oob in SDL_InvalidateMap

[SDLBugzilla]

This bug report was migrated from our old Bugzilla tracker.

These attachments are available in the static archive:

• poc (oob.tar.gz, application/gzip, 2019-05-05 03:04:23 +0000, 12032 bytes)
• [PATCH] pcx: cast size and check calloc return value (CVE-2019-12222.patch, text/plain, 2019-05-28 14:23:35 +0000, 1594 bytes)

Reported in version: unspecified
Reported for operating system, platform: Linux, x86_64

Comments on the original bug report:

On 2019-05-05 03:04:23 +0000, pwd wrote:

Created attachment 3775
poc

SDL_InvalidateMap@SDL_pixels.c:979-13___SEGV_UNKNOW

description

An issue was discovered in libsdl2 2.0.9 with SDL2_image-2.0.4, There is a/an SEGV_UNKNOW in function SDL_InvalidateMap at SDL_pixels.c:979-13

commandline

loadtif  @@ 

source

 975         return;
 976     }
 977     if (map->dst) {
 978         /* Release our reference to the surface - see the note below */
> 979         if (--map->dst->refcount <= 0) {
 980             SDL_FreeSurface(map->dst);
 981         }
 982     }
 983     map->dst = NULL;
 984     map->src_palette_version = 0;

// loadtif.c
// #include <stdio.h>
// #include <SDL.h>
// #include <SDL_image.h>
//
// int main(int argc, char * argv[]){
//         IMG_Init(IMG_INIT_TIF);//IMG_INIT_JPG);IMG_INIT_PNG
//         while(__AFL_LOOP(1000)){
//               SDL_Surface * image = IMG_Load(argv[1]);
//               if (image){
//                 SDL_FreeSurface(image);
//               }
//         }
//         IMG_Quit();
// }

bug report

ASAN:DEADLYSIGNAL
=================================================================
==14609==ERROR: AddressSanitizer: SEGV on unknown address 0x1b60013f810b (pc 0x7f7210f5a77a bp 0x7ffc84c8bd80 sp 0x7ffc84c8ba80 T0)
    # 0 0x7f7210f5a779 in SDL_InvalidateMap /src/libsdl2/src/video/SDL_pixels.c:979:13
    # 1 0x7f7210f68c3a in SDL_FreeSurface_REAL /src/libsdl2/src/video/SDL_surface.c:1221:5
    # 2 0x7f72112bc0b1 in IMG_LoadPCX_RW /src/SDL2_image-2.0.4/IMG_pcx.c:252:13
    # 3 0x7f72112a99bd in IMG_LoadTyped_RW /src/SDL2_image-2.0.4/IMG.c:195:17
    # 4 0x7f72112a8f41 in IMG_Load /src/SDL2_image-2.0.4/IMG.c:136:12
    # 5 0x4ea0f0 in main /src/loadtif.c:8:37
    # 6 0x7f720fdaf82f in __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x2082f)
    # 7 0x4189e8 in _start (/src/aflbuild/installed/bin/loadtif+0x4189e8)

AddressSanitizer can not provide additional info.
SUMMARY: AddressSanitizer: SEGV /src/libsdl2/src/video/SDL_pixels.c:979:13 in SDL_InvalidateMap
==14609==ABORTING

others

from fuzz project pwd-libsdl2-loadtif-00
crash name pwd-libsdl2-loadtif-00-00000007-20190419.tif
Auto-generated by pyspider at 2019-04-19 00:37:05

On 2019-05-28 14:23:35 +0000, Hugo Lefeuvre wrote:

Created attachment 3799
[PATCH] pcx: cast size and check calloc return value

Patch in attachment.

summary:

bpl is stored as a signed integer. If it happens to be negative, calloc
will be called with surface->pitch (since bpl < surface->pitch). Later we
call SDL_RWread(src, buf, bpl, 1). bpl is thus cast to size_t (becoming a
very large positive value), leading to obvious oob write.

We should fail early in this case. It doesn't make sense to continue
processing such files with corrupted bpl.

• (size_t) cast bpl in SDL_max so that it is preferred over surface->pitch
  if it is negative
• check calloc return value to catch allocation failures
• make sure we don't free unallocated buf in done section

On 2019-05-28 14:24:30 +0000, Hugo Lefeuvre wrote:

This is a bug in SDL_Image.

This issue was assigned CVE-2019-12222.

On 2019-05-28 17:58:41 +0000, Ryan C. Gordon wrote:

(In reply to Hugo Lefeuvre from comment # 1)

• make sure we don't free unallocated buf in done section

Minor nitpick:

+    if (buf) {
+        SDL_free(buf);
+    }

There's no need to check if buf != NULL here; SDL_free correctly ignores NULLs.

Other than that, this patch looks good to me.

--ryan.

On 2019-05-28 20:02:14 +0000, Hugo Lefeuvre wrote:

Minor nitpick:

+    if (buf) {
+        SDL_free(buf);
+    }

There's no need to check if buf != NULL here; SDL_free correctly ignores
NULLs.

Other than that, this patch looks good to me.

Thanks for the review. Should I submit an updated version?

On 2019-06-10 23:21:21 +0000, Sam Lantinga wrote:

This is fixed, thanks!

On 2019-06-19 10:00:35 +0000, Castro B wrote:

May i know the solution?

Castro B,
https://sparpedia.ch

On 2019-06-19 13:50:08 +0000, Sam Lantinga wrote:

https://hg.libsdl.org/SDL_image/rev/e7e9786a1a34