locale: Handle loading a missing locale twice (Bug 14247)

Delay setting file->decided until the data has been successfully loaded
by _nl_load_locale().  If the function fails to load the data then we
must return and error and leave decided untouched to allow the caller to
attempt to load the data again at a later time.  We should not set
decided to 1 early in the function since doing so may prevent attempting
to load it again. We want to try loading it again because that allows an
open to fail and set errno correctly.

On the other side of this problem is that if we are called again with
the same inputs we will fetch the cached version of the object and carry
out no open syscalls and that fails to set errno so we must set errno to
ENOENT in that case.  There is a second code path that has to be handled
where the name of the locale matches but the codeset doesn't match.

These changes ensure that errno is correctly set on failure in all the
return paths in _nl_find_locale().

Adds tst-locale-loadlocale to cover the bug.

No regressions on x86_64.

Co-authored-by: Jeff Law <law@redhat.com>
Reviewed-by: Adhemerval Zanella  <adhemerval.zanella@linaro.org>

6 files changed, 119 insertions, 10 deletions

diff --git a/gen-locales.mk b/gen-locales.mk
index 9c523d2a05..b005a72866 100644
--- a/gen-locales.mk
+++ b/gen-locales.mk
@@ -1,8 +1,19 @@
 # defines target $(gen-locales) that generates the locales given in $(LOCALES)
 
 LOCALE_SRCS := $(shell echo "$(LOCALES)"|sed 's/\([^ .]*\)[^@ ]*\(@[^ ]*\)\?/\1\2/g')
+# The CHARMAPS dependency handling must be able to process:
+# 1. No character map e.g. eo, en_US
+# 2. Character maps e.g. en_US.UTF-8
+# 3. Character maps with modifier e.g. tt_RU.UTF-8@iqtelif
+# This complicates the processing slightly so we do it in multiple edits,
+# the first captures the character map with the anchoring period while
+# the rest of the edits remove the period to get a valid file or adjust
+# the name to match the true name.
 CHARMAPS := $(shell echo "$(LOCALES)" | \
-		    sed -e 's/[^ .]*[.]\([^@ ]*\)\(@[^@ ]*\)*/\1/g' -e s/SJIS/SHIFT_JIS/g)
+		    sed -e 's/\([^ .]*\)\([^@ ]*\)\(@[^@ ]*\)*/\2/g' \
+			-e 's/^\./ /g' \
+			-e 's/ \./ /g' \
+			-e s/SJIS/SHIFT_JIS/g)
 CTYPE_FILES = $(addsuffix /LC_CTYPE,$(LOCALES))
 gen-locales := $(addprefix $(common-objpfx)localedata/,$(CTYPE_FILES))
 
diff --git a/locale/findlocale.c b/locale/findlocale.c
index 8d6e4e33e3..43ff7201c1 100644
--- a/locale/findlocale.c
+++ b/locale/findlocale.c
@@ -244,7 +244,14 @@ _nl_find_locale (const char *locale_path, size_t locale_path_len,
       locale_file = locale_file->successor[cnt];
 
       if (locale_file == NULL)
-	return NULL;
+	{
+	  /* If this is the second time we tried to load a failed
+	     locale then the locale_file value comes from the cache
+	     and we will not carry out any actual filesystem
+	     operations so we must set ENOENT here.  */
+	  __set_errno (ENOENT);
+	  return NULL;
+	}
     }
 
   /* The LC_CTYPE category allows to check whether a locale is really
@@ -291,8 +298,14 @@ _nl_find_locale (const char *locale_path, size_t locale_path_len,
       if (__gconv_compare_alias (upstr (ccodeset, ccodeset),
 				 upstr (clocale_codeset,
 					clocale_codeset)) != 0)
-	/* The codesets are not identical, don't use the locale.  */
-	return NULL;
+	{
+	  /* The codesets are not identical, don't use the locale.
+	     If this is the second time we tried to load a locale
+	     whose codeset doesn't match then the result came from
+	     the cache and must set ENOENT here.  */
+	  __set_errno (ENOENT);
+	  return NULL;
+	}
     }
 
   /* Determine the locale name for which loading succeeded.  This
diff --git a/locale/loadlocale.c b/locale/loadlocale.c
index 1e5f93e927..991c0591e9 100644
--- a/locale/loadlocale.c
+++ b/locale/loadlocale.c
@@ -237,7 +237,6 @@ _nl_load_locale (struct loaded_l10nfile *file, int category)
   int save_err;
   int alloc = ld_mapped;
 
-  file->decided = 1;
   file->data = NULL;
 
   fd = __open_nocancel (file->filename, O_RDONLY | O_CLOEXEC);
@@ -345,6 +344,7 @@ _nl_load_locale (struct loaded_l10nfile *file, int category)
   newdata->alloc = alloc;
 
   file->data = newdata;
+  file->decided = 1;
 }
 
 void
diff --git a/localedata/Makefile b/localedata/Makefile
index 713e7aebad..6d0bac225b 100644
--- a/localedata/Makefile
+++ b/localedata/Makefile
@@ -239,6 +239,7 @@ tests = \
   tst-iconv-emojis-trans \
   tst-iconv-math-trans \
   tst-leaks \
+  tst-locale-loadlocale \
   tst-mbswcs1 \
   tst-mbswcs2 \
   tst-mbswcs3 \
@@ -325,6 +326,7 @@ LOCALES := \
   dsb_DE.UTF-8 \
   dz_BT.UTF-8 \
   en_GB.UTF-8 \
+  en_US \
   en_US.ANSI_X3.4-1968 \
   en_US.ISO-8859-1\
   en_US.UTF-8 \
@@ -406,6 +408,8 @@ include ../gen-locales.mk
 $(objpfx)tst-iconv-emojis-trans.out: $(gen-locales)
 
 $(objpfx)tst-iconv-math-trans.out: $(gen-locales)
+# tst-locale-loadlocale: Needs an en_US-named locale for the test.
+$(objpfx)tst-locale-loadlocale.out: $(gen-locales)
 endif
 
 include ../Rules
diff --git a/localedata/gen-locale.sh b/localedata/gen-locale.sh
index b8c422ad13..e400d02b82 100644
--- a/localedata/gen-locale.sh
+++ b/localedata/gen-locale.sh
@@ -48,9 +48,9 @@ generate_locale ()
 }
 
 locfile=`echo $locfile|sed 's|.*/\([^/]*/LC_CTYPE\)|\1|'`
-locale=`echo $locfile|sed 's|\([^.]*\)[.].*/LC_CTYPE|\1|'`
-charmap=`echo $locfile|sed 's|[^.]*[.]\([^@ ]*\)\(@[^ ]*\)\?/LC_CTYPE|\1|'`
-modifier=`echo $locfile|sed 's|[^.]*[.]\([^@ ]*\)\(@[^ ]*\)\?/LC_CTYPE|\2|'`
+locale=`echo $locfile|sed 's|\([^.]*\).*/LC_CTYPE|\1|'`
+charmap=`echo $locfile|sed -e 's|[^.]*\([^@ ]*\)\(@[^ ]*\)\?/LC_CTYPE|\1|' -e 's|^\.||g'`
+modifier=`echo $locfile|sed 's|[^.]*\([^@ ]*\)\(@[^ ]*\)\?/LC_CTYPE|\2|'`
 
 echo "Generating locale $locale.$charmap: this might take a while..."
 
@@ -63,10 +63,16 @@ echo "Generating locale $locale.$charmap: this might take a while..."
 # you to develop in-progress locales.
 flags=""
 
+charmap_real="$charmap"
+
+# If no character map is specified then we fall back to UTF-8.
+if [ -z "$charmap" ]; then
+  charmap_real="UTF-8"
+fi
+
 # For SJIS the charmap is SHIFT_JIS. We just want the locale to have
 # a slightly nicer name instead of using "*.SHIFT_SJIS", but that
 # means we need a mapping here.
-charmap_real="$charmap"
 if [ "$charmap" = "SJIS" ]; then
   charmap_real="SHIFT_JIS"
 fi
@@ -80,4 +86,12 @@ if [ "$charmap_real" = 'SHIFT_JIS' ] \
   flags="$flags --no-warnings=ascii"
 fi
 
-generate_locale $charmap_real $locale$modifier $locale.$charmap$modifier "$flags"
+# If the character map is not specified then we output a locale
+# with the just the name of the locale e.g. eo, en_US. This is
+# used for test locales that act as fallbacks.
+output_file="$locale.$charmap$modifier"
+if [ -z "$charmap" ]; then
+  output_file="$locale"
+fi
+
+generate_locale $charmap_real $locale$modifier $output_file "$flags"
diff --git a/localedata/tst-locale-loadlocale.c b/localedata/tst-locale-loadlocale.c
new file mode 100644
index 0000000000..64796e975d
--- /dev/null
+++ b/localedata/tst-locale-loadlocale.c
@@ -0,0 +1,67 @@
+/* Test for locale loading error handling (Bug 14247)
+   Copyright (C) 2024 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <https://www.gnu.org/licenses/>.  */
+
+#include <stdio.h>
+#include <errno.h>
+#include <locale.h>
+#include <support/check.h>
+#include <support/support.h>
+
+static int
+do_test (void)
+{
+  locale_t loc = NULL;
+
+  /* We must load an en_US-based locale for the final test to fail.
+     The locale loading code will find the en_US locale already loaded
+     but the codesets won't match.  */
+  xsetlocale (LC_ALL, "en_US.UTF-8");
+
+  /* Call newlocale with an invalid locale.  We expect the test system
+     does not have "invalidlocale" locale.  */
+  errno = 0;
+  loc = newlocale (LC_COLLATE_MASK | LC_CTYPE_MASK, "invalidlocale.utf8", 0);
+  TEST_VERIFY (loc == NULL && errno != 0);
+
+  /* Call newlocale again with the same name. This triggers bug 14247 where
+     the second call fails to set errno correctly.  */
+  errno = 0;
+  loc = newlocale (LC_COLLATE_MASK | LC_CTYPE_MASK, "invalidlocale.utf8", 0);
+  TEST_VERIFY (loc == NULL && errno != 0);
+
+  /* Next we attempt to load a locale that exists but whose codeset
+     does not match the codeset of the locale on the system.
+     This is more difficult to test but we rely on the fact that locale
+     processing will normalize the locale name and attempt to open
+     "en_US" with no codeset as a fallback and this will allow us to
+     compare a loaded "en_US" locale with a UTF-8 codeset to the
+     ficiticious "en_US.utf99" and get a codeset match failure.  */
+  errno = 0;
+  loc = newlocale (LC_COLLATE_MASK | LC_CTYPE_MASK, "en_US.utf99", 0);
+  TEST_VERIFY (loc == NULL && errno != 0);
+
+  /* Call newlocale again with the same name. This triggers bug 14247 where
+     the second call fails to set errno correctly.  */
+  errno = 0;
+  loc = newlocale (LC_COLLATE_MASK | LC_CTYPE_MASK, "en_US.utf99", 0);
+  TEST_VERIFY (loc == NULL && errno != 0);
+
+  return 0;
+}
+
+#include <support/test-driver.c>