Re: current boot failure on NVME with and after 8e9fe9996a1f

From: Michael Butler <imb_at_protected-networks.net>
Date: Fri, 14 Aug 2026 12:32:16 UTC 
On 8/14/26 01:24, Kevin Bowling wrote:
> On Thu, Aug 13, 2026 at 12:51 PM Michael Butler
> <imb@protected-networks.net> wrote:
>> On a Dell Latitude 5540 with two NVME disks, the last working commit
>> from bisection is a9752e9ac8a6.
>>
>> Anything beyond this fails on start up with ..
>>
>> d5540 kernel: nvme1: Resetting controller due to a timeout.
>> d5540 kernel: nvme1: event="start"
>> d5540 kernel: nvme1: event="success"
>> d5540 kernel: nvme1: aborting outstanding i/o
>> d5540 kernel: nvme1: done aborting outstanding i/o
>>
>> It seems that the PCI changes don't work for disks attached in this
>> manner ..
> Hi Michael,
>
> I think the regression is 8e9fe9996a1f instead, specific to Intel VMD
> attached devices, which synthesize a virtual pcie endpoint.
>
> Can you try this on the failing kernel from the bootloader?
> set hw.pci.enable_mps_tune=0
> boot
>
> And then try this patch (without setting the bootloader option):
> https://reviews.freebsd.org/D58837

This patch restores normal behavour; thanks!

     Michael