buffile.c

Go to the documentation of this file.

/*-------------------------------------------------------------------------
 *
 * buffile.c
 *    Management of large buffered temporary files.
 *
 * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group
 * Portions Copyright (c) 1994, Regents of the University of California
 *
 * IDENTIFICATION
 *    src/backend/storage/file/buffile.c
 *
 * NOTES:
 *
 * BufFiles provide a very incomplete emulation of stdio atop virtual Files
 * (as managed by fd.c).  Currently, we only support the buffered-I/O
 * aspect of stdio: a read or write of the low-level File occurs only
 * when the buffer is filled or emptied.  This is an even bigger win
 * for virtual Files than for ordinary kernel files, since reducing the
 * frequency with which a virtual File is touched reduces "thrashing"
 * of opening/closing file descriptors.
 *
 * Note that BufFile structs are allocated with palloc(), and therefore
 * will go away automatically at query/transaction end.  Since the underlying
 * virtual Files are made with OpenTemporaryFile, all resources for
 * the file are certain to be cleaned up even if processing is aborted
 * by ereport(ERROR).  The data structures required are made in the
 * palloc context that was current when the BufFile was created, and
 * any external resources such as temp files are owned by the ResourceOwner
 * that was current at that time.
 *
 * BufFile also supports temporary files that exceed the OS file size limit
 * (by opening multiple fd.c temporary files).  This is an essential feature
 * for sorts and hashjoins on large amounts of data.
 *
 * BufFile supports temporary files that can be shared with other backends, as
 * infrastructure for parallel execution.  Such files need to be created as a
 * member of a SharedFileSet that all participants are attached to.
 *
 * BufFile also supports temporary files that can be used by the single backend
 * when the corresponding files need to be survived across the transaction and
 * need to be opened and closed multiple times.  Such files need to be created
 * as a member of a FileSet.
 *-------------------------------------------------------------------------
 */
 
#include "postgres.h"
 
#include "commands/tablespace.h"
#include "executor/instrument.h"
#include "miscadmin.h"
#include "pgstat.h"
#include "storage/buffile.h"
#include "storage/bufmgr.h"
#include "storage/fd.h"
#include "utils/resowner.h"
#include "utils/wait_event.h"
 
/*
 * We break BufFiles into gigabyte-sized segments, regardless of RELSEG_SIZE.
 * The reason is that we'd like large BufFiles to be spread across multiple
 * tablespaces when available.
 */
#define MAX_PHYSICAL_FILESIZE   0x40000000
#define BUFFILE_SEG_SIZE        (MAX_PHYSICAL_FILESIZE / BLCKSZ)
 
/*
 * This data structure represents a buffered file that consists of one or
 * more physical files (each accessed through a virtual file descriptor
 * managed by fd.c).
 */
struct BufFile
{
    int         numFiles;       /* number of physical files in set */
    /* all files except the last have length exactly MAX_PHYSICAL_FILESIZE */
    File       *files;          /* palloc'd array with numFiles entries */
 
    bool        isInterXact;    /* keep open over transactions? */
    bool        dirty;          /* does buffer need to be written? */
    bool        readOnly;       /* has the file been set to read only? */
 
    FileSet    *fileset;        /* space for fileset based segment files */
    const char *name;           /* name of fileset based BufFile */
 
    /*
     * resowner is the ResourceOwner to use for underlying temp files.  (We
     * don't need to remember the memory context we're using explicitly,
     * because after creation we only repalloc our arrays larger.)
     */
    ResourceOwner resowner;
 
    /*
     * "current pos" is position of start of buffer within the logical file.
     * Position as seen by user of BufFile is (curFile, curOffset + pos).
     */
    int         curFile;        /* file index (0..n) part of current pos */
    pgoff_t     curOffset;      /* offset part of current pos */
    int64       pos;            /* next read/write position in buffer */
    int64       nbytes;         /* total # of valid bytes in buffer */
 
    /*
     * XXX Should ideally use PGIOAlignedBlock, but might need a way to avoid
     * wasting per-file alignment padding when some users create many files.
     */
    PGAlignedBlock buffer;
};
 
static BufFile *makeBufFileCommon(int nfiles);
static BufFile *makeBufFile(File firstfile);
static void extendBufFile(BufFile *file);
static void BufFileLoadBuffer(BufFile *file);
static void BufFileDumpBuffer(BufFile *file);
static void BufFileFlush(BufFile *file);
static File MakeNewFileSetSegment(BufFile *buffile, int segment);
 
/*
 * Create BufFile and perform the common initialization.
 */
static BufFile *
makeBufFileCommon(int nfiles)
{
    BufFile    *file = palloc_object(BufFile);
 
    file->numFiles = nfiles;
    file->isInterXact = false;
    file->dirty = false;
    file->resowner = CurrentResourceOwner;
    file->curFile = 0;
    file->curOffset = 0;
    file->pos = 0;
    file->nbytes = 0;
 
    return file;
}
 
/*
 * Create a BufFile given the first underlying physical file.
 * NOTE: caller must set isInterXact if appropriate.
 */
static BufFile *
makeBufFile(File firstfile)
{
    BufFile    *file = makeBufFileCommon(1);
 
    file->files = palloc_object(File);
    file->files[0] = firstfile;
    file->readOnly = false;
    file->fileset = NULL;
    file->name = NULL;
 
    return file;
}
 
/*
 * Add another component temp file.
 */
static void
extendBufFile(BufFile *file)
{
    File        pfile;
    ResourceOwner oldowner;
 
    /* Be sure to associate the file with the BufFile's resource owner */
    oldowner = CurrentResourceOwner;
    CurrentResourceOwner = file->resowner;
 
    if (file->fileset == NULL)
        pfile = OpenTemporaryFile(file->isInterXact);
    else
        pfile = MakeNewFileSetSegment(file, file->numFiles);
 
    Assert(pfile >= 0);
 
    CurrentResourceOwner = oldowner;
 
    file->files = (File *) repalloc(file->files,
                                    (file->numFiles + 1) * sizeof(File));
    file->files[file->numFiles] = pfile;
    file->numFiles++;
}
 
/*
 * Create a BufFile for a new temporary file (which will expand to become
 * multiple temporary files if more than MAX_PHYSICAL_FILESIZE bytes are
 * written to it).
 *
 * If interXact is true, the temp file will not be automatically deleted
 * at end of transaction.
 *
 * Note: if interXact is true, the caller had better be calling us in a
 * memory context, and with a resource owner, that will survive across
 * transaction boundaries.
 */
BufFile *
BufFileCreateTemp(bool interXact)
{
    BufFile    *file;
    File        pfile;
 
    /*
     * Ensure that temp tablespaces are set up for OpenTemporaryFile to use.
     * Possibly the caller will have done this already, but it seems useful to
     * double-check here.  Failure to do this at all would result in the temp
     * files always getting placed in the default tablespace, which is a
     * pretty hard-to-detect bug.  Callers may prefer to do it earlier if they
     * want to be sure that any required catalog access is done in some other
     * resource context.
     */
    PrepareTempTablespaces();
 
    pfile = OpenTemporaryFile(interXact);
    Assert(pfile >= 0);
 
    file = makeBufFile(pfile);
    file->isInterXact = interXact;
 
    return file;
}
 
/*
 * Build the name for a given segment of a given BufFile.
 */
static void
FileSetSegmentName(char *name, const char *buffile_name, int segment)
{
    snprintf(name, MAXPGPATH, "%s.%d", buffile_name, segment);
}
 
/*
 * Create a new segment file backing a fileset based BufFile.
 */
static File
MakeNewFileSetSegment(BufFile *buffile, int segment)
{
    char        name[MAXPGPATH];
    File        file;
 
    /*
     * It is possible that there are files left over from before a crash
     * restart with the same name.  In order for BufFileOpenFileSet() not to
     * get confused about how many segments there are, we'll unlink the next
     * segment number if it already exists.
     */
    FileSetSegmentName(name, buffile->name, segment + 1);
    FileSetDelete(buffile->fileset, name, true);
 
    /* Create the new segment. */
    FileSetSegmentName(name, buffile->name, segment);
    file = FileSetCreate(buffile->fileset, name);
 
    /* FileSetCreate would've errored out */
    Assert(file > 0);
 
    return file;
}
 
/*
 * Create a BufFile that can be discovered and opened read-only by other
 * backends that are attached to the same SharedFileSet using the same name.
 *
 * The naming scheme for fileset based BufFiles is left up to the calling code.
 * The name will appear as part of one or more filenames on disk, and might
 * provide clues to administrators about which subsystem is generating
 * temporary file data.  Since each SharedFileSet object is backed by one or
 * more uniquely named temporary directory, names don't conflict with
 * unrelated SharedFileSet objects.
 */
BufFile *
BufFileCreateFileSet(FileSet *fileset, const char *name)
{
    BufFile    *file;
 
    file = makeBufFileCommon(1);
    file->fileset = fileset;
    file->name = pstrdup(name);
    file->files = palloc_object(File);
    file->files[0] = MakeNewFileSetSegment(file, 0);
    file->readOnly = false;
 
    return file;
}
 
/*
 * Open a file that was previously created in another backend (or this one)
 * with BufFileCreateFileSet in the same FileSet using the same name.
 * The backend that created the file must have called BufFileClose() or
 * BufFileExportFileSet() to make sure that it is ready to be opened by other
 * backends and render it read-only.  If missing_ok is true, which indicates
 * that missing files can be safely ignored, then return NULL if the BufFile
 * with the given name is not found, otherwise, throw an error.
 */
BufFile *
BufFileOpenFileSet(FileSet *fileset, const char *name, int mode,
                   bool missing_ok)
{
    BufFile    *file;
    char        segment_name[MAXPGPATH];
    Size        capacity = 16;
    File       *files;
    int         nfiles = 0;
 
    files = palloc_array(File, capacity);
 
    /*
     * We don't know how many segments there are, so we'll probe the
     * filesystem to find out.
     */
    for (;;)
    {
        /* See if we need to expand our file segment array. */
        if (nfiles + 1 > capacity)
        {
            capacity *= 2;
            files = repalloc_array(files, File, capacity);
        }
        /* Try to load a segment. */
        FileSetSegmentName(segment_name, name, nfiles);
        files[nfiles] = FileSetOpen(fileset, segment_name, mode);
        if (files[nfiles] <= 0)
            break;
        ++nfiles;
 
        CHECK_FOR_INTERRUPTS();
    }
 
    /*
     * If we didn't find any files at all, then no BufFile exists with this
     * name.
     */
    if (nfiles == 0)
    {
        /* free the memory */
        pfree(files);
 
        if (missing_ok)
            return NULL;
 
        ereport(ERROR,
                (errcode_for_file_access(),
                 errmsg("could not open temporary file \"%s\" from BufFile \"%s\": %m",
                        segment_name, name)));
    }
 
    file = makeBufFileCommon(nfiles);
    file->files = files;
    file->readOnly = (mode == O_RDONLY);
    file->fileset = fileset;
    file->name = pstrdup(name);
 
    return file;
}
 
/*
 * Delete a BufFile that was created by BufFileCreateFileSet in the given
 * FileSet using the given name.
 *
 * It is not necessary to delete files explicitly with this function.  It is
 * provided only as a way to delete files proactively, rather than waiting for
 * the FileSet to be cleaned up.
 *
 * Only one backend should attempt to delete a given name, and should know
 * that it exists and has been exported or closed otherwise missing_ok should
 * be passed true.
 */
void
BufFileDeleteFileSet(FileSet *fileset, const char *name, bool missing_ok)
{
    char        segment_name[MAXPGPATH];
    int         segment = 0;
    bool        found = false;
 
    /*
     * We don't know how many segments the file has.  We'll keep deleting
     * until we run out.  If we don't manage to find even an initial segment,
     * raise an error.
     */
    for (;;)
    {
        FileSetSegmentName(segment_name, name, segment);
        if (!FileSetDelete(fileset, segment_name, true))
            break;
        found = true;
        ++segment;
 
        CHECK_FOR_INTERRUPTS();
    }
 
    if (!found && !missing_ok)
        elog(ERROR, "could not delete unknown BufFile \"%s\"", name);
}
 
/*
 * BufFileExportFileSet --- flush and make read-only, in preparation for sharing.
 */
void
BufFileExportFileSet(BufFile *file)
{
    /* Must be a file belonging to a FileSet. */
    Assert(file->fileset != NULL);
 
    /* It's probably a bug if someone calls this twice. */
    Assert(!file->readOnly);
 
    BufFileFlush(file);
    file->readOnly = true;
}
 
/*
 * Close a BufFile
 *
 * Like fclose(), this also implicitly FileCloses the underlying File.
 */
void
BufFileClose(BufFile *file)
{
    int         i;
 
    /* flush any unwritten data */
    BufFileFlush(file);
    /* close and delete the underlying file(s) */
    for (i = 0; i < file->numFiles; i++)
        FileClose(file->files[i]);
    /* release the buffer space */
    pfree(file->files);
    pfree(file);
}
 
/*
 * BufFileLoadBuffer
 *
 * Load some data into buffer, if possible, starting from curOffset.
 * At call, must have dirty = false, pos and nbytes = 0.
 * On exit, nbytes is number of bytes loaded.
 */
static void
BufFileLoadBuffer(BufFile *file)
{
    File        thisfile;
    instr_time  io_start;
    instr_time  io_time;
 
    /*
     * Advance to next component file if necessary and possible.
     */
    if (file->curOffset >= MAX_PHYSICAL_FILESIZE &&
        file->curFile + 1 < file->numFiles)
    {
        file->curFile++;
        file->curOffset = 0;
    }
 
    thisfile = file->files[file->curFile];
 
    if (track_io_timing)
        INSTR_TIME_SET_CURRENT(io_start);
    else
        INSTR_TIME_SET_ZERO(io_start);
 
    /*
     * Read whatever we can get, up to a full bufferload.
     */
    file->nbytes = FileRead(thisfile,
                            file->buffer.data,
                            sizeof(file->buffer.data),
                            file->curOffset,
                            WAIT_EVENT_BUFFILE_READ);
    if (file->nbytes < 0)
    {
        file->nbytes = 0;
        ereport(ERROR,
                (errcode_for_file_access(),
                 errmsg("could not read file \"%s\": %m",
                        FilePathName(thisfile))));
    }
 
    if (track_io_timing)
    {
        INSTR_TIME_SET_CURRENT(io_time);
        INSTR_TIME_ACCUM_DIFF(pgBufferUsage.temp_blk_read_time, io_time, io_start);
    }
 
    /* we choose not to advance curOffset here */
 
    if (file->nbytes > 0)
        pgBufferUsage.temp_blks_read++;
}
 
/*
 * BufFileDumpBuffer
 *
 * Dump buffer contents starting at curOffset.
 * At call, should have dirty = true, nbytes > 0.
 * On exit, dirty is cleared if successful write, and curOffset is advanced.
 */
static void
BufFileDumpBuffer(BufFile *file)
{
    int64       wpos = 0;
    int64       bytestowrite;
    File        thisfile;
 
    /*
     * Unlike BufFileLoadBuffer, we must dump the whole buffer even if it
     * crosses a component-file boundary; so we need a loop.
     */
    while (wpos < file->nbytes)
    {
        int64       availbytes;
        instr_time  io_start;
        instr_time  io_time;
 
        /*
         * Advance to next component file if necessary and possible.
         */
        if (file->curOffset >= MAX_PHYSICAL_FILESIZE)
        {
            while (file->curFile + 1 >= file->numFiles)
                extendBufFile(file);
            file->curFile++;
            file->curOffset = 0;
        }
 
        /*
         * Determine how much we need to write into this file.
         */
        bytestowrite = file->nbytes - wpos;
        availbytes = MAX_PHYSICAL_FILESIZE - file->curOffset;
 
        if (bytestowrite > availbytes)
            bytestowrite = availbytes;
 
        thisfile = file->files[file->curFile];
 
        if (track_io_timing)
            INSTR_TIME_SET_CURRENT(io_start);
        else
            INSTR_TIME_SET_ZERO(io_start);
 
        bytestowrite = FileWrite(thisfile,
                                 file->buffer.data + wpos,
                                 bytestowrite,
                                 file->curOffset,
                                 WAIT_EVENT_BUFFILE_WRITE);
        if (bytestowrite <= 0)
            ereport(ERROR,
                    (errcode_for_file_access(),
                     errmsg("could not write to file \"%s\": %m",
                            FilePathName(thisfile))));
 
        if (track_io_timing)
        {
            INSTR_TIME_SET_CURRENT(io_time);
            INSTR_TIME_ACCUM_DIFF(pgBufferUsage.temp_blk_write_time, io_time, io_start);
        }
 
        file->curOffset += bytestowrite;
        wpos += bytestowrite;
 
        pgBufferUsage.temp_blks_written++;
    }
    file->dirty = false;
 
    /*
     * At this point, curOffset has been advanced to the end of the buffer,
     * ie, its original value + nbytes.  We need to make it point to the
     * logical file position, ie, original value + pos, in case that is less
     * (as could happen due to a small backwards seek in a dirty buffer!)
     */
    file->curOffset -= (file->nbytes - file->pos);
    if (file->curOffset < 0)    /* handle possible segment crossing */
    {
        file->curFile--;
        Assert(file->curFile >= 0);
        file->curOffset += MAX_PHYSICAL_FILESIZE;
    }
 
    /*
     * Now we can set the buffer empty without changing the logical position
     */
    file->pos = 0;
    file->nbytes = 0;
}
 
/*
 * BufFileRead variants
 *
 * Like fread() except we assume 1-byte element size and report I/O errors via
 * ereport().
 *
 * If 'exact' is true, then an error is also raised if the number of bytes
 * read is not exactly 'size' (no short reads).  If 'exact' and 'eofOK' are
 * true, then reading zero bytes is ok.
 */
static size_t
BufFileReadCommon(BufFile *file, void *ptr, size_t size, bool exact, bool eofOK)
{
    size_t      start_size = size;
    size_t      nread = 0;
    size_t      nthistime;
 
    BufFileFlush(file);
 
    while (size > 0)
    {
        if (file->pos >= file->nbytes)
        {
            /* Try to load more data into buffer. */
            file->curOffset += file->pos;
            file->pos = 0;
            file->nbytes = 0;
            BufFileLoadBuffer(file);
            if (file->nbytes <= 0)
                break;          /* no more data available */
        }
 
        nthistime = file->nbytes - file->pos;
        if (nthistime > size)
            nthistime = size;
        Assert(nthistime > 0);
 
        memcpy(ptr, file->buffer.data + file->pos, nthistime);
 
        file->pos += nthistime;
        ptr = (char *) ptr + nthistime;
        size -= nthistime;
        nread += nthistime;
    }
 
    if (exact &&
        (nread != start_size && !(nread == 0 && eofOK)))
        ereport(ERROR,
                errcode_for_file_access(),
                file->name ?
                errmsg("could not read from file set \"%s\": read only %zu of %zu bytes",
                       file->name, nread, start_size) :
                errmsg("could not read from temporary file: read only %zu of %zu bytes",
                       nread, start_size));
 
    return nread;
}
 
/*
 * Legacy interface where the caller needs to check for end of file or short
 * reads.
 */
size_t
BufFileRead(BufFile *file, void *ptr, size_t size)
{
    return BufFileReadCommon(file, ptr, size, false, false);
}
 
/*
 * Require read of exactly the specified size.
 */
void
BufFileReadExact(BufFile *file, void *ptr, size_t size)
{
    BufFileReadCommon(file, ptr, size, true, false);
}
 
/*
 * Require read of exactly the specified size, but optionally allow end of
 * file (in which case 0 is returned).
 */
size_t
BufFileReadMaybeEOF(BufFile *file, void *ptr, size_t size, bool eofOK)
{
    return BufFileReadCommon(file, ptr, size, true, eofOK);
}
 
/*
 * BufFileWrite
 *
 * Like fwrite() except we assume 1-byte element size and report errors via
 * ereport().
 */
void
BufFileWrite(BufFile *file, const void *ptr, size_t size)
{
    size_t      nthistime;
 
    Assert(!file->readOnly);
 
    while (size > 0)
    {
        if (file->pos >= BLCKSZ)
        {
            /* Buffer full, dump it out */
            if (file->dirty)
                BufFileDumpBuffer(file);
            else
            {
                /* Hmm, went directly from reading to writing? */
                file->curOffset += file->pos;
                file->pos = 0;
                file->nbytes = 0;
            }
        }
 
        nthistime = BLCKSZ - file->pos;
        if (nthistime > size)
            nthistime = size;
        Assert(nthistime > 0);
 
        memcpy(file->buffer.data + file->pos, ptr, nthistime);
 
        file->dirty = true;
        file->pos += nthistime;
        if (file->nbytes < file->pos)
            file->nbytes = file->pos;
        ptr = (const char *) ptr + nthistime;
        size -= nthistime;
    }
}
 
/*
 * BufFileFlush
 *
 * Like fflush(), except that I/O errors are reported with ereport().
 */
static void
BufFileFlush(BufFile *file)
{
    if (file->dirty)
        BufFileDumpBuffer(file);
 
    Assert(!file->dirty);
}
 
/*
 * BufFileSeek
 *
 * Like fseek(), except that target position needs two values in order to
 * work when logical filesize exceeds maximum value representable by pgoff_t.
 * We do not support relative seeks across more than that, however.
 * I/O errors are reported by ereport().
 *
 * Result is 0 if OK, EOF if not.  Logical position is not moved if an
 * impossible seek is attempted.
 */
int
BufFileSeek(BufFile *file, int fileno, pgoff_t offset, int whence)
{
    int         newFile;
    pgoff_t     newOffset;
 
    switch (whence)
    {
        case SEEK_SET:
            if (fileno < 0)
                return EOF;
            newFile = fileno;
            newOffset = offset;
            break;
        case SEEK_CUR:
 
            /*
             * Relative seek considers only the signed offset, ignoring
             * fileno.
             */
            newFile = file->curFile;
            newOffset = (file->curOffset + file->pos) + offset;
            break;
        case SEEK_END:
 
            /*
             * The file size of the last file gives us the end offset of that
             * file.
             */
            newFile = file->numFiles - 1;
            newOffset = FileSize(file->files[file->numFiles - 1]);
            if (newOffset < 0)
                ereport(ERROR,
                        (errcode_for_file_access(),
                         errmsg("could not determine size of temporary file \"%s\" from BufFile \"%s\": %m",
                                FilePathName(file->files[file->numFiles - 1]),
                                file->name)));
            break;
        default:
            elog(ERROR, "invalid whence: %d", whence);
            return EOF;
    }
    while (newOffset < 0)
    {
        if (--newFile < 0)
            return EOF;
        newOffset += MAX_PHYSICAL_FILESIZE;
    }
    if (newFile == file->curFile &&
        newOffset >= file->curOffset &&
        newOffset <= file->curOffset + file->nbytes)
    {
        /*
         * Seek is to a point within existing buffer; we can just adjust
         * pos-within-buffer, without flushing buffer.  Note this is OK
         * whether reading or writing, but buffer remains dirty if we were
         * writing.
         */
        file->pos = (int64) (newOffset - file->curOffset);
        return 0;
    }
    /* Otherwise, must reposition buffer, so flush any dirty data */
    BufFileFlush(file);
 
    /*
     * At this point and no sooner, check for seek past last segment. The
     * above flush could have created a new segment, so checking sooner would
     * not work (at least not with this code).
     */
 
    /* convert seek to "start of next seg" to "end of last seg" */
    if (newFile == file->numFiles && newOffset == 0)
    {
        newFile--;
        newOffset = MAX_PHYSICAL_FILESIZE;
    }
    while (newOffset > MAX_PHYSICAL_FILESIZE)
    {
        if (++newFile >= file->numFiles)
            return EOF;
        newOffset -= MAX_PHYSICAL_FILESIZE;
    }
    if (newFile >= file->numFiles)
        return EOF;
    /* Seek is OK! */
    file->curFile = newFile;
    file->curOffset = newOffset;
    file->pos = 0;
    file->nbytes = 0;
    return 0;
}
 
void
BufFileTell(BufFile *file, int *fileno, pgoff_t *offset)
{
    *fileno = file->curFile;
    *offset = file->curOffset + file->pos;
}
 
/*
 * BufFileSeekBlock --- block-oriented seek
 *
 * Performs absolute seek to the start of the n'th BLCKSZ-sized block of
 * the file.  Note that users of this interface will fail if their files
 * exceed BLCKSZ * PG_INT64_MAX bytes, but that is quite a lot; we don't
 * work with tables bigger than that, either...
 *
 * Result is 0 if OK, EOF if not.  Logical position is not moved if an
 * impossible seek is attempted.
 */
int
BufFileSeekBlock(BufFile *file, int64 blknum)
{
    return BufFileSeek(file,
                       (int) (blknum / BUFFILE_SEG_SIZE),
                       (pgoff_t) (blknum % BUFFILE_SEG_SIZE) * BLCKSZ,
                       SEEK_SET);
}
 
/*
 * Returns the amount of data in the given BufFile, in bytes.
 *
 * Returned value includes the size of any holes left behind by BufFileAppend.
 * ereport()s on failure.
 */
int64
BufFileSize(BufFile *file)
{
    int64       lastFileSize;
 
    /* Get the size of the last physical file. */
    lastFileSize = FileSize(file->files[file->numFiles - 1]);
    if (lastFileSize < 0)
        ereport(ERROR,
                (errcode_for_file_access(),
                 errmsg("could not determine size of temporary file \"%s\" from BufFile \"%s\": %m",
                        FilePathName(file->files[file->numFiles - 1]),
                        file->name)));
 
    return ((file->numFiles - 1) * (int64) MAX_PHYSICAL_FILESIZE) +
        lastFileSize;
}
 
/*
 * Append the contents of the source file to the end of the target file.
 *
 * Note that operation subsumes ownership of underlying resources from
 * "source".  Caller should never call BufFileClose against source having
 * called here first.  Resource owners for source and target must match,
 * too.
 *
 * This operation works by manipulating lists of segment files, so the
 * file content is always appended at a MAX_PHYSICAL_FILESIZE-aligned
 * boundary, typically creating empty holes before the boundary.  These
 * areas do not contain any interesting data, and cannot be read from by
 * caller.
 *
 * Returns the block number within target where the contents of source
 * begins.  Caller should apply this as an offset when working off block
 * positions that are in terms of the original BufFile space.
 */
int64
BufFileAppend(BufFile *target, BufFile *source)
{
    int64       startBlock = (int64) target->numFiles * BUFFILE_SEG_SIZE;
    int         newNumFiles = target->numFiles + source->numFiles;
    int         i;
 
    Assert(source->readOnly);
    Assert(!source->dirty);
 
    if (target->resowner != source->resowner)
        elog(ERROR, "could not append BufFile with non-matching resource owner");
 
    target->files = (File *)
        repalloc(target->files, sizeof(File) * newNumFiles);
    for (i = target->numFiles; i < newNumFiles; i++)
        target->files[i] = source->files[i - target->numFiles];
    target->numFiles = newNumFiles;
 
    return startBlock;
}
 
/*
 * Truncate a BufFile created by BufFileCreateFileSet up to the given fileno
 * and the offset.
 */
void
BufFileTruncateFileSet(BufFile *file, int fileno, pgoff_t offset)
{
    int         numFiles = file->numFiles;
    int         newFile = fileno;
    pgoff_t     newOffset = file->curOffset;
    char        segment_name[MAXPGPATH];
    int         i;
 
    /*
     * Loop over all the files up to the given fileno and remove the files
     * that are greater than the fileno and truncate the given file up to the
     * offset. Note that we also remove the given fileno if the offset is 0
     * provided it is not the first file in which we truncate it.
     */
    for (i = file->numFiles - 1; i >= fileno; i--)
    {
        if ((i != fileno || offset == 0) && i != 0)
        {
            FileSetSegmentName(segment_name, file->name, i);
            FileClose(file->files[i]);
            if (!FileSetDelete(file->fileset, segment_name, true))
                ereport(ERROR,
                        (errcode_for_file_access(),
                         errmsg("could not delete fileset \"%s\": %m",
                                segment_name)));
            numFiles--;
            newOffset = MAX_PHYSICAL_FILESIZE;
 
            /*
             * This is required to indicate that we have deleted the given
             * fileno.
             */
            if (i == fileno)
                newFile--;
        }
        else
        {
            if (FileTruncate(file->files[i], offset,
                             WAIT_EVENT_BUFFILE_TRUNCATE) < 0)
                ereport(ERROR,
                        (errcode_for_file_access(),
                         errmsg("could not truncate file \"%s\": %m",
                                FilePathName(file->files[i]))));
            newOffset = offset;
        }
    }
 
    file->numFiles = numFiles;
 
    /*
     * If the truncate point is within existing buffer then we can just adjust
     * pos within buffer.
     */
    if (newFile == file->curFile &&
        newOffset >= file->curOffset &&
        newOffset <= file->curOffset + file->nbytes)
    {
        /* No need to reset the current pos if the new pos is greater. */
        if (newOffset <= file->curOffset + file->pos)
            file->pos = (int64) newOffset - file->curOffset;
 
        /* Adjust the nbytes for the current buffer. */
        file->nbytes = (int64) newOffset - file->curOffset;
    }
    else if (newFile == file->curFile &&
             newOffset < file->curOffset)
    {
        /*
         * The truncate point is within the existing file but prior to the
         * current position, so we can forget the current buffer and reset the
         * current position.
         */
        file->curOffset = newOffset;
        file->pos = 0;
        file->nbytes = 0;
    }
    else if (newFile < file->curFile)
    {
        /*
         * The truncate point is prior to the current file, so need to reset
         * the current position accordingly.
         */
        file->curFile = newFile;
        file->curOffset = newOffset;
        file->pos = 0;
        file->nbytes = 0;
    }
    /* Nothing to do, if the truncate point is beyond current file. */
}