Re: [PATCH v4 3/3] md/raid10: bound reused r10bio devs[] walks by used_nr_devs

在 2026/6/22 15:53, yu kuai 写道:
> Hi，
> 
> 在 2026/6/22 14:34, Chen Cheng 写道:
>> The true dangerous is in-flight r10bio which use old geometry.
> 
> You didn't explain why there can be inflight r10bio when array is suspended,
> if this can happen, this is the root cause need to be fixed.
> 

one scenario is .

CPU A (softirq, raid_end_bio_io)         CPU B (action_store)
================================        ===============================
bio_endio(master_bio)
md_end_clone_io
         percpu_ref_put → 0
                                         wait_event wakeup,
						mddev_suspend return
                                         raid10_start_reshape:
                                           setup_geo(&conf->geo, new)
                                           ...
                                           mempool_destroy(old_pool)
                                           conf->r10bio_pool = new_pool
allow_barrier(conf)
free_r10bio(r10_bio)
put_all_bios:
for (i=0; i<conf->geo.raid_disks; i++)
         ==> old obj, new geo, OOB
mempool_free(r10_bio, conf->r10bio_pool)
         ==> old geo obj free into new pool




The fix,  as your advise , raid_end_bio_io:
1. put_all_bios, mempool_free
2. bio_endio if returned
3. allow_barrier



would be replace with patch 03.