2022-01-11 16:35:47

by Håkon Bugge

Subject: [RFC for-rc] IB/mlx4: Fix the fix for reading flow-counters

It is not logical to call iboe_process_mad() when the link-layer is
Infiniband. Nevertheless, the commit message in commit 43bfb9729ea8
("IB/mlx4: Fix use of flow-counters for process_mad") explicitly state
that iboe_process_mad() shall be called.

Without this fix, reading:

yields "Invalid argument", whereas with this commit, said counter can
be read.

Please note that mlx4_1 is a VF.

Fixes: 43bfb9729ea8 ("IB/mlx4: Fix use of flow-counters for process_mad")
Signed-off-by: Håkon Bugge <[email protected]>
---
drivers/infiniband/hw/mlx4/mad.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/infiniband/hw/mlx4/mad.c b/drivers/infiniband/hw/mlx4/mad.c
index d13ecbd..cc83782 100644
--- a/drivers/infiniband/hw/mlx4/mad.c
+++ b/drivers/infiniband/hw/mlx4/mad.c
@@ -998,8 +998,8 @@ int mlx4_ib_process_mad(struct ib_device *ibdev, int mad_flags, u32 port_num,
(in->mad_hdr.attr_id == IB_PMA_PORT_COUNTERS ||
in->mad_hdr.attr_id == IB_PMA_PORT_COUNTERS_EXT ||
in->mad_hdr.attr_id == IB_PMA_CLASS_PORT_INFO)))
- return iboe_process_mad(ibdev, mad_flags, port_num,
- in_wc, in_grh, in, out);
+ return ib_process_mad(ibdev, mad_flags, port_num,
+ in_wc, in_grh, in, out);

return ib_process_mad(ibdev, mad_flags, port_num, in_wc, in_grh,
in, out);
--
1.8.3.1

2022-01-11 16:38:42

by Håkon Bugge

Subject: Re: [RFC for-rc] IB/mlx4: Fix the fix for reading flow-counters

> On 11 Jan 2022, at 17:35, Håkon Bugge <[email protected]> wrote:
>
> It is not logical to call iboe_process_mad() when the link-layer is
> Infiniband. Nevertheless, the commit message in commit 43bfb9729ea8
> ("IB/mlx4: Fix use of flow-counters for process_mad") explicitly state
> that iboe_process_mad() shall be called.
>
> Without this fix, reading:

# cat /sys/class/infiniband/mlx4_0/ports/2/counters_ext/port_xmit_data_64

(sorry, this line was of course taken as comment and was removed).


Thxs, Håkon


>
> yields "Invalid argument", whereas with this commit, said counter can
> be read.
>
> Please note that mlx4_1 is a VF.
>
> Fixes: 43bfb9729ea8 ("IB/mlx4: Fix use of flow-counters for process_mad")
> Signed-off-by: Håkon Bugge <[email protected]>
> ---
> drivers/infiniband/hw/mlx4/mad.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/infiniband/hw/mlx4/mad.c b/drivers/infiniband/hw/mlx4/mad.c
> index d13ecbd..cc83782 100644
> --- a/drivers/infiniband/hw/mlx4/mad.c
> +++ b/drivers/infiniband/hw/mlx4/mad.c
> @@ -998,8 +998,8 @@ int mlx4_ib_process_mad(struct ib_device *ibdev, int mad_flags, u32 port_num,
> (in->mad_hdr.attr_id == IB_PMA_PORT_COUNTERS ||
> in->mad_hdr.attr_id == IB_PMA_PORT_COUNTERS_EXT ||
> in->mad_hdr.attr_id == IB_PMA_CLASS_PORT_INFO)))
> - return iboe_process_mad(ibdev, mad_flags, port_num,
> - in_wc, in_grh, in, out);
> + return ib_process_mad(ibdev, mad_flags, port_num,
> + in_wc, in_grh, in, out);
>
> return ib_process_mad(ibdev, mad_flags, port_num, in_wc, in_grh,
> in, out);
> --
> 1.8.3.1
>

2022-01-11 16:39:22

by Håkon Bugge

Subject: Re: [RFC for-rc] IB/mlx4: Fix the fix for reading flow-counters

> On 11 Jan 2022, at 17:38, Haakon Bugge <[email protected]> wrote:
>
>
>
>> On 11 Jan 2022, at 17:35, Håkon Bugge <[email protected]> wrote:
>>
>> It is not logical to call iboe_process_mad() when the link-layer is
>> Infiniband. Nevertheless, the commit message in commit 43bfb9729ea8
>> ("IB/mlx4: Fix use of flow-counters for process_mad") explicitly state
>> that iboe_process_mad() shall be called.
>>
>> Without this fix, reading:
>
> # cat /sys/class/infiniband/mlx4_0/ports/2/counters_ext/port_xmit_data_64

should read:

# cat /sys/class/infiniband/mlx4_1/ports/1/counters_ext/port_xmit_data_64


>
> (sorry, this line was of course taken as comment and was removed).
>
>
> Thxs, Håkon
>
>
>>
>> yields "Invalid argument", whereas with this commit, said counter can
>> be read.
>>
>> Please note that mlx4_1 is a VF.
>>
>> Fixes: 43bfb9729ea8 ("IB/mlx4: Fix use of flow-counters for process_mad")
>> Signed-off-by: Håkon Bugge <[email protected]>
>> ---
>> drivers/infiniband/hw/mlx4/mad.c | 4 ++--
>> 1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/infiniband/hw/mlx4/mad.c b/drivers/infiniband/hw/mlx4/mad.c
>> index d13ecbd..cc83782 100644
>> --- a/drivers/infiniband/hw/mlx4/mad.c
>> +++ b/drivers/infiniband/hw/mlx4/mad.c
>> @@ -998,8 +998,8 @@ int mlx4_ib_process_mad(struct ib_device *ibdev, int mad_flags, u32 port_num,
>> (in->mad_hdr.attr_id == IB_PMA_PORT_COUNTERS ||
>> in->mad_hdr.attr_id == IB_PMA_PORT_COUNTERS_EXT ||
>> in->mad_hdr.attr_id == IB_PMA_CLASS_PORT_INFO)))
>> - return iboe_process_mad(ibdev, mad_flags, port_num,
>> - in_wc, in_grh, in, out);
>> + return ib_process_mad(ibdev, mad_flags, port_num,
>> + in_wc, in_grh, in, out);
>>
>> return ib_process_mad(ibdev, mad_flags, port_num, in_wc, in_grh,
>> in, out);
>> --
>> 1.8.3.1

2022-01-11 18:35:14

by Håkon Bugge

Subject: Re: [RFC for-rc] IB/mlx4: Fix the fix for reading flow-counters

> On 11 Jan 2022, at 17:39, Haakon Bugge <[email protected]> wrote:
>
>
>
>> On 11 Jan 2022, at 17:38, Haakon Bugge <[email protected]> wrote:
>>
>>
>>
>>> On 11 Jan 2022, at 17:35, Håkon Bugge <[email protected]> wrote:
>>>
>>> It is not logical to call iboe_process_mad() when the link-layer is
>>> Infiniband. Nevertheless, the commit message in commit 43bfb9729ea8
>>> ("IB/mlx4: Fix use of flow-counters for process_mad") explicitly state
>>> that iboe_process_mad() shall be called.
>>>
>>> Without this fix, reading:
>>
>> # cat /sys/class/infiniband/mlx4_0/ports/2/counters_ext/port_xmit_data_64
>
> should read:
>
> # cat /sys/class/infiniband/mlx4_1/ports/1/counters_ext/port_xmit_data_64

Pleas ignore this RFC, as it is plain wrong. Sorry for the noise.


Thxs, Håkon

>
>
>>
>> (sorry, this line was of course taken as comment and was removed).
>>
>>
>> Thxs, Håkon
>>
>>
>>>
>>> yields "Invalid argument", whereas with this commit, said counter can
>>> be read.
>>>
>>> Please note that mlx4_1 is a VF.
>>>
>>> Fixes: 43bfb9729ea8 ("IB/mlx4: Fix use of flow-counters for process_mad")
>>> Signed-off-by: Håkon Bugge <[email protected]>
>>> ---
>>> drivers/infiniband/hw/mlx4/mad.c | 4 ++--
>>> 1 file changed, 2 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/drivers/infiniband/hw/mlx4/mad.c b/drivers/infiniband/hw/mlx4/mad.c
>>> index d13ecbd..cc83782 100644
>>> --- a/drivers/infiniband/hw/mlx4/mad.c
>>> +++ b/drivers/infiniband/hw/mlx4/mad.c
>>> @@ -998,8 +998,8 @@ int mlx4_ib_process_mad(struct ib_device *ibdev, int mad_flags, u32 port_num,
>>> (in->mad_hdr.attr_id == IB_PMA_PORT_COUNTERS ||
>>> in->mad_hdr.attr_id == IB_PMA_PORT_COUNTERS_EXT ||
>>> in->mad_hdr.attr_id == IB_PMA_CLASS_PORT_INFO)))
>>> - return iboe_process_mad(ibdev, mad_flags, port_num,
>>> - in_wc, in_grh, in, out);
>>> + return ib_process_mad(ibdev, mad_flags, port_num,
>>> + in_wc, in_grh, in, out);
>>>
>>> return ib_process_mad(ibdev, mad_flags, port_num, in_wc, in_grh,
>>> in, out);
>>> --
>>> 1.8.3.1