DM-Crypt usage on i.MX Platforms without CAAM hardware IP

[wooosaiiii]

Hi,

I am following a "DM-Crypt usage on i.MX Platforms without CAAM hardware IP" guide from NXP on i.MX93 board.

I am using the following steps:

modprobe dm-crypt
modprobe tee_crypto
modprobe trusted

export DEV=/dev/loop0
dd if=/dev/zero of=/data/encrypted.img bs=1M count=512
losetup -P $DEV /data/encrypted.img

export KEYNAME=dm_trustedkey
export KEY="$(keyctl add trusted $KEYNAME 'new 32' @s)"
keyctl pipe $KEY >/data/$KEYNAME.blob
keyctl list @s

export ALGO="capi:cbc-aes-tee-plain"
export BLOCKS=$(blockdev --getsz /dev/loop0)
export SECTOR_SIZE=4096
export TABLE="0 $BLOCKS crypt $ALGO :32:trusted:$KEYNAME 0 $DEV 0 1 sector_size:$SECTOR_SIZE"

dmsetup -v create encrypted --table "$TABLE"

However, my kernel crashes with the following OOPS:

[  713.174934] Unable to handle kernel paging request at virtual address ffff8000a1fca858
[  713.182908] Mem abort info:
[  713.185716]   ESR = 0x0000000096000006
[  713.189477]   EC = 0x25: DABT (current EL), IL = 32 bits
[  713.194786]   SET = 0, FnV = 0
[  713.197848]   EA = 0, S1PTW = 0
[  713.200993]   FSC = 0x06: level 2 translation fault
[  713.205873] Data abort info:
[  713.208762]   ISV = 0, ISS = 0x00000006, ISS2 = 0x00000000
[  713.214246]   CM = 0, WnR = 0, TnD = 0, TagAccess = 0
[  713.219296]   GCS = 0, Overlay = 0, DirtyBit = 0, Xs = 0
[  713.224611] swapper pgtable: 4k pages, 48-bit VAs, pgdp=0000000089bcd000
[  713.231309] [ffff8000a1fca858] pgd=10000000fffff003, p4d=10000000fffff003, pud=10000000ffffe003, pmd=0000000000000000
[  713.241943] Internal error: Oops: 0000000096000006 [#1] PREEMPT SMP
[  713.248199] Modules linked in: tee_crypto dm_crypt crct10dif_ce polyval_ce polyval_generic layerscape_edac_mod rtc_rv8803 at24 btnxpuart flexcan can_dev cfg80211 fuse overlay trusted
[  713.264459] CPU: 0 PID: 532 Comm: dmsetup Not tainted 6.6.52-lts-next-07235-gfdd32c7240b4 #1
[  713.272880] Hardware name: EVVA i.MX93 Gateway (DT)
[  713.277743] pstate: a0400009 (NzCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[  713.284694] pc : osq_lock+0x5c/0x134
[  713.288270] lr : __mutex_lock.constprop.0+0x1f0/0x540
[  713.293317] sp : ffff800083543750
[  713.296616] x29: ffff800083543750 x28: ffff000001e3417a x27: 0000000000000001
[  713.303743] x26: ffff000000a43e00 x25: ffff000005b7dbf0 x24: 00000000ffffffff
[  713.310864] x23: 0000000000000002 x22: fffffc000002e4c0 x21: fffffc000002e884
[  713.317988] x20: ffff800083543768 x19: fffffc000002e878 x18: 0000000000000001
[  713.325112] x17: 0000000000000000 x16: 0000000000000000 x15: 0000000000000000
[  713.332239] x14: 0000000000000000 x13: 01485ce3a37a7a58 x12: ed352f5eedb722c0
[  713.339360] x11: 0101010101010101 x10: fffffffffa67a3b3 x9 : 0000000000000000
[  713.346484] x8 : ffff800083543920 x7 : 0000000000000000 x6 : 000000000000003f
[  713.353608] x5 : 0000000000000040 x4 : 0000000003fffbff x3 : ffff800081fcc860
[  713.360732] x2 : ffff800081cf0d00 x1 : ffff00007fb98d00 x0 : fffffc000002e884
[  713.367859] Call trace:
[  713.370294]  osq_lock+0x5c/0x134
[  713.373518]  __mutex_lock.constprop.0+0x1f0/0x540
[  713.378215]  __mutex_lock_slowpath+0x14/0x20
[  713.382479]  mutex_lock+0x48/0x54
[  713.385787]  tee_device_get+0x20/0x6c
[  713.389444]  register_shm_helper+0x3c/0x2e8
[  713.393624]  tee_shm_register_kernel_buf+0x18/0x24
[  713.398408]  skcipher_setkey+0xe8/0x21c [tee_crypto]
[  713.403374]  cbc_skcipher_setkey+0x38/0x7c [tee_crypto]
[  713.408589]  crypto_skcipher_setkey+0x6c/0x124
[  713.413029]  crypt_setkey+0x104/0x22c [dm_crypt]
[  713.417658]  crypt_set_key+0x248/0x360 [dm_crypt]
[  713.422354]  crypt_ctr+0x634/0xfb8 [dm_crypt]
[  713.426705]  dm_table_add_target+0x218/0x380
[  713.430969]  table_load+0x140/0x3f0
[  713.434453]  ctl_ioctl+0x378/0x648
[  713.437851]  dm_ctl_ioctl+0x10/0x20
[  713.441334]  __arm64_sys_ioctl+0xac/0xf0
[  713.445252]  invoke_syscall+0x48/0x114
[  713.448996]  el0_svc_common.constprop.0+0xc0/0xe0
[  713.453693]  do_el0_svc+0x1c/0x28
[  713.457001]  el0_svc+0x40/0xe4
[  713.460055]  el0t_64_sync_handler+0x120/0x12c
[  713.464402]  el0t_64_sync+0x190/0x194
[  713.468066] Code: 340005c4 51000484 d000f723 91218063 (f864d863) 
[  713.474147] ---[ end trace 0000000000000000 ]---
[  713.478800] note: dmsetup[532] exited with preempt_count 1

I am using 6.6.52_2.2.2 scarthgap release.

I think the problematic driver is drivers/tee/crypto/tee_skcipher.c.

Here is the output when loading the kernel module:

[  665.731673] tee_client_open_session failed, err: ffff0008
[  665.737206] tee_crypt algorithms registered in /proc/crypto
[  665.742835] driver 1.0 loaded.

The prerequisites from the document have been fulfilled in our BSP:

Prerequisites:
Ensure that a region of OCRAM is reserved to be accessed by Secure World only. This region is used to save cryptographic keys. Current OCRAM reserved regions:

For i.MX 93: 0x20518000 - 0x2051C000
For i.MX 95: 0x204BC000 - 0x204C0000
For i.MX 91: 0x204A0000 - 0x204A4000
For i.MX 943: 0x204BC000 - 0x204C0000

Make sure the following configurations are enabled in the kernel:
CONFIG_TEE_CRYPTO = m
CONFIG_DM_CRYPT = m
CONFIG_TRUSTED_KEYS = m
CONFIG_TRUSTED_KEYS_CAAM = n
CONFIG_TRUSTED_KEYS_TEE = y

In OP-TEE, check whether the following flags in the core/arch/arm/plat-imx/conf.mk platform specific section for which DM-crypt are enabled or not:
CFG_IMX_TRUSTED_ARM_CE = y
CFG_IN_TREE_EARLY_TAS += trusted_keys/f04a0fe7-1f5d-4b9b-abf7-619b85b4ce8c

Any ideas what might be an issue here?

[wooosaiiii]

Fixed it. I forgot to update optee blob when rebuilding with CFG_IMX_TRUSTED_ARM_CE = y.

Can be closed.