Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB
CVE-2026-98004 Overview
CVE-2026-98004 is a race condition in the Linux kernel's RISC-V Input-Output Memory Management Unit (IOMMU) driver. The flaw stems from improper serialization of command queue publishing. Software producer state could advance before a command was fully written and the hardware tail pointer updated. The upstream fix also moves hardware consumption polling outside the queue lock when the command queue is full, preventing other CPUs from blocking behind a long poll.
Critical Impact
Concurrent CPU producers to the RISC-V IOMMU command queue could observe inconsistent queue state, potentially leading to command corruption, stalled IOMMU operations, or system instability on affected RISC-V platforms.
Affected Products
- Linux kernel RISC-V IOMMU driver (drivers/iommu/riscv/)
- Linux stable branches prior to commits 6d4c12ab9ab8 and ca58afa40946
- RISC-V systems using the in-kernel IOMMU driver for DMA remapping
Discovery Timeline
- 2026-09-25 - CVE CVE-2026-98004 published to NVD
- 2026-09-25 - Last updated in NVD database
Technical Details for CVE-2026-98004
Vulnerability Analysis
The RISC-V IOMMU driver maintains a command queue shared between software producers and the IOMMU hardware consumer. Software writes commands into a ring buffer and advances a producer index. The hardware consumes commands as the driver updates the tail register. Prior to the fix, the producer-side state could advance before the command slot was fully written or before the hardware tail register was updated. On multi-CPU systems, a concurrent producer could observe a stale or partially written slot and treat it as valid.
A second issue compounded the problem. When the command queue was full, the driver polled for hardware consumption while holding the queue lock. This serialized every other CPU attempting to submit IOMMU commands behind a single long-running poll, degrading responsiveness across the system.
Root Cause
The root cause is a missing ordering guarantee between three operations: writing a command to the ring buffer, updating the hardware tail register, and advancing the software producer index. Without serialization, the producer index could become visible to other CPUs before the slot content and hardware tail reflected the new command. This is a classic kernel concurrency defect in a shared ring buffer protocol.
Attack Vector
The NVD record does not list a CVSS vector or confirmed attacker exploitation path. The defect is a kernel concurrency bug reachable through normal IOMMU command submission paths, such as DMA mapping and invalidation operations triggered by device drivers. Any workload that drives concurrent IOMMU command submission on a RISC-V platform can expose the race.
No verified exploit code or proof-of-concept is available for CVE-2026-98004. The upstream commits documented in the kernel git tree describe the defect and its fix in prose. Refer to the Kernel Git Commit 6d4c12a and Kernel Git Commit ca58afa4 for the authoritative technical description.
Detection Methods for CVE-2026-98004
Indicators of Compromise
- No file-based or network IOCs apply. The defect is a kernel race, not an implant.
- Unexplained IOMMU command timeouts or DMA faults logged by the RISC-V IOMMU driver may indicate exposure on unpatched kernels.
- Soft lockups or CPU stall warnings referencing IOMMU command submission paths on RISC-V hosts.
Detection Strategies
- Inventory RISC-V Linux hosts and compare running kernel versions against the patched commits 6d4c12ab9ab8 and ca58afa40946.
- Query /proc/version and distribution package metadata to confirm whether the backport has been applied.
- Review kernel ring buffer logs (dmesg) for RISC-V IOMMU warnings, command queue full events, or DMA translation faults.
Monitoring Recommendations
- Centralize kernel logs from RISC-V hosts and alert on IOMMU subsystem errors and soft lockup messages.
- Track kernel version drift across the RISC-V fleet and flag hosts running pre-patch builds.
- Monitor host availability metrics for RISC-V systems to detect stalls that correlate with heavy DMA workloads.
How to Mitigate CVE-2026-98004
Immediate Actions Required
- Identify RISC-V Linux systems in the environment, including development boards, edge devices, and server-class RISC-V hardware.
- Apply vendor kernel updates that include commits 6d4c12ab9ab8 and ca58afa40946 or rebuild from a stable branch that incorporates both commits.
- Reboot affected hosts after kernel replacement so the patched IOMMU driver loads.
Patch Information
The fix is published in the Linux stable tree. Apply the kernel updates referenced in Kernel Git Commit 6d4c12a and Kernel Git Commit ca58afa4. Distribution maintainers typically backport stable commits into supported kernel packages. Confirm availability with the Linux distribution vendor used on each RISC-V host.
Workarounds
- No vendor-supplied workaround is documented. The upstream patch is the recommended remediation.
- Reduce concurrent IOMMU command pressure where feasible by limiting high-throughput DMA workloads on unpatched hosts until the kernel update is applied.
- Non-RISC-V systems are not affected and do not require action for this CVE.
# Verify the running kernel version on a RISC-V host
uname -a
# Check whether the patched commits are present in the kernel source tree
git -C /usr/src/linux log --oneline | grep -E '6d4c12ab9ab8|ca58afa40946'
Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB

