Brad Bishop | d7bf8c1 | 2018-02-25 22:55:05 -0500 | [diff] [blame] | 1 | From 042421e9e3d266ad0bb7805132041ef51ad3234d Mon Sep 17 00:00:00 2001 |
| 2 | From: Adrian Johnson <ajohnson@redneon.com> |
| 3 | Date: Wed, 16 Aug 2017 22:52:35 -0400 |
| 4 | Subject: [PATCH] cairo: Fix CVE-2017-9814 |
| 5 | |
| 6 | The bug happens because in some scenarios the variable size can |
| 7 | have a value of 0 at line 1288. And malloc(0) is not returning |
| 8 | NULL as some people could expect: |
| 9 | |
| 10 | https://stackoverflow.com/questions/1073157/zero-size-malloc |
| 11 | |
| 12 | malloc(0) returns the smallest chunk possible. So the line 1290 |
| 13 | with the return is not execute. And the execution continues with |
| 14 | an invalid map. |
| 15 | |
| 16 | Since the size is 0 the variable map is not initialized correctly |
| 17 | at load_trutype_table. So, later when the variable map is accessed |
| 18 | previous values from a freed chunk are used. This could allows an |
| 19 | attacker to control the variable map. |
| 20 | |
| 21 | This patch have not merge in upstream now. |
| 22 | |
| 23 | Upstream-Status: Backport [https://bugs.freedesktop.org/show_bug.cgi?id=101547] |
| 24 | CVE: CVE-2017-9814 |
| 25 | Signed-off-by: Dengke Du <dengke.du@windriver.com> |
| 26 | --- |
| 27 | src/cairo-truetype-subset.c | 2 +- |
| 28 | 1 file changed, 1 insertion(+), 1 deletion(-) |
| 29 | |
| 30 | diff --git a/src/cairo-truetype-subset.c b/src/cairo-truetype-subset.c |
| 31 | index e3449a0..f77d11c 100644 |
| 32 | --- a/src/cairo-truetype-subset.c |
| 33 | +++ b/src/cairo-truetype-subset.c |
| 34 | @@ -1285,7 +1285,7 @@ _cairo_truetype_reverse_cmap (cairo_scaled_font_t *scaled_font, |
| 35 | return CAIRO_INT_STATUS_UNSUPPORTED; |
| 36 | |
| 37 | size = be16_to_cpu (map->length); |
| 38 | - map = malloc (size); |
| 39 | + map = _cairo_malloc (size); |
| 40 | if (unlikely (map == NULL)) |
| 41 | return _cairo_error (CAIRO_STATUS_NO_MEMORY); |
| 42 | |
| 43 | -- |
| 44 | 2.8.1 |
| 45 | |