https://lists.ubuntu.com/archives/kernel-team/2022-January/127401.html
** Description changed: [Impact] There is an issue with hypervisor host side mlx5 driver operation on Bluefield devices in kernels <=5.16 From patch description: When using libvirt to passthrough VF to VM it will always set the VF vlan to 0 even if user didn’t request it, this will cause libvirt to fail to boot in case the PF isn't eswitch owner. Example of such case is the DPU host PF which isn't eswitch manager, so any attempt to passthrough VF of it using libvirt will fail. [Fix] * 7846665d net/mlx5e: Unblock setting vid 0 for VF in case PF isn't - eswitch manager + eswitch manager This patch can be cherry-picked into Impish and newer kernels, but it requires some backport work on F/H due to they're missing commit b55b3538 that split the legacy code out. + + The ESW_ALLOWED marco was later converted into a helper routine in + commit b16f2bb6. [Test] Test kernels can be found here: F: https://people.canonical.com/~phlin/kernel/lp-1957753-mlx5e/F/ H: https://people.canonical.com/~phlin/kernel/lp-1957753-mlx5e/H/ I: https://people.canonical.com/~phlin/kernel/lp-1957753-mlx5e/I/ J: https://people.canonical.com/~phlin/kernel/lp-1957753-mlx5e/J/ Kernels tested on a SUT with this device by Dmitrii Shcherbakov, test passed with positive feedback. [Where problems could occur] If this patch is erroneous, it might cause issue to this mlx5 driver with certain operations (setting vid 0). And other programs such as libvirt that depends on this return value might be affected as well. - [Original Bug Description] There is an issue with hypervisor host side mlx5 driver operation on Bluefield devices in kernels <=5.16 that was recently fixed in master with a one-liner: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=7846665d3504812acaebf920d1141851379a7f37 It would be good to have this fix in Focal+ kernels. This is not needed in https://bugs.launchpad.net/ubuntu/+source/linux- bluefield since it's used at the DPU side whereas the issue in question affects the hypervisor side. -- You received this bug notification because you are a member of Kernel Packages, which is subscribed to linux in Ubuntu. https://bugs.launchpad.net/bugs/1957753 Title: net/mlx5e: EPERM on vlan 0 programming Status in linux package in Ubuntu: In Progress Status in linux source package in Focal: In Progress Status in linux source package in Hirsute: In Progress Status in linux source package in Impish: In Progress Status in linux source package in Jammy: In Progress Bug description: [Impact] There is an issue with hypervisor host side mlx5 driver operation on Bluefield devices in kernels <=5.16 From patch description: When using libvirt to passthrough VF to VM it will always set the VF vlan to 0 even if user didn’t request it, this will cause libvirt to fail to boot in case the PF isn't eswitch owner. Example of such case is the DPU host PF which isn't eswitch manager, so any attempt to passthrough VF of it using libvirt will fail. [Fix] * 7846665d net/mlx5e: Unblock setting vid 0 for VF in case PF isn't eswitch manager This patch can be cherry-picked into Impish and newer kernels, but it requires some backport work on F/H due to they're missing commit b55b3538 that split the legacy code out. The ESW_ALLOWED marco was later converted into a helper routine in commit b16f2bb6. [Test] Test kernels can be found here: F: https://people.canonical.com/~phlin/kernel/lp-1957753-mlx5e/F/ H: https://people.canonical.com/~phlin/kernel/lp-1957753-mlx5e/H/ I: https://people.canonical.com/~phlin/kernel/lp-1957753-mlx5e/I/ J: https://people.canonical.com/~phlin/kernel/lp-1957753-mlx5e/J/ Kernels tested on a SUT with this device by Dmitrii Shcherbakov, test passed with positive feedback. [Where problems could occur] If this patch is erroneous, it might cause issue to this mlx5 driver with certain operations (setting vid 0). And other programs such as libvirt that depends on this return value might be affected as well. [Original Bug Description] There is an issue with hypervisor host side mlx5 driver operation on Bluefield devices in kernels <=5.16 that was recently fixed in master with a one-liner: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=7846665d3504812acaebf920d1141851379a7f37 It would be good to have this fix in Focal+ kernels. This is not needed in https://bugs.launchpad.net/ubuntu/+source/linux- bluefield since it's used at the DPU side whereas the issue in question affects the hypervisor side. To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1957753/+subscriptions -- Mailing list: https://launchpad.net/~kernel-packages Post to : kernel-packages@lists.launchpad.net Unsubscribe : https://launchpad.net/~kernel-packages More help : https://help.launchpad.net/ListHelp
