iSharkFly-Open
/
fis-gtm
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
fis-gtm/sr_port/t_end.c

1696 lines
72 KiB
C
Raw Blame History

/****************************************************************
* *
* Copyright 2001, 2012 Fidelity Information Services, Inc *
* *
* This source code contains the intellectual property *
* of its copyright holder(s), and is made available *
* under a license. If you do not know the terms of *
* the license, please stop and do not read further. *
* *
****************************************************************/
#include "mdef.h"
#include <stddef.h>
#include <signal.h> /* for VSIG_ATOMIC_T type */
#include "gtm_time.h"
#include "gtm_inet.h"
#ifdef VMS
#include <descrip.h> /* Required for gtmsource.h */
#endif
#include "cdb_sc.h"
#include "gdsroot.h"
#include "gtm_facility.h"
#include "gdskill.h"
#include "fileinfo.h"
#include "gdsbt.h"
#include "gdsblk.h"
#include "gdsfhead.h"
#include "filestruct.h"
#include "gdscc.h"
#include "gdsbml.h"
#include "ccp.h"
#include "error.h"
#include "iosp.h"
#include "interlock.h"
#include "jnl.h"
#include "buddy_list.h" /* needed for tp.h */
#include "hashtab_int4.h" /* needed for tp.h and cws_insert.h */
#include "tp.h"
#include "gdsbgtr.h"
#include "repl_msg.h"
#include "gtmsource.h"
#include "mupipbckup.h"
#include "cache.h"
#include "gt_timer.h"
#include "longset.h" /* needed for cws_insert.h */
#include "cws_insert.h"
#include "min_max.h"
#include "gtmimagename.h"
#ifdef UNIX
#include "gtmrecv.h"
#include "deferred_signal_handler.h"
#include "repl_instance.h"
#endif
/* Include prototypes */
#include "t_qread.h"
#include "t_retry.h"
#include "t_commit_cleanup.h"
#include "send_msg.h"
#include "bm_getfree.h"
#include "rc_cpt_ops.h"
#include "rel_quant.h"
#include "wcs_flu.h"
#include "mm_update.h"
#include "bg_update.h"
#include "wcs_get_space.h"
#include "wcs_timer_start.h"
#include "process_deferred_stale.h"
#include "t_end.h"
#include "add_inter.h"
#include "jnl_write_pblk.h"
#include "jnl_write_aimg_rec.h"
#include "memcoherency.h"
#include "jnl_get_checksum.h"
#include "wbox_test_init.h"
#include "have_crit.h"
#ifdef GTM_SNAPSHOT
#include "db_snapshot.h"
#endif
#include "shmpool.h"
#include "bml_status_check.h"
#include "is_proc_alive.h"
GBLREF bool rc_locked;
GBLREF unsigned char t_fail_hist[CDB_MAX_TRIES];
GBLREF cache_rec_ptr_t cr_array[((MAX_BT_DEPTH * 2) - 1) * 2]; /* Maximum number of blocks that can be in transaction */
GBLREF unsigned int cr_array_index;
GBLREF boolean_t block_saved;
GBLREF uint4 update_trans;
GBLREF cw_set_element cw_set[]; /* create write set. */
GBLREF gd_region *gv_cur_region;
GBLREF gv_namehead *gv_target;
GBLREF sgmnt_addrs *cs_addrs;
GBLREF sgmnt_data_ptr_t cs_data;
GBLREF uint4 dollar_tlevel;
GBLREF trans_num start_tn;
GBLREF unsigned int t_tries;
GBLREF uint4 t_err, process_id;
GBLREF unsigned char cw_set_depth, cw_map_depth;
GBLREF unsigned char rdfail_detail;
GBLREF jnlpool_addrs jnlpool;
GBLREF jnlpool_ctl_ptr_t jnlpool_ctl, temp_jnlpool_ctl;
GBLREF boolean_t is_updproc;
GBLREF seq_num seq_num_one;
GBLREF boolean_t mu_reorg_process;
GBLREF boolean_t unhandled_stale_timer_pop;
GBLREF jnl_format_buffer *non_tp_jfb_ptr;
GBLREF sgmnt_addrs *kip_csa;
GBLREF boolean_t need_kip_incr;
GBLREF boolean_t write_after_image;
GBLREF boolean_t is_replicator;
GBLREF seq_num seq_num_zero;
GBLREF jnl_gbls_t jgbl;
GBLREF jnl_fence_control jnl_fence_ctl;
GBLREF boolean_t gvdupsetnoop; /* if TRUE, duplicate SETs update journal but not database (except for curr_tn++) */
GBLREF boolean_t is_dollar_incr; /* valid only if gvcst_put is in the call-stack.
* is a copy of "in_gvcst_incr" just before it got reset to FALSE */
GBLREF boolean_t mu_reorg_upgrd_dwngrd_in_prog; /* TRUE if MUPIP REORG UPGRADE/DOWNGRADE is in progress */
GBLREF boolean_t mu_reorg_nosafejnl; /* TRUE if NOSAFEJNL explicitly specified */
GBLREF trans_num mu_reorg_upgrd_dwngrd_blktn; /* tn in blkhdr of current block processed by REORG UP/DOWNGRADE */
GBLREF inctn_opcode_t inctn_opcode;
GBLREF inctn_detail_t inctn_detail; /* holds detail to fill in to inctn jnl record */
GBLREF boolean_t block_is_free;
GBLREF boolean_t gv_play_duplicate_kills;
#ifdef GTM_TRIGGER
GBLREF boolean_t skip_dbtriggers; /* see gbldefs.c for description of this global */
#endif
#ifdef UNIX
GBLREF recvpool_addrs recvpool;
GBLREF int4 strm_index;
#endif
#ifdef DEBUG
GBLREF boolean_t mupip_jnl_recover;
#endif
error_def(ERR_GVKILLFAIL);
error_def(ERR_GVPUTFAIL);
error_def(ERR_JNLFILOPN);
error_def(ERR_JNLFLUSH);
error_def(ERR_JRTNULLFAIL);
error_def(ERR_NOTREPLICATED);
error_def(ERR_TEXT);
#define BLOCK_FLUSHING(x) (csa->hdr->clustered && x->flushing && !CCP_SEGMENT_STATE(cs_addrs->nl,CCST_MASK_HAVE_DIRTY_BUFFERS))
#define RESTORE_CURRTN_IF_NEEDED(csa, write_inctn, decremented_currtn) \
{ \
if (write_inctn && decremented_currtn) \
{ /* decremented curr_tn above; need to restore to original state due to the restart */ \
assert(csa->now_crit); \
if (csa->now_crit) \
{ /* need crit to update curr_tn and early_tn */ \
csa->ti->curr_tn++; \
csa->ti->early_tn++; \
} \
decremented_currtn = FALSE; \
} \
}
/* This macro isn't enclosed in parantheses to allow for optimizations */
#define VALIDATE_CYCLE(is_mm, history) \
if (history) \
{ \
for (t1 = history->h; t1->blk_num; t1++) \
{ \
if (!is_mm && (t1->cr->cycle != t1->cycle)) \
{ /* cache slot has been stolen */ \
assert(!csa->now_crit || csa->hold_onto_crit); \
status = cdb_sc_cyclefail; \
goto failed_skip_revert; \
} \
n_blks_validated++; \
} \
}
trans_num t_end(srch_hist *hist1, srch_hist *hist2, trans_num ctn)
{
srch_hist *hist;
bt_rec_ptr_t bt;
boolean_t blk_used;
cache_rec cr_save;
cache_rec_ptr_t cr;
cw_set_element *cs, *cs_top, *cs1;
enum cdb_sc status;
int int_depth, tmpi;
uint4 jnl_status;
jnl_private_control *jpc;
jnl_buffer_ptr_t jbp, jbbp; /* jbp is non-NULL if journaling, jbbp is non-NULL only if before-image journaling */
sgmnt_addrs *csa, *repl_csa;
sgmnt_data_ptr_t csd;
node_local_ptr_t cnl;
sgm_info *dummysi = NULL; /* needed as a dummy parameter for {mm,bg}_update */
srch_blk_status *t1;
trans_num valid_thru, tnque_earliest_tn, dbtn, blktn, temp_tn, epoch_tn, old_block_tn;
unsigned char cw_depth, cw_bmp_depth;
jnldata_hdr_ptr_t jnl_header;
uint4 total_jnl_rec_size, tmp_cumul_jnl_rec_len, tmp_cw_set_depth, prev_cw_set_depth;
DEBUG_ONLY(unsigned int tot_jrec_size;)
jnlpool_ctl_ptr_t jpl, tjpl;
boolean_t replication = FALSE;
# ifdef UNIX
boolean_t supplementary = FALSE; /* this variable is initialized ONLY if "replication" is TRUE. */
seq_num strm_seqno, next_strm_seqno;
# endif
boolean_t is_mm;
boolean_t read_before_image; /* TRUE if before-image journaling or online backup in progress
* This is used to read before-images of blocks whose cs->mode is gds_t_create */
boolean_t write_inctn = FALSE; /* set to TRUE in case writing an inctn record is necessary */
boolean_t decremented_currtn, retvalue, busy2free_seen, recompute_cksum, cksum_needed;
boolean_t recycled2free_seen;
boolean_t in_mu_truncate = FALSE;
blk_hdr_ptr_t old_block;
unsigned int bsiz, crindex;
jnl_tm_t save_gbl_jrec_time;
enum gds_t_mode mode;
uint4 prev_cr_array_index;
# ifdef DEBUG
boolean_t ready2signal_gvundef_lcl;
enum cdb_sc prev_status;
GTMCRYPT_ONLY(
blk_hdr_ptr_t save_old_block;
)
# endif
int n_blks_validated;
boolean_t before_image_needed, lcl_ss_in_prog = FALSE, reorg_ss_in_prog = FALSE;
boolean_t ss_need_to_restart, new_bkup_started;
# ifdef GTM_TRIGGER
uint4 cycle;
# endif
DCL_THREADGBL_ACCESS;
SETUP_THREADGBL_ACCESS;
/* Currently, the only callers of t_end with NULL histories are the update process and journal recovery when they
* are about to process a JRT_NULL record. Assert that.
*/
assert((hist1 != hist2) || (ERR_JRTNULLFAIL == t_err) && (NULL == hist1)
&& update_trans && (is_updproc || jgbl.forw_phase_recovery));
DEBUG_ONLY(
/* Store global variable ready2signal_gvundef in a local variable and reset the global right away to ensure that
* the global value does not incorrectly get carried over to the next call of "t_end".
*/
ready2signal_gvundef_lcl = TREF(ready2signal_gvundef);
TREF(ready2signal_gvundef) = FALSE;
)
csa = cs_addrs;
csd = csa->hdr;
cnl = csa->nl;
is_mm = (dba_mm == csd->acc_meth);
# ifdef GTM_TRUNCATE
in_mu_truncate = (cnl != NULL && process_id == cnl->trunc_pid);
# endif
status = cdb_sc_normal;
/* The only cases where we set csa->hold_onto_crit to TRUE are the following :
* (a) jgbl.onlnrlbk
* (b) DSE CRIT -SEIZE (and any command that follows it), DSE CHANGE -BLOCK, DSE ALL -SEIZE (and any command that follows)
* and DSE MAPS -RESTORE_ALL. Since we cannot distinguish between different DSE qualifiers, we use IS_DSE_IMAGE.
* (c) gvcst_redo_root_search in the final retry.
*
* Since we don't expect hold_onto_crit to be set by any other utility/function, the below assert is valid and is intended
* to catch cases where the field is inadvertently set to TRUE.
*/
assert(!csa->hold_onto_crit || IS_DSE_IMAGE UNIX_ONLY(|| jgbl.onlnrlbk || TREF(in_gvcst_redo_root_search)));
assert(cs_data == csd);
assert((t_tries < CDB_STAGNATE) || csa->now_crit);
assert(!dollar_tlevel);
/* whenever cw_set_depth is non-zero, ensure update_trans is also non-zero */
assert(!cw_set_depth || (UPDTRNS_DB_UPDATED_MASK == update_trans));
/* whenever cw_set_depth is zero, ensure that update_trans is FALSE except when it is a duplicate set or a duplicate kill
* or a NULL journal record.
*/
assert(cw_set_depth || !update_trans || ((ERR_GVPUTFAIL == t_err) && gvdupsetnoop)
|| (ERR_JRTNULLFAIL == t_err) || ((ERR_GVKILLFAIL == t_err) && gv_play_duplicate_kills));
assert(0 == cr_array_index);
assert(!gv_cur_region->read_only || !update_trans);
cr_array_index = 0; /* be safe and reset it in PRO even if it is not zero */
if (csd->wc_blocked || (is_mm && (csa->total_blks != csa->ti->total_blks)))
{ /* If blocked, or we have MM and file has been extended, force repair */
status = cdb_sc_helpedout; /* force retry with special status so philanthropy isn't punished */
goto failed_skip_revert;
} else
{
/* See if we can take a fast path for read transactions based on the following conditions :
* 1. If the transaction number hasn't changed since we read the blocks from the disk or cache
* 2. If NO concurrent online rollback is running. This is needed because we don't want read transactions
* to succeed. The issue with this check is that for a rollback that was killed, the PID will be non-zero.
* In that case, we might skip the fast path and go ahead and do the validation. The validation logic
* gets crit anyways and so will salvage the lock and do the necessary recovery and issue DBFLCORRP if
* it notices that csd->file_corrupt is TRUE.
*/
if (!update_trans && (start_tn == csa->ti->early_tn) UNIX_ONLY(&& (0 == csa->nl->onln_rlbk_pid)))
{ /* read with no change to the transaction history */
n_blks_validated = 0;
VALIDATE_CYCLE(is_mm, hist1); /* updates n_blks_validated */
VALIDATE_CYCLE(is_mm, hist2); /* updates n_blks_validated */
/* Assert that if gtm_gvundef_fatal is non-zero, then we better not be about to signal a GVUNDEF */
assert(!TREF(gtm_gvundef_fatal) || !ready2signal_gvundef_lcl);
if (csa->now_crit && !csa->hold_onto_crit)
rel_crit(gv_cur_region);
if (unhandled_stale_timer_pop)
process_deferred_stale();
CWS_RESET;
assert(!csa->now_crit || csa->hold_onto_crit); /* shouldn't hold crit unless asked to */
t_tries = 0; /* commit was successful so reset t_tries */
INCR_GVSTATS_COUNTER(csa, cnl, n_nontp_readonly, 1);
INCR_GVSTATS_COUNTER(csa, cnl, n_nontp_blkread, n_blks_validated);
return csa->ti->curr_tn;
}
}
busy2free_seen = FALSE;
recycled2free_seen = FALSE;
cw_depth = cw_set_depth;
cw_bmp_depth = cw_depth;
if (0 != cw_set_depth)
{
if (gds_t_writemap == cw_set[0].mode)
{
cw_depth = 0;
cw_bmp_depth = 0;
} else if ((gds_t_busy2free == cw_set[0].mode) && (gds_t_writemap == cw_set[1].mode))
{
cw_depth = 0;
busy2free_seen = TRUE;
cw_bmp_depth = 1;
} else if ((gds_t_recycled2free == cw_set[0].mode) && (gds_t_writemap == cw_set[1].mode))
{
assert(process_id == cnl->trunc_pid && in_mu_truncate); /* In phase 1 of MUPIP REORG -TRUNCATE */
cw_depth = 1;
recycled2free_seen = TRUE;
cw_bmp_depth = 1;
}
}
# ifdef GTM_SNAPSHOT
if (update_trans && SNAPSHOTS_IN_PROG(cnl))
{
/* If snapshot context is not already created, then create one now to be used by this transaction. If context
* creation failed (for instance, on snapshot file open fail), then SS_INIT_IF_NEEDED sets csa->snapshot_in_prog
* to FALSE.
*/
SS_INIT_IF_NEEDED(csa, cnl);
} else
csa->snapshot_in_prog = FALSE;
lcl_ss_in_prog = SNAPSHOTS_IN_PROG(csa); /* store in local variable to avoid pointer access */
reorg_ss_in_prog = (mu_reorg_process && lcl_ss_in_prog); /* store in local variable if both snapshots and MUPIP REORG
* are in progress */
# endif
if (0 != cw_depth)
{ /* Caution : since csa->backup_in_prog and read_before_image are initialized below
* only if (cw_depth), these variables should be used below only within an if (cw_depth).
*/
assert(SIZEOF(bsiz) == SIZEOF(old_block->bsiz));
assert(update_trans);
csa->backup_in_prog = (BACKUP_NOT_IN_PROGRESS != cnl->nbb);
jbbp = (JNL_ENABLED(csa) && csa->jnl_before_image) ? csa->jnl->jnl_buff : NULL;
read_before_image = ((NULL != jbbp) || csa->backup_in_prog || lcl_ss_in_prog);
for (cs = cw_set, cs_top = cs + cw_depth; cs < cs_top; cs++)
{
assert(0 == cs->jnl_freeaddr); /* ensure haven't missed out resetting jnl_freeaddr for any cse in
* t_write/t_create/t_write_map/t_write_root/mu_write_map [D9B11-001991] */
if (gds_t_create == cs->mode)
{
assert(0 == cs->blk_checksum);
int_depth = (int)cw_set_depth;
if (0 > (cs->blk = bm_getfree(cs->blk, &blk_used, cw_depth, cw_set, &int_depth)))
{
if (FILE_EXTENDED == cs->blk)
{
status = cdb_sc_helpedout;
assert(is_mm);
} else
{
GET_CDB_SC_CODE(cs->blk, status); /* code is set in status */
}
goto failed_skip_revert;
}
cs->was_free = !blk_used;
BEFORE_IMAGE_NEEDED(read_before_image, cs, csa, csd, cs->blk, before_image_needed);
if (!before_image_needed)
cs->old_block = NULL;
else
{
block_is_free = cs->was_free;
cs->old_block = t_qread(cs->blk, (sm_int_ptr_t)&cs->cycle, &cs->cr);
old_block = (blk_hdr_ptr_t)cs->old_block;
if (NULL == old_block)
{
status = (enum cdb_sc)rdfail_detail;
goto failed_skip_revert;
}
if (!cs->was_free && (NULL != jbbp) && (old_block->tn < jbbp->epoch_tn))
{ /* Compute CHECKSUM for writing PBLK record before getting crit.
* It is possible that we are reading a block that is actually marked free in
* the bitmap (due to concurrency issues at this point). Therefore we might be
* actually reading uninitialized block headers and in turn a bad value of
* "old_block->bsiz". Restart if we ever access a buffer whose size is greater
* than the db block size.
*/
bsiz = old_block->bsiz;
if (bsiz > csd->blk_size)
{
assert(CDB_STAGNATE > t_tries);
status = cdb_sc_lostbmlcr;
goto failed_skip_revert;
}
JNL_GET_CHECKSUM_ACQUIRED_BLK(cs, csd, csa, old_block, bsiz);
}
}
/* assert that the block that we got from bm_getfree is less than the total blocks.
* if we do not have crit in this region, then it is possible that bm_getfree can return
* a cs->blk that is >= csa->ti->total_blks (i.e. if the bitmap buffer gets recycled).
* adjust assert accordingly.
* note that checking for crit is equivalent to checking if we are in the final retry.
*/
assert((CDB_STAGNATE > t_tries) || (cs->blk < csa->ti->total_blks));
cs->mode = gds_t_acquired;
assert(GDSVCURR == cs->ondsk_blkver);
} else if (reorg_ss_in_prog && cs->was_free)
{
assert((gds_t_acquired == cs->mode) && (NULL == cs->old_block));
/* If snapshots are in progress, we might want to read the before images of the FREE blocks also.
* Since mu_swap_blk mimics a small part of t_end, it sets cse->mode to gds_t_acquired and hence
* will not read the before images of the FREE blocks in t_end. To workaround this, set
* cse->was_free to TRUE so that in t_end, this condition can be used to read the before images of
* the FREE blocks if needed.
*/
BEFORE_IMAGE_NEEDED(read_before_image, cs, csa, csd, cs->blk, before_image_needed);
if (before_image_needed)
{
block_is_free = TRUE; /* To tell t_qread that the block it's trying to read is
* actually a FREE block */
cs->old_block = t_qread(cs->blk, (sm_int_ptr_t)&cs->cycle, &cs->cr);
if (NULL == cs->old_block)
{
status = (enum cdb_sc)rdfail_detail;
goto failed_skip_revert;
}
}
}
}
}
if (update_trans && JNL_ENABLED(csa))
{ /* compute the total journal record size requirements before grab_crit.
* there is code later that will check for state changes from now to then and if so do a recomputation
*/
assert(!cw_map_depth || cw_set_depth < cw_map_depth);
tmp_cw_set_depth = cw_map_depth ? cw_map_depth : cw_set_depth;
TOTAL_NONTPJNL_REC_SIZE(total_jnl_rec_size, non_tp_jfb_ptr, csa, tmp_cw_set_depth);
/* For a non-tp update maximum journal space we may need is total size of
* 1) space for maximum CDB_CW_SET_SIZE PBLKs, that is, MAX_JNL_REC_SIZE * CDB_CW_SET_SIZE
* 2) space for a logical record itself, that is, MAX_LOGI_JNL_REC_SIZE and
* 3) overhead records (MIN_TOTAL_NONTPJNL_REC_SIZE + JNL_FILE_TAIL_PRESERVE)
* This requirement is less than the minimum autoswitchlimit size (JNL_AUTOSWITCHLIMIT_MIN) as asserted below.
* Therefore we do not need any check to issue JNLTRANS2BIG error like is being done in tp_tend.c
*/
assert((CDB_CW_SET_SIZE * MAX_JNL_REC_SIZE + MAX_LOGI_JNL_REC_SIZE +
MIN_TOTAL_NONTPJNL_REC_SIZE + JNL_FILE_TAIL_PRESERVE) <= (JNL_AUTOSWITCHLIMIT_MIN * DISK_BLOCK_SIZE));
DEBUG_ONLY(tot_jrec_size = MAX_REQD_JNL_FILE_SIZE(total_jnl_rec_size));
assert(tot_jrec_size <= csd->autoswitchlimit);
/* The SET_GBL_JREC_TIME done below should be done before any journal writing activity
* on this region's journal file. This is because all the jnl record writing routines assume
* jgbl.gbl_jrec_time is initialized appropriately.
*/
assert(!jgbl.forw_phase_recovery || jgbl.dont_reset_gbl_jrec_time);
if (!jgbl.dont_reset_gbl_jrec_time)
SET_GBL_JREC_TIME; /* initializes jgbl.gbl_jrec_time */
assert(jgbl.gbl_jrec_time);
}
block_saved = FALSE;
ESTABLISH_RET(t_ch, 0);
assert(!csa->hold_onto_crit || csa->now_crit);
if (!csa->now_crit)
{
if (update_trans)
{ /* Get more space if needed. This is done outside crit so that
* any necessary IO has a chance of occurring outside crit.
* The available space must be double-checked inside crit. */
if (!is_mm && (cnl->wc_in_free < (int4)(cw_set_depth + 1))
&& !wcs_get_space(gv_cur_region, cw_set_depth + 1, NULL))
assert(FALSE); /* wcs_get_space should have returned TRUE unconditionally in this case */
for (;;)
{
grab_crit(gv_cur_region);
if (FALSE == csd->freeze)
break;
rel_crit(gv_cur_region);
/* We are about to wait for freeze. Assert that we are not in phase2 of a bitmap free operation
* (part of an M-kill or REORG operation). The freeze must have waited for the phase2 to complete.
*/
assert((inctn_bmp_mark_free_gtm != inctn_opcode) && (inctn_bmp_mark_free_mu_reorg != inctn_opcode));
while (csd->freeze)
hiber_start(1000);
}
} else
grab_crit(gv_cur_region);
} else
{ /* We expect the process to be in its final retry as it is holding crit. The only exception is if hold_onto_crit
* is TRUE but in that case we dont expect csd->freeze to be TRUE so we dont care much about that case. The other
* exception is if this is DSE which gets crit even without being in the final retry. In that case, skip the check
* about whether we are about to update a frozen db. DSE is the only utility allowed to update frozen databases.
*/
assert((CDB_STAGNATE == t_tries) || csa->hold_onto_crit || IS_DSE_IMAGE);
if (csd->freeze && update_trans && !IS_DSE_IMAGE)
{ /* We are about to update a frozen database. This is possible in rare cases even though
* we waited for the freeze to be lifted in t_retry (see GTM-7004). Restart in this case.
*/
status = cdb_sc_needcrit;
goto failed;
}
}
/* We should never proceed to update a frozen database. Only exception is DSE */
assert(!update_trans || !csd->freeze || IS_DSE_IMAGE);
# ifdef UNIX
/* We never expect to come here with file_corrupt set to TRUE (in case of an online rollback) because
* grab_crit done above will make sure of that. The only exception is RECOVER/ROLLBACK itself coming
* here in the forward phase
*/
assert(!csd->file_corrupt || mupip_jnl_recover);
if (csa->onln_rlbk_cycle != csa->nl->onln_rlbk_cycle)
{ /* A concurrent Online Rollback occurred. Restart to be safe. */
assert(!mupip_jnl_recover);
/* Note: We don't assert that CDB_STAGNATE > t_tries because we can detect an online rollback even in the final
* retry.
*/
status = cdb_sc_onln_rlbk1;
if (csa->db_onln_rlbkd_cycle != csa->nl->db_onln_rlbkd_cycle)
status = cdb_sc_onln_rlbk2; /* database was rolled back to a different logical state */
SYNC_ONLN_RLBK_CYCLES;
goto failed;
}
/* We should never proceed to commit if the global variable - only_reset_clues_if_onln_rlbk - is TRUE AND if the prior
* retry was due to ONLINE ROLLBACK. This way, we ensure that, whoever set the global variable knows to handle ONLINE
* ROLLBACK and resets it before returning control to the application.
*/
DEBUG_ONLY(prev_status = (0 < t_tries) ? t_fail_hist[t_tries - 1] : cdb_sc_normal);
assert((cdb_sc_normal == prev_status) || ((cdb_sc_onln_rlbk1 != prev_status) && (cdb_sc_onln_rlbk2 != prev_status))
|| (!TREF(only_reset_clues_if_onln_rlbk)));
# endif
if (is_mm && ((csa->hdr != csd) || (csa->total_blks != csa->ti->total_blks)))
{ /* If MM, check if wcs_mm_recover was invoked as part of the grab_crit done above OR if
* the file has been extended. If so, restart.
*/
status = cdb_sc_helpedout; /* force retry with special status so philanthropy isn't punished */
goto failed;
}
# ifdef GTM_TRUNCATE
assert(csa->total_blks);
if (csa->ti->total_blks < csa->total_blks)
{
/* File has been truncated since this process entered t_end or last called gdsfilext. I.e., the file is smaller
* than its last known size and we might have allocated blocks beyond csa->ti->total_blks. Restart. */
assert(dba_mm != csd->acc_meth);
assert(CDB_STAGNATE > t_tries); /* On the final retry, should have crit and truncate can't happen */
status = cdb_sc_truncate;
goto failed;
}
# endif
assert(!cw_depth || update_trans);
# ifdef GTM_TRIGGER
if (!skip_dbtriggers)
{
cycle = csd->db_trigger_cycle;
if (csa->db_trigger_cycle != cycle)
{ /* the process' view of the triggers could be potentially stale. restart to be safe. */
/* On an originating instance, in addition to the run-time, utilities can collide with
* with concurrent triggers definition updates
* The following asserts verify that:
* (1) Activities on a replicating instance don't see concurrent trigger changes as update
* process is the only updater in the replicating instance. The only exception is if this
* is a supplementary root primary instance. In that case, the update process coexists with
* GT.M processes and hence can see restarts due to concurrent trigger changes.
* (2) Journal recover operates in standalone mode. So, it should NOT see any concurrent
* trigger changes as well
*/
assert(!is_updproc || jnlpool.repl_inst_filehdr->is_supplementary && !jnlpool.jnlpool_ctl->upd_disabled);
assert(!jgbl.forw_phase_recovery);
assert(cycle > csa->db_trigger_cycle);
/* csa->db_trigger_cycle will be set to csd->db_trigger_cycle in t_retry */
status = cdb_sc_triggermod;
goto failed;
}
}
# endif
/* If inctn_opcode has a valid value, then we better be doing an update. The only exception to this rule is if we are
* in MUPIP REORG UPGRADE/DOWNGRADE (mu_reorg_upgrd_dwngrd.c) where update_trans is explicitly set to 0 in some cases.
*/
assert((inctn_invalid_op == inctn_opcode) || mu_reorg_upgrd_dwngrd_in_prog || update_trans);
if (update_trans)
{
if (JNL_ALLOWED(csa))
{
if ((csa->jnl_state != csd->jnl_state) || (csa->jnl_before_image != csd->jnl_before_image))
{ /* csd->jnl_state or csd->jnl_before_image changed since last time
* csa->jnl_before_image and csa->jnl_state got set */
csa->jnl_before_image = csd->jnl_before_image;
csa->jnl_state = csd->jnl_state;
/* jnl_file_lost causes a jnl_state transition from jnl_open to jnl_closed
* and additionally causes a repl_state transition from repl_open to repl_closed
* all without standalone access. This means that csa->repl_state might be repl_open
* while csd->repl_state might be repl_closed. update csa->repl_state in this case
* as otherwise the rest of the code might look at csa->repl_state and incorrectly
* conclude replication is on and generate sequence numbers when actually no journal
* records are being generated. [C9D01-002219]
*/
csa->repl_state = csd->repl_state;
status = cdb_sc_jnlstatemod;
goto failed;
}
}
/* Flag retry, if other mupip activities like BACKUP, INTEG or FREEZE are in progress.
* If in final retry, go ahead with kill. BACKUP/INTEG/FREEZE will wait for us to be done.
*/
if (need_kip_incr && (0 < cnl->inhibit_kills) && (CDB_STAGNATE > t_tries))
{
status = cdb_sc_inhibitkills;
goto failed;
}
ss_need_to_restart = new_bkup_started = FALSE;
GTM_SNAPSHOT_ONLY(
if (update_trans)
CHK_AND_UPDATE_SNAPSHOT_STATE_IF_NEEDED(csa, cnl, ss_need_to_restart);
)
if (cw_depth)
{
assert(update_trans);
CHK_AND_UPDATE_BKUP_STATE_IF_NEEDED(cnl, csa, new_bkup_started);
/* recalculate based on the new values of snapshot_in_prog and backup_in_prog. Since read_before_image used
* only in the context of acquired blocks, recalculation should happen only for non-zero cw_depth
*/
read_before_image = ((JNL_ENABLED(csa) && csa->jnl_before_image)
|| csa->backup_in_prog
|| SNAPSHOTS_IN_PROG(csa));
}
if ((cw_depth && new_bkup_started) || (update_trans && ss_need_to_restart))
{
if (ss_need_to_restart || (new_bkup_started && !(JNL_ENABLED(csa) && csa->jnl_before_image)))
{
/* If online backup is in progress now and before-image journaling is not enabled,
* we would not have read before-images for created blocks. Although it is possible
* that this transaction might not have blocks with gds_t_create at all, we expect
* this backup_in_prog state change to be so rare that it is ok to restart.
*/
status = cdb_sc_bkupss_statemod;
goto failed;
}
}
/* in crit, ensure cache-space is available. the out-of-crit check done above might not have been enough */
if (!is_mm && (cnl->wc_in_free < (int4)(cw_set_depth + 1))
&& !wcs_get_space(gv_cur_region, cw_set_depth + 1, NULL))
{
assert(csd->wc_blocked); /* only reason we currently know why wcs_get_space could fail */
assert(gtm_white_box_test_case_enabled);
SET_TRACEABLE_VAR(csd->wc_blocked, TRUE);
BG_TRACE_PRO_ANY(csa, wc_blocked_t_end_hist);
status = cdb_sc_cacheprob;
goto failed;
}
if (inctn_invalid_op != inctn_opcode)
{
assert(cw_set_depth || mu_reorg_process);
write_inctn = TRUE; /* mupip reorg or gvcst_bmp_mark_free or extra block split in gvcstput */
decremented_currtn = FALSE;
if (jgbl.forw_phase_recovery && !JNL_ENABLED(csa))
{ /* forward recovery (deduced above from the fact that journaling is not enabled) is supposed
* to accurately simulate GT.M runtime activity for every transaction number. The way it does
* this is by incrementing transaction numbers for all inctn records that GT.M wrote and not
* incrementing transaction number for any inctn activity that forward recovery internally needs
* to do. all inctn activity done outside of t_end has already been protected against incrementing
* transaction number in case of forward recovery. t_end is a little bit tricky since in this case
* a few database blocks get modified with the current transaction number and not incrementing the
* transaction number might result in the database transaction number being lesser than the block
* transaction number. we work around this problem by decrementing the database transaction number
* just before the commit so the database block updates for the inctn transaction get the
* transaction number of the previous transaction effectively merging the inctn transaction with
* the previous transaction.
*/
/* cw_set_depth is 1 for all INCTN operations except for block free operations when it can be 2 */
assert((inctn_gvcstput_extra_blk_split == inctn_opcode)
|| (inctn_bmp_mark_free_gtm == inctn_opcode)
|| (inctn_bmp_mark_free_mu_reorg == inctn_opcode) || (1 == cw_set_depth));
csa->ti->curr_tn--;
csa->ti->early_tn--;
decremented_currtn = TRUE;
}
}
}
assert(csd == csa->hdr);
valid_thru = dbtn = csa->ti->curr_tn;
if (!is_mm)
tnque_earliest_tn = ((th_rec_ptr_t)((sm_uc_ptr_t)csa->th_base + csa->th_base->tnque.fl))->tn;
if (update_trans)
valid_thru++;
n_blks_validated = 0;
for (hist = hist1; (NULL != hist); hist = (hist == hist1) ? hist2 : NULL)
{
for (t1 = hist->h; t1->blk_num; t1++)
{
if (is_mm)
{
if (t1->tn <= ((blk_hdr_ptr_t)(t1->buffaddr))->tn)
{
assert(CDB_STAGNATE > t_tries);
status = cdb_sc_blkmod;
goto failed;
}
t1->cse = NULL; /* reset for next transaction */
} else
{
bt = bt_get(t1->blk_num);
if (NULL == bt)
{
if (t1->tn <= tnque_earliest_tn)
{
assert(CDB_STAGNATE > t_tries);
status = cdb_sc_losthist;
goto failed;
}
cr = db_csh_get(t1->blk_num);
} else
{
if (BLOCK_FLUSHING(bt))
{
assert(CDB_STAGNATE > t_tries);
status = cdb_sc_blockflush;
goto failed;
}
if (CR_NOTVALID == bt->cache_index)
cr = db_csh_get(t1->blk_num);
else
{
cr = (cache_rec_ptr_t)GDS_REL2ABS(bt->cache_index);
if (cr->blk != bt->blk)
{
assert(FALSE);
SET_TRACEABLE_VAR(csd->wc_blocked, TRUE);
BG_TRACE_PRO_ANY(csa, wc_blocked_t_end_crbtmismatch1);
status = cdb_sc_crbtmismatch;
goto failed;
}
}
assert(bt->killtn <= bt->tn);
if (t1->tn <= bt->tn)
{
assert((CDB_STAGNATE > t_tries) || (cdb_sc_helpedout == t_fail_hist[t_tries]));
assert(!IS_DOLLAR_INCREMENT || !write_inctn);
/* If the current operation is a $INCREMENT, then try to optimize by recomputing
* the update array. Do this only as long as ALL the following conditions are met.
* a) the current history's block is a leaf level block we intend to modify
* b) update is restricted to the data block
* c) no block splits are involved.
* d) this does not end up creating a new global variable tree
* e) this is not a case of $INCR about to signal a GVUNDEF
* cw_set_depth is 0 in that case
* Conveniently enough, a simple check of (1 != cw_set_depth) is enough
* to categorize conditions (c) to (e)
*
* Future optimization : It is possible that M-SETs (not just $INCR)
* that update only one data block can benefit from this optimization.
* But that has to be carefully thought out.
*/
if (!IS_DOLLAR_INCREMENT
|| t1->level || (1 != cw_set_depth) || (t1->blk_num != cw_set[0].blk))
{
status = cdb_sc_blkmod;
goto failed;
} else
{
status = gvincr_recompute_upd_array(t1, cw_set, cr);
if (cdb_sc_normal != status)
{
status = cdb_sc_blkmod;
goto failed;
}
}
}
}
if ((cache_rec_ptr_t)CR_NOTVALID == cr)
{
SET_TRACEABLE_VAR(csd->wc_blocked, TRUE);
BG_TRACE_PRO_ANY(csa, wc_blocked_t_end_hist);
status = cdb_sc_cacheprob;
goto failed;
}
if ((NULL == cr) || (cr->cycle != t1->cycle)
|| ((sm_long_t)GDS_REL2ABS(cr->buffaddr) != (sm_long_t)t1->buffaddr))
{
if ((NULL != cr) && (NULL != bt) && (cr->blk != bt->blk))
{
assert(FALSE);
SET_TRACEABLE_VAR(csd->wc_blocked, TRUE);
BG_TRACE_PRO_ANY(csa, wc_blocked_t_end_crbtmismatch2);
status = cdb_sc_crbtmismatch;
goto failed;
}
assert(CDB_STAGNATE > t_tries);
status = cdb_sc_lostcr;
goto failed;
}
assert(0 == cr->in_tend);
/* Pin those buffers that we are planning on updating. Those are easily identified as the ones
* where the history has a non-zero cw-set-element.
*/
cs = t1->cse;
if (cs)
{
if (n_gds_t_op > cs->mode && gds_t_recycled2free != cs->mode) /* don't pass histories... */
{
assert(update_trans);
assert(gds_t_busy2free > gds_t_committed);
assert(gds_t_busy2free < n_gds_t_op);
assert(gds_t_write_root > gds_t_committed);
assert(gds_t_write_root < n_gds_t_op);
assert(gds_t_recycled2free > gds_t_committed);
assert(gds_t_recycled2free < n_gds_t_op);
assert((gds_t_committed > cs->mode)
|| (busy2free_seen && (gds_t_busy2free == cs->mode))
|| (recycled2free_seen && gds_t_recycled2free == cs->mode));
PIN_CACHE_RECORD(cr, cr_array, cr_array_index);
/* If cs->mode is gds_t_busy2free, then the corresponding cache-record needs
* to be pinned to write the before-image right away but this cse is not going
* to go through bg_update. So remember to unpin the cache-record before phase2
* as otherwise the pre-phase2 check (that we have pinned only those cache-records
* that we are planning to update) will fail. But to do that, we rely on the fact
* that the cache-record corresponding to the gds_t_busy2free cse is always the
* first one in the cr_array.
*/
assert((gds_t_busy2free != cs->mode) || (1 == cr_array_index));
assert((gds_t_recycled2free != cs->mode) || (1 == cr_array_index));
}
t1->cse = NULL; /* reset for next transaction */
}
}
t1->tn = valid_thru;
n_blks_validated++;
}
assert((hist != hist2) || (t1 != hist->h));
}
# ifdef DEBUG
/* If clue is non-zero, validate it (BEFORE this could be used in a future transaction). The only exception is reorg
* where we could have an invalid clue (e.g. last_rec < first_rec etc.). This is because reorg shuffles records around
* heavily and therefore it is hard to maintain an uptodate clue. reorg therefore handles this situation by actually
* resetting the clue just before doing the next gvcst_search. The mu_reorg* routines already take care of this reset
* (in fact, this is asserted in gvcst_search too). So we can allow invalid clues here in that special case.
*/
if (!mu_reorg_process && (NULL != gv_target) && gv_target->clue.end && !in_mu_truncate)
/* gv_target can be NULL in case of DSE MAPS etc. */
DEBUG_GVT_CLUE_VALIDATE(gv_target); /* Validate that gvt has valid first_rec, clue & last_rec fields */
# endif
/* Assert that if gtm_gvundef_fatal is non-zero, then we better not be about to signal a GVUNDEF */
assert(!TREF(gtm_gvundef_fatal) || !ready2signal_gvundef_lcl);
/* check bit maps for usage */
if (0 != cw_map_depth)
{ /* Bit maps on end from mu_reorg (from a call to mu_swap_blk) or mu_reorg_upgrd_dwngrd */
prev_cw_set_depth = cw_set_depth;
prev_cr_array_index = cr_array_index; /* note down current depth of pinned cache-records */
cw_set_depth = cw_map_depth;
}
for (cs = &cw_set[cw_bmp_depth], cs_top = &cw_set[cw_set_depth]; cs < cs_top; cs++)
{
assert(0 == cs->jnl_freeaddr); /* ensure haven't missed out resetting jnl_freeaddr for any cse in
* t_write/t_create/{t,mu}_write_map/t_write_root [D9B11-001991]
*/
/* A bitmap block update will cause us to restart with "cdb_sc_bmlmod". TP transactions on the other hand
* try reallocating blocks using the function "reallocate_bitmap". That is not presently used here because
* there are cases like MUPIP REORG or MUPIP REORG UPGRADE etc. where we do not want this functionality.
* Also, non-TP restarts due to bitmap collisions are currently assumed to be negligible. Hence no
* reallocation done in the non-TP case. Reconsider if this assumption is invalidated.
*/
if (is_mm)
{
if (cs->tn <= ((blk_hdr_ptr_t)(cs->old_block))->tn)
{
assert(CDB_STAGNATE > t_tries);
status = cdb_sc_bmlmod;
goto failed;
}
} else
{
bt = bt_get(cs->blk);
if (NULL == bt)
{
if (cs->tn <= tnque_earliest_tn)
{
assert(CDB_STAGNATE > t_tries);
status = cdb_sc_lostbmlhist;
goto failed;
}
cr = db_csh_get(cs->blk);
} else
{
if (BLOCK_FLUSHING(bt))
{
assert(CDB_STAGNATE > t_tries);
status = cdb_sc_blockflush;
goto failed;
}
if (cs->tn <= bt->tn)
{
assert(CDB_STAGNATE > t_tries);
status = cdb_sc_bmlmod;
goto failed;
}
if (CR_NOTVALID == bt->cache_index)
cr = db_csh_get(cs->blk);
else
{
cr = (cache_rec_ptr_t)GDS_REL2ABS(bt->cache_index);
if (cr->blk != bt->blk)
{
assert(FALSE);
SET_TRACEABLE_VAR(csd->wc_blocked, TRUE);
BG_TRACE_PRO_ANY(csa, wc_blocked_t_end_crbtmismatch3);
status = cdb_sc_crbtmismatch;
goto failed;
}
}
}
if ((cache_rec_ptr_t)CR_NOTVALID == cr)
{
SET_TRACEABLE_VAR(csd->wc_blocked, TRUE);
BG_TRACE_PRO_ANY(csa, wc_blocked_t_end_bitmap_nullbt);
status = cdb_sc_cacheprob;
goto failed;
}
if ((NULL == cr) || (cr->cycle != cs->cycle) ||
((sm_long_t)GDS_REL2ABS(cr->buffaddr) != (sm_long_t)cs->old_block))
{
assert(CDB_STAGNATE > t_tries);
status = cdb_sc_lostbmlcr;
goto failed;
}
PIN_CACHE_RECORD(cr, cr_array, cr_array_index);
}
}
if ((0 != cw_map_depth) && mu_reorg_upgrd_dwngrd_in_prog)
{ /* Bit maps on end from mu_reorg_upgrd_dwngrd. Bitmap history has been validated.
* But we do not want bitmap cse to be considered for bg_update. Reset cw_set_depth accordingly.
*/
cw_set_depth = prev_cw_set_depth;
assert(1 >= cw_set_depth);
assert(2 >= cw_map_depth);
/* UNPIN the bitmap cache record we no longer need */
assert(prev_cr_array_index <= cr_array_index);
if (prev_cr_array_index < cr_array_index)
{
cr_array_index--;
assert(prev_cr_array_index == cr_array_index);
assert(process_id == cr_array[cr_array_index]->in_cw_set);
UNPIN_CACHE_RECORD(cr_array[cr_array_index]);
}
}
assert(csd == csa->hdr);
assert(!need_kip_incr || update_trans);
if (update_trans)
{
if (cw_depth && read_before_image && !is_mm)
{
assert(!(JNL_ENABLED(csa) && csa->jnl_before_image) || jbbp == csa->jnl->jnl_buff);
assert((JNL_ENABLED(csa) && csa->jnl_before_image) || (NULL == jbbp));
for (cs = cw_set, cs_top = cs + cw_set_depth; cs < cs_top; ++cs)
{ /* have already read old block for creates before we got crit, make sure
* cache record still has correct block. if not, reset "cse" fields to
* point to correct cache-record. this is ok to do since we only need the
* prior content of the block (for online backup or before-image journaling)
* and did not rely on it for constructing the transaction. Restart if
* block is not present in cache now or is being read in currently.
*/
if ((gds_t_acquired == cs->mode || gds_t_recycled2free == cs->mode) && (NULL != cs->old_block))
{
assert(read_before_image == ((JNL_ENABLED(csa) && csa->jnl_before_image)
|| csa->backup_in_prog
|| SNAPSHOTS_IN_PROG(csa)));
cr = db_csh_get(cs->blk);
if ((cache_rec_ptr_t)CR_NOTVALID == cr)
{
SET_TRACEABLE_VAR(csd->wc_blocked, TRUE);
BG_TRACE_PRO_ANY(csa, wc_blocked_t_end_jnl_cwset);
status = cdb_sc_cacheprob;
goto failed;
}
/* It is possible that cr->in_cw_set is non-zero in case a concurrent MUPIP REORG
* UPGRADE/DOWNGRADE is in PHASE2 touching this very same block. In that case,
* we cannot reuse this block so we restart. We could try finding a different block
* to acquire instead and avoid a restart (tracked as part of C9E11-002651).
* Note that in_cw_set is set to 0 ahead of in_tend in bg_update_phase2. Therefore
* it is possible that we see in_cw_set 0 but in_tend is still non-zero. In that case,
* we cannot proceed with pinning this cache-record as the cr is still locked by
* the other process. We can choose to wait here but instead decide to restart.
*/
if ((NULL == cr) || (0 <= cr->read_in_progress)
|| (0 != cr->in_cw_set) || (0 != cr->in_tend))
{
assert(CDB_STAGNATE > t_tries);
status = cdb_sc_lostbefor;
goto failed;
}
PIN_CACHE_RECORD(cr, cr_array, cr_array_index);
cs->ondsk_blkver = cr->ondsk_blkver;
old_block = (blk_hdr_ptr_t)GDS_REL2ABS(cr->buffaddr);
assert((cs->cr != cr) || (cs->old_block == (sm_uc_ptr_t)old_block));
old_block_tn = old_block->tn;
/* Need checksums if before imaging and if a PBLK record is going to be written. However,
* while doing the bm_getfree, if we got a free block, then no need to compute checksum
* as we would NOT be writing before images of free blocks to journal files
*/
cksum_needed = (!cs->was_free && (NULL != jbbp) && (old_block_tn < jbbp->epoch_tn));
if ((cs->cr != cr) || (cs->cycle != cr->cycle))
{ /* Block has relocated in the cache. Adjust pointers to new location. */
cs->cr = cr;
cs->cycle = cr->cycle;
cs->old_block = (sm_uc_ptr_t)old_block;
/* PBLK checksum was computed outside-of-crit when block was read but
* block has relocated in the cache since then so recompute the checksum
* if this block needs a checksum in the first place (cksum_needed is TRUE).
*/
recompute_cksum = cksum_needed;
} else if (cksum_needed)
{ /* We have determined that a checksum is needed for this block. If we have not
* previously computed one outside crit OR if the block contents have changed
* since the checksum was previously computed, we need to recompute it.
* Otherwise, the out-of-crit computed value can be safely used.
* Note that cs->tn is valid only if a checksum was computed outside of crit.
* So make sure it is used only if checksum is non-zero. There is a rare chance
* that the computed checksum could be zero in which case we will recompute
* unnecessarily. Since that is expected to be very rare, it is considered ok.
*/
recompute_cksum = (!cs->blk_checksum || (cs->tn <= old_block_tn));
}
if (!cksum_needed)
cs->blk_checksum = 0; /* zero any out-of-crit computed checksum */
else if (recompute_cksum)
{ /* We hold crit at this point so we are guaranteed valid bsiz field.
* Hence we do not need to take MIN(bsiz, csd->blk_size) like we did
* in the earlier call to jnl_get_checksum.
*/
assert(NULL != jbbp);
assert(SIZEOF(bsiz) == SIZEOF(old_block->bsiz));
bsiz = old_block->bsiz;
assert(bsiz <= csd->blk_size);
cs->blk_checksum = jnl_get_checksum((uint4*)old_block, csa, bsiz);
}
DEBUG_ONLY(
else
assert(cs->blk_checksum == jnl_get_checksum((uint4 *)old_block,
csa, old_block->bsiz));
)
assert(cs->cr->blk == cs->blk);
}
}
}
/* if we are not writing an INCTN record, we better have a non-zero cw_depth.
* the only known exceptions are
* a) if we were being called from gvcst_put for a duplicate SET
* b) if we were being called from gvcst_kill for a duplicate KILL
* c) if we were called from DSE MAPS
* d) if we were being called from gvcst_jrt_null.
* in case (a) and (b), we want to write logical SET or KILL journal records and replicate them.
* in case (c), we do not want to replicate them. we want to assert that is_replicator is FALSE in this case.
* the following assert achieves that purpose.
*/
assert((inctn_invalid_op != inctn_opcode) || cw_depth
|| !is_replicator /* exception case (c) */
|| (ERR_GVPUTFAIL == t_err) && gvdupsetnoop /* exception case (a) */
|| (ERR_JRTNULLFAIL == t_err) /* exception case (d) */
|| (ERR_GVKILLFAIL == t_err) && gv_play_duplicate_kills); /* exception case (b) */
if (REPL_ALLOWED(csa) && is_replicator && (inctn_invalid_op == inctn_opcode))
{
replication = TRUE;
jpl = jnlpool_ctl;
tjpl = temp_jnlpool_ctl;
repl_csa = &FILE_INFO(jnlpool.jnlpool_dummy_reg)->s_addrs;
/* No concurrent online rollbacks can happen at this point due to grab_crit (done above). So, the jnlpool
* should be in sync. There is one exception though. If this is GT.CM GNP Server and the last client
* disconnected, the server invokes gtcmd_rundown which in-turn invokes gds_rundown thereby running down
* all active databases at this point but leaves the journal pool up and running. Now, if an online rollback
* is attempted, it increments the onln_rlbk_cycle in the journal pool, but repl_csa->onln_rlbk_cycle is not
* synced yet. So, the grab_crit done above will NOT detect a concurrent online rollback and it doesn't need
* to because the rollback happened AFTER the rundown. So, in this case sync the repl_csa->onln_rlbk_cycle
* with the journal pool cycle (now that we have crit). Also, assert that this is the only case we know of
* for the cycles to be out-of-sync. This is all needed in DBG only, as in PRO jnlpool_ctl->onln_rlbk_cycle
* is used only by the replication servers (which GT.CM is not) and so even if it continues with an
* out-of-sync repl_csa->onln_rlbk_cycle, t_end logic does the right thing.
* Note: The below assert is better kept in the GRAB_LOCK macro but is not currently done because image_type
* variable is needed which is of type enum gtmImageTypes which is NOT available wherever GRAB_LOCK is used.
* Until we move the GRAB_LOCK macro to a separate header, keep the assert here.
*/
UNIX_ONLY(
assert((repl_csa->onln_rlbk_cycle == jnlpool_ctl->onln_rlbk_cycle) || IS_GTCM_GNP_SERVER_IMAGE);
DEBUG_ONLY(repl_csa->onln_rlbk_cycle = jnlpool_ctl->onln_rlbk_cycle);
)
if (!repl_csa->hold_onto_crit)
GRAB_LOCK(jnlpool.jnlpool_dummy_reg, ASSERT_NO_ONLINE_ROLLBACK);
assert(repl_csa->now_crit);
tjpl->write_addr = jpl->write_addr;
tjpl->write = jpl->write;
tjpl->jnl_seqno = jpl->jnl_seqno;
# ifdef UNIX
if (INVALID_SUPPL_STRM != strm_index)
{ /* Need to also update supplementary stream seqno */
supplementary = TRUE;
assert(0 <= strm_index);
/* assert(strm_index < ARRAYSIZE(tjpl->strm_seqno)); */
strm_seqno = jpl->strm_seqno[strm_index];
ASSERT_INST_FILE_HDR_HAS_HISTREC_FOR_STRM(strm_index);
} else
supplementary = FALSE;
# endif
INT8_ONLY(assert(tjpl->write == tjpl->write_addr % tjpl->jnlpool_size));
assert(jgbl.cumul_jnl_rec_len);
tmp_cumul_jnl_rec_len = (uint4)(jgbl.cumul_jnl_rec_len + SIZEOF(jnldata_hdr_struct));
tjpl->write += SIZEOF(jnldata_hdr_struct);
if (tjpl->write >= tjpl->jnlpool_size)
{
assert(tjpl->write == tjpl->jnlpool_size);
tjpl->write = 0;
}
assert(jpl->early_write_addr == jpl->write_addr);
jpl->early_write_addr = jpl->write_addr + tmp_cumul_jnl_rec_len;
/* Source server does not read in crit. It relies on early_write_addr, the transaction
* data, lastwrite_len, write_addr being updated in that order. To ensure this order,
* we have to force out early_write_addr to its coherency point now. If not, the source
* server may read data that is overwritten (or stale). This is true only on
* architectures and OSes that allow unordered memory access
*/
SHM_WRITE_MEMORY_BARRIER;
}
assert(cw_set_depth < CDB_CW_SET_SIZE);
ASSERT_CURR_TN_EQUALS_EARLY_TN(csa, dbtn);
CHECK_TN(csa, csd, dbtn); /* can issue rts_error TNTOOLARGE */
if (JNL_ENABLED(csa))
{ /* Since we got the system time (jgbl.gbl_jrec_time) outside of crit, it is possible that
* journal records were written concurrently to this file with a timestamp that is future
* relative to what we recorded. In that case, adjust our recorded time to match this.
* This is necessary to ensure that timestamps of successive journal records for each
* database file are in non-decreasing order. A side-effect of this is that our recorded
* time might not accurately reflect the current system time but that is considered not
* an issue since we dont expect to be off by more than a second or two if at all.
* Another side effect is that even if the system time went back, we will never write
* out-of-order timestamped journal records in the lifetime of this database shared memory.
*/
jpc = csa->jnl;
jbp = jpc->jnl_buff;
/* Before writing to jnlfile, adjust jgbl.gbl_jrec_time if needed to maintain time order of jnl
* records. This needs to be done BEFORE the jnl_ensure_open as that could write journal records
* (if it decides to switch to a new journal file)
*/
ADJUST_GBL_JREC_TIME(jgbl, jbp);
if (replication)
{ /* Make sure timestamp of this seqno is >= timestamp of previous seqno. Note: The below macro
* invocation should be done AFTER the ADJUST_GBL_JREC_TIME call as the below resets
* jpl->prev_jnlseqno_time. Doing it the other way around would mean the reset will happen
* with a potentially lower value than the final adjusted time written in the jnl record.
*/
ADJUST_GBL_JREC_TIME_JNLPOOL(jgbl, jpl);
}
/* Note that jnl_ensure_open can call cre_jnl_file which
* in turn assumes jgbl.gbl_jrec_time is set. Also jnl_file_extend can call
* jnl_write_epoch_rec which in turn assumes jgbl.gbl_jrec_time is set.
* In case of forw-phase-recovery, mur_output_record would have already set this.
*/
assert(jgbl.gbl_jrec_time);
jnl_status = jnl_ensure_open();
GTM_WHITE_BOX_TEST(WBTEST_T_END_JNLFILOPN, jnl_status, ERR_JNLFILOPN);
if (jnl_status == 0)
{ /* tmp_cw_set_depth was used to do TOTAL_NONTPJNL_REC_SIZE calculation earlier in this function.
* It is now though that the actual jnl record write occurs. Ensure that the current value of
* cw_set_depth does not entail any change in journal record size than was calculated.
* Same case with csa->jnl_before_images & jbp->before_images.
* The only exception is that in case of mu_reorg_upgrd_dwngrd_in_prog cw_set_depth will be
* LESSER than tmp_cw_set_depth (this is still fine as there is more size allocated than used).
*/
assert(cw_set_depth == tmp_cw_set_depth
|| mu_reorg_upgrd_dwngrd_in_prog && cw_map_depth && cw_set_depth < tmp_cw_set_depth);
assert(jbp->before_images == csa->jnl_before_image);
assert((csa->jnl_state == csd->jnl_state) && (csa->jnl_before_image == csd->jnl_before_image));
if (DISK_BLOCKS_SUM(jbp->freeaddr, total_jnl_rec_size) > jbp->filesize)
{ /* Moved as part of change to prevent journal records splitting
* across multiple generation journal files. */
if (SS_NORMAL != (jnl_status = jnl_flush(jpc->region)))
{
send_msg(VARLSTCNT(9) ERR_JNLFLUSH, 2, JNL_LEN_STR(csd),
ERR_TEXT, 2, RTS_ERROR_TEXT("Error with journal flush during t_end"),
jnl_status);
assert((!JNL_ENABLED(csd)) && (JNL_ENABLED(csa)));
status = cdb_sc_jnlclose;
goto failed;
} else if (EXIT_ERR == jnl_file_extend(jpc, total_jnl_rec_size))
{
assert(csd == csa->hdr); /* jnl_file_extend() shouldn't reset csd in MM */
assert((!JNL_ENABLED(csd)) && (JNL_ENABLED(csa)));
status = cdb_sc_jnlclose;
goto failed;
}
assert(csd == csa->hdr); /* If MM, csd shouldn't have been reset */
}
assert(jgbl.gbl_jrec_time >= jbp->prev_jrec_time);
if (0 == jpc->pini_addr)
jnl_put_jrt_pini(csa);
if (JNL_HAS_EPOCH(jbp))
{
if ((jbp->next_epoch_time <= jgbl.gbl_jrec_time) UNCONDITIONAL_EPOCH_ONLY(|| TRUE))
{ /* Flush the cache. Since we are in crit, defer syncing epoch */
if (!wcs_flu(WCSFLU_FLUSH_HDR | WCSFLU_WRITE_EPOCH | WCSFLU_IN_COMMIT))
{
SET_WCS_FLU_FAIL_STATUS(status, csd);
SET_TRACEABLE_VAR(csd->wc_blocked, TRUE);
BG_TRACE_PRO_ANY(csa, wc_blocked_t_end_jnl_wcsflu);
goto failed;
}
assert(csd == csa->hdr);
VMS_ONLY(
if (csd->clustered &&
!CCP_SEGMENT_STATE(cnl, CCST_MASK_HAVE_DIRTY_BUFFERS))
{
CCP_FID_MSG(gv_cur_region, CCTR_FLUSHLK);
ccp_userwait(gv_cur_region,
CCST_MASK_HAVE_DIRTY_BUFFERS, 0, cnl->ccp_cycle);
}
)
}
}
} else
{
if (SS_NORMAL != jpc->status)
rts_error(VARLSTCNT(7) jnl_status, 4, JNL_LEN_STR(csd), DB_LEN_STR(gv_cur_region),
jpc->status);
else
rts_error(VARLSTCNT(6) jnl_status, 4, JNL_LEN_STR(csd), DB_LEN_STR(gv_cur_region));
}
}
assert(!TREF(donot_commit)); /* We should never commit a transaction that was determined restartable */
assert(TN_NOT_SPECIFIED > MAX_TN_V5); /* Ensure TN_NOT_SPECIFIED isn't a valid TN number */
blktn = (TN_NOT_SPECIFIED == ctn) ? dbtn : ctn;
csa->ti->early_tn = dbtn + 1;
if (JNL_ENABLED(csa))
{
DEBUG_ONLY(save_gbl_jrec_time = jgbl.gbl_jrec_time;)
if (jbp->before_images)
{ /* do not write PBLKs if MUPIP REORG UPGRADE/DOWNGRADE with -NOSAFEJNL */
if (!mu_reorg_upgrd_dwngrd_in_prog || !mu_reorg_nosafejnl)
{
epoch_tn = jbp->epoch_tn; /* store in a local as it is used in a loop below */
for (cs = cw_set, cs_top = cs + cw_set_depth; cs < cs_top; ++cs)
{
/* PBLK computations for FREE blocks are not needed */
if (cs->was_free)
continue;
/* write out before-update journal image records */
mode = cs->mode;
if (gds_t_committed < mode)
{ /* There are two possibilities at this point.
* a) gds_t_write_root : In this case no need to write PBLK.
* b) gds_t_busy2free : This is set by gvcst_bmp_mark_free to indicate
* that a block has to be freed right away instead of taking it
* through the RECYCLED state. This should be done only if
* csd->db_got_to_v5_once has not yet become TRUE. Once it is
* TRUE, block frees will write PBLK only when the block is reused.
*/
assert((gds_t_write_root == mode) || (gds_t_busy2free == mode)
|| (gds_t_recycled2free == mode));
if ((gds_t_write_root == mode)
|| (gds_t_busy2free == mode) && csd->db_got_to_v5_once)
continue;
}
old_block = (blk_hdr_ptr_t)cs->old_block;
ASSERT_IS_WITHIN_SHM_BOUNDS((sm_uc_ptr_t)old_block, csa);
DBG_ENSURE_OLD_BLOCK_IS_VALID(cs, is_mm, csa, csd);
if ((NULL != old_block) && (old_block->tn < epoch_tn))
{
bsiz = old_block->bsiz;
assert((bsiz <= csd->blk_size) || IS_DSE_IMAGE);
assert(bsiz >= SIZEOF(blk_hdr) || IS_DSE_IMAGE);
/* For acquired or gds_t_busy2free blocks, we should have computed
* checksum already. The only exception is if we found no need to
* compute checksum outside of crit but before we got crit, an
* EPOCH got written concurrently so we have to write a PBLK (and
* hence compute the checksum as well) when earlier we thought none
* was necessary. An easy way to check this is that an EPOCH was
* written AFTER we started this transaction.
*/
assert((gds_t_acquired != cs->mode) || (gds_t_busy2free != cs->mode)
|| cs->blk_checksum || (epoch_tn >= start_tn));
/* It is possible that the block has a bad block-size.
* Before computing checksum ensure bsiz passed is safe.
* The checks done here for "bsiz" assignment are
* similar to those done in jnl_write_pblk/jnl_write_aimg.
*/
if (IS_DSE_IMAGE)
bsiz = MIN(bsiz, csd->blk_size);
assert(!cs->blk_checksum ||
(cs->blk_checksum == jnl_get_checksum((uint4 *)old_block,
csa,
bsiz)));
if (!cs->blk_checksum)
cs->blk_checksum = jnl_get_checksum((uint4 *)old_block,
csa,
bsiz);
# ifdef GTM_CRYPT
if (csd->is_encrypted)
{
DBG_ENSURE_PTR_IS_VALID_GLOBUFF(csa, csd, (sm_uc_ptr_t)old_block);
DEBUG_ONLY(save_old_block = old_block;)
old_block = (blk_hdr_ptr_t)GDS_ANY_ENCRYPTGLOBUF(old_block,
csa);
/* Ensure that the unencrypted block and it's twin counterpart are
* in sync.
*/
assert(save_old_block->tn == old_block->tn);
assert(save_old_block->bsiz == old_block->bsiz);
assert(save_old_block->levl == old_block->levl);
DBG_ENSURE_PTR_IS_VALID_ENCTWINGLOBUFF(csa,
csd,
(sm_uc_ptr_t)old_block);
}
# endif
jnl_write_pblk(csa, cs, old_block);
cs->jnl_freeaddr = jbp->freeaddr;
}
DEBUG_ONLY(
else
assert(0 == cs->jnl_freeaddr);
)
}
}
}
if (write_after_image)
{ /* either DSE or MUPIP RECOVER playing an AIMG record */
assert(1 == cw_set_depth); /* only one block at a time */
assert(!replication);
cs = cw_set;
jnl_write_aimg_rec(csa, cs);
} else if (write_inctn)
{
assert(!replication);
if ((inctn_blkupgrd == inctn_opcode) || (inctn_blkdwngrd == inctn_opcode))
{
assert(1 == cw_set_depth); /* upgrade/downgrade one block at a time */
cs = cw_set;
assert(inctn_detail.blknum_struct.blknum == cs->blk);
assert(mu_reorg_upgrd_dwngrd_blktn < dbtn);
if (mu_reorg_nosafejnl)
{
blktn = mu_reorg_upgrd_dwngrd_blktn;
/* if NOSAFEJNL and there is going to be a block format change
* as a result of this update, note it down in the inctn opcode
* (for recovery) as there is no PBLK record for it to rely on.
*/
if (cs->ondsk_blkver != csd->desired_db_format)
inctn_opcode = (inctn_opcode == inctn_blkupgrd)
? inctn_blkupgrd_fmtchng : inctn_blkdwngrd_fmtchng;
}
}
jnl_write_inctn_rec(csa);
} else if (0 == jnl_fence_ctl.level)
{
assert(!replication || !jgbl.forw_phase_recovery);
if (replication)
{
jnl_fence_ctl.token = tjpl->jnl_seqno;
UNIX_ONLY(
if (supplementary)
jnl_fence_ctl.strm_seqno = SET_STRM_INDEX(strm_seqno, strm_index);
)
} else if (!jgbl.forw_phase_recovery)
jnl_fence_ctl.token = seq_num_zero;
/* In case of forw-phase of recovery, token would have been set by mur_output_record */
jnl_write_logical(csa, non_tp_jfb_ptr);
} else
jnl_write_ztp_logical(csa, non_tp_jfb_ptr);
/* Ensure jgbl.gbl_jrec_time did not get reset by any of the jnl writing functions */
assert(save_gbl_jrec_time == jgbl.gbl_jrec_time);
} else if (replication)
{ /* Case where JNL_ENABLED(csa) is FALSE but REPL_WAS_ENABLED(csa) is TRUE and therefore we need to
* write logical jnl records in the journal pool (no need to write in journal buffer or journal file).
*/
assert(!JNL_ENABLED(csa) && REPL_WAS_ENABLED(csa));
if (0 == jnl_fence_ctl.level)
{
jnl_fence_ctl.token = tjpl->jnl_seqno;
UNIX_ONLY(
if (supplementary)
jnl_fence_ctl.strm_seqno = SET_STRM_INDEX(strm_seqno, strm_index);
)
jnl_write_logical(csa, non_tp_jfb_ptr);
} else
jnl_write_ztp_logical(csa, non_tp_jfb_ptr);
}
if (replication)
{
tjpl->jnl_seqno++;
assert(csa->hdr->reg_seqno < tjpl->jnl_seqno);
csa->hdr->reg_seqno = tjpl->jnl_seqno;
UNIX_ONLY(
if (supplementary)
{
next_strm_seqno = strm_seqno + 1;
/* tjpl->strm_seqno[strm_index] = next_strm_seqno; */
csa->hdr->strm_reg_seqno[strm_index] = next_strm_seqno;
}
)
VMS_ONLY(
if (is_updproc)
{
jgbl.max_resync_seqno++;
csa->hdr->resync_seqno = jgbl.max_resync_seqno;
}
)
}
csa->prev_free_blks = csa->ti->free_blocks;
csa->t_commit_crit = T_COMMIT_CRIT_PHASE1;
if (cw_set_depth)
{
if (!is_mm) /* increment counter of # of processes that are actively doing two-phase commit */
INCR_WCS_PHASE2_COMMIT_PIDCNT(csa, cnl);
# ifdef DEBUG
/* Assert that cs->old_mode if uninitialized, never contains a negative value (relied by secshr_db_clnup) */
for (cs = cw_set, cs_top = cs + cw_set_depth; cs < cs_top; ++cs)
assert(0 <= cs->old_mode);
# endif
for (cs = cw_set, cs_top = cs + cw_set_depth; cs < cs_top; ++cs)
{
mode = cs->mode;
assert((gds_t_write_root != mode) || ((cs - cw_set) + 1 == cw_depth));
cs->old_mode = (int4)mode; /* note down before being reset to gds_t_committed */
assert(gds_t_committed < gds_t_write_root);
assert(gds_t_committed < gds_t_busy2free);
assert((n_gds_t_op != mode) && (gds_t_committed != mode));
assert((kill_t_write != mode) && (kill_t_create != mode));
if (gds_t_committed > mode)
{
DEBUG_ONLY(
/* Check bitmap status of block we are about to modify.
* Two exceptions are
* a) DSE which can modify bitmaps at will.
* b) MUPIP RECOVER writing an AIMG. In this case it is playing
* forward a DSE action so is effectively like DSE doing it.
*/
if (!IS_DSE_IMAGE && !write_after_image)
bml_status_check(cs);
)
if (is_mm)
status = mm_update(cs, dbtn, blktn, dummysi);
else
{
if (csd->dsid)
{
if (ERR_GVKILLFAIL == t_err)
{
if (cs == cw_set)
{
if ((gds_t_acquired == mode) ||
((cw_set_depth > 1) && (0 == cw_set[1].level)))
rc_cpt_inval();
else
rc_cpt_entry(cs->blk);
}
} else if (0 == cs->level)
rc_cpt_entry(cs->blk);
}
/* Do phase1 of bg_update while holding crit on the database.
* This will lock the buffers that need to be changed.
* Once crit is released, invoke phase2 which will update those locked buffers.
* The exception is if it is a bitmap block. In that case we also do phase2
* while holding crit so the next process to use this bitmap will see a
* consistent copy of this bitmap when it gets crit for commit. This avoids
* the reallocate_bitmap routine from restarting or having to wait for a
* concurrent phase2 construction to finish. When the change request C9E11-002651
* (to reduce restarts due to bitmap collisions) is addressed, we can reexamine
* whether it makes sense to move bitmap block builds back to phase2.
*/
status = bg_update_phase1(cs, dbtn, dummysi);
if ((cdb_sc_normal == status) && (gds_t_writemap == mode))
{
status = bg_update_phase2(cs, dbtn, blktn, dummysi);
if (cdb_sc_normal == status)
cs->mode = gds_t_committed;
}
}
if (cdb_sc_normal != status)
{ /* the database is probably in trouble */
INVOKE_T_COMMIT_CLEANUP(status, csa);
assert(cdb_sc_normal == status);
/* At this time "cr_array_index" could be non-zero and a few cache-records might
* have their "in_cw_set" field set to TRUE. We should not reset "in_cw_set" as we
* don't hold crit at this point and also because we might still need those buffers
* pinned until their before-images are backed up in wcs_recover (in case an
* online backup was running while secshr_db_clnup did its job). Reset the
* local variable "cr_array_index" though so we do not accidentally reset the
* "in_cw_set" fields ourselves before the wcs_recover.
*/
cr_array_index = 0;
goto skip_cr_array; /* hence skip until past "cr_array_index" processing */
}
}
}
}
/* signal secshr_db_clnup/t_commit_cleanup, roll-back is no longer possible */
update_trans |= UPDTRNS_TCOMMIT_STARTED_MASK;
assert(cdb_sc_normal == status);
assert(!csd->freeze || IS_DSE_IMAGE); /* should never increment curr_tn on a frozen database except if DSE */
INCREMENT_CURR_TN(csd);
csa->t_commit_crit = T_COMMIT_CRIT_PHASE2; /* set this BEFORE releasing crit but AFTER incrementing curr_tn */
/* If db is journaled, then db header is flushed periodically when writing the EPOCH record,
* otherwise do it here every HEADER_UPDATE_COUNT transactions.
*/
assert(!JNL_ENABLED(csa) || (jbp == csa->jnl->jnl_buff));
if ((!JNL_ENABLED(csa) || !JNL_HAS_EPOCH(jbp)) && !(csd->trans_hist.curr_tn & (HEADER_UPDATE_COUNT - 1)))
fileheader_sync(gv_cur_region);
if (need_kip_incr) /* increment kill_in_prog */
{
INCR_KIP(csd, csa, kip_csa);
need_kip_incr = FALSE;
}
start_tn = dbtn; /* start_tn temporarily used to store currtn (for bg_update_phase2) before releasing crit */
}
if (!is_mm && busy2free_seen)
{
assert((gds_t_busy2free == cw_set[0].mode) && cr_array_index && (cw_set[0].blk == cr_array[0]->blk));
assert(process_id == cr_array[0]->in_cw_set);
UNPIN_CACHE_RECORD(cr_array[0]); /* need to do this BEFORE releasing crit as we have no other lock on this buffer */
} else if (recycled2free_seen)
{
assert(gds_t_recycled2free == cw_set[0].mode);
assert(!is_mm);
if (read_before_image)
{
assert(cr_array_index == 2);
assert(cw_set[0].blk = cr_array[1]->blk);
assert(process_id == cr_array[1]->in_cw_set);
UNPIN_CACHE_RECORD(cr_array[1]);
} else
{
assert(cr_array_index == 1);
/* Never pinned it in the first place */
}
}
if (!csa->hold_onto_crit)
rel_crit(gv_cur_region);
assert(!replication || update_trans);
if (replication)
{
assert(jpl->early_write_addr > jpl->write_addr);
assert(tmp_cumul_jnl_rec_len == (tjpl->write - jpl->write + (tjpl->write > jpl->write ? 0 : jpl->jnlpool_size)));
/* the following statements should be atomic */
jnl_header = (jnldata_hdr_ptr_t)(jnlpool.jnldata_base + jpl->write);
jnl_header->jnldata_len = tmp_cumul_jnl_rec_len;
jnl_header->prev_jnldata_len = jpl->lastwrite_len;
/* The following assert should be an == rather than a >= (as in tp_tend) because, we have
* either one or no update. If no update, we would have no cw_depth and we wouldn't enter
* this path. If there is an update, then both the indices should be 1.
*/
INT8_ONLY(assert(jgbl.cumul_index == jgbl.cu_jnl_index));
UNIX_ONLY(
if (supplementary)
jpl->strm_seqno[strm_index] = next_strm_seqno;
)
jpl->lastwrite_len = jnl_header->jnldata_len;
/* For systems with UNORDERED memory access (example, ALPHA, POWER4, PA-RISC 2.0), on a
* multi processor system, it is possible that the source server notices the change in
* write_addr before seeing the change to jnlheader->jnldata_len, leading it to read an
* invalid transaction length. To avoid such conditions, we should commit the order of
* shared memory updates before we update write_addr. This ensures that the source server
* sees all shared memory updates related to a transaction before the change in write_addr
*/
SHM_WRITE_MEMORY_BARRIER;
jpl->write = tjpl->write;
/* jpl->write_addr should be updated before updating jpl->jnl_seqno as secshr_db_clnup relies on this */
jpl->write_addr += jnl_header->jnldata_len;
assert(jpl->early_write_addr == jpl->write_addr);
jpl->jnl_seqno = tjpl->jnl_seqno;
if (!repl_csa->hold_onto_crit)
rel_lock(jnlpool.jnlpool_dummy_reg);
}
/* If BG, check that we have not pinned any more buffers than we are updating */
DBG_CHECK_PINNED_CR_ARRAY_CONTENTS(is_mm, cr_array, cr_array_index, csd->bplmap);
if (cw_set_depth)
{ /* Finish 2nd phase of commit for BG (updating the buffers in phase1) now that CRIT has been released.
* For MM, only thing needed is to set cs->mode to gds_t_committed.
*/
for (cs = cw_set, cs_top = cs + cw_set_depth; cs < cs_top; ++cs)
{
mode = cs->mode;
assert((gds_t_write_root != mode) || ((cs - cw_set) + 1 == cw_depth));
assert((kill_t_write != mode) && (kill_t_create != mode));
if (gds_t_committed > mode)
{
if (!is_mm)
{ /* Validate old_mode noted down in first phase is the same as the current mode.
* Note that cs->old_mode is negated by bg_update_phase1 (to help secshr_db_clnup).
*/
assert(-cs->old_mode == mode);
status = bg_update_phase2(cs, dbtn, blktn, dummysi);
if (cdb_sc_normal != status)
{ /* the database is probably in trouble */
INVOKE_T_COMMIT_CLEANUP(status, csa);
assert(cdb_sc_normal == status);
/* At this time "cr_array_index" could be non-zero and a few cache-records might
* have their "in_cw_set" field set to TRUE. We should not reset "in_cw_set" as we
* don't hold crit at this point and also because we might still need those buffers
* pinned until their before-images are backed up in wcs_recover (in case an
* online backup was running while secshr_db_clnup did its job). Reset the
* local variable "cr_array_index" though so we do not accidentally reset the
* "in_cw_set" fields ourselves before the wcs_recover.
*/
cr_array_index = 0;
/* Note that seshr_db_clnup (invoked by t_commit_cleanup above) would have
* done a lot of cleanup for us including decrementing the wcs_phase2_commit_pidcnt
* so it is ok to skip all that processing below and go directly to skip_cr_array.
*/
goto skip_cr_array; /* hence skip until past "cr_array_index" processing */
}
}
}
cs->mode = gds_t_committed;
}
if (!is_mm) /* now that two-phase commit is done, decrement counter */
DECR_WCS_PHASE2_COMMIT_PIDCNT(csa, cnl);
}
UNPIN_CR_ARRAY_ON_COMMIT(cr_array, cr_array_index);
assert(!cr_array_index);
csa->t_commit_crit = FALSE;
/* Phase 2 commits are completed. See if we had done a snapshot init (csa->snapshot_in_prog == TRUE). If so,
* try releasing the resources obtained while snapshot init.
*/
if (SNAPSHOTS_IN_PROG(csa))
{
assert(update_trans);
SS_RELEASE_IF_NEEDED(csa, cnl);
}
skip_cr_array:
assert(!csa->now_crit || csa->hold_onto_crit);
assert(cdb_sc_normal == status);
REVERT; /* no need for t_ch to be invoked if any errors occur after this point */
DEFERRED_EXIT_HANDLING_CHECK; /* now that all crits are released, check if deferred signal/exit handling needs to be done */
if (block_saved)
backup_buffer_flush(gv_cur_region);
if (unhandled_stale_timer_pop)
process_deferred_stale();
if (update_trans)
{
wcs_timer_start(gv_cur_region, TRUE);
if (REPL_ALLOWED(csa) && IS_DSE_IMAGE)
{
temp_tn = dbtn + 1;
send_msg(VARLSTCNT(6) ERR_NOTREPLICATED, 4, &temp_tn, LEN_AND_LIT("DSE"), process_id);
}
INCR_GVSTATS_COUNTER(csa, cnl, n_nontp_readwrite, 1);
INCR_GVSTATS_COUNTER(csa, cnl, n_nontp_blkread, n_blks_validated);
INCR_GVSTATS_COUNTER(csa, cnl, n_nontp_blkwrite, cw_set_depth);
GVSTATS_SET_CSA_STATISTIC(csa, db_curr_tn, dbtn);
} else
{
INCR_GVSTATS_COUNTER(csa, cnl, n_nontp_readonly, 1);
INCR_GVSTATS_COUNTER(csa, cnl, n_nontp_blkread, n_blks_validated);
}
/* "secshr_db_clnup/t_commit_cleanup" assume an active non-TP transaction if cw_set_depth is non-zero
* or if update_trans is set to T_COMMIT_STARTED. Now that the transaction is complete, reset these fields.
*/
cw_set_depth = 0;
update_trans = 0;
CWS_RESET;
/* although we have the same assert at the beginning of "skip_cr_array" label, the below assert ensures that we did not grab
* crit in between (in process_deferred_stale() or backup_buffer_flush() or wcs_timer_start()). The assert at the beginning
* of the loop is needed as well to ensure that if we are done with the commit, we should have released crit
*/
assert(!csa->now_crit || csa->hold_onto_crit);
t_tries = 0; /* commit was successful so reset t_tries */
assert(0 == cr_array_index);
return dbtn;
failed:
assert(cdb_sc_normal != status);
REVERT;
failed_skip_revert:
RESTORE_CURRTN_IF_NEEDED(csa, write_inctn, decremented_currtn);
retvalue = t_commit_cleanup(status, 0); /* we expect to get a return value indicating update was NOT underway */
assert(!retvalue); /* if it was, then we would have done a "goto skip_cr_array:" instead */
if (NULL != gv_target) /* gv_target can be NULL in case of DSE MAPS command etc. */
gv_target->clue.end = 0;
/* t_commit_cleanup releases crit as long as the transition is from 2nd to 3rd retry or 3rd to 3rd retry. The only exception
* is if hold_onto_crit is set to TRUE in which case t_commit_cleanup honors it. Assert accordingly.
*/
assert(!csa->now_crit || !NEED_TO_RELEASE_CRIT(t_tries) || csa->hold_onto_crit);
DEFERRED_EXIT_HANDLING_CHECK; /* now that all crits are released, check if deferred signal/exit handling needs to be done */
t_retry(status);
/* in the retry case, we do not do a CWS_RESET as cw_stagnate is used only in the
* final retry in which case t_end will succeed and do a CWS_RESET
*/
cw_map_depth = 0;
assert(0 == cr_array_index);
return 0;
}
Reference in New Issue View Git Blame Copy Permalink