rpms/ghostscript
6
0
Fork 0
Code Issues Pull Requests Releases Activity

RHEL-67050 CVE-2024-46952 ghostscript: Buffer Overflow in Ghostscript PDF XRef Stream Handling

Browse Source 
Resolves: RHEL-67050
This commit is contained in:
Zdenek Dohnal 2025-04-16 12:48:51 +02:00
parent e184197e52
commit f5721d2a28
2 changed files with 65 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

62
0001-PDF-interpreter-sanitise-W-array-values-in-Xref-stre.patch Normal file
View File

@ -0,0 +1,62 @@
From b1f0827c30f59a2dcbc8a39e42cace7a1de35f7f Mon Sep 17 00:00:00 2001
From: Ken Sharp <Ken.Sharp@artifex.com>
Date: Mon, 2 Sep 2024 15:14:01 +0100
Subject: [PATCH] PDF interpreter - sanitise W array values in Xref streams
Bug #708001 "Buffer overflow in PDF XRef stream"
See bug report. I've chosen to fix this by checking the values in the
W array; these can (currently at least) only have certain relatively
small values.
As a future proofing fix I've also updated field_size in
pdf_xref_stream_entries() to be a 64-bit integer. This is far bigger
than required, but matches the W array values and so prevents the
mismatch which could lead to a buffer overrun.
CVE-2024-46952
---
pdf/pdf_xref.c | 20 +++++++++++++++++++-
1 file changed, 19 insertions(+), 1 deletion(-)
diff --git a/pdf/pdf_xref.c b/pdf/pdf_xref.c
index fefd08abe..4ad0f548f 100644
--- a/pdf/pdf_xref.c
+++ b/pdf/pdf_xref.c
@@ -53,7 +53,7 @@ static int resize_xref(pdf_context *ctx, uint64_t new_size)
static int read_xref_stream_entries(pdf_context *ctx, pdf_c_stream *s, int64_t first, int64_t last, int64_t *W)
{
uint i, j;
- uint field_width = 0;
+ uint64_t field_width = 0;
uint32_t type = 0;
uint64_t objnum = 0, gen = 0;
byte *Buffer;
@@ -305,6 +305,24 @@ static int pdfi_process_xref_stream(pdf_context *ctx, pdf_stream *stream_obj, pd
}
pdfi_countdown(a);
+ /* W[0] is either:
+ * 0 (no type field) or a single byte with the type.
+ * W[1] is either:
+ * The object number of the next free object, the byte offset of this object in the file or the object5 number of the object stream where this object is stored.
+ * W[2] is either:
+ * The generation number to use if this object is used again, the generation number of the object or the index of this object within the object stream.
+ *
+ * Object and generation numbers are limited to unsigned 64-bit values, as are bytes offsets in the file, indexes of objects within the stream likewise (actually
+ * most of these are generally 32-bit max). So we can limit the field widths to 8 bytes, enough to hold a 64-bit number.
+ * Even if a later version of the spec makes these larger (which seems unlikely!) we still cna't cope with integers > 64-bits.
+ */
+ if (W[0] > 1 || W[1] > 8 || W[2] > 8) {
+ pdfi_close_file(ctx, XRefStrm);
+ pdfi_countdown(ctx->xref_table);
+ ctx->xref_table = NULL;
+ return code;
+ }
+
code = pdfi_dict_get_type(ctx, sdict, "Index", PDF_ARRAY, (pdf_obj **)&a);
if (code == gs_error_undefined) {
code = read_xref_stream_entries(ctx, XRefStrm, 0, size - 1, W);
--
2.49.0

3
ghostscript.spec
View File

@ -131,6 +131,8 @@ Patch: 0001-Uniprint-device-prevent-string-configuration-changes.patch
Patch: 0001-Bug-707691.patch
# RHEL-67044 CVE-2024-46951 ghostscript: Arbitrary Code Execution in Artifex Ghostscript Pattern Color Space
Patch: 0001-PS-interpreter-check-the-type-of-the-Pattern-Impleme.patch
# CVE-2024-46952 ghostscript: Buffer Overflow in Ghostscript PDF XRef Stream Handling
Patch: 0001-PDF-interpreter-sanitise-W-array-values-in-Xref-stre.patch
# Downstream patches -- these should be always included when doing rebase:
# ------------------
@ -449,6 +451,7 @@ done
%changelog
* Tue Apr 15 2025 Zdenek Dohnal <zdohnal@redhat.com> - 10.02.1-15
- RHEL-67044 CVE-2024-46951 ghostscript: Arbitrary Code Execution in Artifex Ghostscript Pattern Color Space
- RHEL-67050 CVE-2024-46952 ghostscript: Buffer Overflow in Ghostscript PDF XRef Stream Handling
* Tue Oct 29 2024 Troy Dawson <tdawson@redhat.com> - 10.02.1-14
- Bump release for October 2024 mass rebuild: