The Design and Implementation of the FreeBSD Operating System, Second Edition
Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]

FreeBSD/Linux Kernel Cross Reference
sys/kern/vfs_journal.c

Version: -  FREEBSD  -  FREEBSD-13-STABLE  -  FREEBSD-13-0  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  l41  -  OPENBSD  -  linux-2.6  -  MK84  -  PLAN9  -  xnu-8792 
SearchContext: -  none  -  3  -  10 

    1 /*
    2  * Copyright (c) 2004-2006 The DragonFly Project.  All rights reserved.
    3  * 
    4  * This code is derived from software contributed to The DragonFly Project
    5  * by Matthew Dillon <dillon@backplane.com>
    6  * 
    7  * Redistribution and use in source and binary forms, with or without
    8  * modification, are permitted provided that the following conditions
    9  * are met:
   10  * 
   11  * 1. Redistributions of source code must retain the above copyright
   12  *    notice, this list of conditions and the following disclaimer.
   13  * 2. Redistributions in binary form must reproduce the above copyright
   14  *    notice, this list of conditions and the following disclaimer in
   15  *    the documentation and/or other materials provided with the
   16  *    distribution.
   17  * 3. Neither the name of The DragonFly Project nor the names of its
   18  *    contributors may be used to endorse or promote products derived
   19  *    from this software without specific, prior written permission.
   20  * 
   21  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
   22  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
   23  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
   24  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
   25  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
   26  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
   27  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
   28  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
   29  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
   30  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
   31  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   32  * SUCH DAMAGE.
   33  *
   34  * $DragonFly: src/sys/kern/vfs_journal.c,v 1.33 2007/05/09 00:53:34 dillon Exp $
   35  */
   36 /*
   37  * The journaling protocol is intended to evolve into a two-way stream
   38  * whereby transaction IDs can be acknowledged by the journaling target
   39  * when the data has been committed to hard storage.  Both implicit and
   40  * explicit acknowledgement schemes will be supported, depending on the
   41  * sophistication of the journaling stream, plus resynchronization and
   42  * restart when a journaling stream is interrupted.  This information will
   43  * also be made available to journaling-aware filesystems to allow better
   44  * management of their own physical storage synchronization mechanisms as
   45  * well as to allow such filesystems to take direct advantage of the kernel's
   46  * journaling layer so they don't have to roll their own.
   47  *
   48  * In addition, the worker thread will have access to much larger 
   49  * spooling areas then the memory buffer is able to provide by e.g. 
   50  * reserving swap space, in order to absorb potentially long interruptions
   51  * of off-site journaling streams, and to prevent 'slow' off-site linkages
   52  * from radically slowing down local filesystem operations.  
   53  *
   54  * Because of the non-trivial algorithms the journaling system will be
   55  * required to support, use of a worker thread is mandatory.  Efficiencies
   56  * are maintained by utilitizing the memory FIFO to batch transactions when
   57  * possible, reducing the number of gratuitous thread switches and taking
   58  * advantage of cpu caches through the use of shorter batched code paths
   59  * rather then trying to do everything in the context of the process
   60  * originating the filesystem op.  In the future the memory FIFO can be
   61  * made per-cpu to remove BGL or other locking requirements.
   62  */
   63 #include <sys/param.h>
   64 #include <sys/systm.h>
   65 #include <sys/buf.h>
   66 #include <sys/conf.h>
   67 #include <sys/kernel.h>
   68 #include <sys/queue.h>
   69 #include <sys/lock.h>
   70 #include <sys/malloc.h>
   71 #include <sys/mount.h>
   72 #include <sys/unistd.h>
   73 #include <sys/vnode.h>
   74 #include <sys/poll.h>
   75 #include <sys/mountctl.h>
   76 #include <sys/journal.h>
   77 #include <sys/file.h>
   78 #include <sys/proc.h>
   79 #include <sys/xio.h>
   80 #include <sys/socket.h>
   81 #include <sys/socketvar.h>
   82 
   83 #include <machine/limits.h>
   84 
   85 #include <vm/vm.h>
   86 #include <vm/vm_object.h>
   87 #include <vm/vm_page.h>
   88 #include <vm/vm_pager.h>
   89 #include <vm/vnode_pager.h>
   90 
   91 #include <sys/file2.h>
   92 #include <sys/thread2.h>
   93 #include <sys/mplock2.h>
   94 #include <sys/spinlock2.h>
   95 
   96 static void journal_wthread(void *info);
   97 static void journal_rthread(void *info);
   98 
   99 static void *journal_reserve(struct journal *jo,
  100                         struct journal_rawrecbeg **rawpp,
  101                         int16_t streamid, int bytes);
  102 static void *journal_extend(struct journal *jo,
  103                         struct journal_rawrecbeg **rawpp,
  104                         int truncbytes, int bytes, int *newstreamrecp);
  105 static void journal_abort(struct journal *jo,
  106                         struct journal_rawrecbeg **rawpp);
  107 static void journal_commit(struct journal *jo,
  108                         struct journal_rawrecbeg **rawpp,
  109                         int bytes, int closeout);
  110 static void jrecord_data(struct jrecord *jrec,
  111                         void *buf, int bytes, int dtype);
  112 
  113 
  114 MALLOC_DEFINE(M_JOURNAL, "journal", "Journaling structures");
  115 MALLOC_DEFINE(M_JFIFO, "journal-fifo", "Journal FIFO");
  116 
  117 void
  118 journal_create_threads(struct journal *jo)
  119 {
  120         jo->flags &= ~(MC_JOURNAL_STOP_REQ | MC_JOURNAL_STOP_IMM);
  121         jo->flags |= MC_JOURNAL_WACTIVE;
  122         lwkt_create(journal_wthread, jo, NULL, &jo->wthread,
  123                     TDF_NOSTART, -1,
  124                     "journal w:%.*s", JIDMAX, jo->id);
  125         lwkt_setpri(&jo->wthread, TDPRI_KERN_DAEMON);
  126         lwkt_schedule(&jo->wthread);
  127 
  128         if (jo->flags & MC_JOURNAL_WANT_FULLDUPLEX) {
  129             jo->flags |= MC_JOURNAL_RACTIVE;
  130             lwkt_create(journal_rthread, jo, NULL, &jo->rthread,
  131                         TDF_NOSTART, -1,
  132                         "journal r:%.*s", JIDMAX, jo->id);
  133             lwkt_setpri(&jo->rthread, TDPRI_KERN_DAEMON);
  134             lwkt_schedule(&jo->rthread);
  135         }
  136 }
  137 
  138 void
  139 journal_destroy_threads(struct journal *jo, int flags)
  140 {
  141     int wcount;
  142 
  143     jo->flags |= MC_JOURNAL_STOP_REQ | (flags & MC_JOURNAL_STOP_IMM);
  144     wakeup(&jo->fifo);
  145     wcount = 0;
  146     while (jo->flags & (MC_JOURNAL_WACTIVE | MC_JOURNAL_RACTIVE)) {
  147         tsleep(jo, 0, "jwait", hz);
  148         if (++wcount % 10 == 0) {
  149             kprintf("Warning: journal %s waiting for descriptors to close\n",
  150                 jo->id);
  151         }
  152     }
  153 
  154     /*
  155      * XXX SMP - threads should move to cpu requesting the restart or
  156      * termination before finishing up to properly interlock.
  157      */
  158     tsleep(jo, 0, "jwait", hz);
  159     lwkt_free_thread(&jo->wthread);
  160     if (jo->flags & MC_JOURNAL_WANT_FULLDUPLEX)
  161         lwkt_free_thread(&jo->rthread);
  162 }
  163 
  164 /*
  165  * The per-journal worker thread is responsible for writing out the
  166  * journal's FIFO to the target stream.
  167  */
  168 static void
  169 journal_wthread(void *info)
  170 {
  171     struct journal *jo = info;
  172     struct journal_rawrecbeg *rawp;
  173     int error;
  174     size_t avail;
  175     size_t bytes;
  176     size_t res;
  177 
  178     /* not MPSAFE yet */
  179     get_mplock();
  180 
  181     for (;;) {
  182         /*
  183          * Calculate the number of bytes available to write.  This buffer
  184          * area may contain reserved records so we can't just write it out
  185          * without further checks.
  186          */
  187         bytes = jo->fifo.windex - jo->fifo.rindex;
  188 
  189         /*
  190          * sleep if no bytes are available or if an incomplete record is
  191          * encountered (it needs to be filled in before we can write it
  192          * out), and skip any pad records that we encounter.
  193          */
  194         if (bytes == 0) {
  195             if (jo->flags & MC_JOURNAL_STOP_REQ)
  196                 break;
  197             tsleep(&jo->fifo, 0, "jfifo", hz);
  198             continue;
  199         }
  200 
  201         /*
  202          * Sleep if we can not go any further due to hitting an incomplete
  203          * record.  This case should occur rarely but may have to be better
  204          * optimized XXX.
  205          */
  206         rawp = (void *)(jo->fifo.membase + (jo->fifo.rindex & jo->fifo.mask));
  207         if (rawp->begmagic == JREC_INCOMPLETEMAGIC) {
  208             tsleep(&jo->fifo, 0, "jpad", hz);
  209             continue;
  210         }
  211 
  212         /*
  213          * Skip any pad records.  We do not write out pad records if we can
  214          * help it. 
  215          */
  216         if (rawp->streamid == JREC_STREAMID_PAD) {
  217             if ((jo->flags & MC_JOURNAL_WANT_FULLDUPLEX) == 0) {
  218                 if (jo->fifo.rindex == jo->fifo.xindex) {
  219                     jo->fifo.xindex += (rawp->recsize + 15) & ~15;
  220                     jo->total_acked += (rawp->recsize + 15) & ~15;
  221                 }
  222             }
  223             jo->fifo.rindex += (rawp->recsize + 15) & ~15;
  224             jo->total_acked += bytes;
  225             KKASSERT(jo->fifo.windex - jo->fifo.rindex >= 0);
  226             continue;
  227         }
  228 
  229         /*
  230          * 'bytes' is the amount of data that can potentially be written out.  
  231          * Calculate 'res', the amount of data that can actually be written
  232          * out.  res is bounded either by hitting the end of the physical
  233          * memory buffer or by hitting an incomplete record.  Incomplete
  234          * records often occur due to the way the space reservation model
  235          * works.
  236          */
  237         res = 0;
  238         avail = jo->fifo.size - (jo->fifo.rindex & jo->fifo.mask);
  239         while (res < bytes && rawp->begmagic == JREC_BEGMAGIC) {
  240             res += (rawp->recsize + 15) & ~15;
  241             if (res >= avail) {
  242                 KKASSERT(res == avail);
  243                 break;
  244             }
  245             rawp = (void *)((char *)rawp + ((rawp->recsize + 15) & ~15));
  246         }
  247 
  248         /*
  249          * Issue the write and deal with any errors or other conditions.
  250          * For now assume blocking I/O.  Since we are record-aware the
  251          * code cannot yet handle partial writes.
  252          *
  253          * We bump rindex prior to issuing the write to avoid racing
  254          * the acknowledgement coming back (which could prevent the ack
  255          * from bumping xindex).  Restarts are always based on xindex so
  256          * we do not try to undo the rindex if an error occurs.
  257          *
  258          * XXX EWOULDBLOCK/NBIO
  259          * XXX notification on failure
  260          * XXX permanent verses temporary failures
  261          * XXX two-way acknowledgement stream in the return direction / xindex
  262          */
  263         bytes = res;
  264         jo->fifo.rindex += bytes;
  265         error = fp_write(jo->fp, 
  266                         jo->fifo.membase +
  267                          ((jo->fifo.rindex - bytes) & jo->fifo.mask),
  268                         bytes, &res, UIO_SYSSPACE);
  269         if (error) {
  270             kprintf("journal_thread(%s) write, error %d\n", jo->id, error);
  271             /* XXX */
  272         } else {
  273             KKASSERT(res == bytes);
  274         }
  275 
  276         /*
  277          * Advance rindex.  If the journal stream is not full duplex we also
  278          * advance xindex, otherwise the rjournal thread is responsible for
  279          * advancing xindex.
  280          */
  281         if ((jo->flags & MC_JOURNAL_WANT_FULLDUPLEX) == 0) {
  282             jo->fifo.xindex += bytes;
  283             jo->total_acked += bytes;
  284         }
  285         KKASSERT(jo->fifo.windex - jo->fifo.rindex >= 0);
  286         if ((jo->flags & MC_JOURNAL_WANT_FULLDUPLEX) == 0) {
  287             if (jo->flags & MC_JOURNAL_WWAIT) {
  288                 jo->flags &= ~MC_JOURNAL_WWAIT; /* XXX hysteresis */
  289                 wakeup(&jo->fifo.windex);
  290             }
  291         }
  292     }
  293     fp_shutdown(jo->fp, SHUT_WR);
  294     jo->flags &= ~MC_JOURNAL_WACTIVE;
  295     wakeup(jo);
  296     wakeup(&jo->fifo.windex);
  297     rel_mplock();
  298 }
  299 
  300 /*
  301  * A second per-journal worker thread is created for two-way journaling
  302  * streams to deal with the return acknowledgement stream.
  303  */
  304 static void
  305 journal_rthread(void *info)
  306 {
  307     struct journal_rawrecbeg *rawp;
  308     struct journal_ackrecord ack;
  309     struct journal *jo = info;
  310     int64_t transid;
  311     int error;
  312     size_t count;
  313     size_t bytes;
  314 
  315     transid = 0;
  316     error = 0;
  317 
  318     /* not MPSAFE yet */
  319     get_mplock();
  320 
  321     for (;;) {
  322         /*
  323          * We have been asked to stop
  324          */
  325         if (jo->flags & MC_JOURNAL_STOP_REQ)
  326                 break;
  327 
  328         /*
  329          * If we have no active transaction id, get one from the return
  330          * stream.
  331          */
  332         if (transid == 0) {
  333             error = fp_read(jo->fp, &ack, sizeof(ack), &count, 
  334                             1, UIO_SYSSPACE);
  335 #if 0
  336             kprintf("fp_read ack error %d count %d\n", error, count);
  337 #endif
  338             if (error || count != sizeof(ack))
  339                 break;
  340             if (error) {
  341                 kprintf("read error %d on receive stream\n", error);
  342                 break;
  343             }
  344             if (ack.rbeg.begmagic != JREC_BEGMAGIC ||
  345                 ack.rend.endmagic != JREC_ENDMAGIC
  346             ) {
  347                 kprintf("bad begmagic or endmagic on receive stream\n");
  348                 break;
  349             }
  350             transid = ack.rbeg.transid;
  351         }
  352 
  353         /*
  354          * Calculate the number of unacknowledged bytes.  If there are no
  355          * unacknowledged bytes then unsent data was acknowledged, report,
  356          * sleep a bit, and loop in that case.  This should not happen 
  357          * normally.  The ack record is thrown away.
  358          */
  359         bytes = jo->fifo.rindex - jo->fifo.xindex;
  360 
  361         if (bytes == 0) {
  362             kprintf("warning: unsent data acknowledged transid %08llx\n",
  363                     (long long)transid);
  364             tsleep(&jo->fifo.xindex, 0, "jrseq", hz);
  365             transid = 0;
  366             continue;
  367         }
  368 
  369         /*
  370          * Since rindex has advanced, the record pointed to by xindex
  371          * must be a valid record.
  372          */
  373         rawp = (void *)(jo->fifo.membase + (jo->fifo.xindex & jo->fifo.mask));
  374         KKASSERT(rawp->begmagic == JREC_BEGMAGIC);
  375         KKASSERT(rawp->recsize <= bytes);
  376 
  377         /*
  378          * The target can acknowledge several records at once.
  379          */
  380         if (rawp->transid < transid) {
  381 #if 1
  382             kprintf("ackskip %08llx/%08llx\n",
  383                     (long long)rawp->transid,
  384                     (long long)transid);
  385 #endif
  386             jo->fifo.xindex += (rawp->recsize + 15) & ~15;
  387             jo->total_acked += (rawp->recsize + 15) & ~15;
  388             if (jo->flags & MC_JOURNAL_WWAIT) {
  389                 jo->flags &= ~MC_JOURNAL_WWAIT; /* XXX hysteresis */
  390                 wakeup(&jo->fifo.windex);
  391             }
  392             continue;
  393         }
  394         if (rawp->transid == transid) {
  395 #if 1
  396             kprintf("ackskip %08llx/%08llx\n",
  397                     (long long)rawp->transid,
  398                     (long long)transid);
  399 #endif
  400             jo->fifo.xindex += (rawp->recsize + 15) & ~15;
  401             jo->total_acked += (rawp->recsize + 15) & ~15;
  402             if (jo->flags & MC_JOURNAL_WWAIT) {
  403                 jo->flags &= ~MC_JOURNAL_WWAIT; /* XXX hysteresis */
  404                 wakeup(&jo->fifo.windex);
  405             }
  406             transid = 0;
  407             continue;
  408         }
  409         kprintf("warning: unsent data(2) acknowledged transid %08llx\n",
  410                 (long long)transid);
  411         transid = 0;
  412     }
  413     jo->flags &= ~MC_JOURNAL_RACTIVE;
  414     wakeup(jo);
  415     wakeup(&jo->fifo.windex);
  416     rel_mplock();
  417 }
  418 
  419 /*
  420  * This builds a pad record which the journaling thread will skip over.  Pad
  421  * records are required when we are unable to reserve sufficient stream space
  422  * due to insufficient space at the end of the physical memory fifo.
  423  *
  424  * Even though the record is not transmitted, a normal transid must be 
  425  * assigned to it so link recovery operations after a failure work properly.
  426  */
  427 static
  428 void
  429 journal_build_pad(struct journal_rawrecbeg *rawp, int recsize, int64_t transid)
  430 {
  431     struct journal_rawrecend *rendp;
  432     
  433     KKASSERT((recsize & 15) == 0 && recsize >= 16);
  434 
  435     rawp->streamid = JREC_STREAMID_PAD;
  436     rawp->recsize = recsize;    /* must be 16-byte aligned */
  437     rawp->transid = transid;
  438     /*
  439      * WARNING, rendp may overlap rawp->transid.  This is necessary to
  440      * allow PAD records to fit in 16 bytes.  Use cpu_ccfence() to
  441      * hopefully cause the compiler to not make any assumptions.
  442      */
  443     rendp = (void *)((char *)rawp + rawp->recsize - sizeof(*rendp));
  444     rendp->endmagic = JREC_ENDMAGIC;
  445     rendp->check = 0;
  446     rendp->recsize = rawp->recsize;
  447 
  448     /*
  449      * Set the begin magic last.  This is what will allow the journal
  450      * thread to write the record out.  Use a store fence to prevent
  451      * compiler and cpu reordering of the writes.
  452      */
  453     cpu_sfence();
  454     rawp->begmagic = JREC_BEGMAGIC;
  455 }
  456 
  457 /*
  458  * Wake up the worker thread if the FIFO is more then half full or if
  459  * someone is waiting for space to be freed up.  Otherwise let the 
  460  * heartbeat deal with it.  Being able to avoid waking up the worker
  461  * is the key to the journal's cpu performance.
  462  */
  463 static __inline
  464 void
  465 journal_commit_wakeup(struct journal *jo)
  466 {
  467     int avail;
  468 
  469     avail = jo->fifo.size - (jo->fifo.windex - jo->fifo.xindex);
  470     KKASSERT(avail >= 0);
  471     if ((avail < (jo->fifo.size >> 1)) || (jo->flags & MC_JOURNAL_WWAIT))
  472         wakeup(&jo->fifo);
  473 }
  474 
  475 /*
  476  * Create a new BEGIN stream record with the specified streamid and the
  477  * specified amount of payload space.  *rawpp will be set to point to the
  478  * base of the new stream record and a pointer to the base of the payload
  479  * space will be returned.  *rawpp does not need to be pre-NULLd prior to
  480  * making this call.  The raw record header will be partially initialized.
  481  *
  482  * A stream can be extended, aborted, or committed by other API calls
  483  * below.  This may result in a sequence of potentially disconnected
  484  * stream records to be output to the journaling target.  The first record
  485  * (the one created by this function) will be marked JREC_STREAMCTL_BEGIN,
  486  * while the last record on commit or abort will be marked JREC_STREAMCTL_END
  487  * (and possibly also JREC_STREAMCTL_ABORTED).  The last record could wind
  488  * up being the same as the first, in which case the bits are all set in
  489  * the first record.
  490  *
  491  * The stream record is created in an incomplete state by setting the begin
  492  * magic to JREC_INCOMPLETEMAGIC.  This prevents the worker thread from
  493  * flushing the fifo past our record until we have finished populating it.
  494  * Other threads can reserve and operate on their own space without stalling
  495  * but the stream output will stall until we have completed operations.  The
  496  * memory FIFO is intended to be large enough to absorb such situations
  497  * without stalling out other threads.
  498  */
  499 static
  500 void *
  501 journal_reserve(struct journal *jo, struct journal_rawrecbeg **rawpp,
  502                 int16_t streamid, int bytes)
  503 {
  504     struct journal_rawrecbeg *rawp;
  505     int avail;
  506     int availtoend;
  507     int req;
  508 
  509     /*
  510      * Add header and trailer overheads to the passed payload.  Note that
  511      * the passed payload size need not be aligned in any way.
  512      */
  513     bytes += sizeof(struct journal_rawrecbeg);
  514     bytes += sizeof(struct journal_rawrecend);
  515 
  516     for (;;) {
  517         /*
  518          * First, check boundary conditions.  If the request would wrap around
  519          * we have to skip past the ending block and return to the beginning
  520          * of the FIFO's buffer.  Calculate 'req' which is the actual number
  521          * of bytes being reserved, including wrap-around dead space.
  522          *
  523          * Neither 'bytes' or 'req' are aligned.
  524          *
  525          * Note that availtoend is not truncated to avail and so cannot be
  526          * used to determine whether the reservation is possible by itself.
  527          * Also, since all fifo ops are 16-byte aligned, we can check
  528          * the size before calculating the aligned size.
  529          */
  530         availtoend = jo->fifo.size - (jo->fifo.windex & jo->fifo.mask);
  531         KKASSERT((availtoend & 15) == 0);
  532         if (bytes > availtoend) 
  533             req = bytes + availtoend;   /* add pad to end */
  534         else
  535             req = bytes;
  536 
  537         /*
  538          * Next calculate the total available space and see if it is
  539          * sufficient.  We cannot overwrite previously buffered data
  540          * past xindex because otherwise we would not be able to restart
  541          * a broken link at the target's last point of commit.
  542          */
  543         avail = jo->fifo.size - (jo->fifo.windex - jo->fifo.xindex);
  544         KKASSERT(avail >= 0 && (avail & 15) == 0);
  545 
  546         if (avail < req) {
  547             /* XXX MC_JOURNAL_STOP_IMM */
  548             jo->flags |= MC_JOURNAL_WWAIT;
  549             ++jo->fifostalls;
  550             tsleep(&jo->fifo.windex, 0, "jwrite", 0);
  551             continue;
  552         }
  553 
  554         /*
  555          * Create a pad record for any dead space and create an incomplete
  556          * record for the live space, then return a pointer to the
  557          * contiguous buffer space that was requested.
  558          *
  559          * NOTE: The worker thread will not flush past an incomplete
  560          * record, so the reserved space can be filled in at-will.  The
  561          * journaling code must also be aware the reserved sections occuring
  562          * after this one will also not be written out even if completed
  563          * until this one is completed.
  564          *
  565          * The transaction id must accomodate real and potential pad creation.
  566          */
  567         rawp = (void *)(jo->fifo.membase + (jo->fifo.windex & jo->fifo.mask));
  568         if (req != bytes) {
  569             journal_build_pad(rawp, availtoend, jo->transid);
  570             ++jo->transid;
  571             rawp = (void *)jo->fifo.membase;
  572         }
  573         rawp->begmagic = JREC_INCOMPLETEMAGIC;  /* updated by abort/commit */
  574         rawp->recsize = bytes;                  /* (unaligned size) */
  575         rawp->streamid = streamid | JREC_STREAMCTL_BEGIN;
  576         rawp->transid = jo->transid;
  577         jo->transid += 2;
  578 
  579         /*
  580          * Issue a memory barrier to guarentee that the record data has been
  581          * properly initialized before we advance the write index and return
  582          * a pointer to the reserved record.  Otherwise the worker thread
  583          * could accidently run past us.
  584          *
  585          * Note that stream records are always 16-byte aligned.
  586          */
  587         cpu_sfence();
  588         jo->fifo.windex += (req + 15) & ~15;
  589         *rawpp = rawp;
  590         return(rawp + 1);
  591     }
  592     /* not reached */
  593     *rawpp = NULL;
  594     return(NULL);
  595 }
  596 
  597 /*
  598  * Attempt to extend the stream record by <bytes> worth of payload space.
  599  *
  600  * If it is possible to extend the existing stream record no truncation
  601  * occurs and the record is extended as specified.  A pointer to the 
  602  * truncation offset within the payload space is returned.
  603  *
  604  * If it is not possible to do this the existing stream record is truncated
  605  * and committed, and a new stream record of size <bytes> is created.  A
  606  * pointer to the base of the new stream record's payload space is returned.
  607  *
  608  * *rawpp is set to the new reservation in the case of a new record but
  609  * the caller cannot depend on a comparison with the old rawp to determine if
  610  * this case occurs because we could end up using the same memory FIFO
  611  * offset for the new stream record.  Use *newstreamrecp instead.
  612  */
  613 static void *
  614 journal_extend(struct journal *jo, struct journal_rawrecbeg **rawpp, 
  615                 int truncbytes, int bytes, int *newstreamrecp)
  616 {
  617     struct journal_rawrecbeg *rawp;
  618     int16_t streamid;
  619     int availtoend;
  620     int avail;
  621     int osize;
  622     int nsize;
  623     int wbase;
  624     void *rptr;
  625 
  626     *newstreamrecp = 0;
  627     rawp = *rawpp;
  628     osize = (rawp->recsize + 15) & ~15;
  629     nsize = (rawp->recsize + bytes + 15) & ~15;
  630     wbase = (char *)rawp - jo->fifo.membase;
  631 
  632     /*
  633      * If the aligned record size does not change we can trivially adjust
  634      * the record size.
  635      */
  636     if (nsize == osize) {
  637         rawp->recsize += bytes;
  638         return((char *)(rawp + 1) + truncbytes);
  639     }
  640 
  641     /*
  642      * If the fifo's write index hasn't been modified since we made the
  643      * reservation and we do not hit any boundary conditions, we can 
  644      * trivially make the record smaller or larger.
  645      */
  646     if ((jo->fifo.windex & jo->fifo.mask) == wbase + osize) {
  647         availtoend = jo->fifo.size - wbase;
  648         avail = jo->fifo.size - (jo->fifo.windex - jo->fifo.xindex) + osize;
  649         KKASSERT((availtoend & 15) == 0);
  650         KKASSERT((avail & 15) == 0);
  651         if (nsize <= avail && nsize <= availtoend) {
  652             jo->fifo.windex += nsize - osize;
  653             rawp->recsize += bytes;
  654             return((char *)(rawp + 1) + truncbytes);
  655         }
  656     }
  657 
  658     /*
  659      * It was not possible to extend the buffer.  Commit the current
  660      * buffer and create a new one.  We manually clear the BEGIN mark that
  661      * journal_reserve() creates (because this is a continuing record, not
  662      * the start of a new stream).
  663      */
  664     streamid = rawp->streamid & JREC_STREAMID_MASK;
  665     journal_commit(jo, rawpp, truncbytes, 0);
  666     rptr = journal_reserve(jo, rawpp, streamid, bytes);
  667     rawp = *rawpp;
  668     rawp->streamid &= ~JREC_STREAMCTL_BEGIN;
  669     *newstreamrecp = 1;
  670     return(rptr);
  671 }
  672 
  673 /*
  674  * Abort a journal record.  If the transaction record represents a stream
  675  * BEGIN and we can reverse the fifo's write index we can simply reverse
  676  * index the entire record, as if it were never reserved in the first place.
  677  *
  678  * Otherwise we set the JREC_STREAMCTL_ABORTED bit and commit the record
  679  * with the payload truncated to 0 bytes.
  680  */
  681 static void
  682 journal_abort(struct journal *jo, struct journal_rawrecbeg **rawpp)
  683 {
  684     struct journal_rawrecbeg *rawp;
  685     int osize;
  686 
  687     rawp = *rawpp;
  688     osize = (rawp->recsize + 15) & ~15;
  689 
  690     if ((rawp->streamid & JREC_STREAMCTL_BEGIN) &&
  691         (jo->fifo.windex & jo->fifo.mask) == 
  692          (char *)rawp - jo->fifo.membase + osize)
  693     {
  694         jo->fifo.windex -= osize;
  695         *rawpp = NULL;
  696     } else {
  697         rawp->streamid |= JREC_STREAMCTL_ABORTED;
  698         journal_commit(jo, rawpp, 0, 1);
  699     }
  700 }
  701 
  702 /*
  703  * Commit a journal record and potentially truncate it to the specified
  704  * number of payload bytes.  If you do not want to truncate the record,
  705  * simply pass -1 for the bytes parameter.  Do not pass rawp->recsize, that
  706  * field includes header and trailer and will not be correct.  Note that
  707  * passing 0 will truncate the entire data payload of the record.
  708  *
  709  * The logical stream is terminated by this function.
  710  *
  711  * If truncation occurs, and it is not possible to physically optimize the
  712  * memory FIFO due to other threads having reserved space after ours,
  713  * the remaining reserved space will be covered by a pad record.
  714  */
  715 static void
  716 journal_commit(struct journal *jo, struct journal_rawrecbeg **rawpp,
  717                 int bytes, int closeout)
  718 {
  719     struct journal_rawrecbeg *rawp;
  720     struct journal_rawrecend *rendp;
  721     int osize;
  722     int nsize;
  723 
  724     rawp = *rawpp;
  725     *rawpp = NULL;
  726 
  727     KKASSERT((char *)rawp >= jo->fifo.membase &&
  728              (char *)rawp + rawp->recsize <= jo->fifo.membase + jo->fifo.size);
  729     KKASSERT(((intptr_t)rawp & 15) == 0);
  730 
  731     /*
  732      * Truncate the record if necessary.  If the FIFO write index as still
  733      * at the end of our record we can optimally backindex it.  Otherwise
  734      * we have to insert a pad record to cover the dead space.
  735      *
  736      * We calculate osize which is the 16-byte-aligned original recsize.
  737      * We calculate nsize which is the 16-byte-aligned new recsize.
  738      *
  739      * Due to alignment issues or in case the passed truncation bytes is
  740      * the same as the original payload, nsize may be equal to osize even
  741      * if the committed bytes is less then the originally reserved bytes.
  742      */
  743     if (bytes >= 0) {
  744         KKASSERT(bytes >= 0 && bytes <= rawp->recsize - sizeof(struct journal_rawrecbeg) - sizeof(struct journal_rawrecend));
  745         osize = (rawp->recsize + 15) & ~15;
  746         rawp->recsize = bytes + sizeof(struct journal_rawrecbeg) +
  747                         sizeof(struct journal_rawrecend);
  748         nsize = (rawp->recsize + 15) & ~15;
  749         KKASSERT(nsize <= osize);
  750         if (osize == nsize) {
  751             /* do nothing */
  752         } else if ((jo->fifo.windex & jo->fifo.mask) == (char *)rawp - jo->fifo.membase + osize) {
  753             /* we are able to backindex the fifo */
  754             jo->fifo.windex -= osize - nsize;
  755         } else {
  756             /* we cannot backindex the fifo, emplace a pad in the dead space */
  757             journal_build_pad((void *)((char *)rawp + nsize), osize - nsize,
  758                                 rawp->transid + 1);
  759         }
  760     }
  761 
  762     /*
  763      * Fill in the trailer.  Note that unlike pad records, the trailer will
  764      * never overlap the header.
  765      */
  766     rendp = (void *)((char *)rawp + 
  767             ((rawp->recsize + 15) & ~15) - sizeof(*rendp));
  768     rendp->endmagic = JREC_ENDMAGIC;
  769     rendp->recsize = rawp->recsize;
  770     rendp->check = 0;           /* XXX check word, disabled for now */
  771 
  772     /*
  773      * Fill in begmagic last.  This will allow the worker thread to proceed.
  774      * Use a memory barrier to guarentee write ordering.  Mark the stream
  775      * as terminated if closeout is set.  This is the typical case.
  776      */
  777     if (closeout)
  778         rawp->streamid |= JREC_STREAMCTL_END;
  779     cpu_sfence();               /* memory and compiler barrier */
  780     rawp->begmagic = JREC_BEGMAGIC;
  781 
  782     journal_commit_wakeup(jo);
  783 }
  784 
  785 /************************************************************************
  786  *                      TRANSACTION SUPPORT ROUTINES                    *
  787  ************************************************************************
  788  *
  789  * JRECORD_*() - routines to create subrecord transactions and embed them
  790  *               in the logical streams managed by the journal_*() routines.
  791  */
  792 
  793 /*
  794  * Initialize the passed jrecord structure and start a new stream transaction
  795  * by reserving an initial build space in the journal's memory FIFO.
  796  */
  797 void
  798 jrecord_init(struct journal *jo, struct jrecord *jrec, int16_t streamid)
  799 {
  800     bzero(jrec, sizeof(*jrec));
  801     jrec->jo = jo;
  802     jrec->streamid = streamid;
  803     jrec->stream_residual = JREC_DEFAULTSIZE;
  804     jrec->stream_reserved = jrec->stream_residual;
  805     jrec->stream_ptr = 
  806         journal_reserve(jo, &jrec->rawp, streamid, jrec->stream_reserved);
  807 }
  808 
  809 /*
  810  * Push a recursive record type.  All pushes should have matching pops.
  811  * The old parent is returned and the newly pushed record becomes the
  812  * new parent.  Note that the old parent's pointer may already be invalid
  813  * or may become invalid if jrecord_write() had to build a new stream
  814  * record, so the caller should not mess with the returned pointer in
  815  * any way other then to save it.
  816  */
  817 struct journal_subrecord *
  818 jrecord_push(struct jrecord *jrec, int16_t rectype)
  819 {
  820     struct journal_subrecord *save;
  821 
  822     save = jrec->parent;
  823     jrec->parent = jrecord_write(jrec, rectype|JMASK_NESTED, 0);
  824     jrec->last = NULL;
  825     KKASSERT(jrec->parent != NULL);
  826     ++jrec->pushcount;
  827     ++jrec->pushptrgood;        /* cleared on flush */
  828     return(save);
  829 }
  830 
  831 /*
  832  * Pop a previously pushed sub-transaction.  We must set JMASK_LAST
  833  * on the last record written within the subtransaction.  If the last 
  834  * record written is not accessible or if the subtransaction is empty,
  835  * we must write out a pad record with JMASK_LAST set before popping.
  836  *
  837  * When popping a subtransaction the parent record's recsize field
  838  * will be properly set.  If the parent pointer is no longer valid
  839  * (which can occur if the data has already been flushed out to the
  840  * stream), the protocol spec allows us to leave it 0.
  841  *
  842  * The saved parent pointer which we restore may or may not be valid,
  843  * and if not valid may or may not be NULL, depending on the value
  844  * of pushptrgood.
  845  */
  846 void
  847 jrecord_pop(struct jrecord *jrec, struct journal_subrecord *save)
  848 {
  849     struct journal_subrecord *last;
  850 
  851     KKASSERT(jrec->pushcount > 0);
  852     KKASSERT(jrec->residual == 0);
  853 
  854     /*
  855      * Set JMASK_LAST on the last record we wrote at the current
  856      * level.  If last is NULL we either no longer have access to the
  857      * record or the subtransaction was empty and we must write out a pad
  858      * record.
  859      */
  860     if ((last = jrec->last) == NULL) {
  861         jrecord_write(jrec, JLEAF_PAD|JMASK_LAST, 0);
  862         last = jrec->last;      /* reload after possible flush */
  863     } else {
  864         last->rectype |= JMASK_LAST;
  865     }
  866 
  867     /*
  868      * pushptrgood tells us how many levels of parent record pointers
  869      * are valid.  The jrec only stores the current parent record pointer
  870      * (and it is only valid if pushptrgood != 0).  The higher level parent
  871      * record pointers are saved by the routines calling jrecord_push() and
  872      * jrecord_pop().  These pointers may become stale and we determine
  873      * that fact by tracking the count of valid parent pointers with 
  874      * pushptrgood.  Pointers become invalid when their related stream
  875      * record gets pushed out.
  876      *
  877      * If no pointer is available (the data has already been pushed out),
  878      * then no fixup of e.g. the length field is possible for non-leaf
  879      * nodes.  The protocol allows for this situation by placing a larger
  880      * burden on the program scanning the stream on the other end.
  881      *
  882      * [parentA]
  883      *    [node X]
  884      *    [parentB]
  885      *       [node Y]
  886      *       [node Z]
  887      *    (pop B)       see NOTE B
  888      * (pop A)          see NOTE A
  889      *
  890      * NOTE B:  This pop sets LAST in node Z if the node is still accessible,
  891      *          else a PAD record is appended and LAST is set in that.
  892      *
  893      *          This pop sets the record size in parentB if parentB is still
  894      *          accessible, else the record size is left 0 (the scanner must
  895      *          deal with that).
  896      *
  897      *          This pop sets the new 'last' record to parentB, the pointer
  898      *          to which may or may not still be accessible.
  899      *
  900      * NOTE A:  This pop sets LAST in parentB if the node is still accessible,
  901      *          else a PAD record is appended and LAST is set in that.
  902      *
  903      *          This pop sets the record size in parentA if parentA is still
  904      *          accessible, else the record size is left 0 (the scanner must
  905      *          deal with that).
  906      *
  907      *          This pop sets the new 'last' record to parentA, the pointer
  908      *          to which may or may not still be accessible.
  909      *
  910      * Also note that the last record in the stream transaction, which in
  911      * the above example is parentA, does not currently have the LAST bit
  912      * set.
  913      *
  914      * The current parent becomes the last record relative to the
  915      * saved parent passed into us.  It's validity is based on 
  916      * whether pushptrgood is non-zero prior to decrementing.  The saved
  917      * parent becomes the new parent, and its validity is based on whether
  918      * pushptrgood is non-zero after decrementing.
  919      *
  920      * The old jrec->parent may be NULL if it is no longer accessible.
  921      * If pushptrgood is non-zero, however, it is guarenteed to not
  922      * be NULL (since no flush occured).
  923      */
  924     jrec->last = jrec->parent;
  925     --jrec->pushcount;
  926     if (jrec->pushptrgood) {
  927         KKASSERT(jrec->last != NULL && last != NULL);
  928         if (--jrec->pushptrgood == 0) {
  929             jrec->parent = NULL;        /* 'save' contains garbage or NULL */
  930         } else {
  931             KKASSERT(save != NULL);
  932             jrec->parent = save;        /* 'save' must not be NULL */
  933         }
  934 
  935         /*
  936          * Set the record size in the old parent.  'last' still points to
  937          * the original last record in the subtransaction being popped,
  938          * jrec->last points to the old parent (which became the last
  939          * record relative to the new parent being popped into).
  940          */
  941         jrec->last->recsize = (char *)last + last->recsize - (char *)jrec->last;
  942     } else {
  943         jrec->parent = NULL;
  944         KKASSERT(jrec->last == NULL);
  945     }
  946 }
  947 
  948 /*
  949  * Write out a leaf record, including associated data.
  950  */
  951 void
  952 jrecord_leaf(struct jrecord *jrec, int16_t rectype, void *ptr, int bytes)
  953 {
  954     jrecord_write(jrec, rectype, bytes);
  955     jrecord_data(jrec, ptr, bytes, JDATA_KERN);
  956 }
  957 
  958 void
  959 jrecord_leaf_uio(struct jrecord *jrec, int16_t rectype,
  960                  struct uio *uio)
  961 {
  962     struct iovec *iov;
  963     int i;
  964 
  965     for (i = 0; i < uio->uio_iovcnt; ++i) {
  966         iov = &uio->uio_iov[i];
  967         if (iov->iov_len == 0)
  968             continue;
  969         if (uio->uio_segflg == UIO_SYSSPACE) {
  970             jrecord_write(jrec, rectype, iov->iov_len);
  971             jrecord_data(jrec, iov->iov_base, iov->iov_len, JDATA_KERN);
  972         } else { /* UIO_USERSPACE */
  973             jrecord_write(jrec, rectype, iov->iov_len);
  974             jrecord_data(jrec, iov->iov_base, iov->iov_len, JDATA_USER);
  975         }
  976     }
  977 }
  978 
  979 void
  980 jrecord_leaf_xio(struct jrecord *jrec, int16_t rectype, xio_t xio)
  981 {
  982     int bytes = xio->xio_npages * PAGE_SIZE;
  983 
  984     jrecord_write(jrec, rectype, bytes);
  985     jrecord_data(jrec, xio, bytes, JDATA_XIO);
  986 }
  987 
  988 /*
  989  * Write a leaf record out and return a pointer to its base.  The leaf
  990  * record may contain potentially megabytes of data which is supplied
  991  * in jrecord_data() calls.  The exact amount must be specified in this
  992  * call.
  993  *
  994  * THE RETURNED SUBRECORD POINTER IS ONLY VALID IMMEDIATELY AFTER THE
  995  * CALL AND MAY BECOME INVALID AT ANY TIME.  ONLY THE PUSH/POP CODE SHOULD
  996  * USE THE RETURN VALUE.
  997  */
  998 struct journal_subrecord *
  999 jrecord_write(struct jrecord *jrec, int16_t rectype, int bytes)
 1000 {
 1001     struct journal_subrecord *last;
 1002     int pusheditout;
 1003 
 1004     /*
 1005      * Try to catch some obvious errors.  Nesting records must specify a
 1006      * size of 0, and there should be no left-overs from previous operations
 1007      * (such as incomplete data writeouts).
 1008      */
 1009     KKASSERT(bytes == 0 || (rectype & JMASK_NESTED) == 0);
 1010     KKASSERT(jrec->residual == 0);
 1011 
 1012     /*
 1013      * Check to see if the current stream record has enough room for
 1014      * the new subrecord header.  If it doesn't we extend the current
 1015      * stream record.
 1016      *
 1017      * This may have the side effect of pushing out the current stream record
 1018      * and creating a new one.  We must adjust our stream tracking fields
 1019      * accordingly.
 1020      */
 1021     if (jrec->stream_residual < sizeof(struct journal_subrecord)) {
 1022         jrec->stream_ptr = journal_extend(jrec->jo, &jrec->rawp,
 1023                                 jrec->stream_reserved - jrec->stream_residual,
 1024                                 JREC_DEFAULTSIZE, &pusheditout);
 1025         if (pusheditout) {
 1026             /*
 1027              * If a pushout occured, the pushed out stream record was
 1028              * truncated as specified and the new record is exactly the
 1029              * extension size specified.
 1030              */
 1031             jrec->stream_reserved = JREC_DEFAULTSIZE;
 1032             jrec->stream_residual = JREC_DEFAULTSIZE;
 1033             jrec->parent = NULL;        /* no longer accessible */
 1034             jrec->pushptrgood = 0;      /* restored parents in pops no good */
 1035         } else {
 1036             /*
 1037              * If no pushout occured the stream record is NOT truncated and
 1038              * IS extended.
 1039              */
 1040             jrec->stream_reserved += JREC_DEFAULTSIZE;
 1041             jrec->stream_residual += JREC_DEFAULTSIZE;
 1042         }
 1043     }
 1044     last = (void *)jrec->stream_ptr;
 1045     last->rectype = rectype;
 1046     last->reserved = 0;
 1047 
 1048     /*
 1049      * We may not know the record size for recursive records and the 
 1050      * header may become unavailable due to limited FIFO space.  Write
 1051      * -1 to indicate this special case.
 1052      */
 1053     if ((rectype & JMASK_NESTED) && bytes == 0)
 1054         last->recsize = -1;
 1055     else
 1056         last->recsize = sizeof(struct journal_subrecord) + bytes;
 1057     jrec->last = last;
 1058     jrec->residual = bytes;             /* remaining data to be posted */
 1059     jrec->residual_align = -bytes & 7;  /* post-data alignment required */
 1060     jrec->stream_ptr += sizeof(*last);  /* current write pointer */
 1061     jrec->stream_residual -= sizeof(*last); /* space remaining in stream */
 1062     return(last);
 1063 }
 1064 
 1065 /*
 1066  * Write out the data associated with a leaf record.  Any number of calls
 1067  * to this routine may be made as long as the byte count adds up to the
 1068  * amount originally specified in jrecord_write().
 1069  *
 1070  * The act of writing out the leaf data may result in numerous stream records
 1071  * being pushed out.   Callers should be aware that even the associated
 1072  * subrecord header may become inaccessible due to stream record pushouts.
 1073  */
 1074 static void
 1075 jrecord_data(struct jrecord *jrec, void *buf, int bytes, int dtype)
 1076 {
 1077     int pusheditout;
 1078     int extsize;
 1079     int xio_offset = 0;
 1080 
 1081     KKASSERT(bytes >= 0 && bytes <= jrec->residual);
 1082 
 1083     /*
 1084      * Push out stream records as long as there is insufficient room to hold
 1085      * the remaining data.
 1086      */
 1087     while (jrec->stream_residual < bytes) {
 1088         /*
 1089          * Fill in any remaining space in the current stream record.
 1090          */
 1091         switch (dtype) {
 1092         case JDATA_KERN:
 1093             bcopy(buf, jrec->stream_ptr, jrec->stream_residual);
 1094             break;
 1095         case JDATA_USER:
 1096             copyin(buf, jrec->stream_ptr, jrec->stream_residual);
 1097             break;
 1098         case JDATA_XIO:
 1099             xio_copy_xtok((xio_t)buf, xio_offset, jrec->stream_ptr,
 1100                           jrec->stream_residual);
 1101             xio_offset += jrec->stream_residual;
 1102             break;
 1103         }
 1104         if (dtype != JDATA_XIO)
 1105             buf = (char *)buf + jrec->stream_residual;
 1106         bytes -= jrec->stream_residual;
 1107         /*jrec->stream_ptr += jrec->stream_residual;*/
 1108         jrec->residual -= jrec->stream_residual;
 1109         jrec->stream_residual = 0;
 1110 
 1111         /*
 1112          * Try to extend the current stream record, but no more then 1/4
 1113          * the size of the FIFO.
 1114          */
 1115         extsize = jrec->jo->fifo.size >> 2;
 1116         if (extsize > bytes)
 1117             extsize = (bytes + 15) & ~15;
 1118 
 1119         jrec->stream_ptr = journal_extend(jrec->jo, &jrec->rawp,
 1120                                 jrec->stream_reserved - jrec->stream_residual,
 1121                                 extsize, &pusheditout);
 1122         if (pusheditout) {
 1123             jrec->stream_reserved = extsize;
 1124             jrec->stream_residual = extsize;
 1125             jrec->parent = NULL;        /* no longer accessible */
 1126             jrec->last = NULL;          /* no longer accessible */
 1127             jrec->pushptrgood = 0;      /* restored parents in pops no good */
 1128         } else {
 1129             jrec->stream_reserved += extsize;
 1130             jrec->stream_residual += extsize;
 1131         }
 1132     }
 1133 
 1134     /*
 1135      * Push out any remaining bytes into the current stream record.
 1136      */
 1137     if (bytes) {
 1138         switch (dtype) {
 1139         case JDATA_KERN:
 1140             bcopy(buf, jrec->stream_ptr, bytes);
 1141             break;
 1142         case JDATA_USER:
 1143             copyin(buf, jrec->stream_ptr, bytes);
 1144             break;
 1145         case JDATA_XIO:
 1146             xio_copy_xtok((xio_t)buf, xio_offset, jrec->stream_ptr, bytes);
 1147             break;
 1148         }
 1149         jrec->stream_ptr += bytes;
 1150         jrec->stream_residual -= bytes;
 1151         jrec->residual -= bytes;
 1152     }
 1153 
 1154     /*
 1155      * Handle data alignment requirements for the subrecord.  Because the
 1156      * stream record's data space is more strictly aligned, it must already
 1157      * have sufficient space to hold any subrecord alignment slop.
 1158      */
 1159     if (jrec->residual == 0 && jrec->residual_align) {
 1160         KKASSERT(jrec->residual_align <= jrec->stream_residual);
 1161         bzero(jrec->stream_ptr, jrec->residual_align);
 1162         jrec->stream_ptr += jrec->residual_align;
 1163         jrec->stream_residual -= jrec->residual_align;
 1164         jrec->residual_align = 0;
 1165     }
 1166 }
 1167 
 1168 /*
 1169  * We are finished with the transaction.  This closes the transaction created
 1170  * by jrecord_init().
 1171  *
 1172  * NOTE: If abortit is not set then we must be at the top level with no
 1173  *       residual subrecord data left to output.
 1174  *
 1175  *       If abortit is set then we can be in any state, all pushes will be 
 1176  *       popped and it is ok for there to be residual data.  This works 
 1177  *       because the virtual stream itself is truncated.  Scanners must deal
 1178  *       with this situation.
 1179  *
 1180  * The stream record will be committed or aborted as specified and jrecord
 1181  * resources will be cleaned up.
 1182  */
 1183 void
 1184 jrecord_done(struct jrecord *jrec, int abortit)
 1185 {
 1186     KKASSERT(jrec->rawp != NULL);
 1187 
 1188     if (abortit) {
 1189         journal_abort(jrec->jo, &jrec->rawp);
 1190     } else {
 1191         KKASSERT(jrec->pushcount == 0 && jrec->residual == 0);
 1192         journal_commit(jrec->jo, &jrec->rawp, 
 1193                         jrec->stream_reserved - jrec->stream_residual, 1);
 1194     }
 1195 
 1196     /*
 1197      * jrec should not be used beyond this point without another init,
 1198      * but clean up some fields to ensure that we panic if it is.
 1199      *
 1200      * Note that jrec->rawp is NULLd out by journal_abort/journal_commit.
 1201      */
 1202     jrec->jo = NULL;
 1203     jrec->stream_ptr = NULL;
 1204 }
 1205 
 1206 /************************************************************************
 1207  *                      LOW LEVEL RECORD SUPPORT ROUTINES               *
 1208  ************************************************************************
 1209  *
 1210  * These routine create low level recursive and leaf subrecords representing
 1211  * common filesystem structures.
 1212  */
 1213 
 1214 /*
 1215  * Write out a filename path relative to the base of the mount point.
 1216  * rectype is typically JLEAF_PATH{1,2,3,4}.
 1217  */
 1218 void
 1219 jrecord_write_path(struct jrecord *jrec, int16_t rectype, struct namecache *ncp)
 1220 {
 1221     char buf[64];       /* local buffer if it fits, else malloced */
 1222     char *base;
 1223     int pathlen;
 1224     int index;
 1225     struct namecache *scan;
 1226 
 1227     /*
 1228      * Pass 1 - figure out the number of bytes required.  Include terminating
 1229      *         \0 on last element and '/' separator on other elements.
 1230      *
 1231      * The namecache topology terminates at the root of the filesystem
 1232      * (the normal lookup code would then continue by using the mount
 1233      * structure to figure out what it was mounted on).
 1234      */
 1235 again:
 1236     pathlen = 0;
 1237     for (scan = ncp; scan; scan = scan->nc_parent) {
 1238         if (scan->nc_nlen > 0)
 1239             pathlen += scan->nc_nlen + 1;
 1240     }
 1241 
 1242     if (pathlen <= sizeof(buf))
 1243         base = buf;
 1244     else
 1245         base = kmalloc(pathlen, M_TEMP, M_INTWAIT);
 1246 
 1247     /*
 1248      * Pass 2 - generate the path buffer
 1249      */
 1250     index = pathlen;
 1251     for (scan = ncp; scan; scan = scan->nc_parent) {
 1252         if (scan->nc_nlen == 0)
 1253             continue;
 1254         if (scan->nc_nlen >= index) {
 1255             if (base != buf)
 1256                 kfree(base, M_TEMP);
 1257             goto again;
 1258         }
 1259         if (index == pathlen)
 1260             base[--index] = 0;
 1261         else
 1262             base[--index] = '/';
 1263         index -= scan->nc_nlen;
 1264         bcopy(scan->nc_name, base + index, scan->nc_nlen);
 1265     }
 1266     jrecord_leaf(jrec, rectype, base + index, pathlen - index);
 1267     if (base != buf)
 1268         kfree(base, M_TEMP);
 1269 }
 1270 
 1271 /*
 1272  * Write out a file attribute structure.  While somewhat inefficient, using
 1273  * a recursive data structure is the most portable and extensible way.
 1274  */
 1275 void
 1276 jrecord_write_vattr(struct jrecord *jrec, struct vattr *vat)
 1277 {
 1278     void *save;
 1279 
 1280     save = jrecord_push(jrec, JTYPE_VATTR);
 1281     if (vat->va_type != VNON)
 1282         jrecord_leaf(jrec, JLEAF_VTYPE, &vat->va_type, sizeof(vat->va_type));
 1283     if (vat->va_mode != (mode_t)VNOVAL)
 1284         jrecord_leaf(jrec, JLEAF_MODES, &vat->va_mode, sizeof(vat->va_mode));
 1285     if (vat->va_nlink != VNOVAL)
 1286         jrecord_leaf(jrec, JLEAF_NLINK, &vat->va_nlink, sizeof(vat->va_nlink));
 1287     if (vat->va_uid != VNOVAL)
 1288         jrecord_leaf(jrec, JLEAF_UID, &vat->va_uid, sizeof(vat->va_uid));
 1289     if (vat->va_gid != VNOVAL)
 1290         jrecord_leaf(jrec, JLEAF_GID, &vat->va_gid, sizeof(vat->va_gid));
 1291     if (vat->va_fsid != VNOVAL)
 1292         jrecord_leaf(jrec, JLEAF_FSID, &vat->va_fsid, sizeof(vat->va_fsid));
 1293     if (vat->va_fileid != VNOVAL)
 1294         jrecord_leaf(jrec, JLEAF_INUM, &vat->va_fileid, sizeof(vat->va_fileid));
 1295     if (vat->va_size != VNOVAL)
 1296         jrecord_leaf(jrec, JLEAF_SIZE, &vat->va_size, sizeof(vat->va_size));
 1297     if (vat->va_atime.tv_sec != VNOVAL)
 1298         jrecord_leaf(jrec, JLEAF_ATIME, &vat->va_atime, sizeof(vat->va_atime));
 1299     if (vat->va_mtime.tv_sec != VNOVAL)
 1300         jrecord_leaf(jrec, JLEAF_MTIME, &vat->va_mtime, sizeof(vat->va_mtime));
 1301     if (vat->va_ctime.tv_sec != VNOVAL)
 1302         jrecord_leaf(jrec, JLEAF_CTIME, &vat->va_ctime, sizeof(vat->va_ctime));
 1303     if (vat->va_gen != VNOVAL)
 1304         jrecord_leaf(jrec, JLEAF_GEN, &vat->va_gen, sizeof(vat->va_gen));
 1305     if (vat->va_flags != VNOVAL)
 1306         jrecord_leaf(jrec, JLEAF_FLAGS, &vat->va_flags, sizeof(vat->va_flags));
 1307     if (vat->va_rmajor != VNOVAL) {
 1308         udev_t rdev = makeudev(vat->va_rmajor, vat->va_rminor);
 1309         jrecord_leaf(jrec, JLEAF_UDEV, &rdev, sizeof(rdev));
 1310         jrecord_leaf(jrec, JLEAF_UMAJOR, &vat->va_rmajor, sizeof(vat->va_rmajor));
 1311         jrecord_leaf(jrec, JLEAF_UMINOR, &vat->va_rminor, sizeof(vat->va_rminor));
 1312     }
 1313 #if 0
 1314     if (vat->va_filerev != VNOVAL)
 1315         jrecord_leaf(jrec, JLEAF_FILEREV, &vat->va_filerev, sizeof(vat->va_filerev));
 1316 #endif
 1317     jrecord_pop(jrec, save);
 1318 }
 1319 
 1320 /*
 1321  * Write out the creds used to issue a file operation.  If a process is
 1322  * available write out additional tracking information related to the 
 1323  * process.
 1324  *
 1325  * XXX additional tracking info
 1326  * XXX tty line info
 1327  */
 1328 void
 1329 jrecord_write_cred(struct jrecord *jrec, struct thread *td, struct ucred *cred)
 1330 {
 1331     void *save;
 1332     struct proc *p;
 1333 
 1334     save = jrecord_push(jrec, JTYPE_CRED);
 1335     jrecord_leaf(jrec, JLEAF_UID, &cred->cr_uid, sizeof(cred->cr_uid));
 1336     jrecord_leaf(jrec, JLEAF_GID, &cred->cr_gid, sizeof(cred->cr_gid));
 1337     if (td && (p = td->td_proc) != NULL) {
 1338         jrecord_leaf(jrec, JLEAF_PID, &p->p_pid, sizeof(p->p_pid));
 1339         jrecord_leaf(jrec, JLEAF_COMM, p->p_comm, sizeof(p->p_comm));
 1340     }
 1341     jrecord_pop(jrec, save);
 1342 }
 1343 
 1344 /*
 1345  * Write out information required to identify a vnode
 1346  *
 1347  * XXX this needs work.  We should write out the inode number as well,
 1348  * and in fact avoid writing out the file path for seqential writes
 1349  * occuring within e.g. a certain period of time.
 1350  */
 1351 void
 1352 jrecord_write_vnode_ref(struct jrecord *jrec, struct vnode *vp)
 1353 {
 1354     struct nchandle nch;
 1355 
 1356     nch.mount = vp->v_mount;
 1357     spin_lock(&vp->v_spin);
 1358     TAILQ_FOREACH(nch.ncp, &vp->v_namecache, nc_vnode) {
 1359         if ((nch.ncp->nc_flag & (NCF_UNRESOLVED|NCF_DESTROYED)) == 0)
 1360             break;
 1361     }
 1362     if (nch.ncp) {
 1363         cache_hold(&nch);
 1364         spin_unlock(&vp->v_spin);
 1365         jrecord_write_path(jrec, JLEAF_PATH_REF, nch.ncp);
 1366         cache_drop(&nch);
 1367     } else {
 1368         spin_unlock(&vp->v_spin);
 1369     }
 1370 }
 1371 
 1372 void
 1373 jrecord_write_vnode_link(struct jrecord *jrec, struct vnode *vp, 
 1374                          struct namecache *notncp)
 1375 {
 1376     struct nchandle nch;
 1377 
 1378     nch.mount = vp->v_mount;
 1379     spin_lock(&vp->v_spin);
 1380     TAILQ_FOREACH(nch.ncp, &vp->v_namecache, nc_vnode) {
 1381         if (nch.ncp == notncp)
 1382             continue;
 1383         if ((nch.ncp->nc_flag & (NCF_UNRESOLVED|NCF_DESTROYED)) == 0)
 1384             break;
 1385     }
 1386     if (nch.ncp) {
 1387         cache_hold(&nch);
 1388         spin_unlock(&vp->v_spin);
 1389         jrecord_write_path(jrec, JLEAF_PATH_REF, nch.ncp);
 1390         cache_drop(&nch);
 1391     } else {
 1392         spin_unlock(&vp->v_spin);
 1393     }
 1394 }
 1395 
 1396 /*
 1397  * Write out the data represented by a pagelist
 1398  */
 1399 void
 1400 jrecord_write_pagelist(struct jrecord *jrec, int16_t rectype,
 1401                         struct vm_page **pglist, int *rtvals, int pgcount,
 1402                         off_t offset)
 1403 {
 1404     struct xio xio;
 1405     int error;
 1406     int b;
 1407     int i;
 1408 
 1409     i = 0;
 1410     xio_init(&xio);
 1411     while (i < pgcount) {
 1412         /*
 1413          * Find the next valid section.  Skip any invalid elements
 1414          */
 1415         if (rtvals[i] != VM_PAGER_OK) {
 1416             ++i;
 1417             offset += PAGE_SIZE;
 1418             continue;
 1419         }
 1420 
 1421         /*
 1422          * Figure out how big the valid section is, capping I/O at what the
 1423          * MSFBUF can represent.
 1424          */
 1425         b = i;
 1426         while (i < pgcount && i - b != XIO_INTERNAL_PAGES && 
 1427                rtvals[i] == VM_PAGER_OK
 1428         ) {
 1429             ++i;
 1430         }
 1431 
 1432         /*
 1433          * And write it out.
 1434          */
 1435         if (i - b) {
 1436             error = xio_init_pages(&xio, pglist + b, i - b, XIOF_READ);
 1437             if (error == 0) {
 1438                 jrecord_leaf(jrec, JLEAF_SEEKPOS, &offset, sizeof(offset));
 1439                 jrecord_leaf_xio(jrec, rectype, &xio);
 1440             } else {
 1441                 kprintf("jrecord_write_pagelist: xio init failure\n");
 1442             }
 1443             xio_release(&xio);
 1444             offset += (off_t)(i - b) << PAGE_SHIFT;
 1445         }
 1446     }
 1447 }
 1448 
 1449 /*
 1450  * Write out the data represented by a UIO.
 1451  */
 1452 void
 1453 jrecord_write_uio(struct jrecord *jrec, int16_t rectype, struct uio *uio)
 1454 {
 1455     if (uio->uio_segflg != UIO_NOCOPY) {
 1456         jrecord_leaf(jrec, JLEAF_SEEKPOS, &uio->uio_offset, 
 1457                      sizeof(uio->uio_offset));
 1458         jrecord_leaf_uio(jrec, rectype, uio);
 1459     }
 1460 }
 1461 
 1462 void
 1463 jrecord_file_data(struct jrecord *jrec, struct vnode *vp, 
 1464                   off_t off, off_t bytes)
 1465 {
 1466     const int bufsize = 8192;
 1467     char *buf;
 1468     int error;
 1469     int n;
 1470 
 1471     buf = kmalloc(bufsize, M_JOURNAL, M_WAITOK);
 1472     jrecord_leaf(jrec, JLEAF_SEEKPOS, &off, sizeof(off));
 1473     while (bytes) {
 1474         n = (bytes > bufsize) ? bufsize : (int)bytes;
 1475         error = vn_rdwr(UIO_READ, vp, buf, n, off, UIO_SYSSPACE, IO_NODELOCKED,
 1476                         proc0.p_ucred, NULL);
 1477         if (error) {
 1478             jrecord_leaf(jrec, JLEAF_ERROR, &error, sizeof(error));
 1479             break;
 1480         }
 1481         jrecord_leaf(jrec, JLEAF_FILEDATA, buf, n);
 1482         bytes -= n;
 1483         off += n;
 1484     }
 1485     kfree(buf, M_JOURNAL);
 1486 }
 1487 

Cache object: a8e547e19612a98f0f45425e914369a1


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]


This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.