Dear Linux folks,
Am 05.11.21 um 12:56 schrieb Paul Menzel:
On a PowerEdge T440/021KCD, BIOS 2.11.2 04/22/2021, Linux 5.10.70 takes
almost five seconds to initialize PCI. According to the timestamps, 1.5
s are from assigning the PCI devices to the 142 IOMMU groups.
```
$ lspci | wc -l
281
$ dmesg
[…]
[ 2.918411] PCI: Using host bridge windows from ACPI; if necessary, use "pci=nocrs" and report a bug
[ 2.933841] ACPI: Enabled 5 GPEs in block 00 to 7F
[ 2.973739] ACPI: PCI Root Bridge [PC00] (domain 0000 [bus 00-16])
[ 2.980398] acpi PNP0A08:00: _OSC: OS supports [ExtendedConfig ASPM ClockPM Segments MSI HPX-Type3]
[ 2.989457] acpi PNP0A08:00: _OSC: platform does not support [LTR]
[ 2.995451] acpi PNP0A08:00: _OSC: OS now controls [PME PCIeCapability]
[ 3.001394] acpi PNP0A08:00: FADT indicates ASPM is unsupported,
using BIOS configuration
[ 3.010511] PCI host bridge to bus 0000:00
[…]
[ 6.233508] system 00:05: [io 0x1000-0x10fe] has been reserved
[ 6.239420] system 00:05: Plug and Play ACPI device, IDs PNP0c02 (active)
[ 6.239906] pnp: PnP ACPI: found 6 devices
[…]
[ 6.989016] pci 0000:d7:05.0: disabled boot interrupts on device [8086:2034]
[ 6.996063] PCI: CLS 0 bytes, default 64
[ 7.000008] Trying to unpack rootfs image as initramfs...
[ 7.065281] Freeing initrd memory: 5136K
[…]
[ 7.079098] DMAR: dmar7: Using Queued invalidation
[ 7.083983] pci 0000:00:00.0: Adding to iommu group 0
[…]
[ 8.537808] pci 0000:d7:17.1: Adding to iommu group 141
[ 8.571191] DMAR: Intel(R) Virtualization Technology for Directed I/O
[ 8.577618] PCI-DMA: Using software bounce buffering for IO (SWIOTLB)
[…]
```
Is there anything that could be done to reduce the time?
I created an issue at the Kernel.org Bugzilla, and attached the output
of `dmesg` there [1].
Kind regards,
Paul
[1]: https://bugzilla.kernel.org/show_bug.cgi?id=214953
