<!--This is the upstream QEMU issue tracker. If you are able to, it will greatly facilitate bug triage if you attempt to reproduce the problem with the latest qemu.git master built from source. See https://www.qemu.org/download/#source for instructions on how to do this. QEMU generally supports the last two releases advertised on https://www.qemu.org/. Problems with distro-packaged versions of QEMU older than this should be reported to the distribution instead. See https://www.qemu.org/contribute/report-a-bug/ for additional guidance. If this is a security issue, ensure this ticket is marked 'confidential' before submission. See https://www.qemu.org/contribute/security-process/ for additional guidance--> ## Host environment - Operating system: PVE - OS/kernel version: 7.0.14-4-pve - Architecture: x86 - QEMU flavor: qemu-kvm - QEMU version: 11.0.0 - QEMU command line: /usr/bin/kvm \\ \-id 100 \ -name name \ -no-shutdown \ -pidfile /var/run/qemu-server/100.pid \ -daemonize \ -smp '60,sockets=1,cores=60,maxcpus=60' \ -nodefaults \ -cpu host,+kvm_pv_eoi,+kvm_pv_unhalt \ -m 180224 \ -object 'memory-backend-file,id=ram-node0,size=180224M,mem-path=/run/hugepages/kvm/1048576kB,share=on,prealloc=yes' \ -numa 'node,nodeid=0,cpus=0-59,memdev=ram-node0' \ -global 'ICH9-LPC.disable_s3=1' \ -global 'ICH9-LPC.disable_s4=1' \ -device 'intel-iommu,intremap=on,caching-mode=on' \ -readconfig /usr/share/qemu-server/pve-q35-4.0.cfg \ -device 'vfio-pci,host=0000:c1:00.0,id=hostpci0,bus=ich9-pcie-port-1,addr=0x0,rombar=0' \ -machine 'pflash0=pflash0,pflash1=drive-efidisk0,hpet=off,type=q35+pve0,kernel-irqchip=split,mem-merge=off' ``` ./qemu-system-x86_64 -M q35 -m 4096 -enable-kvm -hda fedora32.qcow2 ``` ## Emulated/Virtualized environment - Operating system: Ubuntu 22.04 LTS - OS/kernel version: 6.8.0-124-generic - Architecture: x86 ## Description of problem qemu-kvm v11 no longer announces ECAP.PT capability on emulated intel-iommu (vIOMMU) ## Steps to reproduce 1. Start kvm with viommu 2. DMAR ecap f00f1a (On qemu-kvm v10 it is f00f5a) 3. <!--Note: if this issue was discovered with the assistance of automated tooling LLM, static analysis, fuzzers), the reporter must disclose that in the description. The steps to reproduce, and any other findings, must be fully validated by the user of the tool prior to submission.--> ## Additional information This bug would cause a serious performance regression. On a VM with viommu enabled, a pci device pass-through, and guest kernel configured with iommu=pt, the guest kernel need to create an identity mapping for the pcie device by default. With ECAP.PT feature, the kernel simply informs vIOMMU for an identity mapping. However with this regression, the guest kernel has to manually create a 1:1 RAM mapping for each pci device and program it to vIOMMU. On a vm with large RAM (176GB in this case), or worse, multiple pci devices, this would take a long time and allocate multiple GB in additional to VM RAM. This is slow (multi minutes) and will stall the guest. Note that this would happen every time the guest kernel re-instates an identity mapping, which is not once-per-boot. <!--The line below ensures that proper tags are added to the issue. Please do not remove it.-->