nvmet-fcloop: unblock module removal
Subsystems (4)
v3 cover nvmet-fcloop: unblock module removal
I got it somewhat working with this version here and the blktest fcloop cleanup fix[1].
There is a BIG if though. I have to disable manually the auto connect udev rules from the host system. I've played a bit around how to disable them during runtime. There is a way to inject a runtime global variable into udev rules via 'udevadm control -p BLKTESTS=1'. The udev rule file can then be extended with a rule like 'ENV{BLKTESTS}="1", GOTO=...' but I was not able to get this working reliable. It looks like the update is not getting promotted to all udev daemon threads. Even a 'sleep 5' or 'udevadm settle' didn't help.
The only way I found so far is to do
ln -s /dev/null /etc/udev/rules.d/70-nvmf-autoconnect.rules
which obviously sucks. IIRC, Martin B has the same issue with nvme-stats and nvme-cli...
Anyway, it doesn't creash anymore and there are not warning from lockdep or KASAN. The first few tests cases are passing. So the failing ones could actually be 'real' bugs, who knows :)
# nvme_trtype=fc ./check nvme
nvme/002 (create many subsystems and test discovery) [not run]
nvme_trtype=fc is not supported in this test
nvme/003 (test if we're sending keep-alives to a discovery controller) nvme/003 (test if we're sending keep-alives to a discovery controller) [passed]
runtime 10.277s ... 10.479srt: No such file or directory
nvme/004 (test nvme and nvmet UUID NS descriptors) [passed]
runtime 1.976s ... 1.723s
nvme/005 (reset local loopback target) [passed]
runtime 2.203s ... 1.829s
nvme/006 (create an NVMeOF target with a block device-backed ns) [passed]
runtime 0.248s ... 0.221s
nvme/007 (create an NVMeOF target with a file-backed ns) [passed]
runtime 0.156s ... 0.127s
nvme/008 (create an NVMeOF host with a block device-backed ns) [passed]
runtime 1.945s ... 1.658s
nvme/009 (create an NVMeOF host with a file-backed ns) [passed]
runtime 1.854s ... 1.711s
nvme/010 (run data verification fio job on NVMeOF block device-backed ns) [passed]
runtime 44.555s ... 41.625s
nvme/011 (run data verification fio job on NVMeOF file-backed ns) [passed]
runtime 89.795s ... 71.931s
nvme/012 (run mkfs and data verification fio job on NVMeOF block device-backed ns) [passed]
runtime 61.607s ... 49.917s
nvme/013 (run mkfs and data verification fio job on NVMeOF file-backed ns) [passed]
runtime 109.725s ... 79.189s
nvme/014 (flush a NVMeOF block device-backed ns) [passed]
runtime 7.902s ... 5.977s
nvme/015 (unit test for NVMe flush for file backed ns) [passed]
runtime 7.229s ... 5.734snvme/016 (create/delete many NVMeOF block device-backed ns and test discovery) [not run]
runtime 44.446s ...
nvme_trtype=fc is not supported in this test
nvme/017 (create/delete many file-ns and test discovery) [not run]
runtime 45.306s ...
nvme_trtype=fc is not supported in this test
nvme/018 (unit test NVMe-oF out of range access on a file backend) [passed]
runtime 1.834s ... 1.721s
nvme/019 (test NVMe DSM Discard command on NVMeOF block-device ns) [passed]
runtime 1.832s ... 1.804s
nvme/020 (test NVMe DSM Discard command on NVMeOF file-backed ns) [passed]
runtime 1.811s ... 1.717s
nvme/021 (test NVMe list command on NVMeOF file-backed ns) [passed]
runtime 1.807s ... 1.703s
nvme/022 (test NVMe reset command on NVMeOF file-backed ns) [passed]
runtime 1.914s ... 1.784s
nvme/023 (test NVMe smart-log command on NVMeOF block-device ns) [passed]
runtime 1.852s ... 1.730s
nvme/024 (test NVMe smart-log command on NVMeOF file-backed ns) [passed]
runtime 1.730s ... 1.754s
nvme/025 (test NVMe effects-log command on NVMeOF file-backed ns) [passed]
runtime 1.759s ... 1.719s
nvme/026 (test NVMe ns-descs command on NVMeOF file-backed ns) [passed]
runtime 1.764s ... 1.675s
nvme/027 (test NVMe ns-rescan command on NVMeOF file-backed ns) [passed]
runtime 1.734s ... 1.703s
nvme/028 (test NVMe list-subsys command on NVMeOF file-backed ns) [passed]
runtime 1.831s ... 1.732s
nvme/029 (test userspace IO via nvme-cli read/write interface) [passed]
runtime 2.388s ... 2.065s
nvme/030 (ensure the discovery generation counter is updated appropriately) [passed]
runtime 0.756s ... 0.631s
nvme/031 (test deletion of NVMeOF controllers immediately after setup) [passed]
runtime 4.566s ... 3.441s
nvme/038 (test deletion of NVMeOF subsystem without enabling) [passed]
runtime 0.054s ... 0.055s
nvme/040 (test nvme fabrics controller reset/disconnect operation during I/O)
nvme/040 (test nvme fabrics controller reset/disconnect operation during I/O) [passed]
runtime 7.948s ... 7.866srk/blktests/results/tmpdir.nvme.040.4lT': Directory not emptymodprobe: FATAL: Module loop is in use.
modprobe: FATAL: Module loop is in use.
modprobe: FATAL: Module loop is in use.
modprobe: FATAL: Module loop is in use.
modprobe: FATAL: Module loop is in use.
modprobe: FATAL: Module loop is in use.
modprobe: FATAL: Module loop is in use.
modprobe: FATAL: Module loop is in use.
modprobe: FATAL: Module loop is in use.
modprobe: FATAL: Module loop is in use.
nvme/041 (Create authenticated connections)
runtime 0.964s ...
WARNING: Test did not clean up fc device: nvme0
nvme/041 (Create authenticated connections) [failed]
runtime 0.964s ... 0.311s
--- tests/nvme/041.out 2023-02-20 10:31:10.953935278 +0100
+++ /home/wagi/work/blktests/results/nodev/nvme/041.out.bad 2023-04-18 14:31:05.122062907 +0200 @@ -1,6 +1,8 @@
Running nvme/041
Test unauthenticated connection (should fail)
+failed to lookup subsystem for controller nvme0
NQN:blktests-subsystem-1 disconnected 0 controller(s)
Test authenticated connection
-NQN:blktests-subsystem-1 disconnected 1 controller(s)
+failed to lookup subsystem for controller nvme0
...
(Run 'diff -u tests/nvme/041.out /home/wagi/work/blktests/results/nodev/nvme/041.out.bad' to see the entire diff)
WARNING: Test did not clean up fc device: nvme0
failed to lookup subsystem for controller nvme0
Did not find device nvme0
nvme/042 (Test dhchap key types for authenticated connections) [failed]
runtime 4.495s ... 4.226s
--- tests/nvme/042.out 2022-08-30 10:20:14.174819528 +0200
+++ /home/wagi/work/blktests/results/nodev/nvme/042.out.bad 2023-04-18 14:31:09.654086133 +0200 @@ -1,7 +1,9 @@
Running nvme/042
Testing hmac 0
+failed to lookup subsystem for controller nvme0
NQN:blktests-subsystem-1 disconnected 1 controller(s)
Testing hmac 1
+failed to lookup subsystem for controller nvme0
NQN:blktests-subsystem-1 disconnected 1 controller(s)
...
(Run 'diff -u tests/nvme/042.out /home/wagi/work/blktests/results/nodev/nvme/042.out.bad' to see the entire diff)
nvme/043 (Test hash and DH group variations for authenticated connections) [passed]
runtime 3.592s ... 8.750s
nvme/044 (Test bi-directional authentication)
runtime 1.099s ...WARNING: Test did not clean up fc device: nvme0
nvme/044 (Test bi-directional authentication) [failed]
runtime 1.099s ... 1.064s
--- tests/nvme/044.out 2023-02-20 10:31:10.953935278 +0100
+++ /home/wagi/work/blktests/results/nodev/nvme/044.out.bad 2023-04-18 14:31:20.042139370 +0200 @@ -2,9 +2,12 @@
Test host authentication
NQN:blktests-subsystem-1 disconnected 1 controller(s)
Test invalid ctrl authentication (should fail)
+failed to lookup subsystem for controller nvme0
NQN:blktests-subsystem-1 disconnected 0 controller(s)
Test valid ctrl authentication
-NQN:blktests-subsystem-1 disconnected 1 controller(s)
...
(Run 'diff -u tests/nvme/044.out /home/wagi/work/blktests/results/nodev/nvme/044.out.bad' to see the entire diff)
WARNING: Test did not clean up fc device: nvme0
failed to lookup subsystem for controller nvme0
Did not find device nvme0
nvme/045 (Test re-authentication) [passed]
runtime 5.582s ... 6.262s
nvme/047 (test different queue types for fabric transports) [not run]
nvme_trtype=fc is not supported in this test
nvme/048 (Test queue count changes on reconnect) [failed]
runtime 15.586s ... 16.545s
--- tests/nvme/048.out 2023-04-06 10:12:58.333064747 +0200
+++ /home/wagi/work/blktests/results/nodev/nvme/048.out.bad 2023-04-18 14:31:43.614260172 +0200
@@ -1,3 +1,7 @@
Running nvme/048
+expected state "connecting" not reached within 5 seconds
+FAIL
+expected state "connecting" not reached within 5 seconds
+FAIL
NQN:blktests-subsystem-1 disconnected 1 controller(s)
Test complete[1] https://lore.kernel.org/linux-nvme/[email protected]/
changes:
v3:
- do not unlink rport twice
v2:
- added additional fixes
- https://lore.kernel.org/linux-nvme/[email protected]/
v1:
- initial version
- https://lore.kernel.org/linux-nvme/[email protected]/
Daniel Wagner (4):
nvmet-fcloop: Remove remote port from list when unlinking
nvmet-fcloop: Do not wait on completion when unregister fails
nvmet-fc: Do not wait in vain when unloading module
nvmet-fc: Release reference on target port
drivers/nvme/host/fc.c | 20 +++++++++++++------- drivers/nvme/target/fc.c | 1 + drivers/nvme/target/fcloop.c | 10 ++++------ 3 files changed, 18 insertions(+), 13 deletions(-)
Signature · 1 line
2.40.0
PATCH v3 1/4 nvmet-fcloop: Remove remote port from list when unlinking
The remote port is removed too late from fcloop_nports list. Remove it when port is unregistered.
This prevents a busy loop in fcloop_exit, because it is possible the remote port is found in the list and thus we will never progress.
The kernel log will be spammed with
nvme_fcloop: fcloop_exit: Failed deleting remote port
nvme_fcloop: fcloop_exit: Failed deleting target port
--- drivers/nvme/target/fcloop.c | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-)
drivers/nvme/target/fcloop.c +2 -5
@@ -997,11 +997,6 @@ fcloop_nport_free(struct kref *ref) {struct fcloop_nport *nport = container_of(ref, struct fcloop_nport, ref); -unsigned long flags; - -spin_lock_irqsave (&fcloop_lock, flags); -list_del (&nport->nport_list); -spin_unlock_irqrestore (&fcloop_lock, flags);kfree (nport); }
@@ -1358,6 +1353,8 @@ __unlink_remote_port(struct fcloop_nport *nport) nport->tport->remoteport =NULL ; nport->rport =NULL ; +list_del (&nport->nport_list); +return rport; }
Signature · 1 line
2.40.0
PATCH v3 2/4 nvmet-fcloop: Do not wait on completion when unregister fails
In reply to Daniel WagnerThe nvme_fc_unregister_localport() returns an error code in case that the locaport pointer is NULL or has already been unegisterd. localport is is either in the ONLINE state (all resources allocated) or has already been put into DELETED state.
In this case we will never receive an wakeup call and thus any caller will hang, e.g. module unload.
--- drivers/nvme/target/fcloop.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/nvme/target/fcloop.c +2 -1
@@ -1163,7 +1163,8 @@ __wait_localport_unreg(struct fcloop_lport *lport) ret =nvme_fc_unregister_localport (lport->localport); -wait_for_completion (&lport->unreg_done); +if (!ret) +wait_for_completion (&lport->unreg_done);kfree (lport);
Signature · 1 line
2.40.0
PATCH v3 3/4 nvmet-fc: Do not wait in vain when unloading module
In reply to Daniel WagnerWhen there is no controller to be deleted the module unload path will still wait on the nvme_fc_unload_proceed completion. Because this will will never happen the caller will hang forever.
--- drivers/nvme/host/fc.c | 20 +++++++++++++------- 1 file changed, 13 insertions(+), 7 deletions(-)
drivers/nvme/host/fc.c +13 -7
@@ -3933,10 +3933,11 @@ static int __init nvme_fc_init_module(void)return ret; } -static void +static bool nvme_fc_delete_controllers (struct nvme_fc_rport *rport) {struct nvme_fc_ctrl *ctrl; +bool cleanup = false;spin_lock (&rport->lock);list_for_each_entry (ctrl, &rport->ctrl_list, ctrl_list) {
@@ -3944,21 +3945,28 @@ nvme_fc_delete_controllers(struct nvme_fc_rport *rport)"NVME-FC{%d}: transport unloading: deleting ctrl\n" , ctrl->cnum);nvme_delete_ctrl (&ctrl->ctrl); + cleanup = true; }spin_unlock (&rport->lock); + +return cleanup; } -static void +static bool nvme_fc_cleanup_for_unload (void ) {struct nvme_fc_lport *lport;struct nvme_fc_rport *rport; +bool cleanup = false;list_for_each_entry (lport, &nvme_fc_lport_list, port_list) {list_for_each_entry (rport, &lport->endp_list, endp_list) { -nvme_fc_delete_controllers (rport); +if (nvme_fc_delete_controllers (rport)) + cleanup = true; } } + +return cleanup; } static void __exit nvme_fc_exit_module(void)
@@ -3968,10 +3976,8 @@ static void __exit nvme_fc_exit_module(void)spin_lock_irqsave (&nvme_fc_lock, flags); nvme_fc_waiting_to_unload = true; -if (!list_empty (&nvme_fc_lport_list)) { - need_cleanup = true; -nvme_fc_cleanup_for_unload (); - } +if (!list_empty (&nvme_fc_lport_list)) + need_cleanup =nvme_fc_cleanup_for_unload ();spin_unlock_irqrestore (&nvme_fc_lock, flags);if (need_cleanup) {pr_info ("%s: waiting for ctlr deletes\n" , __func__);
Signature · 1 line
2.40.0
PATCH v3 4/4 nvmet-fc: Release reference on target port
In reply to Daniel WagnerIn case we return early out of __nvmet_fc_finish_ls_req() we still have to release the reference on the target port.
--- drivers/nvme/target/fc.c | 1 + 1 file changed, 1 insertion(+)
drivers/nvme/target/fc.c +1 -0
@@ -359,6 +359,7 @@ __nvmet_fc_finish_ls_req(struct nvmet_fc_ls_req_op *lsop)if (!lsop->req_queued) {spin_unlock_irqrestore (&tgtport->lock, flags); +nvmet_fc_tgtport_put (tgtport);return ; }
Signature · 1 line
2.40.0
Quoted 5 lines · Daniel Wagner
nvme/041 (Create authenticated connections) [failed]
nvme/042 (Test dhchap key types for authenticated connections) [failed] nvme/043 (Test hash and DH group variations for authenticated connections) [passed]
nvme/044 (Test bi-directional authentication) [failed]
nvme/045 (Test re-authentication) [passed]
I suppose these should be disabled for fc as all this is tcp specific.
Quoted 8 lines · Daniel Wagner
Quoted 5 lines · Daniel Wagner
nvme/041 (Create authenticated connections) [failed]
nvme/042 (Test dhchap key types for authenticated connections) [failed] nvme/043 (Test hash and DH group variations for authenticated connections) [passed]
nvme/044 (Test bi-directional authentication) [failed]
nvme/045 (Test re-authentication) [passed]I suppose these should be disabled for fc as all this is tcp specific.
After a fresh reboot the deleter tport, lport and rport trouble is back...
nvme/003 (test if we're sending keep-alives to a discovery controller) [passed]
runtime 10.265s ... 10.365s
tests/nvme/rc: line 198: /sys/class/fcloop/ctl/del_target_port: No such file or directory tests/nvme/rc: line 190: /sys/class/fcloop/ctl/del_local_port: No such file or directory tests/nvme/rc: line 182: /sys/class/fcloop/ctl/del_remote_port: No such file or directory
Quoted 10 lines · Daniel Wagner
Quoted 8 lines · Daniel Wagner
Quoted 5 lines · Daniel Wagner
nvme/041 (Create authenticated connections) [failed]
nvme/042 (Test dhchap key types for authenticated connections) [failed] nvme/043 (Test hash and DH group variations for authenticated connections) [passed]
nvme/044 (Test bi-directional authentication) [failed]
nvme/045 (Test re-authentication) [passed]I suppose these should be disabled for fc as all this is tcp specific.
Quoted 7 lines
After a fresh reboot the deleter tport, lport and rport trouble is back...
nvme/003 (test if we're sending keep-alives to a discovery controller) [passed]
runtime 10.265s ... 10.365stests/nvme/rc: line 198:
/sys/class/fcloop/ctl/del_target_port: No such file or directory tests/nvme/rc: line 190:/sys/class/fcloop/ctl/del_local_port: No such file or directory tests/nvme/rc: line 182:/sys/class/fcloop/ctl/del_remote_port: No such file or directory
Eventually, I figured the out the root problem. The modules got unloaded before the resource were freed. This explains a lot of the nasty problems I saw. Anyway, I posted an updated blktests fixes but I think we should still consider these patches here.
https://lore.kernel.org/linux-nvme/[email protected]/
BTW, the authentication tests fail for fc, but not for the rest. And after reading up on it, it supposed to work on fc as well. So here we go first real bugs found.
Sagi Grimberg
maintainerQuoted 8 lines
Quoted 6 lines
nvme/041 (Create authenticated connections) [failed]
nvme/042 (Test dhchap key types for authenticated connections) [failed] nvme/043 (Test hash and DH group variations for authenticated connections) [passed]
nvme/044 (Test bi-directional authentication) [failed]
nvme/045 (Test re-authentication) [passed]I suppose these should be disabled for fc as all this is tcp specific.
Umm, no their not...