Skip to contentSkip to conversation

Add test for CVE 2018-7566

PATCH New Martin Doucha 2 participants 2 messages Apr 8, 2020

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Add test for CVE 2018-7566

Fixes #606
---

There are two reproducers for this CVE:
- ioctl()/ioctl() race
- ioctl()/write() race

I've implemented the first one because it reliably crashes SLE-12SP3 GM kernel. The other reproducer gets stuck in one of the syscalls but I couldn't get it to trigger an actual crash. Should I implement the other one as well?

 runtest/cve                        |   1 +
 testcases/kernel/sound/.gitignore  |   1 +
 testcases/kernel/sound/Makefile    |   2 +
 testcases/kernel/sound/snd_seq01.c | 107 +++++++++++++++++++++++++++++
 4 files changed, 111 insertions(+)
 create mode 100644 testcases/kernel/sound/snd_seq01.c
runtest/cve +1 -0
index dbd065fd1..7d754085e 100644
@@ -40,6 +40,7 @@ cve-2017-17053 cve-2017-17053
 cve-2017-18075 pcrypt_aead01
 cve-2017-1000380 snd_timer01
 cve-2018-5803 sctp_big_chunk
+cve-2018-7566 snd_seq01
 cve-2018-1000001 realpath01
 cve-2018-1000199 ptrace08
 cve-2018-1000204 ioctl_sg01
testcases/kernel/sound/.gitignore +1 -0
index 33692e690..e33480fb6 100644
@@ -1 +1,2 @@
 snd_timer01
+/snd_seq01
testcases/kernel/sound/Makefile +2 -0
index 9abcc2b27..022e69197 100644
@@ -8,5 +8,7 @@ CPPFLAGS		+= -D_GNU_SOURCE
 
 snd_timer01: CFLAGS+=-pthread
 snd_timer01: LDLIBS+=-lrt
+snd_seq01:	CFLAGS += -pthread
+snd_seq01:	LDLIBS += -lrt
 
 include $(top_srcdir)/include/mk/generic_leaf_target.mk
testcases/kernel/sound/snd_seq01.c new file +107 -0

107 lines not on this page · raw message

Signature · 1 line
2.26.0
#1 lore raw

PATCH 1/1 Add test for CVE 2018-7566

Hi!

Quoted 1 line

Fixes #606

Add your signed-off-by here and pushed, thanks.

Quoted 9 lines

---

There are two reproducers for this CVE:
- ioctl()/ioctl() race
- ioctl()/write() race

I've implemented the first one because it reliably crashes SLE-12SP3 GM kernel. The other reproducer gets stuck in one of the syscalls but I couldn't get it to trigger an actual crash. Should I implement the other one as well?

Yes please, stuck tests is easy to detect as well, moreover the behavior of these tests is unpredictable anyways so we cannot really tell what will happen when the bug is triggered.

Signature · 2 lines
Cyril Hrubis
[email protected]