linux-kernel - [PATCH 5.15 084/107] s390: fix nospec table alignments

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 for Android: free password hash cracker in your pocket

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

Date:   Tue,  6 Sep 2022 15:31:05 +0200
From:   Greg Kroah-Hartman <gregkh@...uxfoundation.org>
To:     linux-kernel@...r.kernel.org
Cc:     Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        stable@...r.kernel.org, Josh Poimboeuf <jpoimboe@...nel.org>,
        Heiko Carstens <hca@...ux.ibm.com>,
        Vasily Gorbik <gor@...ux.ibm.com>
Subject: [PATCH 5.15 084/107] s390: fix nospec table alignments

From: Josh Poimboeuf <jpoimboe@...nel.org>

commit c9305b6c1f52060377c72aebe3a701389e9f3172 upstream.

Add proper alignment for .nospec_call_table and .nospec_return_table in
vmlinux.

[hca@...ux.ibm.com]: The problem with the missing alignment of the nospec
tables exist since a long time, however only since commit e6ed91fd0768
("s390/alternatives: remove padding generation code") and with
CONFIG_RELOCATABLE=n the kernel may also crash at boot time.

The above named commit reduced the size of struct alt_instr by one byte,
so its new size is 11 bytes. Therefore depending on the number of cpu
alternatives the size of the __alt_instructions array maybe odd, which
again also causes that the addresses of the nospec tables will be odd.

If the address of __nospec_call_start is odd and the kernel is compiled
With CONFIG_RELOCATABLE=n the compiler may generate code that loads the
address of __nospec_call_start with a 'larl' instruction.

This will generate incorrect code since the 'larl' instruction only works
with even addresses. In result the members of the nospec tables will be
accessed with an off-by-one offset, which subsequently may lead to
addressing exceptions within __nospec_revert().

Fixes: f19fbd5ed642 ("s390: introduce execute-trampolines for branches")
Signed-off-by: Josh Poimboeuf <jpoimboe@...nel.org>
Link: https://lore.kernel.org/r/8719bf1ce4a72ebdeb575200290094e9ce047bcc.1661557333.git.jpoimboe@kernel.org
Cc: <stable@...r.kernel.org> # 4.16
Reviewed-by: Heiko Carstens <hca@...ux.ibm.com>
Signed-off-by: Heiko Carstens <hca@...ux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@...ux.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@...uxfoundation.org>
---
 arch/s390/kernel/vmlinux.lds.S |    1 +
 1 file changed, 1 insertion(+)

--- a/arch/s390/kernel/vmlinux.lds.S
+++ b/arch/s390/kernel/vmlinux.lds.S
@@ -132,6 +132,7 @@ SECTIONS
 	/*
 	 * Table with the patch locations to undo expolines
 	*/
+	. = ALIGN(4);
 	.nospec_call_table : {
 		__nospec_call_start = . ;
 		*(.s390_indirect*)