Re: [PATCH v3 00/38] vfs, nfsd: implement directory delegations

[Jeff Layton <jlayton@kernel.org>]

On Thu, 2025-09-25 at 17:02 -0400, Jeff Layton wrote:
> On Thu, 2025-09-25 at 16:45 -0400, Anna Schumaker wrote:
> > Hi Jeff,
> > 
> > (I trimmed off most of the extra people CC-ed, since this appears to mostly be an
> > NFS issue)
> > 
> > I hit this crash while stepping through my client side code and fixing it up against
> > your latest branch. I'm at a point where the client only requests the delegation, but
> > without any requested notifications, if that helps:
> > 
> > 
> > [  643.888646] BUG: kernel NULL pointer dereference, address: 0000000000000168
> > [  643.889045] #PF: supervisor read access in kernel mode
> > [  643.889314] #PF: error_code(0x0000) - not-present page
> > [  643.889591] PGD 0 P4D 0 
> > [  643.889733] Oops: Oops: 0000 [#1] SMP NOPTI
> > [  643.889960] CPU: 3 UID: 0 PID: 1003 Comm: nfsd Not tainted 6.17.0-rc7-00095-gf6fa32f97c33 #47188 PREEMPT(full)  b859994234adae648e07409684697d13c51d22ee
> > [  643.890665] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS unknown 02/02/2022
> > [  643.891076] RIP: 0010:nfsd_handle_dir_event+0x49/0x2c0 [nfsd]
> > [  643.891490] Code: 24 04 83 f9 01 74 11 83 f9 04 74 19 83 f9 02 75 11 4d 8b 64 24 08 eb 0d 49 8b 04 24 4c 8b 60 08 eb 03 45 31 e4 0f 1f 44 00 00 <4c> 8b b6 68 01 00 00 4d 85 f6 0f 84 dc 01 00 00 49 8d 5e 28 49 8b
> > [  643.892432] RSP: 0018:ffffd0d202333a80 EFLAGS: 00010246
> > [  643.892711] RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000004
> > [  643.893076] RDX: ffff8e3b55ad4480 RSI: 0000000000000000 RDI: 0000000040000004
> > [  643.893441] RBP: 0000000040000004 R08: 0000000000000000 R09: 0000000000000000
> > [  643.893816] R10: fffffffffffffffb R11: ffffffffc0a80500 R12: ffff8e3b55ad4480
> > [  643.894183] R13: 0000000000000000 R14: 0000000000000004 R15: 0000000000000004
> > [  643.894554] FS:  0000000000000000(0000) GS:ffff8e3f10206000(0000) knlGS:0000000000000000
> > [  643.894961] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> > [  643.895258] CR2: 0000000000000168 CR3: 0000000112571002 CR4: 0000000000772ef0
> > [  643.895644] PKRU: 55555554
> > [  643.895792] Call Trace:
> > [  643.895931]  <TASK>
> > [  643.896050]  fsnotify+0x69a/0x9a0
> > [  643.896233]  fsnotify_change+0xad/0xc0
> > [  643.896431]  notify_change+0x34f/0x380
> > [  643.896633]  nfsd_setattr+0x314/0x6f0 [nfsd 0b3fc8b3981bb65e36d518f9edccb6d9f216a09a]
> > [  643.897073]  ? nfsd_setuser_and_check_port+0xdd/0x120 [nfsd 0b3fc8b3981bb65e36d518f9edccb6d9f216a09a]
> > [  643.897576]  nfsd4_setattr+0x254/0x370 [nfsd 0b3fc8b3981bb65e36d518f9edccb6d9f216a09a]
> > [  643.898014]  ? nfsd4_encode_operation+0x207/0x2b0 [nfsd 0b3fc8b3981bb65e36d518f9edccb6d9f216a09a]
> > [  643.898502]  nfsd4_proc_compound+0x337/0x600 [nfsd 0b3fc8b3981bb65e36d518f9edccb6d9f216a09a]
> > [  643.898964]  nfsd_dispatch+0xc1/0x210 [nfsd 0b3fc8b3981bb65e36d518f9edccb6d9f216a09a]
> > [  643.899390]  svc_process_common+0x567/0x6a0 [sunrpc c8ffd8e151f2f4e7c45ca22edc099dc57603df52]
> > [  643.899881]  ? __pfx_nfsd_dispatch+0x10/0x10 [nfsd 0b3fc8b3981bb65e36d518f9edccb6d9f216a09a]
> > [  643.900341]  svc_process+0x117/0x200 [sunrpc c8ffd8e151f2f4e7c45ca22edc099dc57603df52]
> > [  643.900790]  svc_recv+0xa7d/0xbc0 [sunrpc c8ffd8e151f2f4e7c45ca22edc099dc57603df52]
> > [  643.901208]  nfsd+0xb6/0xf0 [nfsd 0b3fc8b3981bb65e36d518f9edccb6d9f216a09a]
> > [  643.901598]  ? __pfx_nfsd+0x10/0x10 [nfsd 0b3fc8b3981bb65e36d518f9edccb6d9f216a09a]
> > [  643.902024]  kthread+0x215/0x250
> > [  643.902201]  ? __pfx_kthread+0x10/0x10
> > [  643.902394]  ret_from_fork+0x106/0x1d0
> > [  643.902606]  ? __pfx_kthread+0x10/0x10
> > [  643.902800]  ret_from_fork_asm+0x1a/0x30
> > [  643.903013]  </TASK>
> > [  643.903133] Modules linked in: rpcsec_gss_krb5 rpcrdma rdma_cm ib_cm iw_cm ib_core cfg80211 rfkill 8021q mrp garp stp llc ext4 mbcache crc16 jbd2 vfat fat intel_rapl_msr intel_rapl_common intel_uncore_frequency_common intel_pmc_core intel_pmc_ssram_telemetry pmt_telemetry snd_hda_codec_generic pmt_discovery pmt_class intel_vsec snd_hda_intel snd_intel_dspcfg kvm_intel snd_hda_codec snd_hwdep kvm snd_hda_core irqbypass snd_pcm polyval_clmulni iTCO_wdt ghash_clmulni_intel intel_pmc_bxt iTCO_vendor_support aesni_intel snd_timer psmouse rapl i2c_i801 pcspkr snd lpc_ich i2c_smbus soundcore joydev mousedev mac_hid btrfs raid6_pq xor nfsd nfs_acl lockd grace nfs_localio auth_rpcgss usbip_host usbip_core dm_mod loop sunrpc nfnetlink vsock_loopback vmw_vsock_virtio_transport_common vmw_vsock_vmci_transport vsock vmw_vmci qemu_fw_cfg xfs serio_raw virtio_gpu virtio_dma_buf virtio_rng virtio_scsi virtio_balloon virtio_net intel_agp net_failover failover intel_gtt
> > [  643.909087] CR2: 0000000000000168
> > [  643.909503] ---[ end trace 0000000000000000 ]---
> > 
> > 
> > Any guesses for what could be causing this?
> > Anna
> > 
> 
> Yes. It happened in a SETATTR and I forgot to handle that case in
> nfsd_handle_dir_event(). Will fix.
> 
> I suspect that we probably just want to just recall the delegation in
> this case. Another option would be to send a DIR_ATTRS update to the
> client, but it's possible that the dir_delegation isn't set up to send
> the attrs that changed, and trying to sort that out sounds like a mess.

Actually, the delegation should have been recalled. We just want to
ignore the update in this case. The latest patches in my dir-deleg
branch should have a fix.

Thanks for helping test!
-- 
Jeff Layton <jlayton@kernel.org>