| Message ID | 20231226062612.97313-1-chengen.du@canonical.com |
|---|---|
| Headers | show |
| Series | Hotplugging SCSI disk in QEMU VM fails | expand |
On 12/25/23 11:26 PM, Chengen Du wrote: > BugLink: https://bugs.launchpad.net/bugs/2047382 > > SRU Justification: > > [Impact] > Hot-plugging SCSI disks for QEMU VMs booting with UEFI and using guest kernels >= 6.5 might be broken. > The inconsistency suggests a potential race condition. > > [Fix] > Reverting the two specified commits appears to restore reliability. > ========== > cc22522fd55e2 PCI: acpiphp: Use pci_assign_unassigned_bridge_resources() only for non-root bus > 40613da52b13f PCI: acpiphp: Reassign resources on bridge if necessary > ========== > > Upstream has resolved the issue by reverting both commits through a commit placed below. > ========== > 5df12742b7e3a Revert "PCI: acpiphp: Reassign resources on bridge if necessary > ========== > > [Test Plan] > 1. Create a VM > virt-install --name scsi_test --memory 2048 --vcpus 2 --boot uefi --disk path=/home/ubuntu/mantic-server-cloudimg-amd64.img,size=20,format=qcow2,bus=virtio -os-variant=ubuntu18.04 --graphics spice --network bridge=virbr0 > 2. Prepare two XMLs for adding two disk devices > # cat disk_X.xml > <disk type='file' device='disk'> > <driver name='qemu' type='raw'/> > <source file='XXX.img'/> > <target dev='vdX' bus='virtio'/> > </disk> > 3. Hotplug disks and check for success > virsh attach-device scsi_test disk_X.xml > > [Where problems could occur] > The patch reverts two commits that previously addressed long-standing issues related to hotplugging devices with large BARs. > While this reversion may reintroduce legacy issues, it does not introduce new problems. > The associated risks are deemed low, making it worthwhile to address timing issues during the hotplugging of SCSI disks in QEMU VM. > > Bjorn Helgaas (1): > Revert "PCI: acpiphp: Reassign resources on bridge if necessary" > > drivers/pci/hotplug/acpiphp_glue.c | 9 +++------ > 1 file changed, 3 insertions(+), 6 deletions(-) > You mention reverting 2 patches, but I only see one.
Hi Tim, Upstream has reverted two commits in a single commit, and you can find it here: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=5df12742b7e3aae2594a30a9d14d5d6e9e7699f4 I have directly backported this commit to address the revert of the related commits. Best regards, Chengen Du On Thu, Jan 4, 2024 at 12:54 AM Tim Gardner <tim.gardner@canonical.com> wrote: > > On 12/25/23 11:26 PM, Chengen Du wrote: > > BugLink: https://bugs.launchpad.net/bugs/2047382 > > > > SRU Justification: > > > > [Impact] > > Hot-plugging SCSI disks for QEMU VMs booting with UEFI and using guest kernels >= 6.5 might be broken. > > The inconsistency suggests a potential race condition. > > > > [Fix] > > Reverting the two specified commits appears to restore reliability. > > ========== > > cc22522fd55e2 PCI: acpiphp: Use pci_assign_unassigned_bridge_resources() only for non-root bus > > 40613da52b13f PCI: acpiphp: Reassign resources on bridge if necessary > > ========== > > > > Upstream has resolved the issue by reverting both commits through a commit placed below. > > ========== > > 5df12742b7e3a Revert "PCI: acpiphp: Reassign resources on bridge if necessary > > ========== > > > > [Test Plan] > > 1. Create a VM > > virt-install --name scsi_test --memory 2048 --vcpus 2 --boot uefi --disk path=/home/ubuntu/mantic-server-cloudimg-amd64.img,size=20,format=qcow2,bus=virtio -os-variant=ubuntu18.04 --graphics spice --network bridge=virbr0 > > 2. Prepare two XMLs for adding two disk devices > > # cat disk_X.xml > > <disk type='file' device='disk'> > > <driver name='qemu' type='raw'/> > > <source file='XXX.img'/> > > <target dev='vdX' bus='virtio'/> > > </disk> > > 3. Hotplug disks and check for success > > virsh attach-device scsi_test disk_X.xml > > > > [Where problems could occur] > > The patch reverts two commits that previously addressed long-standing issues related to hotplugging devices with large BARs. > > While this reversion may reintroduce legacy issues, it does not introduce new problems. > > The associated risks are deemed low, making it worthwhile to address timing issues during the hotplugging of SCSI disks in QEMU VM. > > > > Bjorn Helgaas (1): > > Revert "PCI: acpiphp: Reassign resources on bridge if necessary" > > > > drivers/pci/hotplug/acpiphp_glue.c | 9 +++------ > > 1 file changed, 3 insertions(+), 6 deletions(-) > > > > You mention reverting 2 patches, but I only see one. > -- > ----------- > Tim Gardner > Canonical, Inc >
On 1/3/24 7:42 PM, Chengen Du wrote: > Hi Tim, > > Upstream has reverted two commits in a single commit, and you can find it here: > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=5df12742b7e3aae2594a30a9d14d5d6e9e7699f4 > > I have directly backported this commit to address the revert of the > related commits. > Indeed. Must read for content... > Best regards, > Chengen Du > > On Thu, Jan 4, 2024 at 12:54 AM Tim Gardner <tim.gardner@canonical.com> wrote: >> >> On 12/25/23 11:26 PM, Chengen Du wrote: >>> BugLink: https://bugs.launchpad.net/bugs/2047382 >>> >>> SRU Justification: >>> >>> [Impact] >>> Hot-plugging SCSI disks for QEMU VMs booting with UEFI and using guest kernels >= 6.5 might be broken. >>> The inconsistency suggests a potential race condition. >>> >>> [Fix] >>> Reverting the two specified commits appears to restore reliability. >>> ========== >>> cc22522fd55e2 PCI: acpiphp: Use pci_assign_unassigned_bridge_resources() only for non-root bus >>> 40613da52b13f PCI: acpiphp: Reassign resources on bridge if necessary >>> ========== >>> >>> Upstream has resolved the issue by reverting both commits through a commit placed below. >>> ========== >>> 5df12742b7e3a Revert "PCI: acpiphp: Reassign resources on bridge if necessary >>> ========== >>> >>> [Test Plan] >>> 1. Create a VM >>> virt-install --name scsi_test --memory 2048 --vcpus 2 --boot uefi --disk path=/home/ubuntu/mantic-server-cloudimg-amd64.img,size=20,format=qcow2,bus=virtio -os-variant=ubuntu18.04 --graphics spice --network bridge=virbr0 >>> 2. Prepare two XMLs for adding two disk devices >>> # cat disk_X.xml >>> <disk type='file' device='disk'> >>> <driver name='qemu' type='raw'/> >>> <source file='XXX.img'/> >>> <target dev='vdX' bus='virtio'/> >>> </disk> >>> 3. Hotplug disks and check for success >>> virsh attach-device scsi_test disk_X.xml >>> >>> [Where problems could occur] >>> The patch reverts two commits that previously addressed long-standing issues related to hotplugging devices with large BARs. >>> While this reversion may reintroduce legacy issues, it does not introduce new problems. >>> The associated risks are deemed low, making it worthwhile to address timing issues during the hotplugging of SCSI disks in QEMU VM. >>> >>> Bjorn Helgaas (1): >>> Revert "PCI: acpiphp: Reassign resources on bridge if necessary" >>> >>> drivers/pci/hotplug/acpiphp_glue.c | 9 +++------ >>> 1 file changed, 3 insertions(+), 6 deletions(-) >>> >> >> You mention reverting 2 patches, but I only see one. >> -- >> ----------- >> Tim Gardner >> Canonical, Inc >> Acked-by: Tim Gardner <tim.gardner@canonical.com>
On Tue, Dec 26, 2023 at 02:26:11PM +0800, Chengen Du wrote: > BugLink: https://bugs.launchpad.net/bugs/2047382 > > SRU Justification: > > [Impact] > Hot-plugging SCSI disks for QEMU VMs booting with UEFI and using guest kernels >= 6.5 might be broken. > The inconsistency suggests a potential race condition. > > [Fix] > Reverting the two specified commits appears to restore reliability. > ========== > cc22522fd55e2 PCI: acpiphp: Use pci_assign_unassigned_bridge_resources() only for non-root bus > 40613da52b13f PCI: acpiphp: Reassign resources on bridge if necessary > ========== > > Upstream has resolved the issue by reverting both commits through a commit placed below. > ========== > 5df12742b7e3a Revert "PCI: acpiphp: Reassign resources on bridge if necessary > ========== > > [Test Plan] > 1. Create a VM > virt-install --name scsi_test --memory 2048 --vcpus 2 --boot uefi --disk path=/home/ubuntu/mantic-server-cloudimg-amd64.img,size=20,format=qcow2,bus=virtio -os-variant=ubuntu18.04 --graphics spice --network bridge=virbr0 > 2. Prepare two XMLs for adding two disk devices > # cat disk_X.xml > <disk type='file' device='disk'> > <driver name='qemu' type='raw'/> > <source file='XXX.img'/> > <target dev='vdX' bus='virtio'/> > </disk> > 3. Hotplug disks and check for success > virsh attach-device scsi_test disk_X.xml > > [Where problems could occur] > The patch reverts two commits that previously addressed long-standing issues related to hotplugging devices with large BARs. > While this reversion may reintroduce legacy issues, it does not introduce new problems. > The associated risks are deemed low, making it worthwhile to address timing issues during the hotplugging of SCSI disks in QEMU VM. > > Bjorn Helgaas (1): > Revert "PCI: acpiphp: Reassign resources on bridge if necessary" > > drivers/pci/hotplug/acpiphp_glue.c | 9 +++------ > 1 file changed, 3 insertions(+), 6 deletions(-) > > -- > 2.40.1 > > > -- > kernel-team mailing list > kernel-team@lists.ubuntu.com > https://lists.ubuntu.com/mailman/listinfo/kernel-team Acked-by: Manuel Diewald <manuel.diewald@canonical.com>
On 26/12/2023 07:26, Chengen Du wrote: > BugLink: https://bugs.launchpad.net/bugs/2047382 > > SRU Justification: > > [Impact] > Hot-plugging SCSI disks for QEMU VMs booting with UEFI and using guest kernels >= 6.5 might be broken. > The inconsistency suggests a potential race condition. > > [Fix] > Reverting the two specified commits appears to restore reliability. > ========== > cc22522fd55e2 PCI: acpiphp: Use pci_assign_unassigned_bridge_resources() only for non-root bus > 40613da52b13f PCI: acpiphp: Reassign resources on bridge if necessary > ========== > > Upstream has resolved the issue by reverting both commits through a commit placed below. > ========== > 5df12742b7e3a Revert "PCI: acpiphp: Reassign resources on bridge if necessary > ========== > > [Test Plan] > 1. Create a VM > virt-install --name scsi_test --memory 2048 --vcpus 2 --boot uefi --disk path=/home/ubuntu/mantic-server-cloudimg-amd64.img,size=20,format=qcow2,bus=virtio -os-variant=ubuntu18.04 --graphics spice --network bridge=virbr0 > 2. Prepare two XMLs for adding two disk devices > # cat disk_X.xml > <disk type='file' device='disk'> > <driver name='qemu' type='raw'/> > <source file='XXX.img'/> > <target dev='vdX' bus='virtio'/> > </disk> > 3. Hotplug disks and check for success > virsh attach-device scsi_test disk_X.xml > > [Where problems could occur] > The patch reverts two commits that previously addressed long-standing issues related to hotplugging devices with large BARs. > While this reversion may reintroduce legacy issues, it does not introduce new problems. > The associated risks are deemed low, making it worthwhile to address timing issues during the hotplugging of SCSI disks in QEMU VM. > > Bjorn Helgaas (1): > Revert "PCI: acpiphp: Reassign resources on bridge if necessary" > > drivers/pci/hotplug/acpiphp_glue.c | 9 +++------ > 1 file changed, 3 insertions(+), 6 deletions(-) > Applied to mantic, lunar, jammy, focal master-next branch. Thanks!
On 26.12.23 07:26, Chengen Du wrote: > BugLink: https://bugs.launchpad.net/bugs/2047382 > > SRU Justification: > > [Impact] > Hot-plugging SCSI disks for QEMU VMs booting with UEFI and using guest kernels >= 6.5 might be broken. > The inconsistency suggests a potential race condition. > > [Fix] > Reverting the two specified commits appears to restore reliability. > ========== > cc22522fd55e2 PCI: acpiphp: Use pci_assign_unassigned_bridge_resources() only for non-root bus > 40613da52b13f PCI: acpiphp: Reassign resources on bridge if necessary > ========== > > Upstream has resolved the issue by reverting both commits through a commit placed below. > ========== > 5df12742b7e3a Revert "PCI: acpiphp: Reassign resources on bridge if necessary > ========== > > [Test Plan] > 1. Create a VM > virt-install --name scsi_test --memory 2048 --vcpus 2 --boot uefi --disk path=/home/ubuntu/mantic-server-cloudimg-amd64.img,size=20,format=qcow2,bus=virtio -os-variant=ubuntu18.04 --graphics spice --network bridge=virbr0 > 2. Prepare two XMLs for adding two disk devices > # cat disk_X.xml > <disk type='file' device='disk'> > <driver name='qemu' type='raw'/> > <source file='XXX.img'/> > <target dev='vdX' bus='virtio'/> > </disk> > 3. Hotplug disks and check for success > virsh attach-device scsi_test disk_X.xml > > [Where problems could occur] > The patch reverts two commits that previously addressed long-standing issues related to hotplugging devices with large BARs. > While this reversion may reintroduce legacy issues, it does not introduce new problems. > The associated risks are deemed low, making it worthwhile to address timing issues during the hotplugging of SCSI disks in QEMU VM. > > Bjorn Helgaas (1): > Revert "PCI: acpiphp: Reassign resources on bridge if necessary" > > drivers/pci/hotplug/acpiphp_glue.c | 9 +++------ > 1 file changed, 3 insertions(+), 6 deletions(-) > Rejected for the following reasons: - The change is already included in Noble with 6.7. -Stefan