DPDK patches and discussions
 help / color / mirror / Atom feed
* [BUG] [bonding] bonding member delete bug
@ 2023-12-18  2:51 Simon Jones
  2023-12-18  6:37 ` Simon Jones
  0 siblings, 1 reply; 5+ messages in thread
From: Simon Jones @ 2023-12-18  2:51 UTC (permalink / raw)
  To: dev

[-- Attachment #1: Type: text/plain, Size: 2762 bytes --]

Hi all,

I'm using DPDK-21.11 in ovs-dpdk.

I found a "bonding member delete bug" .

1. How to reproduce

```
NOTICE: bondctl is a tool I develop, it's to control DPDK.

### step 1, Add bonding device bond0.
bondctl add bond0 mode active-backup

### step 2, Add member m1 into bond0.
bondctl set 0000:00:0a.0 master bond0

### step 3, Add bond0 into ovs bridge.
ovs-vsctl add-port brp0 bond0 -- set interface bond0 type=dpdk
options:dpdk-devargs=net_bonding-bond0
(this command call @bond_ethdev_start at last.)

### step 4, Delete bond0 from ovs bridge.
ovs-vsctl del-port br-phy bond0
(this command call @bond_ethdev_stop at last.)

### step 5, Delete m1 from bond0.
bondctl set 0000:00:0a.0 nomaster

### step 6, Delete bond0.
bondctl del bond0

### step 7, Add bond0.
bondctl add bond0 mode active-backup

### step 8, Add member m1 into bond0.
bondctl set 0000:00:0a.0 master bond0
(this command call @bond_ethdev_start at last.)

### Then got error message.
2023-12-15T08:24:04.153Z|00017|dpdk|ERR|Port 0 must be stopped to allow
configurr
ation
2023-12-15T08:24:04.153Z|00018|dpdk|ERR|bond_cmd_set_master(581) - can not
confii
g slave 0000:00:0a.0!
```

2. Debug

I found the reason is, when member port is DOWN, then add operation will
call "eth_dev->data->dev_started = 1;", but no one add active member port,
so when delete bond0, will NOT call @rte_eth_dev_stop, then add bond0
again, got error. Detail is:
```
### After step 1-3, add bond0 into ovs-dpdk
bond_ethdev_start
    eth_dev->data->dev_started = 1;
    for (i = 0; i < internals->slave_count; i++) {
        if (slave_configure(eth_dev, slave_ethdev) != 0) {
        if (slave_start(eth_dev, slave_ethdev) != 0) {
            rte_eth_dev_start

### NOTICE, as member port is DOWN, so will NOT call @activate_slave,
so @active_slave_count is 0.
bond_ethdev_lsc_event_callback
    activate_slave(bonded_eth_dev, port_id);

### After step 4, delete bond0 from ovs-dpdk, NOTICE,
as @active_slave_count is 0, so will NOT call @rte_eth_dev_stop
bond_ethdev_stop
    for (i = 0; i < internals->slave_count; i++) {
        if (find_slave_by_id(internals->active_slaves,
                internals->active_slave_count, slave_id) !=
                        internals->active_slave_count) {
            ret = rte_eth_dev_stop(slave_id);

### After step 5-7, delete bond0 and then add bond0

### After step 8, add bond0, as it's NOT call @rte_eth_dev_stop, so
call @rte_eth_dev_start
again will got error.
2023-12-15T08:24:04.153Z|00017|dpdk|ERR|Port 0 must be stopped to allow
configurr
ation

```

3. My question

Is this bug fixed ? Which commit ?

If NOT, how to fix this bug? I think it's better to call @rte_eth_dev_stop
for every member, even it's DOWN. How about this?

Thanks~


----
Simon Jones

[-- Attachment #2: Type: text/html, Size: 4259 bytes --]

^ permalink raw reply	[flat|nested] 5+ messages in thread
end of thread, other threads:[~2024-01-09 11:30 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-12-18  2:51 [BUG] [bonding] bonding member delete bug Simon Jones
2023-12-18  6:37 ` Simon Jones
2024-01-08 15:55   ` Ferruh Yigit
2024-01-08 16:07     ` Kevin Traynor
2024-01-09 11:30       ` Ferruh Yigit

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).