On Mon, Nov 14, 2022, Oliver Upton wrote:
> Hi Marek,
>
> On Mon, Nov 14, 2022 at 03:29:14PM +0100, Marek Szyprowski wrote:
> > This patch landed in today's linux-next (20221114) as commit
> > c3119ae45dfb ("KVM: arm64: Protect stage-2 traversal with RCU").
> > Unfortunately it introduces a following warning:
>
> Thanks for the bug report :) I had failed to test nVHE in the past few
> revisions of this series.
>
> > --->8---
> >
> > kvm [1]: IPA Size Limit: 40 bits
> > BUG: sleeping function called from invalid context at
> > include/linux/sched/mm.h:274
> > in_atomic(): 0, irqs_disabled(): 0, non_block: 0, pid: 1, name: swapper/0
> > preempt_count: 0, expected: 0
> > RCU nest depth: 1, expected: 0
> > 2 locks held by swapper/0/1:
> > #0: ffff80000a8a44d0 (kvm_hyp_pgd_mutex){+.+.}-{3:3}, at:
> > __create_hyp_mappings+0x80/0xc4
> > #1: ffff80000a927720 (rcu_read_lock){....}-{1:2}, at:
> > kvm_pgtable_walk+0x0/0x1f4
> > CPU: 2 PID: 1 Comm: swapper/0 Not tainted 6.1.0-rc3+ #5918
> > Hardware name: Raspberry Pi 3 Model B (DT)
> > Call trace:
> > dump_backtrace.part.0+0xe4/0xf0
> > show_stack+0x18/0x40
> > dump_stack_lvl+0x8c/0xb8
> > dump_stack+0x18/0x34
> > __might_resched+0x178/0x220
> > __might_sleep+0x48/0xa0
> > prepare_alloc_pages+0x178/0x1a0
> > __alloc_pages+0x9c/0x109c
> > alloc_page_interleave+0x1c/0xc4
> > alloc_pages+0xec/0x160
> > get_zeroed_page+0x1c/0x44
> > kvm_hyp_zalloc_page+0x14/0x20
> > hyp_map_walker+0xd4/0x134
> > kvm_pgtable_visitor_cb.isra.0+0x38/0x5c
> > __kvm_pgtable_walk+0x1a4/0x220
> > kvm_pgtable_walk+0x104/0x1f4
> > kvm_pgtable_hyp_map+0x80/0xc4
> > __create_hyp_mappings+0x9c/0xc4
> > kvm_mmu_init+0x144/0x1cc
> > kvm_arch_init+0xe4/0xef4
> > kvm_init+0x3c/0x3d0
> > arm_init+0x20/0x30
> > do_one_initcall+0x74/0x400
> > kernel_init_freeable+0x2e0/0x350
> > kernel_init+0x24/0x130
> > ret_from_fork+0x10/0x20
> > kvm [1]: Hyp mode initialized successfully
> >
> > --->8----
> >
> > I looks that more changes in the KVM code are needed to use RCU for that
> > code.
>
> Right, the specific issue is that while the stage-2 walkers preallocate
> any table memory they may need, the hyp walkers do not and allocate
> inline.
>
> As hyp stage-1 is protected by a spinlock there is no actual need for
> RCU in that case. I'll post something later on today that addresses the
> issue.
>
For each stage-2 page table walk, KVM will use
kvm_mmu_topup_memory_cache() before taking the mmu lock. This ensures
whoever holding the mmu lock won't sleep. hyp walkers seems to
miss this notion completely, whic makes me puzzeled. Using a spinlock
only ensures functionality but seems quite inefficient if the one who
holds the spinlock try to allocate pages and sleep...
But that seems to be a separate problem for nvhe. Why do we need an RCU
lock here. Oh is it for batching?
