Whenever I try to setup LUKS on my CentOS install on a Supermicro A2SDi-8C-HLN4F motherboard (Intel ATOM CPU C3758), everything seems to work fine until the process completely freezes when creating the filesystem (ext4) or trying to mount the filesystem (xfs). It doesn’t matter if I try to put it on using the —encrypted directive in my kickstart config, or try to do it on a physical drive (/dev/sde), RAID device (/dev/md127), physical volume or logical volume in LVM, or even a loopback-mounted file (/dev/loop0). The same problem shows up both in CentOS 7.7 and 8.1.
I read somewhere that memory might be an issue, so I switched my two DIMMs and ran MemTest86+ with 0 errors.
Steps used when trying to create the filesystem on the devices above:
cryptsetup --force-password luksFormat <device>cryptsetup luksOpen <device> <testname>mkfs -t ext4 /dev/mapper/<testname>
During the mkfs command, the process invariably hangs, and the process (or rather the mkfs.ext4 child process) can’t be killed, even with kill -9.
After a little while, the following shows up in my dmesg:
[ 492.528687] Not tainted 4.18.0-147.5.1.el8_1.x86_64 #1
[ 492.528719] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 492.528764] kworker/u17:0 D 0 4238 2 0x80000080
[ 492.528777] Workqueue: kcryptd/253:3 kcryptd_crypt [dm_crypt]
[ 492.528778] Call Trace:
[ 492.528787] ? __schedule+0x253/0x830
[ 492.528791] ? mempool_alloc+0x67/0x190
[ 492.528793] schedule+0x28/0x70
[ 492.528795] schedule_timeout+0x26d/0x390
[ 492.528803] ? qat_alg_sgl_to_bufl.isra.11+0x456/0x770 [intel_qat]
[ 492.528807] ? dma_direct_unmap_page+0x7a/0x80
[ 492.528809] wait_for_completion+0x11f/0x190
[ 492.528811] ? wake_up_q+0x70/0x70
[ 492.528814] crypt_convert+0xa13/0xf00 [dm_crypt]
[ 492.528818] ? bio_alloc_bioset+0xdc/0x210
[ 492.528820] ? __switch_to_asm+0x41/0x70
[ 492.528822] ? __switch_to_asm+0x35/0x70
[ 492.528825] kcryptd_crypt+0x2f3/0x3b0 [dm_crypt]
[ 492.528828] process_one_work+0x1a7/0x3b0
[ 492.528831] worker_thread+0x30/0x390
[ 492.528833] ? create_worker+0x1a0/0x1a0
[ 492.528835] kthread+0x112/0x130
[ 492.528837] ? kthread_flush_work_fn+0x10/0x10
[ 492.528839] ret_from_fork+0x35/0x40
Starting the command using strace mkfs ... always stops at exactly the same character in the output:
pwrite64(3, ”\3...”) = 4096
pwrite64(3, ”\3...”) = 4096
fsync(3
I don’t know the relevance of the missing closing parenthesis on the last line, but it always stops at this exact place.
How would I go about identifying more exactly what is going on and where the problem might lie?
