> From: linux-rdma-owner@xxxxxxxxxxxxxxx [mailto:linux-rdma- > owner@xxxxxxxxxxxxxxx] On Behalf Of Kalderon, Michal > > > From: Chuck Lever [mailto:chuck.lever@xxxxxxxxxx] > > Sent: Wednesday, February 14, 2018 6:58 PM > > > > > > > On Feb 14, 2018, at 11:49 AM, Kalderon, Michal > > <Michal.Kalderon@xxxxxxxxxx> wrote: > > > > > >> From: Leon Romanovsky [mailto:leon@xxxxxxxxxx] > > >> Sent: Wednesday, February 14, 2018 6:34 PM > > >> To: Chuck Lever <chuck.lever@xxxxxxxxxx> > > >> Cc: Kalderon, Michal <Michal.Kalderon@xxxxxxxxxx>; Le, Thong > > >> <Thong.Le@xxxxxxxxxx>; linux-rdma@xxxxxxxxxxxxxxx > > >> Subject: Re: rdma resource warning on 4.16-rc1 when unloading qedr > > >> after NFS mount > > >> > > >> On Wed, Feb 14, 2018 at 11:20:39AM -0500, Chuck Lever wrote: > > >>> > > >>> > > >>>> On Feb 14, 2018, at 11:00 AM, Kalderon, Michal > > >> <Michal.Kalderon@xxxxxxxxxx> wrote: > > >>>> > > >>>> Hi Leon, Chuck, > > >>>> > > >>>> We ran nfs mount over qedr using 4.16-rc1 When unloading qedr we > > >>>> get a WARNING from the resource tracker ( pasted below) > > >>>> > > >>>> Can you please advise on the best way to debug this? How can we > > >>>> get > > >> more info on the resource not being freed? > > >>> > > >>> I haven't seen this kind of report before, so I can't directly > > >>> answer your questions. But can you tell us more about reproducing it: > > >> > > >> It is resource tracking which was entered in last merge window. > > >> > > >>> > > >>> - Is there a workload running on the NFS mount point when the > > >>> module is unloaded? > > > no > > >>> > > >>> - Is the issue 100% reproducible, or intermittent? > > > Seems to be > > >>> > > >>> - Have you tried bisecting? > > > No, bisecting is a tough one here since we ran this scenario to > > > verify the last Two related nfs fixes > > > e89e8d8 xprtrdma: Fix BUG after a device removal 1179e2c xprtrdma: > > > Fix calculation of ri_max_send_sges > > > > > >> > > >> It will be one of three patches: > > >> 9d5f8c209b3f RDMA/core: Add resource tracking for create and > > >> destroy PDs 08f294a1524b RDMA/core: Add resource tracking for > > >> create and destroy CQs > > >> 78a0cd648a80 RDMA/core: Add resource tracking for create and > > >> destroy QPs > > > Do you think these could lead to a resource not being freed? Or only > > > issues > > with tracking? > > > > > >> > > >>> > > >>> - iWARP, RoCE, or both? > > > Only tested over RoCE for now > > >>> > > >>> - Have you tried reproducing with a different model of device? > > > no > > >> > > >> I doubt that it is related to device, it looks like a resource leak > > >> while removing rpcrdma. > > >> > > >> We definitely need to add more information to this warning to > > >> understand which one of three available resources wasn't freed. > > > > > > Missed an output from our driver saying there's a PD not freed. As > > > mentioned, due to other Issues we're not sure whether we've seen > > > this > > message from our driver in the past. > > > > When I've tested device unload with rpcrdma.ko, the unload hangs if > > rpcrdma.ko doesn't release all resources. > > > > rpcrdma_ia_remove() releases transport resources. It destroys the QP > > and CQs, but leaves the ID and PD to be destroyed by the device driver or > core. > > The CM event handler returns 1 to signal this is the case. > > > > I suspect it could be a driver bug. > Our driver doesn't take care of releasing PDs, it counts on layers above to do > so. > Why should the PD be treated differently than the CQs/QPs in this case? > we will look into this further to understand whether this is newly introduced. > thanks Hi Chuck, the PD that is not freed here by rpcrdma is freed if we issue a umount. Mount: this is the creation of the pd: [ 1162.401116] ? rpcrdma_create_id+0x20b/0x270 [rpcrdma] [ 1162.401124] rpcrdma_ia_open+0x40/0xe0 [rpcrdma] [ 1162.401132] xprt_setup_rdma+0x110/0x3a0 [rpcrdma] [ 1162.401147] xprt_create_transport+0x7d/0x210 [sunrpc] [ 1162.401161] rpc_create+0xc5/0x1c0 [sunrpc] Umount: [ 1011.602701] qedr_dealloc_pd+0x18/0x90 [qedr] [ 1011.602709] ib_dealloc_pd+0x45/0x80 [ib_core] [ 1011.602716] rpcrdma_ia_close+0x57/0x70 [rpcrdma] [ 1011.602719] xprt_rdma_destroy+0x4d/0xb0 [rpcrdma] Why not call rpcrdma_ia_close from rpcrdma_ia_remove Thanks, Michal > > > > > > > >>>> Thanks, > > >>>> Michal > > >>>> > > >>>> GAD17990 login: [ 300.480137] ib_srpt srpt_remove_one(qedr0): > > >>>> nothing > > >> to do. > > >>>> [ 300.515527] ib_srpt srpt_remove_one(qedr1): nothing to do. > > >>>> [ 300.542182] rpcrdma: removing device qedr1 for > > >>>> 192.168.110.146:20049 [ 300.573789] WARNING: CPU: 12 PID: 3545 > > >>>> at > > >>>> drivers/infiniband/core/restrack.c:20 > > >>>> rdma_restrack_clean+0x25/0x30 [ib_core] [ 300.625985] Modules > > >>>> linked in: rpcsec_gss_krb5 nfsv4 dns_resolver nfs fscache rpcrdma > > >>>> ib_isert iscsi_target_mod ib_iser libiscsi scsi_transport_iscsi > > >>>> ib_srpt target_core_mod ib_srp scsi_transport_srp ib_ipoib > > >>>> rdma_ucm ib_ucm ib_uverbs ib_umad rdma_cm ib_cm iw_cm 8021q > garp > > >>>> mrp qedr(-) ib_core xt_CHECKSUM iptable_mangle > ipt_MASQUERADE > > >>>> nf_nat_masquerade_ipv4 > > iptable_nat > > >>>> nf_nat_ipv4 nf_nat nf_conntrack_ipv4 nf_defrag_ipv4 xt_conntrack > > >>>> nf_conntrack ipt_REJECT nf_reject_ipv4 tun bridge stp llc > > >>>> ebtable_filter ebtables fuse ip6table_filter ip6_tables > > >>>> iptable_filter dm_mirror dm_region_hash dm_log dm_mod vfat fat > > >>>> dax intel_rapl sb_edac x86_pkg_temp_thermal intel_powerclamp > > coretemp > > >>>> kvm_intel kvm irqbypass crct10dif_pclmul crc32_pclmul > > >>>> ghash_clmulni_intel pcbc aesni_intel crypto_simd glue_helper > > >>>> cryptd ipmi_si [ 300.972993] iTCO_wdt ipmi_devintf sg pcspkr > > >> iTCO_vendor_support hpwdt hpilo lpc_ich ipmi_msghandler > pcc_cpufreq > > >> ioatdma i2c_i801 mfd_core wmi shpchp dca acpi_power_meter i2c_core > > >> nfsd auth_rpcgss nfs_acl lockd grace sunrpc ip_tables xfs libcrc32c > > >> sd_mod qede qed crc32c_intel tg3 hpsa scsi_transport_sas crc8 [ > > 301.109036] CPU: 12 PID: > > >> 3545 Comm: rmmod Not tainted 4.16.0-rc1 #1 [ 301.139518] Hardware > > name: > > >> HP ProLiant DL380 Gen9/ProLiant DL380 Gen9, BIOS P89 02/17/2017 [ > > >> 301.180411] RIP: 0010:rdma_restrack_clean+0x25/0x30 [ib_core] [ > > >> 301.208350] RSP: 0018:ffffb1820478fe88 EFLAGS: 00010286 [ > > >> 301.233241] > > >> RAX: 0000000000000000 RBX: ffffa099ed1b4070 RCX: ffffdf02a193c800 [ > > >> 301.268001] RDX: ffffa095ed12d7a0 RSI: 0000000000025900 RDI: > > >> ffffa099ed1b47d0 [ 301.302530] RBP: ffffa099ed1b4070 R08: > > >> ffffa095de9dd000 R09: 0000000180080007 [ 301.337245] R10: > > >> 0000000000000001 R11: ffffa095de9dd000 R12: ffffa099ed1b4000 [ > > >> 301.372151] R13: ffffa099ed1b405c R14: 0000000000e231c0 R15: > > >> 0000000000e23010 [ 301.407384] FS: 00007f2b0c854740(0000) > > >> GS:ffffa099ff700000(0000) knlGS:0000000000000000 [ 301.447026] CS: > > >> 0010 > > >> DS: 0000 ES: 0000 CR0: 0000000080050033 [ 301.475409] CR2: > > >> 0000000000e2caf8 CR3: 0000000865c0d006 CR4: 00000000001606e0 [ > > >> 301.510892] Call Trace: > > >>>> [ 301.522715] ib_unregister_device+0xf5/0x190 [ib_core] [ > > >>>> 301.547966] qedr_remove+0x37/0x60 [qedr] [ 301.568393] > > >>>> qede_rdma_unregister_driver+0x4b/0x90 [qede] [ 301.594980] > > >>>> SyS_delete_module+0x168/0x240 [ 301.615057] > > >>>> do_syscall_64+0x6f/0x1a0 [ 301.633588] > > >>>> entry_SYSCALL_64_after_hwframe+0x21/0x86 > > >>>> [ 301.658657] RIP: 0033:0x7f2b0bd33707 [ 301.676005] RSP: > > >>>> 002b:00007ffdefa29d98 EFLAGS: 00000202 ORIG_RAX: > > 00000000000000b0 > > >> [ > > >>>> 301.713324] RAX: ffffffffffffffda RBX: 0000000000e231c0 RCX: > > >>>> 00007f2b0bd33707 [ 301.748186] RDX: 00007f2b0bda3a80 RSI: > > >>>> 0000000000000800 RDI: 0000000000e23228 [ 301.782960] RBP: > > >>>> 0000000000000000 R08: 00007f2b0bff8060 R09: 00007f2b0bda3a80 [ > > >>>> 301.818142] R10: 00007ffdefa29b20 R11: 0000000000000202 R12: > > >>>> 00007ffdefa2b70d [ 301.853290] R13: 0000000000000000 R14: > > >>>> 0000000000e231c0 R15: 0000000000e23010 [ 301.888138] Code: 84 00 > > >>>> 00 > > >>>> 00 00 00 0f 1f 44 00 00 48 83 c7 28 31 c0 eb 0c 48 83 c0 08 48 3d > > >>>> 00 > > >>>> 08 00 00 74 0f 48 8d 14 07 48 8b 12 48 85 d2 74 e8 <0f> ff c3 f3 > > >>>> c3 > > >>>> 66 0f 1f 44 00 00 0f 1f 44 00 00 53 48 8b 47 28 [ 301.981140] > > >>>> ---[ end trace 28dec8f15205789a ]--- > > >>> > > >>> -- > > >>> Chuck Lever > > >>> > > >>> > > >>> > > > -- > > > To unsubscribe from this list: send the line "unsubscribe linux-rdma" > > > in the body of a message to majordomo@xxxxxxxxxxxxxxx More > > majordomo > > > info at http://vger.kernel.org/majordomo-info.html > > > > -- > > Chuck Lever > > > > > > -- > To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the > body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at > http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
