lkml.org 
[lkml]   [2023]   [Mar]   [1]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: [PATCH 1/2] platform/x86/amd: pmc: Add a helper for checking minimum SMU version
From
On 3/1/23 20:50, Shyam Sundar S K wrote:
>
>
> On 3/2/2023 8:12 AM, Mario Limonciello wrote:
>> On 3/1/23 20:39, Shyam Sundar S K wrote:
>>> Hi Mario,
>>>
>>> On 3/1/2023 9:01 PM, Limonciello, Mario wrote:
>>>> [Public]
>>>>
>>>>
>>>>
>>>>> -----Original Message-----
>>>>> From: Hans de Goede <hdegoede@redhat.com>
>>>>> Sent: Wednesday, March 1, 2023 09:28
>>>>> To: Limonciello, Mario <Mario.Limonciello@amd.com>; S-k, Shyam-sundar
>>>>> <Shyam-sundar.S-k@amd.com>
>>>>> Cc: Mark Gross <markgross@kernel.org>; platform-driver-
>>>>> x86@vger.kernel.org; linux-kernel@vger.kernel.org
>>>>> Subject: Re: [PATCH 1/2] platform/x86/amd: pmc: Add a helper for
>>>>> checking
>>>>> minimum SMU version
>>>>>
>>>>> Hi,
>>>>>
>>>>> On 3/1/23 16:08, Mario Limonciello wrote:
>>>>>> In a few locations there is some boilerplate code for checking
>>>>>> minimum SMU version.  Switch this to a helper for this check.
>>>>>>
>>>>>> No intended functional changes.
>>>>>>
>>>>>> Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
>>>>>> ---
>>>>>>   drivers/platform/x86/amd/pmc.c | 49
>>>>>> +++++++++++++++++-----------------
>>>>>>   1 file changed, 24 insertions(+), 25 deletions(-)
>>>>>>
>>>>>> diff --git a/drivers/platform/x86/amd/pmc.c
>>>>> b/drivers/platform/x86/amd/pmc.c
>>>>>> index 2edaae04a691..c42fa47381c3 100644
>>>>>> --- a/drivers/platform/x86/amd/pmc.c
>>>>>> +++ b/drivers/platform/x86/amd/pmc.c
>>>>>> @@ -418,6 +418,22 @@ static int amd_pmc_get_smu_version(struct
>>>>> amd_pmc_dev *dev)
>>>>>>       return 0;
>>>>>>   }
>>>>>>
>>>>>> +static bool amd_pmc_verify_min_version(struct amd_pmc_dev *pdev,
>>>>> int major, int minor)
>>>>>> +{
>>>>>> +    if (!pdev->major) {
>>>>>> +        int rc = amd_pmc_get_smu_version(pdev);
>>>>>> +
>>>>>> +        if (rc) {
>>>>>> +            dev_warn(pdev->dev, "failed to read SMU version:
>>>>> %d\n", rc);
>>>>>> +            return false;
>>>>>> +        }
>>>>>> +    }
>>>>>> +    if (pdev->major > major)
>>>>>> +        return true;
>>>>>> +
>>>>>> +    return pdev->major == major && pdev->minor >= minor;
>>>>>> +}
>>>>>> +
>>>>>>   static ssize_t smu_fw_version_show(struct device *d, struct
>>>>> device_attribute *attr,
>>>>>>                      char *buf)
>>>>>>   {
>>>>>> @@ -526,14 +542,7 @@ static int amd_pmc_idlemask_show(struct seq_file
>>>>> *s, void *unused)
>>>>>>       struct amd_pmc_dev *dev = s->private;
>>>>>>       int rc;
>>>>>>
>>>>>> -    /* we haven't yet read SMU version */
>>>>>> -    if (!dev->major) {
>>>>>> -        rc = amd_pmc_get_smu_version(dev);
>>>>>> -        if (rc)
>>>>>> -            return rc;
>>>>>> -    }
>>>>>> -
>>>>>> -    if (dev->major > 56 || (dev->major >= 55 && dev->minor >= 37)) {
>>>>>
>>>>> The 2 major checks here originally were not in sync, so for major == 55
>>>>> *and* major == 56 it would only succeed if minor >= 37.
>>>>>
>>>>> Where as after this patch for major == 56 it will now always succeed.
>>>>>
>>>>> This feels like a bug in the original code, but might have been
>>>>> intentional ? Please verify this.
>>>>
>>>> @S-k, Shyam-sundar as the original author of that, can you please
>>>> confirm?
>>>
>>> I cannot completely recall :-) It was something like if the major
>>> version is greater than 56, there is no need to check the other part of
>>> the "OR".
>>>
>>> which is kind of similar to what you are now doing in
>>> amd_pmc_verify_min_version().
>>
>> OK yeah, then I'll split this correction of the logic off to that in a
>> separate patch to make this one "really no intended functional changes".
>>
>>>
>>> Like we discussed off-list, we should have this boilerplate in place, so
>>> that the future checks would not be duplicated.
>>
>> Something else I noticed that we probably need to consider is that there
>> is no examination for the "program" version which may be important.
>>
>> We don't have any version checks for YC, but if we did for example YC A0
>> and YC B0 use program "0" or program "4" respectively so version checks
>> could fall over.
>
> Checking for "program" version may not be required as A0/B0 are never
> meant for production and IMO its a logical overhead.
>
> Do you have a specific case, were you felt the real usage of "program"
> version?

A0/B0 of YC was just one example. The reason I mention it is I think
technically it's possible that A0 is leveraged for an embedded product.

I know similar thing had occurred for Renoir leveraged products too.

>
>>
>> I'll add something like this in for v2 of the patch as well.
>>
>>>
>>> Thanks,
>>> Shyam
>>>
>>>>
>>>>>
>>>>> After verifying please post a v2 updating the commit message to
>>>>> point out this functional change.
>>>>>
>>>>
>>>> Sure, thanks.
>>>>
>>>>>> +    if (amd_pmc_verify_min_version(dev, 55, 37)) {
>>>>>>           rc = amd_pmc_idlemask_read(dev, NULL, s);
>>>>>>           if (rc)
>>>>>>               return rc;
>>>>>> @@ -686,15 +695,8 @@ static int amd_pmc_get_os_hint(struct
>>>>> amd_pmc_dev *dev)
>>>>>>   static int amd_pmc_czn_wa_irq1(struct amd_pmc_dev *pdev)
>>>>>>   {
>>>>>>       struct device *d;
>>>>>> -    int rc;
>>>>>>
>>>>>> -    if (!pdev->major) {
>>>>>> -        rc = amd_pmc_get_smu_version(pdev);
>>>>>> -        if (rc)
>>>>>> -            return rc;
>>>>>> -    }
>>>>>> -
>>>>>> -    if (pdev->major > 64 || (pdev->major == 64 && pdev->minor > 65))
>>>>>> +    if (amd_pmc_verify_min_version(pdev, 64, 66))
>>>>>>           return 0;
>>>>>>
>>>>>>       d = bus_find_device_by_name(&serio_bus, NULL, "serio0");
>>>>>> @@ -718,14 +720,10 @@ static int amd_pmc_verify_czn_rtc(struct
>>>>> amd_pmc_dev *pdev, u32 *arg)
>>>>>>       struct rtc_time tm;
>>>>>>       int rc;
>>>>>>
>>>>>> -    /* we haven't yet read SMU version */
>>>>>> -    if (!pdev->major) {
>>>>>> -        rc = amd_pmc_get_smu_version(pdev);
>>>>>> -        if (rc)
>>>>>> -            return rc;
>>>>>> -    }
>>>>>> +    if (disable_workarounds)
>>>>>> +        return 0;
>>>>>>
>>>>>> -    if (pdev->major < 64 || (pdev->major == 64 && pdev->minor < 53))
>>>>>> +    if (!amd_pmc_verify_min_version(pdev, 64, 53))
>>>>>>           return 0;
>>>>>>
>>>>>>       rtc_device = rtc_class_open("rtc0");
>>>>>> @@ -772,13 +770,14 @@ static void amd_pmc_s2idle_prepare(void)
>>>>>>       /* Reset and Start SMU logging - to monitor the s0i3 stats */
>>>>>>       amd_pmc_setup_smu_logging(pdev);
>>>>>>
>>>>>> -    /* Activate CZN specific platform bug workarounds */
>>>>>> -    if (pdev->cpu_id == AMD_CPU_ID_CZN && !disable_workarounds) {
>>>>>> +    switch (pdev->cpu_id) {
>>>>>> +    case AMD_CPU_ID_CZN:
>>>>>>           rc = amd_pmc_verify_czn_rtc(pdev, &arg);
>>>>>>           if (rc) {
>>>>>>               dev_err(pdev->dev, "failed to set RTC: %d\n", rc);
>>>>>>               return;
>>>>>>           }
>>>>>> +        break;
>>>>>>       }
>>>>>>
>>>>>>       msg = amd_pmc_get_os_hint(pdev);
>>>>>
>>>>>
>>>>> Patch 2/2 looks good to me.
>>>>>
>>>>> Should I queue v2 (once posted) up as a fix for 6.3-rc#  ?
>>>>
>>>> Yes please.  If it makes it easier I can re-order the series so that
>>>> we add a check in 1/2 and switch to the helper as 2/2.
>>>>
>>>> This might make it easier to take the LTS kernel too for stable,
>>>> but I don't feel strongly.
>>>>
>>>>>
>>>>> Regards,
>>>>>
>>>>> Hans
>>

\
 
 \ /
  Last update: 2023-03-27 00:41    [W:0.089 / U:0.124 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site