Hi,
The mlxbf-pmc driver fails to load when the firmware reports a new but not
yet implemented performance block. I can reproduce this today with a
Bluefield-3 card and UEFI version 4.6.0-18-g7d063bb-BId13035, since this
reports the new clock_measure performance block.
This[1] patch from Shravan implements the clock_measure support and will
solve the issue. But this series avoids the situation by ignoring and
logging unsupported performance blocks.
NOTE: This series is based on latest linux-next which contains new changes
to mlxbf-pmc.
[1] https://lore.kernel.org/lkml/1c2f1b6da51523fe0f338f9ddce9e3903148f604.1707808180.git.shravankr@nvidia.com/
Luiz Capitulino (2):
platform/mellanox: mlxbf-pmc: mlxbf_pmc_event_list(): make size ptr
optional
platform/mellanox: mlxbf-pmc: Ignore unsupported performance blocks
drivers/platform/mellanox/mlxbf-pmc.c | 56 +++++++++++++++++----------
1 file changed, 36 insertions(+), 20 deletions(-)
--
2.43.1
Currently, the driver has two behaviors to deal with new & unsupported
performance blocks reported by the firmware:
1. For register and unknown block types, the driver will fail to load
with the following error message:
[ 4510.956369] mlxbf-pmc: probe of MLNXBFD2:00 failed with error -22
2. For counter and crspace blocks, the driver will load and sysfs files
will be created but getting the contents of event_list or trying to
setup the counter will fail
Instead, let's ignore and log unsupported blocks. This means the driver
will always load and unsupported blocks will never show up in sysfs.
Signed-off-by: Luiz Capitulino <[email protected]>
---
drivers/platform/mellanox/mlxbf-pmc.c | 16 ++++++++++++++--
1 file changed, 14 insertions(+), 2 deletions(-)
diff --git a/drivers/platform/mellanox/mlxbf-pmc.c b/drivers/platform/mellanox/mlxbf-pmc.c
index b71636eb3db1..746567767e5b 100644
--- a/drivers/platform/mellanox/mlxbf-pmc.c
+++ b/drivers/platform/mellanox/mlxbf-pmc.c
@@ -1043,6 +1043,11 @@ static const struct mlxbf_pmc_events *mlxbf_pmc_event_list(const char *blk, size
return events;
}
+static bool mlxbf_pmc_event_supported(const char *blk)
+{
+ return !!mlxbf_pmc_event_list(blk, NULL);
+}
+
/* Get the event number given the name */
static int mlxbf_pmc_get_event_num(const char *blk, const char *evt)
{
@@ -1761,6 +1766,9 @@ static int mlxbf_pmc_init_perftype_counter(struct device *dev, unsigned int blk_
struct mlxbf_pmc_attribute *attr;
unsigned int i = 0, j = 0;
+ if (!mlxbf_pmc_event_supported(pmc->block_name[blk_num]))
+ return -ENOENT;
+
/* "event_list" sysfs to list events supported by the block */
attr = &pmc->block[blk_num].attr_event_list;
attr->dev_attr.attr.mode = 0444;
@@ -1840,7 +1848,7 @@ static int mlxbf_pmc_init_perftype_reg(struct device *dev, unsigned int blk_num)
events = mlxbf_pmc_event_list(pmc->block_name[blk_num], &count);
if (!events)
- return -EINVAL;
+ return -ENOENT;
pmc->block[blk_num].attr_event = devm_kcalloc(
dev, count, sizeof(struct mlxbf_pmc_attribute), GFP_KERNEL);
@@ -1878,7 +1886,7 @@ static int mlxbf_pmc_create_groups(struct device *dev, unsigned int blk_num)
else if (pmc->block[blk_num].type == MLXBF_PMC_TYPE_REGISTER)
err = mlxbf_pmc_init_perftype_reg(dev, blk_num);
else
- err = -EINVAL;
+ err = -ENOENT;
if (err)
return err;
@@ -1983,6 +1991,10 @@ static int mlxbf_pmc_map_counters(struct device *dev)
return -ENOMEM;
ret = mlxbf_pmc_create_groups(dev, i);
+ if (ret == -ENOENT) {
+ dev_warn(dev, "ignoring unsupported block: '%s'\n", pmc->block_name[i]);
+ continue;
+ }
if (ret)
return ret;
}
--
2.43.1
Hi,
On 2/22/24 21:57, Luiz Capitulino wrote:
> Currently, the driver has two behaviors to deal with new & unsupported
> performance blocks reported by the firmware:
>
> 1. For register and unknown block types, the driver will fail to load
> with the following error message:
>
> [ 4510.956369] mlxbf-pmc: probe of MLNXBFD2:00 failed with error -22
>
> 2. For counter and crspace blocks, the driver will load and sysfs files
> will be created but getting the contents of event_list or trying to
> setup the counter will fail
>
> Instead, let's ignore and log unsupported blocks. This means the driver
> will always load and unsupported blocks will never show up in sysfs.
>
> Signed-off-by: Luiz Capitulino <[email protected]>
Thanks, patch looks good to me:
Reviewed-by: Hans de Goede <[email protected]>
Regards,
Hans
> ---
> drivers/platform/mellanox/mlxbf-pmc.c | 16 ++++++++++++++--
> 1 file changed, 14 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/platform/mellanox/mlxbf-pmc.c b/drivers/platform/mellanox/mlxbf-pmc.c
> index b71636eb3db1..746567767e5b 100644
> --- a/drivers/platform/mellanox/mlxbf-pmc.c
> +++ b/drivers/platform/mellanox/mlxbf-pmc.c
> @@ -1043,6 +1043,11 @@ static const struct mlxbf_pmc_events *mlxbf_pmc_event_list(const char *blk, size
> return events;
> }
>
> +static bool mlxbf_pmc_event_supported(const char *blk)
> +{
> + return !!mlxbf_pmc_event_list(blk, NULL);
> +}
> +
> /* Get the event number given the name */
> static int mlxbf_pmc_get_event_num(const char *blk, const char *evt)
> {
> @@ -1761,6 +1766,9 @@ static int mlxbf_pmc_init_perftype_counter(struct device *dev, unsigned int blk_
> struct mlxbf_pmc_attribute *attr;
> unsigned int i = 0, j = 0;
>
> + if (!mlxbf_pmc_event_supported(pmc->block_name[blk_num]))
> + return -ENOENT;
> +
> /* "event_list" sysfs to list events supported by the block */
> attr = &pmc->block[blk_num].attr_event_list;
> attr->dev_attr.attr.mode = 0444;
> @@ -1840,7 +1848,7 @@ static int mlxbf_pmc_init_perftype_reg(struct device *dev, unsigned int blk_num)
>
> events = mlxbf_pmc_event_list(pmc->block_name[blk_num], &count);
> if (!events)
> - return -EINVAL;
> + return -ENOENT;
>
> pmc->block[blk_num].attr_event = devm_kcalloc(
> dev, count, sizeof(struct mlxbf_pmc_attribute), GFP_KERNEL);
> @@ -1878,7 +1886,7 @@ static int mlxbf_pmc_create_groups(struct device *dev, unsigned int blk_num)
> else if (pmc->block[blk_num].type == MLXBF_PMC_TYPE_REGISTER)
> err = mlxbf_pmc_init_perftype_reg(dev, blk_num);
> else
> - err = -EINVAL;
> + err = -ENOENT;
>
> if (err)
> return err;
> @@ -1983,6 +1991,10 @@ static int mlxbf_pmc_map_counters(struct device *dev)
> return -ENOMEM;
>
> ret = mlxbf_pmc_create_groups(dev, i);
> + if (ret == -ENOENT) {
> + dev_warn(dev, "ignoring unsupported block: '%s'\n", pmc->block_name[i]);
> + continue;
> + }
> if (ret)
> return ret;
> }
On Thu, 22 Feb 2024 15:57:28 -0500, Luiz Capitulino wrote:
> The mlxbf-pmc driver fails to load when the firmware reports a new but not
> yet implemented performance block. I can reproduce this today with a
> Bluefield-3 card and UEFI version 4.6.0-18-g7d063bb-BId13035, since this
> reports the new clock_measure performance block.
>
> This[1] patch from Shravan implements the clock_measure support and will
> solve the issue. But this series avoids the situation by ignoring and
> logging unsupported performance blocks.
>
> [...]
Thank you for your contribution, it has been applied to my local
review-ilpo branch. Note it will show up in the public
platform-drivers-x86/review-ilpo branch only once I've pushed my
local branch there, which might take a while.
The list of commits applied:
[1/2] platform/mellanox: mlxbf-pmc: mlxbf_pmc_event_list(): make size ptr optional
commit: c5b649996ac63d43f1d4185de177c90d664b2230
[2/2] platform/mellanox: mlxbf-pmc: Ignore unsupported performance blocks
commit: 4e39d7be4123f65adf78b0a466cbaf1169d7cedb
--
i.
On 2024-02-26 08:27, Ilpo Järvinen wrote:
> On Thu, 22 Feb 2024 15:57:28 -0500, Luiz Capitulino wrote:
>
>> The mlxbf-pmc driver fails to load when the firmware reports a new but not
>> yet implemented performance block. I can reproduce this today with a
>> Bluefield-3 card and UEFI version 4.6.0-18-g7d063bb-BId13035, since this
>> reports the new clock_measure performance block.
>>
>> This[1] patch from Shravan implements the clock_measure support and will
>> solve the issue. But this series avoids the situation by ignoring and
>> logging unsupported performance blocks.
>>
>> [...]
>
>
> Thank you for your contribution, it has been applied to my local
> review-ilpo branch. Note it will show up in the public
> platform-drivers-x86/review-ilpo branch only once I've pushed my
> local branch there, which might take a while.
Thank you Ilpo and thanks Hans for the review.
The only detail is that we probably want this merged for 6.8 since
the driver doesn't currently load with the configuration mentioned above.
- Luiz
>
> The list of commits applied:
> [1/2] platform/mellanox: mlxbf-pmc: mlxbf_pmc_event_list(): make size ptr optional
> commit: c5b649996ac63d43f1d4185de177c90d664b2230
> [2/2] platform/mellanox: mlxbf-pmc: Ignore unsupported performance blocks
> commit: 4e39d7be4123f65adf78b0a466cbaf1169d7cedb
>
> --
> i.
>
>
On 2024-02-26 11:04, Ilpo Järvinen wrote:
> On Mon, 26 Feb 2024, Luiz Capitulino wrote:
>
>> On 2024-02-26 08:27, Ilpo Järvinen wrote:
>>> On Thu, 22 Feb 2024 15:57:28 -0500, Luiz Capitulino wrote:
>>>
>>>> The mlxbf-pmc driver fails to load when the firmware reports a new but not
>>>> yet implemented performance block. I can reproduce this today with a
>>>> Bluefield-3 card and UEFI version 4.6.0-18-g7d063bb-BId13035, since this
>>>> reports the new clock_measure performance block.
>>>>
>>>> This[1] patch from Shravan implements the clock_measure support and will
>>>> solve the issue. But this series avoids the situation by ignoring and
>>>> logging unsupported performance blocks.
>>>>
>>>> [...]
>>>
>>>
>>> Thank you for your contribution, it has been applied to my local
>>> review-ilpo branch. Note it will show up in the public
>>> platform-drivers-x86/review-ilpo branch only once I've pushed my
>>> local branch there, which might take a while.
>>
>> Thank you Ilpo and thanks Hans for the review.
>>
>> The only detail is that we probably want this merged for 6.8 since
>> the driver doesn't currently load with the configuration mentioned above.
>
> Oh, sorry, I missed the mention in the coverletter.
>
> So you'd want I drop these from review-ilpo branch as there they end
> up into for-next branch, and they should go through Hans instead who
> handles fixes branch for this cycle?
If that's the path to get this series merged for this cycle then yes,
but let's see if Hans agrees (sorry that I didn't know this before
posting).
One additional detail is that this series is on top of linux-next, which
has two additional mlxbf-pmc changes:
*
https://lore.kernel.org/lkml/39be055af3506ce6f843d11e45d71620f2a96e26.1707808180.git.shravankr@nvidia.com/
*
https://lore.kernel.org/lkml/d8548c70339a29258a906b2b518e5c48f669795c.1707808180.git.shravankr@nvidia.com/
Maybe those two should be included for 6.8 as well?
- Luiz
Hi Luiz,
On 2/26/24 17:10, Luiz Capitulino wrote:
> On 2024-02-26 11:04, Ilpo Järvinen wrote:
>> On Mon, 26 Feb 2024, Luiz Capitulino wrote:
>>
>>> On 2024-02-26 08:27, Ilpo Järvinen wrote:
>>>> On Thu, 22 Feb 2024 15:57:28 -0500, Luiz Capitulino wrote:
>>>>
>>>>> The mlxbf-pmc driver fails to load when the firmware reports a new but not
>>>>> yet implemented performance block. I can reproduce this today with a
>>>>> Bluefield-3 card and UEFI version 4.6.0-18-g7d063bb-BId13035, since this
>>>>> reports the new clock_measure performance block.
>>>>>
>>>>> This[1] patch from Shravan implements the clock_measure support and will
>>>>> solve the issue. But this series avoids the situation by ignoring and
>>>>> logging unsupported performance blocks.
>>>>>
>>>>> [...]
>>>>
>>>>
>>>> Thank you for your contribution, it has been applied to my local
>>>> review-ilpo branch. Note it will show up in the public
>>>> platform-drivers-x86/review-ilpo branch only once I've pushed my
>>>> local branch there, which might take a while.
>>>
>>> Thank you Ilpo and thanks Hans for the review.
>>>
>>> The only detail is that we probably want this merged for 6.8 since
>>> the driver doesn't currently load with the configuration mentioned above.
>>
>> Oh, sorry, I missed the mention in the coverletter.
>>
>> So you'd want I drop these from review-ilpo branch as there they end
>> up into for-next branch, and they should go through Hans instead who
>> handles fixes branch for this cycle?
>
> If that's the path to get this series merged for this cycle then yes,
> but let's see if Hans agrees (sorry that I didn't know this before
> posting).
Hmm, new hw enablement typically goes through -next and not to
the current fixes branch. And AFAICT this is new hw enablement,
not a regression / bug-fix.
Is there any special reason why this needs to be in 6.8 ?
For RHEL kernels you can cherry-pick patches from -next
as necessary.
> One additional detail is that this series is on top of linux-next, which
> has two additional mlxbf-pmc changes:
>
> * https://lore.kernel.org/lkml/39be055af3506ce6f843d11e45d71620f2a96e26.1707808180.git.shravankr@nvidia.com/
> * https://lore.kernel.org/lkml/d8548c70339a29258a906b2b518e5c48f669795c.1707808180.git.shravankr@nvidia.com/
Hmm, those are not small patches, any other reason
why this really should go to -next IMHO.
Regards,
Hans
On 2024-02-26 11:57, Hans de Goede wrote:
> Hi Luiz,
>
> On 2/26/24 17:10, Luiz Capitulino wrote:
>> On 2024-02-26 11:04, Ilpo Järvinen wrote:
>>> On Mon, 26 Feb 2024, Luiz Capitulino wrote:
>>>
>>>> On 2024-02-26 08:27, Ilpo Järvinen wrote:
>>>>> On Thu, 22 Feb 2024 15:57:28 -0500, Luiz Capitulino wrote:
>>>>>
>>>>>> The mlxbf-pmc driver fails to load when the firmware reports a new but not
>>>>>> yet implemented performance block. I can reproduce this today with a
>>>>>> Bluefield-3 card and UEFI version 4.6.0-18-g7d063bb-BId13035, since this
>>>>>> reports the new clock_measure performance block.
>>>>>>
>>>>>> This[1] patch from Shravan implements the clock_measure support and will
>>>>>> solve the issue. But this series avoids the situation by ignoring and
>>>>>> logging unsupported performance blocks.
>>>>>>
>>>>>> [...]
>>>>>
>>>>>
>>>>> Thank you for your contribution, it has been applied to my local
>>>>> review-ilpo branch. Note it will show up in the public
>>>>> platform-drivers-x86/review-ilpo branch only once I've pushed my
>>>>> local branch there, which might take a while.
>>>>
>>>> Thank you Ilpo and thanks Hans for the review.
>>>>
>>>> The only detail is that we probably want this merged for 6.8 since
>>>> the driver doesn't currently load with the configuration mentioned above.
>>>
>>> Oh, sorry, I missed the mention in the coverletter.
>>>
>>> So you'd want I drop these from review-ilpo branch as there they end
>>> up into for-next branch, and they should go through Hans instead who
>>> handles fixes branch for this cycle?
>>
>> If that's the path to get this series merged for this cycle then yes,
>> but let's see if Hans agrees (sorry that I didn't know this before
>> posting).
>
> Hmm, new hw enablement typically goes through -next and not to
> the current fixes branch. And AFAICT this is new hw enablement,
> not a regression / bug-fix.
>
> Is there any special reason why this needs to be in 6.8 ?
Since the new firmware feature is causing the driver not to load,
I'm seeing this more as a bug than new enablement. But it's fine
with me if you decide on not having them on 6.8.
> For RHEL kernels you can cherry-pick patches from -next
> as necessary.
I know :)
>> One additional detail is that this series is on top of linux-next, which
>> has two additional mlxbf-pmc changes:
>>
>> * https://lore.kernel.org/lkml/39be055af3506ce6f843d11e45d71620f2a96e26.1707808180.git.shravankr@nvidia.com/
>> * https://lore.kernel.org/lkml/d8548c70339a29258a906b2b518e5c48f669795c.1707808180.git.shravankr@nvidia.com/
>
> Hmm, those are not small patches, any other reason
> why this really should go to -next IMHO.
OK.
- Luiz
On 2024-02-26 11:59, Ilpo Järvinen wrote:
> On Mon, 26 Feb 2024, Luiz Capitulino wrote:
>
>> On 2024-02-26 11:04, Ilpo Järvinen wrote:
>>> On Mon, 26 Feb 2024, Luiz Capitulino wrote:
>>>
>>>> On 2024-02-26 08:27, Ilpo Järvinen wrote:
>>>>> On Thu, 22 Feb 2024 15:57:28 -0500, Luiz Capitulino wrote:
>>>>>
>>>>>> The mlxbf-pmc driver fails to load when the firmware reports a new but
>>>>>> not
>>>>>> yet implemented performance block. I can reproduce this today with a
>>>>>> Bluefield-3 card and UEFI version 4.6.0-18-g7d063bb-BId13035, since
>>>>>> this
>>>>>> reports the new clock_measure performance block.
>>>>>>
>>>>>> This[1] patch from Shravan implements the clock_measure support and
>>>>>> will
>>>>>> solve the issue. But this series avoids the situation by ignoring and
>>>>>> logging unsupported performance blocks.
>>>>>>
>>>>>> [...]
>>>>>
>>>>>
>>>>> Thank you for your contribution, it has been applied to my local
>>>>> review-ilpo branch. Note it will show up in the public
>>>>> platform-drivers-x86/review-ilpo branch only once I've pushed my
>>>>> local branch there, which might take a while.
>>>>
>>>> Thank you Ilpo and thanks Hans for the review.
>>>>
>>>> The only detail is that we probably want this merged for 6.8 since
>>>> the driver doesn't currently load with the configuration mentioned above.
>>>
>>> Oh, sorry, I missed the mention in the coverletter.
>>>
>>> So you'd want I drop these from review-ilpo branch as there they end
>>> up into for-next branch, and they should go through Hans instead who
>>> handles fixes branch for this cycle?
>>
>> If that's the path to get this series merged for this cycle then yes,
>> but let's see if Hans agrees (sorry that I didn't know this before
>> posting).
>>
>> One additional detail is that this series is on top of linux-next, which
>> has two additional mlxbf-pmc changes:
>>
>> *
>> https://lore.kernel.org/lkml/39be055af3506ce6f843d11e45d71620f2a96e26.1707808180.git.shravankr@nvidia.com/
>> *
>> https://lore.kernel.org/lkml/d8548c70339a29258a906b2b518e5c48f669795c.1707808180.git.shravankr@nvidia.com/
>>
>> Maybe those two should be included for 6.8 as well?
>
> Those look a new feature to me so they belong to for-next. So no, they
> will not end up into 6.8 (to fixes branch). If the 2 patches in this
> series do not apply without some for-next targetting dependencies, you
> should rebase on top of fixes branch and send a new version.
Understood.
> About those two patches, please also see my reply. I intentionally only 2
> patches of that series because I wanted to see sysfs documentation first
> so you should resend those two patches to for-next with sysfs
> documentation.
I'm actually not author of the other patches :)
- Luiz
On Mon, 26 Feb 2024, Luiz Capitulino wrote:
> On 2024-02-26 11:04, Ilpo J?rvinen wrote:
> > On Mon, 26 Feb 2024, Luiz Capitulino wrote:
> >
> > > On 2024-02-26 08:27, Ilpo J?rvinen wrote:
> > > > On Thu, 22 Feb 2024 15:57:28 -0500, Luiz Capitulino wrote:
> > > >
> > > > > The mlxbf-pmc driver fails to load when the firmware reports a new but
> > > > > not
> > > > > yet implemented performance block. I can reproduce this today with a
> > > > > Bluefield-3 card and UEFI version 4.6.0-18-g7d063bb-BId13035, since
> > > > > this
> > > > > reports the new clock_measure performance block.
> > > > >
> > > > > This[1] patch from Shravan implements the clock_measure support and
> > > > > will
> > > > > solve the issue. But this series avoids the situation by ignoring and
> > > > > logging unsupported performance blocks.
> > > > >
> > > > > [...]
> > > >
> > > >
> > > > Thank you for your contribution, it has been applied to my local
> > > > review-ilpo branch. Note it will show up in the public
> > > > platform-drivers-x86/review-ilpo branch only once I've pushed my
> > > > local branch there, which might take a while.
> > >
> > > Thank you Ilpo and thanks Hans for the review.
> > >
> > > The only detail is that we probably want this merged for 6.8 since
> > > the driver doesn't currently load with the configuration mentioned above.
> >
> > Oh, sorry, I missed the mention in the coverletter.
> >
> > So you'd want I drop these from review-ilpo branch as there they end
> > up into for-next branch, and they should go through Hans instead who
> > handles fixes branch for this cycle?
>
> If that's the path to get this series merged for this cycle then yes,
> but let's see if Hans agrees (sorry that I didn't know this before
> posting).
>
> One additional detail is that this series is on top of linux-next, which
> has two additional mlxbf-pmc changes:
>
> *
> https://lore.kernel.org/lkml/39be055af3506ce6f843d11e45d71620f2a96e26.1707808180.git.shravankr@nvidia.com/
> *
> https://lore.kernel.org/lkml/d8548c70339a29258a906b2b518e5c48f669795c.1707808180.git.shravankr@nvidia.com/
>
> Maybe those two should be included for 6.8 as well?
Those look a new feature to me so they belong to for-next. So no, they
will not end up into 6.8 (to fixes branch). If the 2 patches in this
series do not apply without some for-next targetting dependencies, you
should rebase on top of fixes branch and send a new version.
About those two patches, please also see my reply. I intentionally only 2
patches of that series because I wanted to see sysfs documentation first
so you should resend those two patches to for-next with sysfs
documentation.
--
i.
On Mon, 26 Feb 2024, Luiz Capitulino wrote:
> On 2024-02-26 08:27, Ilpo Järvinen wrote:
> > On Thu, 22 Feb 2024 15:57:28 -0500, Luiz Capitulino wrote:
> >
> > > The mlxbf-pmc driver fails to load when the firmware reports a new but not
> > > yet implemented performance block. I can reproduce this today with a
> > > Bluefield-3 card and UEFI version 4.6.0-18-g7d063bb-BId13035, since this
> > > reports the new clock_measure performance block.
> > >
> > > This[1] patch from Shravan implements the clock_measure support and will
> > > solve the issue. But this series avoids the situation by ignoring and
> > > logging unsupported performance blocks.
> > >
> > > [...]
> >
> >
> > Thank you for your contribution, it has been applied to my local
> > review-ilpo branch. Note it will show up in the public
> > platform-drivers-x86/review-ilpo branch only once I've pushed my
> > local branch there, which might take a while.
>
> Thank you Ilpo and thanks Hans for the review.
>
> The only detail is that we probably want this merged for 6.8 since
> the driver doesn't currently load with the configuration mentioned above.
Oh, sorry, I missed the mention in the coverletter.
So you'd want I drop these from review-ilpo branch as there they end
up into for-next branch, and they should go through Hans instead who
handles fixes branch for this cycle?
--
i.
On Mon, 26 Feb 2024, Hans de Goede wrote:
> Hi Luiz,
>
> On 2/26/24 17:10, Luiz Capitulino wrote:
> > On 2024-02-26 11:04, Ilpo J?rvinen wrote:
> >> On Mon, 26 Feb 2024, Luiz Capitulino wrote:
> >>
> >>> On 2024-02-26 08:27, Ilpo J?rvinen wrote:
> >>>> On Thu, 22 Feb 2024 15:57:28 -0500, Luiz Capitulino wrote:
> >>>>
> >>>>> The mlxbf-pmc driver fails to load when the firmware reports a new but not
> >>>>> yet implemented performance block. I can reproduce this today with a
> >>>>> Bluefield-3 card and UEFI version 4.6.0-18-g7d063bb-BId13035, since this
> >>>>> reports the new clock_measure performance block.
> >>>>>
> >>>>> This[1] patch from Shravan implements the clock_measure support and will
> >>>>> solve the issue. But this series avoids the situation by ignoring and
> >>>>> logging unsupported performance blocks.
> >>>>>
> >>>>> [...]
> >>>>
> >>>>
> >>>> Thank you for your contribution, it has been applied to my local
> >>>> review-ilpo branch. Note it will show up in the public
> >>>> platform-drivers-x86/review-ilpo branch only once I've pushed my
> >>>> local branch there, which might take a while.
> >>>
> >>> Thank you Ilpo and thanks Hans for the review.
> >>>
> >>> The only detail is that we probably want this merged for 6.8 since
> >>> the driver doesn't currently load with the configuration mentioned above.
> >>
> >> Oh, sorry, I missed the mention in the coverletter.
> >>
> >> So you'd want I drop these from review-ilpo branch as there they end
> >> up into for-next branch, and they should go through Hans instead who
> >> handles fixes branch for this cycle?
> >
> > If that's the path to get this series merged for this cycle then yes,
> > but let's see if Hans agrees (sorry that I didn't know this before
> > posting).
>
> Hmm, new hw enablement typically goes through -next and not to
> the current fixes branch. And AFAICT this is new hw enablement,
> not a regression / bug-fix.
>
> Is there any special reason why this needs to be in 6.8 ?
To me it sounded like fix to 1a218d312e65 ("platform/mellanox: mlxbf-pmc:
Add Mellanox BlueField PMC driver") and 423c3361855c ("platform/mellanox:
mlxbf-pmc: Add support for BlueField-3") although not explicitly marked as
such.
But I'm fine with taking these through for-next, it's relatively late into
the cycle already anyway.
> For RHEL kernels you can cherry-pick patches from -next
> as necessary.
It's also possible to send them later directly to stable folks once
Linus' tree has them after the next merge window if you feel they're
useful for stable inclusion.
> > One additional detail is that this series is on top of linux-next, which
> > has two additional mlxbf-pmc changes:
> >
> > * https://lore.kernel.org/lkml/39be055af3506ce6f843d11e45d71620f2a96e26.1707808180.git.shravankr@nvidia.com/
> > * https://lore.kernel.org/lkml/d8548c70339a29258a906b2b518e5c48f669795c.1707808180.git.shravankr@nvidia.com/
>
> Hmm, those are not small patches, any other reason
> why this really should go to -next IMHO.
Those two linked patches are totally unrelated.
--
i.
On Mon, 26 Feb 2024, Luiz Capitulino wrote:
> On 2024-02-26 11:59, Ilpo Järvinen wrote:
> > On Mon, 26 Feb 2024, Luiz Capitulino wrote:
> >
> > > On 2024-02-26 11:04, Ilpo Järvinen wrote:
> > > > On Mon, 26 Feb 2024, Luiz Capitulino wrote:
> > > >
> > > > > On 2024-02-26 08:27, Ilpo Järvinen wrote:
> > > > > > On Thu, 22 Feb 2024 15:57:28 -0500, Luiz Capitulino wrote:
> > > > > >
> > > > > > > The mlxbf-pmc driver fails to load when the firmware reports a new
> > > > > > > but
> > > > > > > not
> > > > > > > yet implemented performance block. I can reproduce this today with
> > > > > > > a
> > > > > > > Bluefield-3 card and UEFI version 4.6.0-18-g7d063bb-BId13035,
> > > > > > > since
> > > > > > > this
> > > > > > > reports the new clock_measure performance block.
> > > > > > >
> > > > > > > This[1] patch from Shravan implements the clock_measure support
> > > > > > > and
> > > > > > > will
> > > > > > > solve the issue. But this series avoids the situation by ignoring
> > > > > > > and
> > > > > > > logging unsupported performance blocks.
> > > > > > >
> > > > > > > [...]
> > > > > >
> > > > > >
> > > > > > Thank you for your contribution, it has been applied to my local
> > > > > > review-ilpo branch. Note it will show up in the public
> > > > > > platform-drivers-x86/review-ilpo branch only once I've pushed my
> > > > > > local branch there, which might take a while.
> > > > >
> > > > > Thank you Ilpo and thanks Hans for the review.
> > > > >
> > > > > The only detail is that we probably want this merged for 6.8 since
> > > > > the driver doesn't currently load with the configuration mentioned
> > > > > above.
> > > >
> > > > Oh, sorry, I missed the mention in the coverletter.
> > > >
> > > > So you'd want I drop these from review-ilpo branch as there they end
> > > > up into for-next branch, and they should go through Hans instead who
> > > > handles fixes branch for this cycle?
> > >
> > > If that's the path to get this series merged for this cycle then yes,
> > > but let's see if Hans agrees (sorry that I didn't know this before
> > > posting).
> > >
> > > One additional detail is that this series is on top of linux-next, which
> > > has two additional mlxbf-pmc changes:
> > >
> > > *
> > > https://lore.kernel.org/lkml/
39be055af3506ce6f843d11e45d71620f2a96e26.1707808180.git.shravankr@nvidia.com/
> > > *
> > > https://lore.kernel.org/lkml/d8548c70339a29258a906b2b518e5c48f669795c.1707808180.git.shravankr@nvidia.com/
> > >
> > > Maybe those two should be included for 6.8 as well?
> >
> > Those look a new feature to me so they belong to for-next. So no, they
> > will not end up into 6.8 (to fixes branch). If the 2 patches in this
> > series do not apply without some for-next targetting dependencies, you
> > should rebase on top of fixes branch and send a new version.
>
> Understood.
>
> > About those two patches, please also see my reply. I intentionally only 2
> > patches of that series because I wanted to see sysfs documentation first
> > so you should resend those two patches to for-next with sysfs
> > documentation.
>
> I'm actually not author of the other patches :)
Ah, sorry. I didn't pay enough attention to that. :-)
--
i.
On 2024-02-27 08:18, Ilpo Järvinen wrote:
> On Mon, 26 Feb 2024, Hans de Goede wrote:
>
>> Hi Luiz,
>>
>> On 2/26/24 17:10, Luiz Capitulino wrote:
>>> On 2024-02-26 11:04, Ilpo Järvinen wrote:
>>>> On Mon, 26 Feb 2024, Luiz Capitulino wrote:
>>>>
>>>>> On 2024-02-26 08:27, Ilpo Järvinen wrote:
>>>>>> On Thu, 22 Feb 2024 15:57:28 -0500, Luiz Capitulino wrote:
>>>>>>
>>>>>>> The mlxbf-pmc driver fails to load when the firmware reports a new but not
>>>>>>> yet implemented performance block. I can reproduce this today with a
>>>>>>> Bluefield-3 card and UEFI version 4.6.0-18-g7d063bb-BId13035, since this
>>>>>>> reports the new clock_measure performance block.
>>>>>>>
>>>>>>> This[1] patch from Shravan implements the clock_measure support and will
>>>>>>> solve the issue. But this series avoids the situation by ignoring and
>>>>>>> logging unsupported performance blocks.
>>>>>>>
>>>>>>> [...]
>>>>>>
>>>>>>
>>>>>> Thank you for your contribution, it has been applied to my local
>>>>>> review-ilpo branch. Note it will show up in the public
>>>>>> platform-drivers-x86/review-ilpo branch only once I've pushed my
>>>>>> local branch there, which might take a while.
>>>>>
>>>>> Thank you Ilpo and thanks Hans for the review.
>>>>>
>>>>> The only detail is that we probably want this merged for 6.8 since
>>>>> the driver doesn't currently load with the configuration mentioned above.
>>>>
>>>> Oh, sorry, I missed the mention in the coverletter.
>>>>
>>>> So you'd want I drop these from review-ilpo branch as there they end
>>>> up into for-next branch, and they should go through Hans instead who
>>>> handles fixes branch for this cycle?
>>>
>>> If that's the path to get this series merged for this cycle then yes,
>>> but let's see if Hans agrees (sorry that I didn't know this before
>>> posting).
>>
>> Hmm, new hw enablement typically goes through -next and not to
>> the current fixes branch. And AFAICT this is new hw enablement,
>> not a regression / bug-fix.
>>
>> Is there any special reason why this needs to be in 6.8 ?
>
> To me it sounded like fix to 1a218d312e65 ("platform/mellanox: mlxbf-pmc:
> Add Mellanox BlueField PMC driver") and 423c3361855c ("platform/mellanox:
> mlxbf-pmc: Add support for BlueField-3") although not explicitly marked as
> such.
>
> But I'm fine with taking these through for-next, it's relatively late into
> the cycle already anyway.
>
>> For RHEL kernels you can cherry-pick patches from -next
>> as necessary.
>
> It's also possible to send them later directly to stable folks once
> Linus' tree has them after the next merge window if you feel they're
> useful for stable inclusion.
Fair enough. Let's proceed with the original plan of having them merged
in the for-next branch. Sorry for the noise this discussion may have
caused.
- Luiz
>
>>> One additional detail is that this series is on top of linux-next, which
>>> has two additional mlxbf-pmc changes:
>>>
>>> * https://lore.kernel.org/lkml/39be055af3506ce6f843d11e45d71620f2a96e26.1707808180.git.shravankr@nvidia.com/
>>> * https://lore.kernel.org/lkml/d8548c70339a29258a906b2b518e5c48f669795c.1707808180.git.shravankr@nvidia.com/
>>
>> Hmm, those are not small patches, any other reason
>> why this really should go to -next IMHO.
>
> Those two linked patches are totally unrelated.
>
>