An assertion failure issue was found in the Network Block Device(NBD) Server in all QEMU versions before QEMU 5.0.1. This flaw occurs when an nbd-client sends a spec-compliant request that is near the boundary of maximum permitted request length. A remote nbd-client could use this flaw to crash the qemu-nbd server resulting in a denial of service.
hw/pci/pci.c in QEMU 4.2.0 allows guest OS users to trigger an out-of-bounds access by providing an address near the end of the PCI configuration space.
In QEMU 5.0.0 and earlier, megasas_lookup_frame in hw/scsi/megasas.c has an out-of-bounds read via a crafted reply_queue_head field from a guest OS user.
In QEMU 5.0.0 and earlier, es1370_transfer_audio in hw/audio/es1370.c does not properly validate the frame count, which allows guest OS users to trigger an out-of-bounds access during an es1370_write() operation.
sd_wp_addr in hw/sd/sd.c in QEMU 4.2.0 uses an unvalidated address, which leads to an out-of-bounds read during sdhci_write() operations. A guest OS user can crash the QEMU process.
An out-of-bounds heap buffer access flaw was found in the way the iSCSI Block driver in QEMU versions 2.12.0 before 4.2.1 handled a response coming from an iSCSI server while checking the status of a Logical Address Block (LBA) in an iscsi_co_block_status() routine. A remote user could use this flaw to crash the QEMU process, resulting in a denial of service or potential execution of arbitrary code with privileges of the QEMU process on the host.
An issue was discovered in ide_dma_cb() in hw/ide/core.c in QEMU 2.4.0 through 4.2.0. The guest system can crash the QEMU process in the host system via a special SCSI_IOCTL_SEND_COMMAND. It hits an assertion that implies that the size of successful DMA transfers there must be a multiple of 512 (the size of a sector). NOTE: a member of the QEMU security team disputes the significance of this issue because a "privileged guest user has many ways to cause similar DoS effect, without triggering this assert.
qemu-bridge-helper.c in QEMU 3.1 and 4.0.0 does not ensure that a network interface name (obtained from bridge.conf or a --br=bridge option) is limited to the IFNAMSIZ size, which can lead to an ACL bypass.