Kernel panic may result when dd commands are run with PowerPath for Linux version 6.2 on Red Hat Enterprise Linux 7.4

Affected products:
EMC Software: PowerPath for Linux 6.2
Operating System: Red Hat Enterprise Linux (RHEL) 7.4

Running dd commands may cause the host kernel to panic, as the following display shows:

[ 107.397908] Kernel panic - not syncing: stack-protector: Kernel stack is corrupted in: ffffffffc06a566b

[ 107.398011] CPU: 1 PID: 0 Comm: swapper/1 Tainted: P OE ------------ 3.10.0-693.el7.x86_64 #1
[ 107.398085] Hardware name: Dell Inc. PowerEdge R730xd/072T6D, BIOS 2.4.3 01/17/2017
[ 107.398144] ffff885fbaa7e800 ffffffffc06a566b ffff885fbddfff90 ffffffff816a3d91
[ 107.398216] ffff885fbde00010 ffffffff8169dc54 ffff885f00000010 ffff885fbde00020
[ 107.398285] ffff885fbddfffc0 ffffffffc06a566b ffff885fbde001a8 ffffffffc06a566b
[ 107.398355] Call Trace:

CPUS: 24
DATE: Tue Aug 15 13:27:10 2017
UPTIME: 00:01:47
LOAD AVERAGE: 0.44, 0.22, 0.08 ===> Load Average Normal
TASKS: 461
NODENAME: host.domain.com
RELEASE: 3.10.0-693.el7.x86_64
VERSION: #1 SMP Thu Jul 6 HH:MM:SS EDT 2017
MACHINE: x86_64 (3400 Mhz)
MEMORY: 383.9 GB
PANIC: "Kernel panic - not syncing: stack-protector: Kernel stack is corrupted in: ffffffffc06a566b" ==> Got panicked here
PID: 0
COMMAND: "swapper/1"
TASK: ffff8830b6c10000 (1 of 24) [THREAD_INFO: ffff8830b76f4000]
CPU: 1
STATE: TASK_RUNNING (PANIC)

crash> ps -S
RU: 24
IN: 436
UN: 1 ===> One task in Hung state

crash> ps -m|grep UN
[0 00:00:00.243] [UN] PID: 5418 TASK: ffff885739381fa0 CPU: 1 COMMAND: "dd" ===> 'dd' task in hung state

PID: 5418 TASK: ffff885739381fa0 CPU: 1 COMMAND: "dd"
#0 [ffff884a58ef38f0] __schedule at ffffffff816a8f45
#1 [ffff884a58ef3958] schedule at ffffffff816a94c9
#2 [ffff884a58ef3968] schedule_timeout at ffffffff816a6fd9
#3 [ffff884a58ef3a10] io_schedule_timeout at ffffffff816a8b4d ====> ioschedule got timed out
#4 [ffff884a58ef3a40] io_schedule at ffffffff816a8be8
#5 [ffff884a58ef3a50] get_request at ffffffff812f7e28
#6 [ffff884a58ef3b00] blk_queue_bio at ffffffff812fa846
#7 [ffff884a58ef3b48] generic_make_request at ffffffff812f8b45
#8 [ffff884a58ef3ba0] submit_bio at ffffffff812f8dc0
#9 [ffff884a58ef3bf8] xfs_add_to_ioend at ffffffffc058d4f5 [xfs]
#10 [ffff884a58ef3c30] xfs_do_writepage at ffffffffc058db65 [xfs]
#11 [ffff884a58ef3cb8] write_cache_pages at ffffffff8118dc91
#12 [ffff884a58ef3dd0] xfs_vm_writepages at ffffffffc058d78d [xfs]
#13 [ffff884a58ef3e48] do_writepages at ffffffff8118effe
#14 [ffff884a58ef3e58] __filemap_fdatawrite_range at ffffffff81183c55
#15 [ffff884a58ef3ea8] filemap_write_and_wait_range at ffffffff81183da1
#16 [ffff884a58ef3ed8] xfs_file_fsync at ffffffffc0598b16 [xfs]
#17 [ffff884a58ef3f30] do_fsync at ffffffff812334b5
#18 [ffff884a58ef3f70] sys_fsync at ffffffff81233790
#19 [ffff884a58ef3f80] system_call_fastpath at ffffffff816b4fc9
RIP: 00007fb9c46c9930 RSP: 00007ffe77f87e00 RFLAGS: 00010246
RAX: 000000000000004a RBX: ffffffff816b4fc9 RCX: ffffffffffffffff
RDX: 000000007ffff000 RSI: 0000000000000000 RDI: 0000000000000001
RBP: 0000000000000001 R8: ffffffffffffffff R9: 0000000080002003
R10: 00007ffe77f87b50 R11: 0000000000000246 R12: ffffffff81233790
R13: ffff884a58ef3f78 R14: 0000000000000000 R15: 0000000000000001
ORIG_RAX: 000000000000004a CS: 0033 SS: 002b

In RHEL 7.4, XFS uses chained Basic Input/Output System(BIOS) to dispatch Input/Output(I/O) when a block size of 2 GB or greater is used.

Nested calls to iodone functions in PowerPath and the bio_chain_endio() kernel function to process Parent/Child chained BIOS may result in kernel stack overrun, which may then lead to stack corruption due to the way Child and Parent BIOS are chained by XFS.

PowerPath for Linux 7.4 is not supported.

Workaround:

1. Use RHEL 7.3 and earlier Red Hat versions.
2. If using RHEL 7.4, use a file system other than XFS.

PowerPath Engineering is investigating. This article will be updated with additional information as appropriate.