Hi Christoph, Jens, and Ming,
While booting a 5.2 SEV-enabled guest we have encountered the following
WARNING that is followed up by a BUG because we are in atomic context
while trying to call set_memory_decrypted:
WARNING: suspicious RCU usage
5.2.0 #1 Not tainted
-----------------------------
include/linux/rcupdate.h:266 Illegal context switch in RCU read-side critical
section!
other info that might help us debug this:
rcu_scheduler_active = 2, debug_locks = 1
3 locks held by kworker/0:1H/97:
#0: 0000000016e1b654 ((wq_completion)kblockd){+.+.}, at:
process_one_work+0x1b5/0x5e0
#1: 00000000002674ff ((work_completion)(&(&hctx->run_work)->work)){+.+.}, at:
process_one_work+0x1b5/0x5e0
#2: 00000000addb6aba (rcu_read_lock){....}, at: hctx_lock+0x17/0xe0
stack backtrace:
CPU: 0 PID: 97 Comm: kworker/0:1H Not tainted 5.2.0 #1
Workqueue: kblockd blk_mq_run_work_fn
Call Trace:
dump_stack+0x67/0x90
___might_sleep+0xfb/0x180
_vm_unmap_aliases+0x3e/0x1f0
__set_memory_enc_dec+0x7b/0x120
dma_direct_alloc_pages+0xcc/0x100
dma_pool_alloc+0xcf/0x1e0
nvme_queue_rq+0x5fb/0x9f0
blk_mq_dispatch_rq_list+0x350/0x5a0
blk_mq_do_dispatch_sched+0x76/0x110
blk_mq_sched_dispatch_requests+0x119/0x170
__blk_mq_run_hw_queue+0x6c/0xf0
process_one_work+0x23b/0x5e0
worker_thread+0x3d/0x390
kthread+0x121/0x140
ret_from_fork+0x27/0x50
hctx_lock() in __blk_mq_run_hw_queue() takes rcu_read_lock or
srcu_read_lock depending on BLK_MQ_F_BLOCKING.
dma_direct_alloc_pages() can then call set_memory_decrypted() which must
be allowed to block.
Any ideas on how to address this?
