diff mbox series

[2/4] crypto4xx: fix cfb and ofb "overran dst buffer" issues

Message ID 4d8f4f483feb713126bbdb789b095936819d9804.1555932334.git.chunkeey@gmail.com (mailing list archive)
State Accepted
Delegated to: Herbert Xu
Headers show
Series [1/4] crypto4xx: fix ctr-aes missing output IV | expand

Commit Message

Christian Lamparter April 22, 2019, 11:25 a.m. UTC
Currently, crypto4xx CFB and OFB AES ciphers are
failing testmgr's test vectors.

|cfb-aes-ppc4xx encryption overran dst buffer on test vector 3, cfg="in-place"
|ofb-aes-ppc4xx encryption overran dst buffer on test vector 1, cfg="in-place"

This is because of a very subtile "bug" in the hardware that
gets indirectly mentioned in 18.1.3.5 Encryption/Decryption
of the hardware spec:

the OFB and CFB modes for AES are listed there as operation
modes for >>> "Block ciphers" <<<. Which kind of makes sense,
but we would like them to be considered as stream ciphers just
like the CTR mode.

To workaround this issue and stop the hardware from causing
"overran dst buffer" on crypttexts that are not a multiple
of 16 (AES_BLOCK_SIZE), we force the driver to use the scatter
buffers as the go-between.

As a bonus this patch also kills redundant pd_uinfo->num_gd
and pd_uinfo->num_sd setters since the value has already been
set before.

Cc: stable@vger.kernel.org
Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
---
 drivers/crypto/amcc/crypto4xx_core.c | 31 +++++++++++++++++++---------
 1 file changed, 21 insertions(+), 10 deletions(-)

Comments

Christian Lamparter April 22, 2019, 4:32 p.m. UTC | #1
On 4/22/19, Sasha Levin <sashal@kernel.org> wrote:
> Hi,
>
> [This is an automated email]
>
> This commit has been processed because it contains a -stable tag.
> The stable tag indicates that it's relevant for the following trees: all
>
> The bot has tested the following trees: v5.0.9, v4.19.36, v4.14.113,
> v4.9.170, v4.4.178, v3.18.138.
>
> v5.0.9: Build OK!
> v4.19.36: Build OK!
> v4.14.113: Failed to apply! Possible dependencies:
[...]
>     f2a13e7cba9e ("crypto: crypto4xx - enable AES RFC3686, ECB, CFB and OFB
> offloads")
>
>
> How should we proceed with this patch?
>
Same as with the aes-ctr fix. I guess, I should have added

Fixes: f2a13e7cba9e ("crypto: crypto4xx - enable AES RFC3686, ECB, CFB
and OFB offloads")

Regards,
Christian
diff mbox series

Patch

diff --git a/drivers/crypto/amcc/crypto4xx_core.c b/drivers/crypto/amcc/crypto4xx_core.c
index 06574a884715..920bd5e720b2 100644
--- a/drivers/crypto/amcc/crypto4xx_core.c
+++ b/drivers/crypto/amcc/crypto4xx_core.c
@@ -714,7 +714,23 @@  int crypto4xx_build_pd(struct crypto_async_request *req,
 	size_t offset_to_sr_ptr;
 	u32 gd_idx = 0;
 	int tmp;
-	bool is_busy;
+	bool is_busy, force_sd;
+
+	/*
+	 * There's a very subtile/disguised "bug" in the hardware that
+	 * gets indirectly mentioned in 18.1.3.5 Encryption/Decryption
+	 * of the hardware spec:
+	 * *drum roll* the AES/(T)DES OFB and CFB modes are listed as
+	 * operation modes for >>> "Block ciphers" <<<.
+	 *
+	 * To workaround this issue and stop the hardware from causing
+	 * "overran dst buffer" on crypttexts that are not a multiple
+	 * of 16 (AES_BLOCK_SIZE), we force the driver to use the
+	 * scatter buffers.
+	 */
+	force_sd = (req_sa->sa_command_1.bf.crypto_mode9_8 == CRYPTO_MODE_CFB
+		|| req_sa->sa_command_1.bf.crypto_mode9_8 == CRYPTO_MODE_OFB)
+		&& (datalen % AES_BLOCK_SIZE);
 
 	/* figure how many gd are needed */
 	tmp = sg_nents_for_len(src, assoclen + datalen);
@@ -732,7 +748,7 @@  int crypto4xx_build_pd(struct crypto_async_request *req,
 	}
 
 	/* figure how many sd are needed */
-	if (sg_is_last(dst)) {
+	if (sg_is_last(dst) && force_sd == false) {
 		num_sd = 0;
 	} else {
 		if (datalen > PPC4XX_SD_BUFFER_SIZE) {
@@ -807,9 +823,10 @@  int crypto4xx_build_pd(struct crypto_async_request *req,
 	pd->sa_len = sa_len;
 
 	pd_uinfo = &dev->pdr_uinfo[pd_entry];
-	pd_uinfo->async_req = req;
 	pd_uinfo->num_gd = num_gd;
 	pd_uinfo->num_sd = num_sd;
+	pd_uinfo->dest_va = dst;
+	pd_uinfo->async_req = req;
 
 	if (iv_len)
 		memcpy(pd_uinfo->sr_va->save_iv, iv, iv_len);
@@ -828,7 +845,6 @@  int crypto4xx_build_pd(struct crypto_async_request *req,
 		/* get first gd we are going to use */
 		gd_idx = fst_gd;
 		pd_uinfo->first_gd = fst_gd;
-		pd_uinfo->num_gd = num_gd;
 		gd = crypto4xx_get_gdp(dev, &gd_dma, gd_idx);
 		pd->src = gd_dma;
 		/* enable gather */
@@ -865,17 +881,14 @@  int crypto4xx_build_pd(struct crypto_async_request *req,
 		 * Indicate gather array is not used
 		 */
 		pd_uinfo->first_gd = 0xffffffff;
-		pd_uinfo->num_gd = 0;
 	}
-	if (sg_is_last(dst)) {
+	if (!num_sd) {
 		/*
 		 * we know application give us dst a whole piece of memory
 		 * no need to use scatter ring.
 		 */
 		pd_uinfo->using_sd = 0;
 		pd_uinfo->first_sd = 0xffffffff;
-		pd_uinfo->num_sd = 0;
-		pd_uinfo->dest_va = dst;
 		sa->sa_command_0.bf.scatter = 0;
 		pd->dest = (u32)dma_map_page(dev->core_dev->device,
 					     sg_page(dst), dst->offset,
@@ -889,9 +902,7 @@  int crypto4xx_build_pd(struct crypto_async_request *req,
 		nbytes = datalen;
 		sa->sa_command_0.bf.scatter = 1;
 		pd_uinfo->using_sd = 1;
-		pd_uinfo->dest_va = dst;
 		pd_uinfo->first_sd = fst_sd;
-		pd_uinfo->num_sd = num_sd;
 		sd = crypto4xx_get_sdp(dev, &sd_dma, sd_idx);
 		pd->dest = sd_dma;
 		/* setup scatter descriptor */