linux-kernel - Re: How to reduce PCI initialization from 5 s (1.5 s adding them to IOMMU groups)

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20211105185304.GA936068@bhelgaas>
Date:   Fri, 5 Nov 2021 13:53:04 -0500
From:   Bjorn Helgaas <helgaas@...nel.org>
To:     Paul Menzel <pmenzel@...gen.mpg.de>
Cc:     Jörg Rödel <joro@...tes.org>,
        Suravee Suthikulpanit <suravee.suthikulpanit@....com>,
        Bjorn Helgaas <bhelgaas@...gle.com>,
        iommu@...ts.linux-foundation.org,
        Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...hat.com>, Borislav Petkov <bp@...en8.de>,
        x86@...nel.org, LKML <linux-kernel@...r.kernel.org>,
        linux-pci@...r.kernel.org
Subject: Re: How to reduce PCI initialization from 5 s (1.5 s adding them to
 IOMMU groups)

On Fri, Nov 05, 2021 at 12:56:09PM +0100, Paul Menzel wrote:
> Dear Linux folks,
> 
> 
> 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

For ~280 PCI devices, (6.24-2.92)/280 = 0.012 s/dev.  On my laptop I
have about (.66-.37)/36 = 0.008 s/dev (on v5.4), so about the same
ballpark.

Faster would always be better, of course.  I assume this is not really
a regression?

> [    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

I don't have this iommu stuff turned on and don't know what's
happening here.

> Is there anything that could be done to reduce the time?
> 
> 
> Kind regards,
> 
> Paul