Re: [PATCH v2 0/2] crypto: mxs_dcp: Fix an Oops on i.MX6ULL

[Herbert Xu <herbert@gondor.apana.org.au>]

On Thu, Jul 01, 2021 at 02:56:36PM -0400, Sean Anderson wrote:
> This fixes at least one oops when using the DCP on ULL. However, I got
> another Oops when running kcapi-dgst-test.sh from the libkcapi test
> suite [1]:
> 
> [ 6961.181777] Unable to handle kernel NULL pointer dereference at virtual address 000008f8
> [ 6961.190143] pgd = e59542a6
> [ 6961.192917] [000008f8] *pgd=00000000
> [ 6961.196586] Internal error: Oops: 5 [#1] ARM
> [ 6961.200877] Modules linked in: crypto_user mxs_dcp cfg80211 rfkill des_generic libdes arc4 libarc4 cbc ecb algif_skcipher sha256_generic libsha256 sha1_generic hmac aes_generic libaes cmac sha512_generic md5 md4 algif_hash af_alg i2c_imx ci_hdrc_imx ci_hdrc i2c_core ulpi roles udc_core imx_sdma usb_common firmware_class usbmisc_imx virt_dma phy_mxs_usb nf_tables nfnetlink ip_tables x_tables ipv6 autofs4 [last unloaded: mxs_dcp]
> [ 6961.239228] CPU: 0 PID: 469 Comm: mxs_dcp_chan/ae Not tainted 5.10.46-315-tiago #315
> [ 6961.246988] Hardware name: Freescale i.MX6 Ultralite (Device Tree)
> [ 6961.253201] PC is at memcpy+0xc0/0x330
> [ 6961.256993] LR is at dcp_chan_thread_aes+0x220/0x94c [mxs_dcp]
> [ 6961.262847] pc : [<c053f1e0>]    lr : [<bf13cda4>]    psr: 800e0013
> [ 6961.269130] sp : cdc09ef4  ip : 00000010  fp : c36e5808
> [ 6961.274370] r10: cdcc3150  r9 : 00000000  r8 : bff46000
> [ 6961.279613] r7 : c36e59d0  r6 : c2e42840  r5 : cdcc3140  r4 : 00000001
> [ 6961.286156] r3 : 000008f9  r2 : 80000000  r1 : 000008f8  r0 : cdc1004f
> [ 6961.292704] Flags: Nzcv  IRQs on  FIQs on  Mode SVC_32  ISA ARM  Segment none
> [ 6961.299860] Control: 10c53c7d  Table: 83358059  DAC: 00000051
> [ 6961.305628] Process mxs_dcp_chan/ae (pid: 469, stack limit = 0xe1efdc80)
> [ 6961.312346] Stack: (0xcdc09ef4 to 0xcdc0a000)
> [ 6961.316726] 9ee0:                                              cdc1004f 00000001 bf13cda4
> [ 6961.324930] 9f00: 00000000 00000000 c23b41a0 00000000 c36e59d0 00000001 00000010 00000000
> [ 6961.333132] 9f20: 00000000 00000000 c13de2fc 000008f9 8dc13080 00000010 cdcc3150 c21e5010
> [ 6961.341335] 9f40: cdc08000 cdc10040 00000001 bf13fa40 cdc11040 c2e42880 00000002 cc861440
> [ 6961.349535] 9f60: ffffe000 c33dbe00 c332cb40 cdc08000 00000000 bf13cb84 00000000 c3353c54
> [ 6961.357736] 9f80: c33dbe44 c0140d34 cdc08000 c332cb40 c0140c00 00000000 00000000 00000000
> [ 6961.365936] 9fa0: 00000000 00000000 00000000 c0100114 00000000 00000000 00000000 00000000
> [ 6961.374138] 9fc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
> [ 6961.382338] 9fe0: 00000000 00000000 00000000 00000000 00000013 00000000 00000000 00000000
> [ 6961.390567] [<c053f1e0>] (memcpy) from [<bf13cda4>] (dcp_chan_thread_aes+0x220/0x94c [mxs_dcp])
> [ 6961.399312] [<bf13cda4>] (dcp_chan_thread_aes [mxs_dcp]) from [<c0140d34>] (kthread+0x134/0x160)
> [ 6961.408137] [<c0140d34>] (kthread) from [<c0100114>] (ret_from_fork+0x14/0x20)
> [ 6961.415377] Exception stack(0xcdc09fb0 to 0xcdc09ff8)
> [ 6961.420448] 9fa0:                                     00000000 00000000 00000000 00000000
> [ 6961.428647] 9fc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
> [ 6961.436845] 9fe0: 00000000 00000000 00000000 00000000 00000013 00000000
> [ 6961.443488] Code: e4808004 e480e004 e8bd01e0 e1b02f82 (14d13001)
> 
> where dcp_chan_thread_aes+0x220 is the line
> 
> 	memcpy(in_buf + actx->fill, src_buf, clen);
> 
> in mxs_dcp_aes_block_crypt. I also tried with the following patch
> instead of the one included in this series:
> 
> ---
> diff --git a/drivers/crypto/mxs-dcp.c b/drivers/crypto/mxs-dcp.c
> index f397cc5bf102..54fd24ba1261 100644
> --- a/drivers/crypto/mxs-dcp.c
> +++ b/drivers/crypto/mxs-dcp.c
> @@ -367,6 +367,7 @@ static int mxs_dcp_aes_block_crypt(struct crypto_async_request *arq)
>                                 last_out_len = actx->fill;
>                                 while (dst && actx->fill) {
>                                         if (!split) {
> +                                               kmap(sg_page(dst));
>                                                 dst_buf = sg_virt(dst);
>                                                 dst_off = 0;
>                                         }
> @@ -379,6 +380,7 @@ static int mxs_dcp_aes_block_crypt(struct crypto_async_request *arq)
>                                         actx->fill -= rem;
> 
>                                         if (dst_off == sg_dma_len(dst)) {
> +                                               kunmap(sg_page(dst));
>                                                 dst = sg_next(dst);
>                                                 split = 0;
>                                         } else {
> --
> 
> but got the same oops. Unfortunately, I don't have the time to
> investigate this oops as well. I'd appreciate if anyone else using this
> device could look into this and see if they encounter the same errors.
> 
> [1] https://github.com/smuellerDD/libkcapi/blob/master/test/kcapi-dgst-test.sh
> 
> Changes in v2:
> - Fix warning when taking the minimum of a u32 and a size_t
> - Use sg_pcopy_from_buffer to properly deal with partial reads.
> 
> Sean Anderson (2):
>   crypto: mxs-dcp: Check for DMA mapping errors
>   crypto: mxs_dcp: Use sg_mapping_iter to copy data
> 
>  drivers/crypto/mxs-dcp.c | 81 ++++++++++++++++++++++------------------
>  1 file changed, 44 insertions(+), 37 deletions(-)

All applied.  Thanks.
-- 
Email: Herbert Xu <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt