On 21.07.23 18:04, Greg Kroah-Hartman wrote:
From: Igor Mammedov imammedo@redhat.com
commit 40613da52b13fb21c5566f10b287e0ca8c12c4e9 upstream.
When using ACPI PCI hotplug, hotplugging a device with large BARs may fail if bridge windows programmed by firmware are not large enough.
Reproducer: $ qemu-kvm -monitor stdio -M q35 -m 4G \ -global ICH9-LPC.acpi-pci-hotplug-with-bridge-support=on \ -device id=rp1,pcie-root-port,bus=pcie.0,chassis=4 \ disk_image
wait till linux guest boots, then hotplug device: (qemu) device_add qxl,bus=rp1
[...]
Greg, just so you know, that patch (which is also queued for 6.1 and 5.15) is known to cause a regression in 6.5-rc. To quote https://lore.kernel.org/all/11fc981c-af49-ce64-6b43-3e282728bd1a@gmail.com/
``` Laptop shows a kernel crash trace after a first suspend to ram, on a second attempt to suspend it becomes frozen solid. This is 100% repeatable with a 6.5-rc2 kernel, not happening with a 6.4 kernel - see the attached dmesg output.
I have bisected the kernel uilds and it points to : [40613da52b13fb21c5566f10b287e0ca8c12c4e9] PCI: acpiphp: Reassign resources on bridge if necessary
Reversing this patch seems to fix the kernel crash problem on my laptop. ```
Ciao, Thorsten
P.S.: I only noticed this by chance and yet again wonder how to handle these situations better. I guess exporting the list of regressions regzbot tracks in some simple format might be a start; then the stable scripts could simply look up commit ids there when a patch is queued and warn if they find it (which won't help in caes the regression is reported after the patch is queued :-/ ).