.oi_id & + 0xFFFFFFFF)) { + OBD_RACE(OBD_FAIL_LLOG_PROCESS_TIMEOUT); } - if (rec->lrh_len == 0 || - rec->lrh_len > LLOG_CHUNK_SIZE) { - CWARN("invalid length %d in llog record for " - "index %d/%d\n", rec->lrh_len, - rec->lrh_index, index); - GOTO(out, rc = -EINVAL); - } - if (rec->lrh_index < index) { - CDEBUG(D_OTHER, "skipping lrh_index %d\n", - rec->lrh_index); - continue; - } + /* the bitmap could be changed during processing + * records from the chunk. For wrapped catalog + * it means we can read deleted record and try to + * process it. Check this case and reread the chunk. + * It is safe to process to lh_last_idx, including + * lh_last_idx if it was synced. We can not do <= + * comparison, cause for wrapped catalog lgh_last_idx + * could be less than index. So we detect last index + * for processing as index == lh_last_idx+1. But when + * catalog is wrapped and full lgh_last_idx=llh_cat_idx, + * the first processing index is llh_cat_idx+1. + */ - CDEBUG(D_OTHER, - "lrh_index: %d lrh_len: %d (%d remains)\n", - rec->lrh_index, rec->lrh_len, - (int)(buf + LLOG_CHUNK_SIZE - (char *)rec)); + if ((index == lh_last_idx && synced_idx != index) || + (index == (lh_last_idx + 1) && + !(index == (llh->llh_cat_idx + 1) && + (llh->llh_flags & LLOG_F_IS_CAT))) || + (rec->lrh_index == 0 && !repeated)) { + + /* save offset inside buffer for the re-read */ + buf_offset = (char *)rec - (char *)buf; + cur_offset = chunk_offset; + repeated = true; + /* We need to be sure lgh_last_idx + * record was saved to disk + */ + down_read(&loghandle->lgh_last_sem); + synced_idx = LLOG_HDR_TAIL(llh)->lrt_index; + up_read(&loghandle->lgh_last_sem); + CDEBUG(D_OTHER, "synced_idx: %d\n", synced_idx); + goto repeat; - loghandle->lgh_cur_idx = rec->lrh_index; - loghandle->lgh_cur_offset = (char *)rec - (char *)buf + - last_offset; + } - /* if set, process the callback on this record */ - if (ext2_test_bit(index, llh->llh_bitmap)) { + repeated = false; + + if (rec->lrh_len == 0 || rec->lrh_len > chunk_size) { + CWARN("%s: invalid length %d in llog "DFID + "record for index %d/%d\n", + loghandle->lgh_ctxt->loc_obd->obd_name, + rec->lrh_len, + PFID(&loghandle->lgh_id.lgl_oi.oi_fid), + rec->lrh_index, index); + + GOTO(out, rc = -EINVAL); + } + + if (rec->lrh_index < index) { + CDEBUG(D_OTHER, "skipping lrh_index %d\n", + rec->lrh_index); + continue; + } + + if (rec->lrh_index != index) { + CERROR("%s: "DFID" Invalid record: index %u" + " but expected %u\n", + loghandle->lgh_ctxt->loc_obd->obd_name, + PFID(&loghandle->lgh_id.lgl_oi.oi_fid), + rec->lrh_index, index); + GOTO(out, rc = -ERANGE); + } + + CDEBUG(D_OTHER, + "lrh_index: %d lrh_len: %d (%d remains)\n", + rec->lrh_index, rec->lrh_len, + (int)(buf + chunk_size - (char *)rec)); + + /* lgh_cur_offset is used only at llog_test_3 */ + loghandle->lgh_cur_offset = (char *)rec - (char *)buf + + chunk_offset; + + /* if set, process the callback on this record */ + if (ext2_test_bit(index, LLOG_HDR_BITMAP(llh))) { + struct llog_cookie *lgc; + __u64 tmp_off; + int tmp_idx; + + CDEBUG(D_OTHER, "index: %d, lh_last_idx: %d " + "synced_idx: %d lgh_last_idx: %d\n", + index, lh_last_idx, synced_idx, + loghandle->lgh_last_idx); + + if (lti != NULL) { + lgc = <i->lgi_cookie; + /* store lu_env for recursive calls */ + tmp_off = lgc->lgc_offset; + tmp_idx = lgc->lgc_index; + + lgc->lgc_offset = (char *)rec - + (char *)buf + chunk_offset; + lgc->lgc_index = rec->lrh_index; + } + /* using lu_env for passing record offset to + * llog_write through various callbacks */ rc = lpi->lpi_cb(lpi->lpi_env, loghandle, rec, lpi->lpi_cbdata); last_called_index = index; + + if (lti != NULL) { + lgc->lgc_offset = tmp_off; + lgc->lgc_index = tmp_idx; + } + if (rc == LLOG_PROC_BREAK) { GOTO(out, rc); } else if (rc == LLOG_DEL_RECORD) { rc = llog_cancel_rec(lpi->lpi_env, loghandle, rec->lrh_index); - } - if (rc) - GOTO(out, rc); - } else { - CDEBUG(D_OTHER, "Skipped index %d\n", index); - } - - /* next record, still in buffer? */ - ++index; - if (index > last_index) - GOTO(out, rc = 0); - } - } + } + if (rc) + GOTO(out, rc); + /* some stupid callbacks directly cancel records + * and delete llog. Check it and stop + * processing. */ + if (loghandle->lgh_hdr == NULL || + loghandle->lgh_hdr->llh_count == 1) + GOTO(out, rc = 0); + } + /* exit if the last index is reached */ + if (index >= last_index) + GOTO(out, rc = 0); + ++index; + } + } out: - if (cd != NULL) - cd->lpcd_last_idx = last_called_index; + if (cd != NULL) + cd->lpcd_last_idx = last_called_index; if (unlikely(rc == -EIO && loghandle->lgh_obj != NULL)) { - /* something bad happened to the processing of a local - * llog file, probably I/O error or the log got corrupted.. - * to be able to finally release the log we discard any - * remaining bits in the header */ - CERROR("Local llog found corrupted\n"); - while (index <= last_index) { - if (ext2_test_bit(index, llh->llh_bitmap) != 0) - llog_cancel_rec(lpi->lpi_env, loghandle, index); - index++; + if (dt_object_remote(loghandle->lgh_obj)) { + /* If it is remote object, then -EIO might means + * disconnection or eviction, let's return -EAGAIN, + * so for update recovery log processing, it will + * retry until the umount or abort recovery, see + * lod_sub_recovery_thread() */ + CERROR("%s retry remote llog process\n", + loghandle->lgh_ctxt->loc_obd->obd_name); + rc = -EAGAIN; + } else { + /* something bad happened to the processing of a local + * llog file, probably I/O error or the log got + * corrupted to be able to finally release the log we + * discard any remaining bits in the header */ + CERROR("%s: Local llog found corrupted #"DOSTID":%x" + " %s index %d count %d\n", + loghandle->lgh_ctxt->loc_obd->obd_name, + POSTID(&loghandle->lgh_id.lgl_oi), + loghandle->lgh_id.lgl_ogen, + ((llh->llh_flags & LLOG_F_IS_CAT) ? "catalog" : + "plain"), index, llh->llh_count); + + while (index <= last_index) { + if (ext2_test_bit(index, + LLOG_HDR_BITMAP(llh)) != 0) + llog_cancel_rec(lpi->lpi_env, loghandle, + index); + index++; + } + rc = 0; } - rc = 0; } - OBD_FREE(buf, LLOG_CHUNK_SIZE); - lpi->lpi_rc = rc; - return 0; + OBD_FREE_LARGE(buf, chunk_size); + lpi->lpi_rc = rc; + return 0; } static int llog_process_thread_daemonize(void *arg) @@ -420,6 +749,23 @@ static int llog_process_thread_daemonize(void *arg) struct llog_process_info *lpi = arg; struct lu_env env; int rc; + struct nsproxy *new_ns, *curr_ns = current->nsproxy; + + task_lock(lpi->lpi_reftask); + new_ns = lpi->lpi_reftask->nsproxy; + if (curr_ns != new_ns) { + get_nsproxy(new_ns); + + current->nsproxy = new_ns; + /* XXX: we should call put_nsproxy() instead of + * atomic_dec(&ns->count) directly. But put_nsproxy() cannot be + * used outside of the kernel itself, because it calls + * free_nsproxy() which is not exported by the kernel + * (defined in kernel/nsproxy.c) */ + if (curr_ns) + atomic_dec(&curr_ns->count); + } + task_unlock(lpi->lpi_reftask); unshare_fs_struct(); @@ -441,21 +787,28 @@ int llog_process_or_fork(const struct lu_env *env, struct llog_handle *loghandle, llog_cb_t cb, void *data, void *catdata, bool fork) { - struct llog_process_info *lpi; - int rc; - - ENTRY; + struct llog_process_info *lpi; + struct llog_process_data *d = data; + struct llog_process_cat_data *cd = catdata; + int rc; - OBD_ALLOC_PTR(lpi); - if (lpi == NULL) { - CERROR("cannot alloc pointer\n"); - RETURN(-ENOMEM); - } - lpi->lpi_loghandle = loghandle; - lpi->lpi_cb = cb; - lpi->lpi_cbdata = data; - lpi->lpi_catdata = catdata; + ENTRY; + OBD_ALLOC_PTR(lpi); + if (lpi == NULL) { + CERROR("cannot alloc pointer\n"); + RETURN(-ENOMEM); + } + lpi->lpi_loghandle = loghandle; + lpi->lpi_cb = cb; + lpi->lpi_cbdata = data; + lpi->lpi_catdata = catdata; + + CDEBUG(D_OTHER, "Processing "DFID" flags 0x%03x startcat %d startidx %d first_idx %d last_idx %d\n", + PFID(&loghandle->lgh_id.lgl_oi.oi_fid), + loghandle->lgh_hdr->llh_flags, d ? d->lpd_startcat : -1, + d ? d->lpd_startidx : -1, cd ? cd->lpcd_first_idx : -1, + cd ? cd->lpcd_last_idx : -1); if (fork) { struct task_struct *task; @@ -463,6 +816,10 @@ int llog_process_or_fork(const struct lu_env *env, * init the new one in llog_process_thread_daemonize. */ lpi->lpi_env = NULL; init_completion(&lpi->lpi_completion); + /* take reference to current, so that + * llog_process_thread_daemonize() can use it to switch to + * namespace associated with current */ + lpi->lpi_reftask = current; task = kthread_run(llog_process_thread_daemonize, lpi, "llog_process_thread"); if (IS_ERR(task)) { @@ -493,6 +850,27 @@ int llog_process(const struct lu_env *env, struct llog_handle *loghandle, } EXPORT_SYMBOL(llog_process); +static inline const struct cred *llog_raise_resource(void) +{ + struct cred *cred = NULL; + + if (cap_raised(current_cap(), CAP_SYS_RESOURCE)) + return cred; + + cred = prepare_creds(); + if (!cred) + return cred; + + cap_raise(cred->cap_effective, CAP_SYS_RESOURCE); + return override_creds(cred); +} + +static inline void llog_restore_resource(const struct cred *old_cred) +{ + if (old_cred) + revert_creds(old_cred); +} + int llog_reverse_process(const struct lu_env *env, struct llog_handle *loghandle, llog_cb_t cb, void *data, void *catdata) @@ -501,36 +879,36 @@ int llog_reverse_process(const struct lu_env *env, struct llog_process_cat_data *cd = catdata; void *buf; int rc = 0, first_index = 1, index, idx; + __u32 chunk_size = llh->llh_hdr.lrh_len; ENTRY; - OBD_ALLOC(buf, LLOG_CHUNK_SIZE); - if (!buf) - RETURN(-ENOMEM); - - if (cd != NULL) - first_index = cd->lpcd_first_idx + 1; - if (cd != NULL && cd->lpcd_last_idx) - index = cd->lpcd_last_idx; - else - index = LLOG_BITMAP_BYTES * 8 - 1; - - while (rc == 0) { - struct llog_rec_hdr *rec; - struct llog_rec_tail *tail; - - /* skip records not set in bitmap */ - while (index >= first_index && - !ext2_test_bit(index, llh->llh_bitmap)) - --index; - - LASSERT(index >= first_index - 1); - if (index == first_index - 1) - break; + OBD_ALLOC_LARGE(buf, chunk_size); + if (buf == NULL) + RETURN(-ENOMEM); - /* get the buf with our target record; avoid old garbage */ - memset(buf, 0, LLOG_CHUNK_SIZE); - rc = llog_prev_block(env, loghandle, index, buf, - LLOG_CHUNK_SIZE); + if (cd != NULL) + first_index = cd->lpcd_first_idx + 1; + if (cd != NULL && cd->lpcd_last_idx) + index = cd->lpcd_last_idx; + else + index = LLOG_HDR_BITMAP_SIZE(llh) - 1; + + while (rc == 0) { + struct llog_rec_hdr *rec; + struct llog_rec_tail *tail; + + /* skip records not set in bitmap */ + while (index >= first_index && + !ext2_test_bit(index, LLOG_HDR_BITMAP(llh))) + --index; + + LASSERT(index >= first_index - 1); + if (index == first_index - 1) + break; + + /* get the buf with our target record; avoid old garbage */ + memset(buf, 0, chunk_size); + rc = llog_prev_block(env, loghandle, index, buf, chunk_size); if (rc) GOTO(out, rc); @@ -546,13 +924,13 @@ int llog_reverse_process(const struct lu_env *env, LASSERT(idx == index); tail = (void *)rec + rec->lrh_len - sizeof(*tail); - /* process records in buffer, starting where we found one */ - while ((void *)tail > buf) { + /* process records in buffer, starting where we found one */ + while ((void *)tail > buf) { if (tail->lrt_index == 0) GOTO(out, rc = 0); /* no more records */ - /* if set, process the callback on this record */ - if (ext2_test_bit(index, llh->llh_bitmap)) { + /* if set, process the callback on this record */ + if (ext2_test_bit(index, LLOG_HDR_BITMAP(llh))) { rec = (void *)tail - tail->lrt_len + sizeof(*tail); @@ -576,8 +954,8 @@ int llog_reverse_process(const struct lu_env *env, } out: - if (buf) - OBD_FREE(buf, LLOG_CHUNK_SIZE); + if (buf != NULL) + OBD_FREE_LARGE(buf, chunk_size); RETURN(rc); } EXPORT_SYMBOL(llog_reverse_process); @@ -617,8 +995,9 @@ EXPORT_SYMBOL(llog_exist); int llog_declare_create(const struct lu_env *env, struct llog_handle *loghandle, struct thandle *th) { + const struct cred *old_cred; struct llog_operations *lop; - int raised, rc; + int rc; ENTRY; @@ -628,21 +1007,18 @@ int llog_declare_create(const struct lu_env *env, if (lop->lop_declare_create == NULL) RETURN(-EOPNOTSUPP); - raised = cfs_cap_raised(CFS_CAP_SYS_RESOURCE); - if (!raised) - cfs_cap_raise(CFS_CAP_SYS_RESOURCE); + old_cred = llog_raise_resource(); rc = lop->lop_declare_create(env, loghandle, th); - if (!raised) - cfs_cap_lower(CFS_CAP_SYS_RESOURCE); + llog_restore_resource(old_cred); RETURN(rc); } -EXPORT_SYMBOL(llog_declare_create); int llog_create(const struct lu_env *env, struct llog_handle *handle, struct thandle *th) { + const struct cred *old_cred; struct llog_operations *lop; - int raised, rc; + int rc; ENTRY; @@ -652,23 +1028,20 @@ int llog_create(const struct lu_env *env, struct llog_handle *handle, if (lop->lop_create == NULL) RETURN(-EOPNOTSUPP); - raised = cfs_cap_raised(CFS_CAP_SYS_RESOURCE); - if (!raised) - cfs_cap_raise(CFS_CAP_SYS_RESOURCE); + old_cred = llog_raise_resource(); rc = lop->lop_create(env, handle, th); - if (!raised) - cfs_cap_lower(CFS_CAP_SYS_RESOURCE); + llog_restore_resource(old_cred); RETURN(rc); } -EXPORT_SYMBOL(llog_create); int llog_declare_write_rec(const struct lu_env *env, struct llog_handle *handle, struct llog_rec_hdr *rec, int idx, struct thandle *th) { + const struct cred *old_cred; struct llog_operations *lop; - int raised, rc; + int rc; ENTRY; @@ -679,63 +1052,72 @@ int llog_declare_write_rec(const struct lu_env *env, if (lop->lop_declare_write_rec == NULL) RETURN(-EOPNOTSUPP); - raised = cfs_cap_raised(CFS_CAP_SYS_RESOURCE); - if (!raised) - cfs_cap_raise(CFS_CAP_SYS_RESOURCE); + old_cred = llog_raise_resource(); rc = lop->lop_declare_write_rec(env, handle, rec, idx, th); - if (!raised) - cfs_cap_lower(CFS_CAP_SYS_RESOURCE); + llog_restore_resource(old_cred); RETURN(rc); } -EXPORT_SYMBOL(llog_declare_write_rec); int llog_write_rec(const struct lu_env *env, struct llog_handle *handle, struct llog_rec_hdr *rec, struct llog_cookie *logcookies, int idx, struct thandle *th) { + const struct cred *old_cred; struct llog_operations *lop; - int raised, rc, buflen; + int rc, buflen; ENTRY; + /* API sanity checks */ + if (handle == NULL) { + CERROR("loghandle is missed\n"); + RETURN(-EPROTO); + } else if (handle->lgh_obj == NULL) { + CERROR("loghandle %p with NULL object\n", + handle); + RETURN(-EPROTO); + } else if (th == NULL) { + CERROR("%s: missed transaction handle\n", + handle->lgh_obj->do_lu.lo_dev->ld_obd->obd_name); + RETURN(-EPROTO); + } else if (handle->lgh_hdr == NULL) { + CERROR("%s: loghandle %p with no header\n", + handle->lgh_obj->do_lu.lo_dev->ld_obd->obd_name, + handle); + RETURN(-EPROTO); + } + rc = llog_handle2ops(handle, &lop); if (rc) RETURN(rc); - LASSERT(lop); if (lop->lop_write_rec == NULL) RETURN(-EOPNOTSUPP); buflen = rec->lrh_len; LASSERT(cfs_size_round(buflen) == buflen); - raised = cfs_cap_raised(CFS_CAP_SYS_RESOURCE); - if (!raised) - cfs_cap_raise(CFS_CAP_SYS_RESOURCE); + old_cred = llog_raise_resource(); rc = lop->lop_write_rec(env, handle, rec, logcookies, idx, th); - if (!raised) - cfs_cap_lower(CFS_CAP_SYS_RESOURCE); + llog_restore_resource(old_cred); RETURN(rc); } -EXPORT_SYMBOL(llog_write_rec); int llog_add(const struct lu_env *env, struct llog_handle *lgh, struct llog_rec_hdr *rec, struct llog_cookie *logcookies, struct thandle *th) { - int raised, rc; + const struct cred *old_cred; + int rc; ENTRY; if (lgh->lgh_logops->lop_add == NULL) RETURN(-EOPNOTSUPP); - raised = cfs_cap_raised(CFS_CAP_SYS_RESOURCE); - if (!raised) - cfs_cap_raise(CFS_CAP_SYS_RESOURCE); + old_cred = llog_raise_resource(); rc = lgh->lgh_logops->lop_add(env, lgh, rec, logcookies, th); - if (!raised) - cfs_cap_lower(CFS_CAP_SYS_RESOURCE); + llog_restore_resource(old_cred); RETURN(rc); } EXPORT_SYMBOL(llog_add); @@ -743,19 +1125,17 @@ EXPORT_SYMBOL(llog_add); int llog_declare_add(const struct lu_env *env, struct llog_handle *lgh, struct llog_rec_hdr *rec, struct thandle *th) { - int raised, rc; + const struct cred *old_cred; + int rc; ENTRY; if (lgh->lgh_logops->lop_declare_add == NULL) RETURN(-EOPNOTSUPP); - raised = cfs_cap_raised(CFS_CAP_SYS_RESOURCE); - if (!raised) - cfs_cap_raise(CFS_CAP_SYS_RESOURCE); + old_cred = llog_raise_resource(); rc = lgh->lgh_logops->lop_declare_add(env, lgh, rec, th); - if (!raised) - cfs_cap_lower(CFS_CAP_SYS_RESOURCE); + llog_restore_resource(old_cred); RETURN(rc); } EXPORT_SYMBOL(llog_declare_add); @@ -785,10 +1165,22 @@ int llog_open_create(const struct lu_env *env, struct llog_ctxt *ctxt, d = lu2dt_dev((*res)->lgh_obj->do_lu.lo_dev); + if (unlikely(unlikely(d->dd_rdonly))) + RETURN(-EROFS); + th = dt_trans_create(env, d); if (IS_ERR(th)) GOTO(out, rc = PTR_ERR(th)); + /* Create update llog object synchronously, which + * happens during inialization process see + * lod_sub_prep_llog(), to make sure the update + * llog object is created before corss-MDT writing + * updates into the llog object */ + if (ctxt->loc_flags & LLOG_CTXT_FLAG_NORMAL_FID) + th->th_sync = 1; + + th->th_wait_submit = 1; rc = llog_declare_create(env, *res, th); if (rc == 0) { rc = dt_trans_start_local(env, d, th); @@ -843,7 +1235,8 @@ int llog_write(const struct lu_env *env, struct llog_handle *loghandle, { struct dt_device *dt; struct thandle *th; - int rc; + bool need_cookie; + int rc; ENTRY; @@ -853,6 +1246,9 @@ int llog_write(const struct lu_env *env, struct llog_handle *loghandle, dt = lu2dt_dev(loghandle->lgh_obj->do_lu.lo_dev); + if (unlikely(unlikely(dt->dd_rdonly))) + RETURN(-EROFS); + th = dt_trans_create(env, dt); if (IS_ERR(th)) RETURN(PTR_ERR(th)); @@ -861,12 +1257,26 @@ int llog_write(const struct lu_env *env, struct llog_handle *loghandle, if (rc) GOTO(out_trans, rc); + th->th_wait_submit = 1; rc = dt_trans_start_local(env, dt, th); if (rc) GOTO(out_trans, rc); + need_cookie = !(idx == LLOG_HEADER_IDX || idx == LLOG_NEXT_IDX); + down_write(&loghandle->lgh_lock); - rc = llog_write_rec(env, loghandle, rec, NULL, idx, th); + if (need_cookie) { + struct llog_thread_info *lti = llog_info(env); + + /* cookie comes from llog_process_thread */ + rc = llog_write_rec(env, loghandle, rec, <i->lgi_cookie, + rec->lrh_index, th); + /* upper layer didn`t pass cookie so change rc */ + rc = (rc == 1 ? 0 : rc); + } else { + rc = llog_write_rec(env, loghandle, rec, NULL, idx, th); + } + up_write(&loghandle->lgh_lock); out_trans: dt_trans_stop(env, dt, th); @@ -878,7 +1288,7 @@ int llog_open(const struct lu_env *env, struct llog_ctxt *ctxt, struct llog_handle **lgh, struct llog_logid *logid, char *name, enum llog_open_param open_param) { - int raised; + const struct cred *old_cred; int rc; ENTRY; @@ -897,12 +1307,9 @@ int llog_open(const struct lu_env *env, struct llog_ctxt *ctxt, (*lgh)->lgh_ctxt = ctxt; (*lgh)->lgh_logops = ctxt->loc_logops; - raised = cfs_cap_raised(CFS_CAP_SYS_RESOURCE); - if (!raised) - cfs_cap_raise(CFS_CAP_SYS_RESOURCE); + old_cred = llog_raise_resource(); rc = ctxt->loc_logops->lop_open(env, *lgh, logid, name, open_param); - if (!raised) - cfs_cap_lower(CFS_CAP_SYS_RESOURCE); + llog_restore_resource(old_cred); if (rc) { llog_free_handle(*lgh); *lgh = NULL; @@ -913,20 +1320,7 @@ EXPORT_SYMBOL(llog_open); int llog_close(const struct lu_env *env, struct llog_handle *loghandle) { - struct llog_operations *lop; - int rc; - - ENTRY; - - rc = llog_handle2ops(loghandle, &lop); - if (rc) - GOTO(out, rc); - if (lop->lop_close == NULL) - GOTO(out, rc = -EOPNOTSUPP); - rc = lop->lop_close(env, loghandle); -out: - llog_handle_put(loghandle); - RETURN(rc); + return llog_handle_put(env, loghandle); } EXPORT_SYMBOL(llog_close); @@ -976,7 +1370,6 @@ int llog_copy_handler(const struct lu_env *env, struct llog_handle *llh, /* Append all records */ return llog_write(env, copy_llh, rec, LLOG_NEXT_IDX); } -EXPORT_SYMBOL(llog_copy_handler); /* backup plain llog */ int llog_backup(const struct lu_env *env, struct obd_device *obd, @@ -1042,3 +1435,22 @@ out_close: RETURN(rc); } EXPORT_SYMBOL(llog_backup); + +/* Get size of llog */ +__u64 llog_size(const struct lu_env *env, struct llog_handle *llh) +{ + int rc; + struct lu_attr la; + + rc = llh->lgh_obj->do_ops->do_attr_get(env, llh->lgh_obj, &la); + if (rc) { + CERROR("%s: attr_get failed for "DFID": rc = %d\n", + llh->lgh_ctxt->loc_obd->obd_name, + PFID(&llh->lgh_id.lgl_oi.oi_fid), rc); + return 0; + } + + return la.la_size; +} +EXPORT_SYMBOL(llog_size); +