An issue was discovered in Xen through 4.14.x. When a Xenstore watch fires, the xenstore client that registered the watch will receive a Xenstore message containing the path of the modified Xenstore entry that triggered the watch, and the tag that was specified when registering the watch. Any communication with xenstored is done via Xenstore messages, consisting of a message header and the payload. The payload length is limited to 4096 bytes. Any request to xenstored resulting in a response with a payload longer than 4096 bytes will result in an error. When registering a watch, the payload length limit applies to the combined length of the watched path and the specified tag. Because watches for a specific path are also triggered for all nodes below that path, the payload of a watch event message can be longer than the payload needed to register the watch. A malicious guest that registers a watch using a very large tag (i.e., with a registration operation payload length close to the 4096 byte limit) can cause the generation of watch events with a payload length larger than 4096 bytes, by writing to Xenstore entries below the watched path. This will result in an error condition in xenstored. This error can result in a NULL pointer dereference, leading to a crash of xenstored. A malicious guest administrator can cause xenstored to crash, leading to a denial of service. Following a xenstored crash, domains may continue to run, but management operations will be impossible. Only C xenstored is affected, oxenstored is not affected.

An issue was discovered in Xen through 4.14.x. Xenstored and guests communicate via a shared memory page using a specific protocol. When a guest violates this protocol, xenstored will drop the connection to that guest. Unfortunately, this is done by just removing the guest from xenstored's internal management, resulting in the same actions as if the guest had been destroyed, including sending an @releaseDomain event. @releaseDomain events do not say that the guest has been removed. All watchers of this event must look at the states of all guests to find the guest that has been removed. When an @releaseDomain is generated due to a domain xenstored protocol violation, because the guest is still running, the watchers will not react. Later, when the guest is actually destroyed, xenstored will no longer have it stored in its internal data base, so no further @releaseDomain event will be sent. This can lead to a zombie domain; memory mappings of that guest's memory will not be removed, due to the missing event. This zombie domain will be cleaned up only after another domain is destroyed, as that will trigger another @releaseDomain event. If the device model of the guest that violated the Xenstore protocol is running in a stub-domain, a use-after-free case could happen in xenstored, after having removed the guest from its internal data base, possibly resulting in a crash of xenstored. A malicious guest can block resources of the host for a period after its own death. Guests with a stub domain device model can eventually crash xenstored, resulting in a more serious denial of service (the prevention of any further domain management operations). Only the C variant of Xenstore is affected; the Ocaml variant is not affected. Only HVM guests with a stubdom device model can cause a serious DoS.

An issue was discovered in Xen through 4.14.x. A guest may access xenstore paths via absolute paths containing a full pathname, or via a relative path, which implicitly includes /local/domain/$DOMID for their own domain id. Management tools must access paths in guests' namespaces, necessarily using absolute paths. oxenstored imposes a pathname limit that is applied solely to the relative or absolute path specified by the client. Therefore, a guest can create paths in its own namespace which are too long for management tools to access. Depending on the toolstack in use, a malicious guest administrator might cause some management tools and debugging operations to fail. For example, a guest administrator can cause "xenstore-ls -r" to fail. However, a guest administrator cannot prevent the host administrator from tearing down the domain. All systems using oxenstored are vulnerable. Building and using oxenstored is the default in the upstream Xen distribution, if the Ocaml compiler is available. Systems using C xenstored are not vulnerable.

An issue was discovered in Xen through 4.14.x. Access rights of Xenstore nodes are per domid. Unfortunately, existing granted access rights are not removed when a domain is being destroyed. This means that a new domain created with the same domid will inherit the access rights to Xenstore nodes from the previous domain(s) with the same domid. Because all Xenstore entries of a guest below /local/domain/<domid> are being deleted by Xen tools when a guest is destroyed, only Xenstore entries of other guests still running are affected. For example, a newly created guest domain might be able to read sensitive information that had belonged to a previously existing guest domain. Both Xenstore implementations (C and Ocaml) are vulnerable.

An issue was discovered in Xen through 4.14.x. Neither xenstore implementation does any permission checks when reporting a xenstore watch event. A guest administrator can watch the root xenstored node, which will cause notifications for every created, modified, and deleted key. A guest administrator can also use the special watches, which will cause a notification every time a domain is created and destroyed. Data may include: number, type, and domids of other VMs; existence and domids of driver domains; numbers of virtual interfaces, block devices, vcpus; existence of virtual framebuffers and their backend style (e.g., existence of VNC service); Xen VM UUIDs for other domains; timing information about domain creation and device setup; and some hints at the backend provisioning of VMs and their devices. The watch events do not contain values stored in xenstore, only key names. A guest administrator can observe non-sensitive domain and device lifecycle events relating to other guests. This information allows some insight into overall system configuration (including the number and general nature of other guests), and configuration of other guests (including the number and general nature of other guests' devices). This information might be commercially interesting or might make other attacks easier. There is not believed to be exposure of sensitive data. Specifically, there is no exposure of VNC passwords, port numbers, pathnames in host and guest filesystems, cryptographic keys, or within-guest data.

An issue was discovered in Xen through 4.14.x. In the Ocaml xenstored implementation, the internal representation of the tree has special cases for the root node, because this node has no parent. Unfortunately, permissions were not checked for certain operations on the root node. Unprivileged guests can get and modify permissions, list, and delete the root node. (Deleting the whole xenstore tree is a host-wide denial of service.) Achieving xenstore write access is also possible. All systems using oxenstored are vulnerable. Building and using oxenstored is the default in the upstream Xen distribution, if the Ocaml compiler is available. Systems using C xenstored are not vulnerable.

An issue was discovered in the Linux kernel through 5.10.1, as used with Xen through 4.14.x. The Linux kernel PV block backend expects the kernel thread handler to reset ring->xenblkd to NULL when stopped. However, the handler may not have time to run if the frontend quickly toggles between the states connect and disconnect. As a consequence, the block backend may re-use a pointer after it was freed. A misbehaving guest can trigger a dom0 crash by continuously connecting / disconnecting a block frontend. Privilege escalation and information leaks cannot be ruled out. This only affects systems with a Linux blkback.

An issue was discovered in Xen through 4.14.x. Some OSes (such as Linux, FreeBSD, and NetBSD) are processing watch events using a single thread. If the events are received faster than the thread is able to handle, they will get queued. As the queue is unbounded, a guest may be able to trigger an OOM in the backend. All systems with a FreeBSD, Linux, or NetBSD (any version) dom0 are vulnerable.

An issue was discovered in Xen through 4.14.x allowing x86 HVM guest OS users to cause a denial of service (stack corruption), cause a data leak, or possibly gain privileges because of an off-by-one error. NOTE: this issue is caused by an incorrect fix for CVE-2020-27671.

Xen through 4.14.x allows guest OS administrators to obtain sensitive information (such as AES keys from outside the guest) via a side-channel attack on a power/energy monitoring interface, aka a "Platypus" attack. NOTE: there is only one logically independent fix: to change the access control for each such interface in Xen.

An issue was discovered in the Linux kernel through 5.9.1, as used with Xen through 4.14.x. Guest OS users can cause a denial of service (host OS hang) via a high rate of events to dom0, aka CID-e99502f76271.

An issue was discovered in Xen through 4.14.x. There is a race condition when migrating timers between x86 HVM vCPUs. When migrating timers of x86 HVM guests between its vCPUs, the locking model used allows for a second vCPU of the same guest (also operating on the timers) to release a lock that it didn't acquire. The most likely effect of the issue is a hang or crash of the hypervisor, i.e., a Denial of Service (DoS). All versions of Xen are affected. Only x86 systems are vulnerable. Arm systems are not vulnerable. Only x86 HVM guests can leverage the vulnerability. x86 PV and PVH cannot leverage the vulnerability. Only guests with more than one vCPU can exploit the vulnerability.

An issue was discovered in Xen through 4.14.x. There are missing memory barriers when accessing/allocating an event channel. Event channels control structures can be accessed lockless as long as the port is considered to be valid. Such a sequence is missing an appropriate memory barrier (e.g., smp_*mb()) to prevent both the compiler and CPU from re-ordering access. A malicious guest may be able to cause a hypervisor crash resulting in a Denial of Service (DoS). Information leak and privilege escalation cannot be excluded. Systems running all versions of Xen are affected. Whether a system is vulnerable will depend on the CPU and compiler used to build Xen. For all systems, the presence and the scope of the vulnerability depend on the precise re-ordering performed by the compiler used to build Xen. We have not been able to survey compilers; consequently we cannot say which compiler(s) might produce vulnerable code (with which code generation options). GCC documentation clearly suggests that re-ordering is possible. Arm systems will also be vulnerable if the CPU is able to re-order memory access. Please consult your CPU vendor. x86 systems are only vulnerable if a compiler performs re-ordering.

An issue was discovered in Xen through 4.14.x. There is a lack of preemption in evtchn_reset() / evtchn_destroy(). In particular, the FIFO event channel model allows guests to have a large number of event channels active at a time. Closing all of these (when resetting all event channels or when cleaning up after the guest) may take extended periods of time. So far, there was no arrangement for preemption at suitable intervals, allowing a CPU to spend an almost unbounded amount of time in the processing of these operations. Malicious or buggy guest kernels can mount a Denial of Service (DoS) attack affecting the entire system. All Xen versions are vulnerable in principle. Whether versions 4.3 and older are vulnerable depends on underlying hardware characteristics.

An issue was discovered in Xen through 4.14.x. x86 PV guest kernels can experience denial of service via SYSENTER. The SYSENTER instruction leaves various state sanitization activities to software. One of Xen's sanitization paths injects a #GP fault, and incorrectly delivers it twice to the guest. This causes the guest kernel to observe a kernel-privilege #GP fault (typically fatal) rather than a user-privilege #GP fault (usually converted into SIGSEGV/etc.). Malicious or buggy userspace can crash the guest kernel, resulting in a VM Denial of Service. All versions of Xen from 3.2 onwards are vulnerable. Only x86 systems are vulnerable. ARM platforms are not vulnerable. Only x86 systems that support the SYSENTER instruction in 64bit mode are vulnerable. This is believed to be Intel, Centaur, and Shanghai CPUs. AMD and Hygon CPUs are not believed to be vulnerable. Only x86 PV guests can exploit the vulnerability. x86 PVH / HVM guests cannot exploit the vulnerability.

An issue was discovered in Xen through 4.14.x. The PCI passthrough code improperly uses register data. Code paths in Xen's MSI handling have been identified that act on unsanitized values read back from device hardware registers. While devices strictly compliant with PCI specifications shouldn't be able to affect these registers, experience shows that it's very common for devices to have out-of-spec "backdoor" operations that can affect the result of these reads. A not fully trusted guest may be able to crash Xen, leading to a Denial of Service (DoS) for the entire system. Privilege escalation and information leaks cannot be excluded. All versions of Xen supporting PCI passthrough are affected. Only x86 systems are vulnerable. Arm systems are not vulnerable. Only guests with passed through PCI devices may be able to leverage the vulnerability. Only systems passing through devices with out-of-spec ("backdoor") functionality can cause issues. Experience shows that such out-of-spec functionality is common; unless you have reason to believe that your device does not have such functionality, it's better to assume that it does.

An issue was discovered in Xen through 4.13.x, allowing Intel guest OS users to gain privileges or cause a denial of service because of non-atomic modification of a live EPT PTE. When mapping guest EPT (nested paging) tables, Xen would in some circumstances use a series of non-atomic bitfield writes. Depending on the compiler version and optimisation flags, Xen might expose a dangerous partially written PTE to the hardware, which an attacker might be able to race to exploit. A guest administrator or perhaps even an unprivileged guest user might be able to cause denial of service, data corruption, or privilege escalation. Only systems using Intel CPUs are vulnerable. Systems using AMD CPUs, and Arm systems, are not vulnerable. Only systems using nested paging (hap, aka nested paging, aka in this case Intel EPT) are vulnerable. Only HVM and PVH guests can exploit the vulnerability. The presence and scope of the vulnerability depends on the precise optimisations performed by the compiler used to build Xen. If the compiler generates (a) a single 64-bit write, or (b) a series of read-modify-write operations in the same order as the source code, the hypervisor is not vulnerable. For example, in one test build using GCC 8.3 with normal settings, the compiler generated multiple (unlocked) read-modify-write operations in source-code order, which did not constitute a vulnerability. We have not been able to survey compilers; consequently we cannot say which compiler(s) might produce vulnerable code (with which code-generation options). The source code clearly violates the C rules, and thus should be considered vulnerable.

An issue was discovered in Xen through 4.13.x, allowing guest OS users to cause a denial of service because of a bad error path in GNTTABOP_map_grant. Grant table operations are expected to return 0 for success, and a negative number for errors. Some misplaced brackets cause one error path to return 1 instead of a negative value. The grant table code in Linux treats this condition as success, and proceeds with incorrectly initialised state. A buggy or malicious guest can construct its grant table in such a way that, when a backend domain tries to map a grant, it hits the incorrect error path. This will crash a Linux based dom0 or backend domain.

An issue was discovered in Xen through 4.13.x, allowing guest OS users to cause a denial of service because of bad continuation handling in GNTTABOP_copy. Grant table operations are expected to return 0 for success, and a negative number for errors. The fix for CVE-2017-12135 introduced a path through grant copy handling where success may be returned to the caller without any action taken. In particular, the status fields of individual operations are left uninitialised, and may result in errant behaviour in the caller of GNTTABOP_copy. A buggy or malicious guest can construct its grant table in such a way that, when a backend domain tries to copy a grant, it hits the incorrect exit path. This returns success to the caller without doing anything, which may cause crashes or other incorrect behaviour.

An issue was discovered in xenoprof in Xen through 4.13.x, allowing guest OS users (with active profiling) to obtain sensitive information about other guests, cause a denial of service, or possibly gain privileges. For guests for which "active" profiling was enabled by the administrator, the xenoprof code uses the standard Xen shared ring structure. Unfortunately, this code did not treat the guest as a potential adversary: it trusts the guest not to modify buffer size information or modify head / tail pointers in unexpected ways. This can crash the host (DoS). Privilege escalation cannot be ruled out.