linux-kernel - [PATCH] Add compile-time check for __ARCH_SI_PREAMBLE

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Date:	Sun, 17 Jan 2016 22:00:46 +0100
From:	Helge Deller <deller@....de>
To:	Linus Torvalds <torvalds@...ux-foundation.org>,
	Andrew Morton <akpm@...ux-foundation.org>,
	linux-kernel@...r.kernel.org, linux-parisc@...r.kernel.org
Cc:	Stephen Rothwell <sfr@...b.auug.org.au>,
	Michael Ellerman <mpe@...erman.id.au>
Subject: [PATCH] Add compile-time check for __ARCH_SI_PREAMBLE_SIZE

The value of __ARCH_SI_PREAMBLE_SIZE defines the size (including padding) of
the part of the struct siginfo that is before the union, and it is then used to
calculate the needed padding (SI_PAD_SIZE) to make the size of struct siginfo
equal to 128 (SI_MAX_SIZE) bytes.

Depending on the target architecture and word width it equals to either
3 or 4 times sizeof int.

Since the very beginning we had __ARCH_SI_PREAMBLE_SIZE wrong on the parisc
architecture for the 64bit kernel build. It's even more frustrating, because it
can easily be checked at compile time if the value was defined correctly.

This patch adds such a check for the correctness of __ARCH_SI_PREAMBLE_SIZE in
the hope that it will prevent existing and future architectures from running
into the same problem.

I refrained from replacing __ARCH_SI_PREAMBLE_SIZE by offsetof() in
copy_siginfo() in include/asm-generic/siginfo.h, because a) it doesn't make any
difference and b) it's used in the Documentation/kmemcheck.txt example.

I ran this patch through the 0-DAY kernel test infrastructure and only the
parisc architecture triggered as expected. That means that this patch should be
OK for all major architectures.

Signed-off-by: Helge Deller <deller@....de>

diff --git a/kernel/signal.c b/kernel/signal.c
index f3f1f7a..1ccd218 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -3579,6 +3579,10 @@ __weak const char *arch_vma_name(struct vm_area_struct *vma)

 void __init signals_init(void)
 {
+	/* If this check fails, the __ARCH_SI_PREAMBLE_SIZE value is wrong! */
+	BUILD_BUG_ON(__ARCH_SI_PREAMBLE_SIZE
+		!= offsetof(struct siginfo, _sifields._pad));
+
 	sigqueue_cachep = KMEM_CACHE(sigqueue, SLAB_PANIC);
 }