From: Philippe Mathieu-Daudé <phi...@redhat.com> When the NVMe block driver was introduced (see commit bdd6a90a9e5, January 2018), Linux VFIO_IOMMU_MAP_DMA ioctl was only returning -ENOMEM in case of error. The driver was correctly handling the error path to recycle its volatile IOVA mappings.
To fix CVE-2019-3882, Linux commit 492855939bdb ("vfio/type1: Limit DMA mappings per container", April 2019) added the -ENOSPC error to signal the user exhausted the DMA mappings available for a container. The block driver started to mis-behave: qemu-system-x86_64: VFIO_MAP_DMA failed: No space left on device (qemu) (qemu) info status VM status: paused (io-error) (qemu) c VFIO_MAP_DMA failed: No space left on device (qemu) c VFIO_MAP_DMA failed: No space left on device (The VM is not resumable from here, hence stuck.) Fix by handling the new -ENOSPC error (when DMA mappings are exhausted) without any distinction to the current -ENOMEM error, so we don't change the behavior on old kernels where the CVE-2019-3882 fix is not present. An easy way to reproduce this bug is to restrict the DMA mapping limit (65535 by default) when loading the VFIO IOMMU module: # modprobe vfio_iommu_type1 dma_entry_limit=666 Cc: qemu-sta...@nongnu.org Cc: Fam Zheng <f...@euphon.net> Cc: Maxim Levitsky <mlevi...@redhat.com> Cc: Alex Williamson <alex.william...@redhat.com> Reported-by: Michal Prívozník <mpriv...@redhat.com> Signed-off-by: Philippe Mathieu-Daudé <phi...@redhat.com> Message-id: 20210723195843.1032825-1-phi...@redhat.com Fixes: bdd6a90a9e5 ("block: Add VFIO based NVMe driver") Buglink: https://bugs.launchpad.net/qemu/+bug/1863333 Resolves: https://gitlab.com/qemu-project/qemu/-/issues/65 Signed-off-by: Philippe Mathieu-Daudé <phi...@redhat.com> Signed-off-by: Stefan Hajnoczi <stefa...@redhat.com> (cherry picked from commit 15a730e7a3aaac180df72cd5730e0617bcf44a5a) Signed-off-by: Michael Roth <michael.r...@amd.com> --- block/nvme.c | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+) diff --git a/block/nvme.c b/block/nvme.c index 2b5421e7aa..e8dbbc2317 100644 --- a/block/nvme.c +++ b/block/nvme.c @@ -1030,7 +1030,29 @@ try_map: r = qemu_vfio_dma_map(s->vfio, qiov->iov[i].iov_base, len, true, &iova); + if (r == -ENOSPC) { + /* + * In addition to the -ENOMEM error, the VFIO_IOMMU_MAP_DMA + * ioctl returns -ENOSPC to signal the user exhausted the DMA + * mappings available for a container since Linux kernel commit + * 492855939bdb ("vfio/type1: Limit DMA mappings per container", + * April 2019, see CVE-2019-3882). + * + * This block driver already handles this error path by checking + * for the -ENOMEM error, so we directly replace -ENOSPC by + * -ENOMEM. Beside, -ENOSPC has a specific meaning for blockdev + * coroutines: it triggers BLOCKDEV_ON_ERROR_ENOSPC and + * BLOCK_ERROR_ACTION_STOP which stops the VM, asking the operator + * to add more storage to the blockdev. Not something we can do + * easily with an IOMMU :) + */ + r = -ENOMEM; + } if (r == -ENOMEM && retry) { + /* + * We exhausted the DMA mappings available for our container: + * recycle the volatile IOVA mappings. + */ retry = false; trace_nvme_dma_flush_queue_wait(s); if (s->dma_map_count) { -- 2.25.1