On 2024-10-10 6:10 a.m., Jinpu Wang wrote:
> Hi Greg,
>
>
> On Thu, Oct 10, 2024 at 11:31 AM Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
>>
>> On Thu, Oct 10, 2024 at 11:13:42AM +0200, Jinpu Wang wrote:
>>> Hi Greg,
>>>
>>> On Thu, Oct 10, 2024 at 11:07 AM Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
>>>>
>>>> On Thu, Oct 10, 2024 at 09:31:37AM +0200, Jinpu Wang wrote:
>>>>> Hello all,
>>>>>
>>>>> We are experiencing a boot hang issue when booting kernel version
>>>>> 6.1.83+ on a Dell Inc. PowerEdge R770 equipped with an Intel Xeon
>>>>> 6710E processor. After extensive testing and use of `git bisect`, we
>>>>> have traced the issue to commit:
>>>>>
>>>>> `586e19c88a0c ("iommu/vt-d: Retrieve IOMMU perfmon capability information")`
>>>>>
>>>>> This commit appears to be part of a larger patchset, which can be found here:
>>>>> [Patchset on lore.kernel.org](https://lore.kernel.org/lkml/7c4b3e4e-1c5d-04f1-1891-84f686c94736@xxxxxxxxxxxxxxx/T/)
>>>>>
>>>>> We attempted to boot with the `intel_iommu=off` option, but the system
>>>>> hangs in the same manner. However, the system boots successfully after
>>>>> disabling `CONFIG_INTEL_IOMMU_PERF_EVENTS`.
>>>>
>>>> Is there any error messages? Does the latest 6.6.y tree work properly?
>>>> If so, why not just use that, no new hardware should be using older
>>>> kernel trees anyway :)
>>> No error, just hang, I've removed "quiet" and added "debug".
>>> Yes, the latest 6.6.y tree works for this, but there are other
>>> problems/dependency we have to solve.
>>
>> Ok, that implies that we need to add some other patch to 6.1.y, OR we
>> can revert it from 6.1.y. Let me know what you think is the better
>> thing to do.
>>
> I think better to revert both:
> 8c91a4bfc7f8 ("iommu: Fix compilation without CONFIG_IOMMU_INTEL")
I'm not sure about this one. May need baolu's comments.
> 586e19c88a0c ("iommu/vt-d: Retrieve IOMMU perfmon capability information")
>
7 patches are required to enable the IOMMU perfmon.
https://lore.kernel.org/all/20230128200428.1459118-1-kan.liang@xxxxxxxxxxxxxxx/
But it looks like only the above 1 patch is back ported to the 6.1.y.
Unless we can back port the rest of 6 patches, I think it should be ok
to revert it for 6.1.y.
Thanks,
Kan
> unless other guys have a different opinon.
>> thanks,
>>
>> greg k-h
> Thanks!
