* [dpdk-dev] [PATCH] net/bond: change link status check to no-wait
@ 2017-06-26 15:13 Declan Doherty
2017-06-26 20:53 ` Chas Williams
2017-06-28 10:15 ` [dpdk-dev] [PATCH v2] " Declan Doherty
0 siblings, 2 replies; 5+ messages in thread
From: Declan Doherty @ 2017-06-26 15:13 UTC (permalink / raw)
To: dev; +Cc: Declan Doherty
In the 802.3ad periodic callback function the link status of all slaves
is checked using rte_eth_link_get function. Depending on the slave
device this function can block for up to 9 seconds and therefore
could cause issues with the LACP Daemon state machine and control
patches handling. This patch changes to the no-wait version of
the link get function.
Fixes: 46fb4368 ("bond:add mode 4")
Signed-off-by: Declan Doherty <declan.doherty@intel.com>
---
drivers/net/bonding/rte_eth_bond_8023ad.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/net/bonding/rte_eth_bond_8023ad.c b/drivers/net/bonding/rte_eth_bond_8023ad.c
index d2b7592..3a97336 100644
--- a/drivers/net/bonding/rte_eth_bond_8023ad.c
+++ b/drivers/net/bonding/rte_eth_bond_8023ad.c
@@ -758,7 +758,7 @@ bond_mode_8023ad_periodic_cb(void *arg)
uint16_t key;
slave_id = internals->active_slaves[i];
- rte_eth_link_get(slave_id, &link_info);
+ rte_eth_link_get_nowait(slave_id, &link_info);
rte_eth_macaddr_get(slave_id, &slave_addr);
if (link_info.link_status != 0) {
--
2.9.4
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [dpdk-dev] [PATCH] net/bond: change link status check to no-wait
2017-06-26 15:13 [dpdk-dev] [PATCH] net/bond: change link status check to no-wait Declan Doherty
@ 2017-06-26 20:53 ` Chas Williams
2017-06-28 9:13 ` Declan Doherty
2017-06-28 10:15 ` [dpdk-dev] [PATCH v2] " Declan Doherty
1 sibling, 1 reply; 5+ messages in thread
From: Chas Williams @ 2017-06-26 20:53 UTC (permalink / raw)
To: Declan Doherty, dev
On Mon, 2017-06-26 at 16:13 +0100, Declan Doherty wrote:
> In the 802.3ad periodic callback function the link status of all slaves
> is checked using rte_eth_link_get function. Depending on the slave
> device this function can block for up to 9 seconds and therefore
> could cause issues with the LACP Daemon state machine and control
> patches handling. This patch changes to the no-wait version of
> the link get function.
>
> Fixes: 46fb4368 ("bond:add mode 4")
>
> Signed-off-by: Declan Doherty <declan.doherty@intel.com>
> ---
> drivers/net/bonding/rte_eth_bond_8023ad.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/net/bonding/rte_eth_bond_8023ad.c b/drivers/net/bonding/rte_eth_bond_8023ad.c
> index d2b7592..3a97336 100644
> --- a/drivers/net/bonding/rte_eth_bond_8023ad.c
> +++ b/drivers/net/bonding/rte_eth_bond_8023ad.c
> @@ -758,7 +758,7 @@ bond_mode_8023ad_periodic_cb(void *arg)
> uint16_t key;
>
> slave_id = internals->active_slaves[i];
> - rte_eth_link_get(slave_id, &link_info);
> + rte_eth_link_get_nowait(slave_id, &link_info);
> rte_eth_macaddr_get(slave_id, &slave_addr);
>
> if (link_info.link_status != 0) {
It seems like this should be done in bandwidth_left() as well.
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [dpdk-dev] [PATCH] net/bond: change link status check to no-wait
2017-06-26 20:53 ` Chas Williams
@ 2017-06-28 9:13 ` Declan Doherty
0 siblings, 0 replies; 5+ messages in thread
From: Declan Doherty @ 2017-06-28 9:13 UTC (permalink / raw)
To: Chas Williams, dev
On 26/06/2017 9:53 PM, Chas Williams wrote:
> On Mon, 2017-06-26 at 16:13 +0100, Declan Doherty wrote:
>> In the 802.3ad periodic callback function the link status of all slaves
>> is checked using rte_eth_link_get function. Depending on the slave
>> device this function can block for up to 9 seconds and therefore
>> could cause issues with the LACP Daemon state machine and control
>> patches handling. This patch changes to the no-wait version of
>> the link get function.
>>
>> Fixes: 46fb4368 ("bond:add mode 4")
>>
>> Signed-off-by: Declan Doherty <declan.doherty@intel.com>
>> ---
...
>
> It seems like this should be done in bandwidth_left() as well.
>
Yep, your right. I'll make that change in the v2.
Thanks
Declan
^ permalink raw reply [flat|nested] 5+ messages in thread
* [dpdk-dev] [PATCH v2] net/bond: change link status check to no-wait
2017-06-26 15:13 [dpdk-dev] [PATCH] net/bond: change link status check to no-wait Declan Doherty
2017-06-26 20:53 ` Chas Williams
@ 2017-06-28 10:15 ` Declan Doherty
2017-06-29 11:34 ` Ferruh Yigit
1 sibling, 1 reply; 5+ messages in thread
From: Declan Doherty @ 2017-06-28 10:15 UTC (permalink / raw)
To: dev; +Cc: Declan Doherty
In 2 modes (802.3ad and TLB) of the link bonding driver, monitoring of
link status is used to determine the active slaves to use. These
functions are currently using the rte_link_link_get which could block
for up to 9 seconds, depending on the slave device.
In the 802.3ad periodic callback and in the TLB slave callback the
link status of slaves are checked using rte_eth_link_get function.
This patch changes to the no-wait version of the link get function to
avoid the possible issues which could be introduced if this call was
to block for an extended period of time.
Fixes: 46fb4368 ("bond:add mode 4")
Fixes: 7c76a747 ("bond:add mode 5")
Signed-off-by: Declan Doherty <declan.doherty@intel.com>
---
drivers/net/bonding/rte_eth_bond_8023ad.c | 2 +-
drivers/net/bonding/rte_eth_bond_pmd.c | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/net/bonding/rte_eth_bond_8023ad.c b/drivers/net/bonding/rte_eth_bond_8023ad.c
index d2b7592..3a97336 100644
--- a/drivers/net/bonding/rte_eth_bond_8023ad.c
+++ b/drivers/net/bonding/rte_eth_bond_8023ad.c
@@ -758,7 +758,7 @@ bond_mode_8023ad_periodic_cb(void *arg)
uint16_t key;
slave_id = internals->active_slaves[i];
- rte_eth_link_get(slave_id, &link_info);
+ rte_eth_link_get_nowait(slave_id, &link_info);
rte_eth_macaddr_get(slave_id, &slave_addr);
if (link_info.link_status != 0) {
diff --git a/drivers/net/bonding/rte_eth_bond_pmd.c b/drivers/net/bonding/rte_eth_bond_pmd.c
index a8d9780..eff2a66 100644
--- a/drivers/net/bonding/rte_eth_bond_pmd.c
+++ b/drivers/net/bonding/rte_eth_bond_pmd.c
@@ -654,7 +654,7 @@ bandwidth_left(uint8_t port_id, uint64_t load, uint8_t update_idx,
{
struct rte_eth_link link_status;
- rte_eth_link_get(port_id, &link_status);
+ rte_eth_link_get_nowait(port_id, &link_status);
uint64_t link_bwg = link_status.link_speed * 1000000ULL / 8;
if (link_bwg == 0)
return;
--
2.9.4
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [dpdk-dev] [PATCH v2] net/bond: change link status check to no-wait
2017-06-28 10:15 ` [dpdk-dev] [PATCH v2] " Declan Doherty
@ 2017-06-29 11:34 ` Ferruh Yigit
0 siblings, 0 replies; 5+ messages in thread
From: Ferruh Yigit @ 2017-06-29 11:34 UTC (permalink / raw)
To: Declan Doherty, dev
On 6/28/2017 11:15 AM, Declan Doherty wrote:
> In 2 modes (802.3ad and TLB) of the link bonding driver, monitoring of
> link status is used to determine the active slaves to use. These
> functions are currently using the rte_link_link_get which could block
> for up to 9 seconds, depending on the slave device.
>
> In the 802.3ad periodic callback and in the TLB slave callback the
> link status of slaves are checked using rte_eth_link_get function.
>
> This patch changes to the no-wait version of the link get function to
> avoid the possible issues which could be introduced if this call was
> to block for an extended period of time.
Fixes: 46fb43683679 ("bond: add mode 4")
Fixes: 7c76a747e68c ("bond: add mode 5")
Cc: stable@dpdk.org
>
> Signed-off-by: Declan Doherty <declan.doherty@intel.com>
Applied to dpdk-next-net/master, thanks.
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2017-06-29 11:34 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-06-26 15:13 [dpdk-dev] [PATCH] net/bond: change link status check to no-wait Declan Doherty
2017-06-26 20:53 ` Chas Williams
2017-06-28 9:13 ` Declan Doherty
2017-06-28 10:15 ` [dpdk-dev] [PATCH v2] " Declan Doherty
2017-06-29 11:34 ` 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).