The fixes for XSA-422 (Branch Type Confusion) and XSA-434 (Speculative Return Stack Overflow) are not IRQ-safe. It was believed that the mitigations always operated in contexts with IRQs disabled. However, the original XSA-254 fix for Meltdown (XPTI) deliberately left interrupts enabled on two entry paths; one unconditionally, and one conditionally on whether XPTI was active. As BTC/SRSO and Meltdown affect different CPU vendors, the mitigations are not active together by default. Therefore, there is a race condition whereby a malicious PV guest can bypass BTC/SRSO protections and launch a BTC/SRSO attack against Xen.

The current setup of the quarantine page tables assumes that the quarantine domain (dom_io) has been initialized with an address width of DEFAULT_DOMAIN_ADDRESS_WIDTH (48) and hence 4 page table levels. However dom_io being a PV domain gets the AMD-Vi IOMMU page tables levels based on the maximum (hot pluggable) RAM address, and hence on systems with no RAM above the 512GB mark only 3 page-table levels are configured in the IOMMU. On systems without RAM above the 512GB boundary amd_iommu_quarantine_init() will setup page tables for the scratch page with 4 levels, while the IOMMU will be configured to use 3 levels only, resulting in the last page table directory (PDE) effectively becoming a page table entry (PTE), and hence a device in quarantine mode gaining write access to the page destined to be a PDE. Due to this page table level mismatch, the sink page the device gets read/write access to is no longer cleared between device assignment, possibly leading to data leaks.

The caching invalidation guidelines from the AMD-Vi specification (48882—Rev 3.07-PUB—Oct 2022) is incorrect on some hardware, as devices will malfunction (see stale DMA mappings) if some fields of the DTE are updated but the IOMMU TLB is not flushed. Such stale DMA mappings can point to memory ranges not owned by the guest, thus allowing access to unindented memory regions.

[This CNA information record relates to multiple CVEs; the text explains which aspects/vulnerabilities correspond to which CVE.] libfsimage contains parsing code for several filesystems, most of them based on grub-legacy code. libfsimage is used by pygrub to inspect guest disks. Pygrub runs as the same user as the toolstack (root in a priviledged domain). At least one issue has been reported to the Xen Security Team that allows an attacker to trigger a stack buffer overflow in libfsimage. After further analisys the Xen Security Team is no longer confident in the suitability of libfsimage when run against guest controlled input with super user priviledges. In order to not affect current deployments that rely on pygrub patches are provided in the resolution section of the advisory that allow running pygrub in deprivileged mode. CVE-2023-4949 refers to the original issue in the upstream grub project ("An attacker with local access to a system (either through a disk or external drive) can present a modified XFS partition to grub-legacy in such a way to exploit a memory corruption in grub’s XFS file system implementation.") CVE-2023-34325 refers specifically to the vulnerabilities in Xen's copy of libfsimage, which is decended from a very old version of grub.

When a transaction is committed, C Xenstored will first check the quota is correct before attempting to commit any nodes. It would be possible that accounting is temporarily negative if a node has been removed outside of the transaction. Unfortunately, some versions of C Xenstored are assuming that the quota cannot be negative and are using assert() to confirm it. This will lead to C Xenstored crash when tools are built without -DNDEBUG (this is the default).

Cortex-A77 cores (r0p0 and r1p0) are affected by erratum 1508412 where software, under certain circumstances, could deadlock a core due to the execution of either a load to device or non-cacheable memory, and either a store exclusive or register read of the Physical Address Register (PAR_EL1) in close proximity.

Oxenstored 32->31 bit integer truncation issues Integers in Ocaml are 63 or 31 bits of signed precision. The Ocaml Xenbus library takes a C uint32_t out of the ring and casts it directly to an Ocaml integer. In 64-bit Ocaml builds this is fine, but in 32-bit builds, it truncates off the most significant bit, and then creates unsigned/signed confusion in the remainder. This in turn can feed a negative value into logic not expecting a negative value, resulting in unexpected exceptions being thrown. The unexpected exception is not handled suitably, creating a busy-loop trying (and failing) to take the bad packet out of the xenstore ring.

lock order inversion in transitive grant copy handling As part of XSA-226 a missing cleanup call was inserted on an error handling path. While doing so, locking requirements were not paid attention to. As a result two cooperating guests granting each other transitive grants can cause locks to be acquired nested within one another, but in respectively opposite order. With suitable timing between the involved grant copy operations this may result in the locking up of a CPU.

Arm: unbounded memory consumption for 2nd-level page tables Certain actions require e.g. removing pages from a guest's P2M (Physical-to-Machine) mapping. When large pages are in use to map guest pages in the 2nd-stage page tables, such a removal operation may incur a memory allocation (to replace a large mapping with individual smaller ones). These memory allocations are taken from the global memory pool. A malicious guest might be able to cause the global memory pool to be exhausted by manipulating its own P2M mappings.

guests may exceed their designated memory limit When a guest is permitted to have close to 16TiB of memory, it may be able to issue hypercalls to increase its memory allocation beyond the administrator established limit. This is a result of a calculation done with 32-bit precision, which may overflow. It would then only be the overflowed (and hence small) number which gets compared against the established upper bound.

Another race in XENMAPSPACE_grant_table handling Guests are permitted access to certain Xen-owned pages of memory. The majority of such pages remain allocated / associated with a guest for its entire lifetime. Grant table v2 status pages, however, are de-allocated when a guest switches (back) from v2 to v1. Freeing such pages requires that the hypervisor enforce that no parallel request can result in the addition of a mapping of such a page to a guest. That enforcement was missing, allowing guests to retain access to pages that were freed and perhaps re-used for other purposes. Unfortunately, when XSA-379 was being prepared, this similar issue was not noticed.

long running loops in grant table handling In order to properly monitor resource use, Xen maintains information on the grant mappings a domain may create to map grants offered by other domains. In the process of carrying out certain actions, Xen would iterate over all such entries, including ones which aren't in use anymore and some which may have been created but never used. If the number of entries for a given domain is large enough, this iterating of the entire table may tie up a CPU for too long, starving other domains or causing issues in the hypervisor itself. Note that a domain may map its own grants, i.e. there is no need for multiple domains to be involved here. A pair of "cooperating" guests may, however, cause the effects to be more severe.

grant table v2 status pages may remain accessible after de-allocation Guest get permitted access to certain Xen-owned pages of memory. The majority of such pages remain allocated / associated with a guest for its entire lifetime. Grant table v2 status pages, however, get de-allocated when a guest switched (back) from v2 to v1. The freeing of such pages requires that the hypervisor know where in the guest these pages were mapped. The hypervisor tracks only one use within guest space, but racing requests from the guest to insert mappings of these pages may result in any of them to become mapped in multiple locations. Upon switching back from v2 to v1, the guest would then retain access to a page that was freed and perhaps re-used for other purposes.

IOMMU page mapping issues on x86 T[his CNA information record relates to multiple CVEs; the text explains which aspects/vulnerabilities correspond to which CVE.] Both AMD and Intel allow ACPI tables to specify regions of memory which should be left untranslated, which typically means these addresses should pass the translation phase unaltered. While these are typically device specific ACPI properties, they can also be specified to apply to a range of devices, or even all devices. On all systems with such regions Xen failed to prevent guests from undoing/replacing such mappings (CVE-2021-28694). On AMD systems, where a discontinuous range is specified by firmware, the supposedly-excluded middle range will also be identity-mapped (CVE-2021-28695). Further, on AMD systems, upon de-assigment of a physical device from a guest, the identity mappings would be left in place, allowing a guest continued access to ranges of memory which it shouldn't have access to anymore (CVE-2021-28696).

IOMMU page mapping issues on x86 T[his CNA information record relates to multiple CVEs; the text explains which aspects/vulnerabilities correspond to which CVE.] Both AMD and Intel allow ACPI tables to specify regions of memory which should be left untranslated, which typically means these addresses should pass the translation phase unaltered. While these are typically device specific ACPI properties, they can also be specified to apply to a range of devices, or even all devices. On all systems with such regions Xen failed to prevent guests from undoing/replacing such mappings (CVE-2021-28694). On AMD systems, where a discontinuous range is specified by firmware, the supposedly-excluded middle range will also be identity-mapped (CVE-2021-28695). Further, on AMD systems, upon de-assigment of a physical device from a guest, the identity mappings would be left in place, allowing a guest continued access to ranges of memory which it shouldn't have access to anymore (CVE-2021-28696).

IOMMU page mapping issues on x86 T[his CNA information record relates to multiple CVEs; the text explains which aspects/vulnerabilities correspond to which CVE.] Both AMD and Intel allow ACPI tables to specify regions of memory which should be left untranslated, which typically means these addresses should pass the translation phase unaltered. While these are typically device specific ACPI properties, they can also be specified to apply to a range of devices, or even all devices. On all systems with such regions Xen failed to prevent guests from undoing/replacing such mappings (CVE-2021-28694). On AMD systems, where a discontinuous range is specified by firmware, the supposedly-excluded middle range will also be identity-mapped (CVE-2021-28695). Further, on AMD systems, upon de-assigment of a physical device from a guest, the identity mappings would be left in place, allowing a guest continued access to ranges of memory which it shouldn't have access to anymore (CVE-2021-28696).

inappropriate x86 IOMMU timeout detection / handling IOMMUs process commands issued to them in parallel with the operation of the CPU(s) issuing such commands. In the current implementation in Xen, asynchronous notification of the completion of such commands is not used. Instead, the issuing CPU spin-waits for the completion of the most recently issued command(s). Some of these waiting loops try to apply a timeout to fail overly-slow commands. The course of action upon a perceived timeout actually being detected is inappropriate: - on Intel hardware guests which did not originally cause the timeout may be marked as crashed, - on AMD hardware higher layer callers would not be notified of the issue, making them continue as if the IOMMU operation succeeded.

x86: Speculative vulnerabilities with bare (non-shim) 32-bit PV guests 32-bit x86 PV guest kernels run in ring 1. At the time when Xen was developed, this area of the i386 architecture was rarely used, which is why Xen was able to use it to implement paravirtualisation, Xen's novel approach to virtualization. In AMD64, Xen had to use a different implementation approach, so Xen does not use ring 1 to support 64-bit guests. With the focus now being on 64-bit systems, and the availability of explicit hardware support for virtualization, fixing speculation issues in ring 1 is not a priority for processor companies. Indirect Branch Restricted Speculation (IBRS) is an architectural x86 extension put together to combat speculative execution sidechannel attacks, including Spectre v2. It was retrofitted in microcode to existing CPUs. For more details on Spectre v2, see: http://xenbits.xen.org/xsa/advisory-254.html However, IBRS does not architecturally protect ring 0 from predictions learnt in ring 1. For more details, see: https://software.intel.com/security-software-guidance/deep-dives/deep-dive-indirect-branch-restricted-speculation Similar situations may exist with other mitigations for other kinds of speculative execution attacks. The situation is quite likely to be similar for speculative execution attacks which have yet to be discovered, disclosed, or mitigated.

Potential floating point value injection in all supported CPU products, in conjunction with software vulnerabilities relating to speculative execution with incorrect floating point results, may cause the use of incorrect data from FPVI and may result in data leakage.

Potential speculative code store bypass in all supported CPU products, in conjunction with software vulnerabilities relating to speculative execution of overwritten instructions, may cause an incorrect speculation and could result in data leakage.