utils/misc/tzparser.c

/*-------------------------------------------------------------------------
 *
 * tzparser.c
 *	  Functions for parsing timezone offset files
 *
 * Note: this code is invoked from the check_hook for the GUC variable
 * timezone_abbreviations.  Therefore, it should report problems using
 * GUC_check_errmsg() and related functions, and try to avoid throwing
 * elog(ERROR).  This is not completely bulletproof at present --- in
 * particular out-of-memory will throw an error.  Could probably fix with
 * PG_TRY if necessary.
 *
 *
 * Portions Copyright (c) 1996-2017, PostgreSQL Global Development Group
 * Portions Copyright (c) 1994, Regents of the University of California
 *
 * IDENTIFICATION
 *	  src/backend/utils/misc/tzparser.c
 *
 *-------------------------------------------------------------------------
 */

#include "postgres.h"

#include <ctype.h>

#include "miscadmin.h"
#include "storage/fd.h"
#include "utils/guc.h"
#include "utils/memutils.h"
#include "utils/tzparser.h"


#define WHITESPACE " \t\n\r"

static bool validateTzEntry(tzEntry *tzentry);
static bool splitTzLine(const char *filename, int lineno,
			char *line, tzEntry *tzentry);
static int addToArray(tzEntry **base, int *arraysize, int n,
		   tzEntry *entry, bool override);
static int ParseTzFile(const char *filename, int depth,
			tzEntry **base, int *arraysize, int n);


/*
 * Apply additional validation checks to a tzEntry
 *
 * Returns TRUE if OK, else false
 */
static bool
validateTzEntry(tzEntry *tzentry)
{
	unsigned char *p;

	/*
	 * Check restrictions imposed by datetkntbl storage format (see
	 * datetime.c)
	 */
	if (strlen(tzentry->abbrev) > TOKMAXLEN)
	{
		GUC_check_errmsg("time zone abbreviation \"%s\" is too long (maximum %d characters) in time zone file \"%s\", line %d",
						 tzentry->abbrev, TOKMAXLEN,
						 tzentry->filename, tzentry->lineno);
		return false;
	}

	/*
	 * Sanity-check the offset: shouldn't exceed 14 hours
	 */
	if (tzentry->offset > 14 * 60 * 60 ||
		tzentry->offset < -14 * 60 * 60)
	{
		GUC_check_errmsg("time zone offset %d is out of range in time zone file \"%s\", line %d",
						 tzentry->offset,
						 tzentry->filename, tzentry->lineno);
		return false;
	}

	/*
	 * Convert abbrev to lowercase (must match datetime.c's conversion)
	 */
	for (p = (unsigned char *) tzentry->abbrev; *p; p++)
		*p = pg_tolower(*p);

	return true;
}

/*
 * Attempt to parse the line as a timezone abbrev spec
 *
 * Valid formats are:
 *	name  zone
 *	name  offset  dst
 *
 * Returns TRUE if OK, else false; data is stored in *tzentry
 */
static bool
splitTzLine(const char *filename, int lineno, char *line, tzEntry *tzentry)
{
	char	   *abbrev;
	char	   *offset;
	char	   *offset_endptr;
	char	   *remain;
	char	   *is_dst;

	tzentry->lineno = lineno;
	tzentry->filename = filename;

	abbrev = strtok(line, WHITESPACE);
	if (!abbrev)
	{
		GUC_check_errmsg("missing time zone abbreviation in time zone file \"%s\", line %d",
						 filename, lineno);
		return false;
	}
	tzentry->abbrev = pstrdup(abbrev);

	offset = strtok(NULL, WHITESPACE);
	if (!offset)
	{
		GUC_check_errmsg("missing time zone offset in time zone file \"%s\", line %d",
						 filename, lineno);
		return false;
	}

	/* We assume zone names don't begin with a digit or sign */
	if (isdigit((unsigned char) *offset) || *offset == '+' || *offset == '-')
	{
		tzentry->zone = NULL;
		tzentry->offset = strtol(offset, &offset_endptr, 10);
		if (offset_endptr == offset || *offset_endptr != '\0')
		{
			GUC_check_errmsg("invalid number for time zone offset in time zone file \"%s\", line %d",
							 filename, lineno);
			return false;
		}

		is_dst = strtok(NULL, WHITESPACE);
		if (is_dst && pg_strcasecmp(is_dst, "D") == 0)
		{
			tzentry->is_dst = true;
			remain = strtok(NULL, WHITESPACE);
		}
		else
		{
			/* there was no 'D' dst specifier */
			tzentry->is_dst = false;
			remain = is_dst;
		}
	}
	else
	{
		/*
		 * Assume entry is a zone name.  We do not try to validate it by
		 * looking up the zone, because that would force loading of a lot of
		 * zones that probably will never be used in the current session.
		 */
		tzentry->zone = pstrdup(offset);
		tzentry->offset = 0;
		tzentry->is_dst = false;
		remain = strtok(NULL, WHITESPACE);
	}

	if (!remain)				/* no more non-whitespace chars */
		return true;

	if (remain[0] != '#')		/* must be a comment */
	{
		GUC_check_errmsg("invalid syntax in time zone file \"%s\", line %d",
						 filename, lineno);
		return false;
	}
	return true;
}

/*
 * Insert entry into sorted array
 *
 * *base: base address of array (changeable if must enlarge array)
 * *arraysize: allocated length of array (changeable if must enlarge array)
 * n: current number of valid elements in array
 * entry: new data to insert
 * override: TRUE if OK to override
 *
 * Returns the new array length (new value for n), or -1 if error
 */
static int
addToArray(tzEntry **base, int *arraysize, int n,
		   tzEntry *entry, bool override)
{
	tzEntry    *arrayptr;
	int			low;
	int			high;

	/*
	 * Search the array for a duplicate; as a useful side effect, the array is
	 * maintained in sorted order.  We use strcmp() to ensure we match the
	 * sort order datetime.c expects.
	 */
	arrayptr = *base;
	low = 0;
	high = n - 1;
	while (low <= high)
	{
		int			mid = (low + high) >> 1;
		tzEntry    *midptr = arrayptr + mid;
		int			cmp;

		cmp = strcmp(entry->abbrev, midptr->abbrev);
		if (cmp < 0)
			high = mid - 1;
		else if (cmp > 0)
			low = mid + 1;
		else
		{
			/*
			 * Found a duplicate entry; complain unless it's the same.
			 */
			if ((midptr->zone == NULL && entry->zone == NULL &&
				 midptr->offset == entry->offset &&
				 midptr->is_dst == entry->is_dst) ||
				(midptr->zone != NULL && entry->zone != NULL &&
				 strcmp(midptr->zone, entry->zone) == 0))
			{
				/* return unchanged array */
				return n;
			}
			if (override)
			{
				/* same abbrev but something is different, override */
				midptr->zone = entry->zone;
				midptr->offset = entry->offset;
				midptr->is_dst = entry->is_dst;
				return n;
			}
			/* same abbrev but something is different, complain */
			GUC_check_errmsg("time zone abbreviation \"%s\" is multiply defined",
							 entry->abbrev);
			GUC_check_errdetail("Entry in time zone file \"%s\", line %d, conflicts with entry in file \"%s\", line %d.",
								midptr->filename, midptr->lineno,
								entry->filename, entry->lineno);
			return -1;
		}
	}

	/*
	 * No match, insert at position "low".
	 */
	if (n >= *arraysize)
	{
		*arraysize *= 2;
		*base = (tzEntry *) repalloc(*base, *arraysize * sizeof(tzEntry));
	}

	arrayptr = *base + low;

	memmove(arrayptr + 1, arrayptr, (n - low) * sizeof(tzEntry));

	memcpy(arrayptr, entry, sizeof(tzEntry));

	return n + 1;
}

/*
 * Parse a single timezone abbrev file --- can recurse to handle @INCLUDE
 *
 * filename: user-specified file name (does not include path)
 * depth: current recursion depth
 * *base: array for results (changeable if must enlarge array)
 * *arraysize: allocated length of array (changeable if must enlarge array)
 * n: current number of valid elements in array
 *
 * Returns the new array length (new value for n), or -1 if error
 */
static int
ParseTzFile(const char *filename, int depth,
			tzEntry **base, int *arraysize, int n)
{
	char		share_path[MAXPGPATH];
	char		file_path[MAXPGPATH];
	FILE	   *tzFile;
	char		tzbuf[1024];
	char	   *line;
	tzEntry		tzentry;
	int			lineno = 0;
	bool		override = false;
	const char *p;

	/*
	 * We enforce that the filename is all alpha characters.  This may be
	 * overly restrictive, but we don't want to allow access to anything
	 * outside the timezonesets directory, so for instance '/' *must* be
	 * rejected.
	 */
	for (p = filename; *p; p++)
	{
		if (!isalpha((unsigned char) *p))
		{
			/* at level 0, just use guc.c's regular "invalid value" message */
			if (depth > 0)
				GUC_check_errmsg("invalid time zone file name \"%s\"",
								 filename);
			return -1;
		}
	}

	/*
	 * The maximal recursion depth is a pretty arbitrary setting. It is hard
	 * to imagine that someone needs more than 3 levels so stick with this
	 * conservative setting until someone complains.
	 */
	if (depth > 3)
	{
		GUC_check_errmsg("time zone file recursion limit exceeded in file \"%s\"",
						 filename);
		return -1;
	}

	get_share_path(my_exec_path, share_path);
	snprintf(file_path, sizeof(file_path), "%s/timezonesets/%s",
			 share_path, filename);
	tzFile = AllocateFile(file_path, "r");
	if (!tzFile)
	{
		/*
		 * Check to see if the problem is not the filename but the directory.
		 * This is worth troubling over because if the installation share/
		 * directory is missing or unreadable, this is likely to be the first
		 * place we notice a problem during postmaster startup.
		 */
		int			save_errno = errno;
		DIR		   *tzdir;

		snprintf(file_path, sizeof(file_path), "%s/timezonesets",
				 share_path);
		tzdir = AllocateDir(file_path);
		if (tzdir == NULL)
		{
			GUC_check_errmsg("could not open directory \"%s\": %m",
							 file_path);
			GUC_check_errhint("This may indicate an incomplete PostgreSQL installation, or that the file \"%s\" has been moved away from its proper location.",
							  my_exec_path);
			return -1;
		}
		FreeDir(tzdir);
		errno = save_errno;

		/*
		 * otherwise, if file doesn't exist and it's level 0, guc.c's
		 * complaint is enough
		 */
		if (errno != ENOENT || depth > 0)
			GUC_check_errmsg("could not read time zone file \"%s\": %m",
							 filename);

		return -1;
	}

	while (!feof(tzFile))
	{
		lineno++;
		if (fgets(tzbuf, sizeof(tzbuf), tzFile) == NULL)
		{
			if (ferror(tzFile))
			{
				GUC_check_errmsg("could not read time zone file \"%s\": %m",
								 filename);
				return -1;
			}
			/* else we're at EOF after all */
			break;
		}
		if (strlen(tzbuf) == sizeof(tzbuf) - 1)
		{
			/* the line is too long for tzbuf */
			GUC_check_errmsg("line is too long in time zone file \"%s\", line %d",
							 filename, lineno);
			return -1;
		}

		/* skip over whitespace */
		line = tzbuf;
		while (*line && isspace((unsigned char) *line))
			line++;

		if (*line == '\0')		/* empty line */
			continue;
		if (*line == '#')		/* comment line */
			continue;

		if (pg_strncasecmp(line, "@INCLUDE", strlen("@INCLUDE")) == 0)
		{
			/* pstrdup so we can use filename in result data structure */
			char	   *includeFile = pstrdup(line + strlen("@INCLUDE"));

			includeFile = strtok(includeFile, WHITESPACE);
			if (!includeFile || !*includeFile)
			{
				GUC_check_errmsg("@INCLUDE without file name in time zone file \"%s\", line %d",
								 filename, lineno);
				return -1;
			}
			n = ParseTzFile(includeFile, depth + 1,
							base, arraysize, n);
			if (n < 0)
				return -1;
			continue;
		}

		if (pg_strncasecmp(line, "@OVERRIDE", strlen("@OVERRIDE")) == 0)
		{
			override = true;
			continue;
		}

		if (!splitTzLine(filename, lineno, line, &tzentry))
			return -1;
		if (!validateTzEntry(&tzentry))
			return -1;
		n = addToArray(base, arraysize, n, &tzentry, override);
		if (n < 0)
			return -1;
	}

	FreeFile(tzFile);

	return n;
}

/*
 * load_tzoffsets --- read and parse the specified timezone offset file
 *
 * On success, return a filled-in TimeZoneAbbrevTable, which must have been
 * malloc'd not palloc'd.  On failure, return NULL, using GUC_check_errmsg
 * and friends to give details of the problem.
 */
TimeZoneAbbrevTable *
load_tzoffsets(const char *filename)
{
	TimeZoneAbbrevTable *result = NULL;
	MemoryContext tmpContext;
	MemoryContext oldContext;
	tzEntry    *array;
	int			arraysize;
	int			n;

	/*
	 * Create a temp memory context to work in.  This makes it easy to clean
	 * up afterwards.
	 */
	tmpContext = AllocSetContextCreate(CurrentMemoryContext,
									   "TZParserMemory",
									   ALLOCSET_SMALL_SIZES);
	oldContext = MemoryContextSwitchTo(tmpContext);

	/* Initialize array at a reasonable size */
	arraysize = 128;
	array = (tzEntry *) palloc(arraysize * sizeof(tzEntry));

	/* Parse the file(s) */
	n = ParseTzFile(filename, 0, &array, &arraysize, 0);

	/* If no errors so far, let datetime.c allocate memory & convert format */
	if (n >= 0)
	{
		result = ConvertTimeZoneAbbrevs(array, n);
		if (!result)
			GUC_check_errmsg("out of memory");
	}

	/* Clean up */
	MemoryContextSwitchTo(oldContext);
	MemoryContextDelete(tmpContext);

	return result;
}