public inbox for libc-alpha@sourceware.org
 help / color / mirror / Atom feed
From: Gao Xiang <gaoxiang@kylinos.cn>
To: schwab@suse.de
Cc: libc-alpha@sourceware.org, Xiang Gao <gaoxiang@kylinos.cn>
Subject: [PATCH v3] libio: Fix wide stream backup buffer leak on fclose [BZ #33999]
Date: Thu,  2 Apr 2026 14:25:44 +0800	[thread overview]
Message-ID: <20260402062544.322321-1-gaoxiang@kylinos.cn> (raw)
In-Reply-To: <mvmtstueqdi.fsf@suse.de>

From: Xiang Gao <gaoxiang@kylinos.cn>

This patch fixes a memory leak when ungetwc is used on a wide oriented stream.
The backup buffer was never freed on fclose, causing a memory leak per
ungetwc/fclose call.

The leak has two causes:

In iofclose.c, for wide streams (fp->mode > 0), _IO_new_fclose never calls
_IO_free_wbackup_area. Fixed by adding the missing call.

In wgenops.c, _IO_wdefault_finish checks fp->_IO_save_base (the narrow field,
always NULL for wide streams) instead of fp->_wide_data->_IO_save_base,
and uses a bare free() that leaves _IO_save_end and _IO_backup_base dangling.
Replace the hand-rolled cleanup with _IO_have_wbackup/_IO_free_wbackup_area,
which handles backup-mode switching and clears all three pointers.

This was independently reported by Rocket Ma [1], whose patch corrects the condition
but still uses the manual free path.

Apply the same _IO_have_backup condition in genops.c for consistency.

Tested by:
make test t=libio/tst-wbackup-leak

[1] https://patchwork.sourceware.org/project/glibc/patch/20260323171742.1039768-1-marocketbd@gmail.com/

Signed-off-by: Xiang Gao <gaoxiang@kylinos.cn>
---
Thanks for the review, Andreas.

The non-breaking spaces in the comment block were caused by a misconfigured
vim setting on my local Fedora submission environment. Fixed now.

I added a note in the test explaining that it relies on open_wmemstream not
setting _IO_NO_READS on the stream. I will keep an eye on any future discussion
or changes around open_wmemstream.

Changes since v2:
 - Use regular spaces in comment block.
 - Notes the open_wmemstream limit in the test comment to make clear the
   assumption the test depends on.
 - Updates condition check on wide stream path in iofclose.c.
 - Adjust surrounding indentation to meet coding style.
---
 libio/Makefile           |  1 +
 libio/genops.c           |  7 ++----
 libio/iofclose.c         |  2 ++
 libio/tst-wbackup-leak.c | 50 ++++++++++++++++++++++++++++++++++++++++
 libio/wgenops.c          |  7 ++----
 5 files changed, 57 insertions(+), 10 deletions(-)
 create mode 100644 libio/tst-wbackup-leak.c

diff --git a/libio/Makefile b/libio/Makefile
index 08e1e0ec25..93656466df 100644
--- a/libio/Makefile
+++ b/libio/Makefile
@@ -135,6 +135,7 @@ tests = \
   tst-swscanf \
   tst-ungetwc1 \
   tst-ungetwc2 \
+  tst-wbackup-leak \
   tst-wfile-sync \
   tst-wfiledoallocate-static \
   tst-widetext \
diff --git a/libio/genops.c b/libio/genops.c
index cc1684e00a..90e08e6571 100644
--- a/libio/genops.c
+++ b/libio/genops.c
@@ -636,11 +636,8 @@ _IO_default_finish (FILE *fp, int dummy)
   for (mark = fp->_markers; mark != NULL; mark = mark->_next)
     mark->_sbuf = NULL;
 
-  if (fp->_IO_save_base)
-    {
-      _IO_free_backup_buf (fp, fp->_IO_save_base);
-      fp->_IO_save_base = NULL;
-    }
+  if (_IO_have_backup (fp))
+    _IO_free_backup_area (fp);
 
   _IO_un_link ((struct _IO_FILE_plus *) fp);
 
diff --git a/libio/iofclose.c b/libio/iofclose.c
index 89782e99d7..0ba85955bf 100644
--- a/libio/iofclose.c
+++ b/libio/iofclose.c
@@ -67,6 +67,8 @@ _IO_new_fclose (FILE *fp)
   _IO_FINISH (fp);
   if (fp->_mode > 0)
     {
+      if (_IO_have_wbackup (fp))
+	_IO_free_wbackup_area (fp);
       /* This stream has a wide orientation.  This means we have to free
 	 the conversion functions.  */
       struct _IO_codecvt *cc = fp->_codecvt;
diff --git a/libio/tst-wbackup-leak.c b/libio/tst-wbackup-leak.c
new file mode 100644
index 0000000000..8a3e0fc1ed
--- /dev/null
+++ b/libio/tst-wbackup-leak.c
@@ -0,0 +1,50 @@
+/* Test _IO_wdefault_finish frees wide backup buffer [BZ #33999].  */
+
+#include <malloc.h>
+#include <stdio.h>
+#include <wchar.h>
+#include <support/check.h>
+
+static void
+one_round (void)
+{
+  wchar_t *buf = NULL;
+  size_t size = 0;
+
+  FILE *fp = open_wmemstream (&buf, &size);
+  TEST_VERIFY_EXIT (fp != NULL);
+  fputwc (L'A', fp);
+  fflush (fp);
+  /* Push back without prior read. read_ptr == read_base, so
+     _IO_wdefault_pbackfail skips the buggy narrow read_ptr access
+     (BZ #33998) and goes straight to allocating a wide backup
+     buffer at fp->_wide_data->_IO_save_base.
+
+     Note: this testcase relies on the fact that open_wmemstream
+     does not set _IO_NO_READS on the stream. If that implementation
+     is changed, this test would need a different stream type to verify
+     the leak.  */
+  ungetwc (L'Z', fp);
+  fclose (fp);
+  free (buf);
+}
+
+static int
+do_test (void)
+{
+  /* Warm up to stabilize allocator state.  */
+  one_round ();
+
+  struct mallinfo2 before = mallinfo2 ();
+  for (int i = 0; i < 1000; i++)
+    one_round ();
+  struct mallinfo2 after = mallinfo2 ();
+
+  /* Each leak is 128 * sizeof(wchar_t) = 512 bytes.
+   * 1000 iterations would leak ~512 KB. Allow 4 KB noise. */
+  TEST_VERIFY (after.uordblks - before.uordblks < 4096);
+
+  return 0;
+}
+
+#include <support/test-driver.c>
diff --git a/libio/wgenops.c b/libio/wgenops.c
index 064d71266d..6829477e0c 100644
--- a/libio/wgenops.c
+++ b/libio/wgenops.c
@@ -181,11 +181,8 @@ _IO_wdefault_finish (FILE *fp, int dummy)
   for (mark = fp->_markers; mark != NULL; mark = mark->_next)
     mark->_sbuf = NULL;
 
-  if (fp->_IO_save_base)
-    {
-      free (fp->_wide_data->_IO_save_base);
-      fp->_IO_save_base = NULL;
-    }
+  if (_IO_have_wbackup (fp))
+    _IO_free_wbackup_area (fp);
 
 #ifdef _IO_MTSAFE_IO
   if (fp->_lock != NULL)
-- 
2.53.0


  reply	other threads:[~2026-04-02  6:26 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-03-31 13:35 [PATCH v2] " Gao Xiang
2026-04-01 13:56 ` Andreas Schwab
2026-04-02  6:25   ` Gao Xiang [this message]
2026-04-02 12:47     ` [PATCH v3] " Andreas Schwab
2026-04-03  1:57       ` [PATCH v4] " Gao Xiang

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20260402062544.322321-1-gaoxiang@kylinos.cn \
    --to=gaoxiang@kylinos.cn \
    --cc=libc-alpha@sourceware.org \
    --cc=schwab@suse.de \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).