On Tue, Oct 8, 2019 at 3:26 AM Dongsheng Yang
<dongsheng.yang@xxxxxxxxxxxx> wrote:
>
>
>
> On 10/07/2019 04:05 AM, Ilya Dryomov wrote:
> > On Fri, Sep 27, 2019 at 5:35 PM Dongsheng Yang
> > <dongsheng.yang@xxxxxxxxxxxx> wrote:
> >> There is a warning message[1] in my test with below steps:
> >> # rbd bench --io-type write --io-size 4K --io-threads 1 --io-pattern rand test &
> >> # sleep 5
> >> # pkill -9 rbd
> >> # rbd map test &
> >> # sleep 5
> >> # pkill rbd
> >>
> >> The reason is that the rbd_add_acquire_lock() is interruptable,
> >> that means, when we kill the waiting on ->acquire_wait, the lock_dwork
> >> could be still running.
> >>
> >> 1. do_rbd_add() 2. lock_dwork
> >> rbd_add_acquire_lock()
> >> - queue_delayed_work()
> >> lock_dwork queued
> >> - wait_for_completion_killable_timeout() <-- kill happen
> >> rbd_dev_image_unlock() <-- UNLOCKED now, nothing to do.
> >> rbd_dev_device_release()
> >> rbd_dev_image_release()
> >> - ...
> >> lock successed here
> >> - cancel_delayed_work_sync(&rbd_dev->lock_dwork)
> >>
> >> Then when we reach the rbd_dev_free(), lock_state is not RBD_LOCK_STATE_UNLOCKED.
> >>
> >> To fix it, this commit make sure the lock_dwork was finished before calling
> >> rbd_dev_image_unlock().
> >>
> >> On the other hand, this would not happend in do_rbd_remove(), because
> >> after rbd mapped, lock_dwork will only be queued for IO request, and
> >> request will continue unless lock_dwork finished. when we call
> >> rbd_dev_image_unlock() in do_rbd_remove(), all requests are done.
> >> that means, lock_state should not be locked again after rbd_dev_image_unlock().
> >>
> >> [1]:
> >> [ 116.043632] rbd: rbd0: no lock owners detected
> >> [ 117.745975] rbd: rbd0: failed to acquire exclusive lock: -512
> >> [ 121.055286] rbd: image test: no lock owners detected
> >> [ 126.066977] rbd: image test: no lock owners detected
> >> [ 131.103870] rbd: image test: no lock owners detected
> >> [ 132.360374] rbd: image test: no lock owners detected
> >> [ 132.369202] rbd: image test: breaking header lock owned by client24204
> >> [ 132.903969] rbd: image test: failed to unwatch: -512
> >> [ 137.905601] ------------[ cut here ]------------
> >> [ 137.906922] WARNING: CPU: 18 PID: 4545 at drivers/block/rbd.c:5576 rbd_dev_free+0xa5/0xc0 [rbd]
> >> [ 137.908957] Modules linked in: nbd(OE) rbd(OE) libceph(OE) dns_resolver tcp_diag udp_diag inet_diag unix_diag af_packet_diag netlink_diag sg cfg80211 rfkill snd_hda_codec_generic ledtrig_audio kvm_intel kvm irqbypass crct10dif_pclmul cirrus crc32_pclmul drm_kms_helper snd_hda_intel ghash_clmulni_intel ext4 syscopyarea snd_hda_codec sysfillrect nfsd mbcache sysimgblt snd_hda_core fb_sys_fops jbd2 drm snd_hwdep aesni_intel snd_seq crypto_simd cryptd glue_helper snd_seq_device auth_rpcgss snd_pcm nfs_acl snd_timer lockd i2c_piix4 snd i2c_core pcspkr virtio_balloon soundcore grace sunrpc ip_tables xfs libcrc32c virtio_console virtio_blk(O) 8139too ata_generic pata_acpi serio_raw ata_piix libata virtio_pci crc32c_intel 8139cp virtio_ring mii floppy(O) virtio dm_mirror dm_region_hash dm_log dm_mod dax [last unloaded: libceph]
> >> [ 137.926270] CPU: 18 PID: 4545 Comm: rbd Tainted: G OE 5.3.0-rc1+ #15
> >> [ 137.928091] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.11.2-0-gf9626ccb91-prebuilt.qemu-project.org 04/01/2014
> >> [ 137.931081] RIP: 0010:rbd_dev_free+0xa5/0xc0 [rbd]
> >> [ 137.932240] Code: 00 00 75 05 e8 1c ad ff ff 48 8b bb e0 00 00 00 e8 c0 99 ba e5 48 89 df 5b e9 b7 99 ba e5 48 c7 c7 98 90 6d c0 e8 69 20 a4 e5 <0f> 0b e9 75 ff ff ff 48 c7 c7 98 90 6d c0 e8 56 20 a4 e5 0f 0b e9
> >> [ 137.936783] RSP: 0018:ffffb0f9c2f53dc8 EFLAGS: 00010282
> >> [ 137.938161] RAX: 0000000000000024 RBX: ffff97f68636f000 RCX: 0000000000000000
> >> [ 137.939922] RDX: 0000000000000000 RSI: ffff97f69fa97b18 RDI: ffff97f69fa97b18
> >> [ 137.941673] RBP: ffff97f68636f000 R08: 0000000000000000 R09: 0000000000000000
> >> [ 137.943469] R10: 0000000000000000 R11: 0000000000000000 R12: ffff97f68636f7c0
> >> [ 137.944970] R13: ffff97f693dad270 R14: 0000000000000000 R15: ffff97f68fc76800
> >> [ 137.946292] FS: 00007f3de0e7bb00(0000) GS:ffff97f69fa80000(0000) knlGS:0000000000000000
> >> [ 137.948207] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> >> [ 137.949720] CR2: 00007f1281623000 CR3: 00000007f0ec2005 CR4: 00000000000206e0
> >> [ 137.951469] Call Trace:
> >> [ 137.952417] rbd_dev_release+0x41/0x60 [rbd]
> >> [ 137.953651] device_release+0x27/0x80
> >> [ 137.954792] kobject_release+0x68/0x190
> >> [ 137.955954] do_rbd_add.isra.67+0x970/0xf60 [rbd]
> >> [ 137.957253] kernfs_fop_write+0x109/0x190
> >> [ 137.958484] vfs_write+0xbe/0x1d0
> >> [ 137.959566] ksys_write+0xa4/0xe0
> >> [ 137.960643] do_syscall_64+0x60/0x270
> >> [ 137.961749] entry_SYSCALL_64_after_hwframe+0x49/0xbe
> >> [ 137.963119] RIP: 0033:0x7f3dd455469d
> >> [ 137.964245] Code: cd 20 00 00 75 10 b8 01 00 00 00 0f 05 48 3d 01 f0 ff ff 73 31 c3 48 83 ec 08 e8 4e fd ff ff 48 89 04 24 b8 01 00 00 00 0f 05 <48> 8b 3c 24 48 89 c2 e8 97 fd ff ff 48 89 d0 48 83 c4 08 48 3d 01
> >> [ 137.968699] RSP: 002b:00007ffd91490010 EFLAGS: 00000293 ORIG_RAX: 0000000000000001
> >> [ 137.970559] RAX: ffffffffffffffda RBX: 0000000000000086 RCX: 00007f3dd455469d
> >> [ 137.972254] RDX: 0000000000000086 RSI: 000055790f5f5338 RDI: 0000000000000009
> >> [ 137.974030] RBP: 00007ffd91490050 R08: 0000000000000004 R09: 000000000000004d
> >> [ 137.975806] R10: 00007ffd9148fbe0 R11: 0000000000000293 R12: 000055790f5c8e58
> >> [ 137.977528] R13: 000055790f3479b8 R14: 0000000000000001 R15: 0000000000000001
> >> [ 137.979250] irq event stamp: 61780
> >> [ 137.980421] hardirqs last enabled at (61779): [<ffffffffa610f403>] console_unlock+0x503/0x5f0
> >> [ 137.982397] hardirqs last disabled at (61780): [<ffffffffa600379a>] trace_hardirqs_off_thunk+0x1a/0x20
> >> [ 137.984495] softirqs last enabled at (61776): [<ffffffffa6c00343>] __do_softirq+0x343/0x447
> >> [ 137.986504] softirqs last disabled at (61769): [<ffffffffa609a1b7>] irq_exit+0xe7/0xf0
> >> [ 137.988417] ---[ end trace 10917ffd2e2f48ca ]---
> >>
> >> Signed-off-by: Dongsheng Yang <dongsheng.yang@xxxxxxxxxxxx>
> >> ---
> >> drivers/block/rbd.c | 1 +
> >> 1 file changed, 1 insertion(+)
> >>
> >> diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c
> >> index 5bb98f5..44dd7a0 100644
> >> --- a/drivers/block/rbd.c
> >> +++ b/drivers/block/rbd.c
> >> @@ -7807,6 +7807,7 @@ static ssize_t do_rbd_add(struct bus_type *bus,
> >> return rc;
> >>
> >> err_out_image_lock:
> >> + cancel_delayed_work_sync(&rbd_dev->lock_dwork);
> >> rbd_dev_image_unlock(rbd_dev);
> >> rbd_dev_device_release(rbd_dev);
> >> err_out_image_probe:
> > Hi Dongsheng,
> >
> > I didn't pay much attention to the interrupt/timeout case because it
> > can't work properly with blocking lock, unlock, etc calls anyway. That
> > said this is a valid race!
>
> Hi Ilya,
> Yes, it can't solve the problem with blocking ceph_osdc_call(). As
> what we discussed, the better solution should be rbd-level timeout.
> But I found the warning in my testing, so I think we should fix it
> currently.
> >
> > I think it would be better to do that in rbd_add_acquire_lock(), only
> > if the wait is interrupted. Please see the attachment.
>
> Looks good to me to avoid some unnecessary callings.
>
> Should I send a V2 for it? or you can fix it by yourself? Either is fine
> to me.
I'll amend it myself and queue it up for the next -rc.
Thanks,
Ilya
