Hi,
在 2022/12/21 0:01, Tejun Heo 写道:
Hello,
On Tue, Dec 20, 2022 at 05:19:12PM +0800, Yu Kuai wrote:
Yes, that sounds good. BTW, queue_lock is also used to protect
pd_alloc_fn/pd_init_fn，and we found that blkcg_activate_policy() is
problematic:
blkcg_activate_policy
spin_lock_irq(&q->queue_lock);
list_for_each_entry_reverse(blkg, &q->blkg_list
pd_alloc_fn(GFP_NOWAIT | __GFP_NOWARN,...) -> failed
spin_unlock_irq(&q->queue_lock);
// release queue_lock here is problematic, this will cause
pd_offline_fn called without pd_init_fn.
pd_alloc_fn(__GFP_NOWARN,...)
So, if a blkg is destroyed while a policy is being activated, right?
Yes, remove cgroup can race with this, for bfq null pointer deference
will be triggered in bfq_pd_offline().
If we are using a mutex to protect rq_qos ops, it seems the right thing
to do do also using the mutex to protect blkcg_policy ops, and this
problem can be fixed because mutex can be held to alloc memroy with
GFP_KERNEL. What do you think?
One worry is that switching to mutex can be more headache due to destroy
path synchronization. Another approach would be using a per-blkg flag to
track whether a blkg has been initialized.
I think perhaps you mean per blkg_policy_data flag? per blkg flag should
not work in this case.
Thanks,
Kuai
Thanks.
