linux-kernel - [PATCH] crypto: s5p-sss.c: Fix kernel Oops in AES-ECB mode

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

Date:   Mon, 05 Feb 2018 18:40:20 +0100
From:   Kamil Konieczny <k.konieczny@...tner.samsung.com>
To:     Herbert Xu <herbert@...dor.apana.org.au>
Cc:     Krzysztof Kozlowski <krzk@...nel.org>,
        Vladimir Zapolskiy <vz@...ia.com>,
        "David S. Miller" <davem@...emloft.net>,
        Bartlomiej Zolnierkiewicz <b.zolnierkie@...sung.com>,
        Marek Szyprowski <m.szyprowski@...sung.com>,
        Anand Moon <linux.amoon@...il.com>,
        linux-crypto@...r.kernel.org, linux-samsung-soc@...r.kernel.org,
        linux-kernel <linux-kernel@...r.kernel.org>
Subject: [PATCH] crypto: s5p-sss.c: Fix kernel Oops in AES-ECB mode


In AES-ECB mode crypt is done with key only, so any use of IV
can cause kernel Oops, as reported by Anand Moon.
Fixed it by using IV only in AES-CBC and AES-CTR.

Signed-off-by: Kamil Konieczny <k.konieczny@...tner.samsung.com>
Reported-by: Anand Moon <linux.amoon@...il.com>
---
Tested on Odroid XU4/HC1, kernel 4.15 with following command:

fallocate -l 128MiB /tmp/test.bin
dd if=/dev/urandom of=/tmp/testkey.key bs=128 count=1
sync
cryptsetup luksFormat --debug -q -d /tmp/testkey.key \
  --cipher aes-cbc-essiv:sha256 -h sha256 -s 128 /tmp/test.bin

The original report by Anand Moon:
https://www.spinics.net/lists/linux-crypto/msg31180.html

Oops reproduced with cryptsetup 2.0.0, kernel 4.15,
in .config in crypto API ECB support was turned off, and s5p-sss AES driver on.

cryptsetup is using aes-ecb and has req->info in aes_ctx set to 0x10,
which caused Oops:

[ 2078.683779] Internal error: Oops: 17 [#1] PREEMPT SMP ARM
[ 2078.689148] Modules linked in: algif_skcipher af_alg sd_mod sg
evdev uas usb_storage scsi_mod gpio_keys fbtft(C) spidev spi_s3c64xx
ipv6
[ 2078.701377] CPU: 1 PID: 15 Comm: ksoftirqd/1 Tainted: G         C
    4.15.0-rc9-xu4krck #1
[ 2078.709861] Hardware name: SAMSUNG EXYNOS (Flattened Device Tree)
[ 2078.715932] PC is at memcpy+0x80/0x330
[ 2078.719652] LR is at s5p_tasklet_cb+0x19c/0x328

 drivers/crypto/s5p-sss.c | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)

diff --git a/drivers/crypto/s5p-sss.c b/drivers/crypto/s5p-sss.c
index 142c6020cec7..5c0496d1ed41 100644
--- a/drivers/crypto/s5p-sss.c
+++ b/drivers/crypto/s5p-sss.c
@@ -1926,15 +1926,21 @@ static void s5p_aes_crypt_start(struct s5p_aes_dev *dev, unsigned long mode)
 	uint32_t aes_control;
 	unsigned long flags;
 	int err;
+	u8 *iv;
 
 	aes_control = SSS_AES_KEY_CHANGE_MODE;
 	if (mode & FLAGS_AES_DECRYPT)
 		aes_control |= SSS_AES_MODE_DECRYPT;
 
-	if ((mode & FLAGS_AES_MODE_MASK) == FLAGS_AES_CBC)
+	if ((mode & FLAGS_AES_MODE_MASK) == FLAGS_AES_CBC) {
 		aes_control |= SSS_AES_CHAIN_MODE_CBC;
-	else if ((mode & FLAGS_AES_MODE_MASK) == FLAGS_AES_CTR)
+		iv = req->info;
+	} else if ((mode & FLAGS_AES_MODE_MASK) == FLAGS_AES_CTR) {
 		aes_control |= SSS_AES_CHAIN_MODE_CTR;
+		iv = req->info;
+	} else {
+		iv = NULL; /* AES_ECB */
+	}
 
 	if (dev->ctx->keylen == AES_KEYSIZE_192)
 		aes_control |= SSS_AES_KEY_SIZE_192;
@@ -1965,7 +1971,7 @@ static void s5p_aes_crypt_start(struct s5p_aes_dev *dev, unsigned long mode)
 		goto outdata_error;
 
 	SSS_AES_WRITE(dev, AES_CONTROL, aes_control);
-	s5p_set_aes(dev, dev->ctx->aes_key, req->info, dev->ctx->keylen);
+	s5p_set_aes(dev, dev->ctx->aes_key, iv, dev->ctx->keylen);
 
 	s5p_set_dma_indata(dev,  dev->sg_src);
 	s5p_set_dma_outdata(dev, dev->sg_dst);
-- 
2.16.0