"value":"In the Linux kernel, the following vulnerability has been resolved:\n\nvirtio_net: Do not send RSS key if it is not supported\n\nThere is a bug when setting the RSS options in virtio_net that can break\nthe whole machine, getting the kernel into an infinite loop.\n\nRunning the following command in any QEMU virtual machine with virtionet\nwill reproduce this problem:\n\n # ethtool -X eth0 hfunc toeplitz\n\nThis is how the problem happens:\n\n1) ethtool_set_rxfh() calls virtnet_set_rxfh()\n\n2) virtnet_set_rxfh() calls virtnet_commit_rss_command()\n\n3) virtnet_commit_rss_command() populates 4 entries for the rss\nscatter-gather\n\n4) Since the command above does not have a key, then the last\nscatter-gatter entry will be zeroed, since rss_key_size == 0.\nsg_buf_size = vi->rss_key_size;\n\n5) This buffer is passed to qemu, but qemu is not happy with a buffer\nwith zero length, and do the following in virtqueue_map_desc() (QEMU\nfunction):\n\n if (!sz) {\n virtio_error(vdev, \"virtio: zero sized buffers are not allowed\");\n\n6) virtio_error() (also QEMU function) set the device as broken\n\n vdev->broken = true;\n\n7) Qemu bails out, and do not repond this crazy kernel.\n\n8) The kernel is waiting for the response to come back (function\nvirtnet_send_command())\n\n9) The kernel is waiting doing the following :\n\n while (!virtqueue_get_buf(vi->cvq, &tmp) &&\n\t !virtqueue_is_broken(vi->cvq))\n\t cpu_relax();\n\n10) None of the following functions above is true, thus, the kernel\nloops here forever. Keeping in mind that virtqueue_is_broken() does\nnot look at the qemu `vdev->broken`, so, it never realizes that the\nvitio is broken at QEMU side.\n\nFix it by not sending RSS commands if the feature is not available in\nthe device."
