Look up another CVE · open CVE-2026-74396 in the interactive tracker
CVE-2026-74396
RDMA/mlx5: Fix UMR XLT cleanup on ODP populate failure
Published 2026-08-15 (2 months ago) by the Linux kernel CNA. Subsystem drivers/infiniband. Debian urgency: unimportant.
What to do, per Debian release
Debian's own status for CVE-2026-74396 in every release this site tracks, and what each one means in practice. Every status below is the one the Debian Security Team publishes for this CVE. source
| Release | Status | Version | What this means |
|---|---|---|---|
| Debian 12bookworm | not affected | 6.1.190-1 | Nothing to do. bookworm never shipped the vulnerable code. |
| Debian 13trixie | not affected | 6.12.113-1 | Nothing to do. trixie never shipped the vulnerable code. |
| Debian 14forky | fixed | 7.1.5-1 | Update the kernel and reboot. Fixed in 7.1.5-1. A running kernel keeps the old code until the machine restarts. |
| Debian unstablesid | fixed | 7.1.5-1 | Update the kernel and reboot. Fixed in 7.1.5-1. A running kernel keeps the old code until the machine restarts. |
The vulnerability
What the bug is
mlx5r_umr_update_xlt() allocates and DMA maps an XLT buffer with mlx5r_umr_create_xlt(). The buffer is released by the common cleanup path through mlx5r_umr_unmap_free_xlt(). After mlx5_odp_populate_xlt() became fallible, its error path returned directly and skipped that cleanup. This leaks the XLT DMA mapping and buffer. If the emergency XLT page was used, it also leaves xlt_emergency_page_mutex locked. Break out of the loop so execution falls through the existing cleanup path.
Where it lives
The fix touches these files. If the subsystem is one you do not use, the practical exposure is lower, though the package is still the vulnerable one.
drivers/infiniband/hw/mlx5/umr.c
- What this is
- MELLANOX MLX5 IB driverMAINTAINERS section matched on drivers/infiniband/hw/mlx5/ · linux-rdma@vger.kernel.org · Supported
- Kernel module
- mlx5_ib
Check it on the machine
grep -wE 'CONFIG_INFINIBAND|CONFIG_MLX5_INFINIBAND' /boot/config-$(uname -r)Reads the booted kernel's own configuration, on the machine and nowhere else. =y is built in, =m is a module, and an is not set line means the code is not there.
lsmod | grep -w mlx5_ibPrints a line if that module is loaded right now, and nothing if it is not. One that is not loaded can still be loaded later by anything that needs it.
Show whether Debian builds this, per release
Debian builds this, but not on every flavour. At least one release below builds the code on some of its kernel flavours and not on others, so the answer depends on the architecture and flavour installed.
- bookwormm
- trixieMnot built on armel/rpi
- forky, sidm
- m
- a loadable module on every flavour - check lsmod, can be blacklisted
- M
- a module where it is built at all; some flavours do not build it
- Build switches
- CONFIG_INFINIBAND, CONFIG_MLX5_INFINIBAND
- Reads as
- all on means the file is compiled; any one off means it is not.
From the configuration Debian publishes for every kernel flavour it builds. It covers only the architectures Debian ships a linux-config package for, and nothing about a kernel you built yourself. Debian's own status above stays authoritative about the source package.MAINTAINERS kbuild Makefiles Debian linux-config packages
Signals
- Published
- 2026-08-15 (2 months ago)source
- CVSS
- 7.5 high
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:Hsource recompute it - Attack vector
- network (AV:N)source
- EPSS
- 0.73%, higher than 53.5% of all CVEssource
- CISA KEV
- not listedthe KEV catalogue
- Debian urgency
- unimportantsource
- Subsystem
- drivers/infiniband
Upstream fixes
The versions the CNA records as introducing and fixing this, and the commit for each. source
| Introduced | Fixed in | Commit |
|---|---|---|
| 6.16 | 6.18.40 | ffa85a2c1979 |
| 6.16 | 7.1.5 | 9619909d4869 |
| 6.16 | 7.2 | 1eae35b37923 |
Why the CNA scored it this way
Quoted from the kernel CNA's own scoring file, word for word. A highlighted metric carries the most severe value CVSS v3.1 defines for it, which is what pushes the score up. source
- AV:NAttack vector, network
- Remote RDMA peers trigger mlx5 ODP hardware page faults over RoCE/InfiniBand (mlx5_ib_mr_rdma_pfault_handler → pagefault_real_mr → mlx5r_umr_update_xlt), reaching the buggy cleanup on mlx5 AI/HPC/cloud servers without local syscalls.
- AC:LAttack complexity, low
- An attacker can repeatedly drive ODP page faults, exhaust memory to force the emergency XLT page path, and retry until mlx5_odp_populate_xlt()/hmm_dma_map_pfn() fails; no uncontrollable race or memory layout is required.
- PR:NPrivileges required, none
- The RDMA responder page-fault path executes in kernel workqueues on packets from a connected remote peer and performs no Linux capability or credential check before calling mlx5r_umr_update_xlt().
- UI:NUser interaction, none
- Exploitation requires only attacker-generated RDMA traffic against exported ODP memory regions once a queue pair exists; no additional victim action beyond normal RDMA service operation is needed.
- S:UScope, unchanged
- The DMA mapping leak and permanent xlt_emergency_page_mutex lock affect mlx5 driver resources within the same host kernel security authority; there is no VM escape or IOMMU-boundary bypass.
- C:NConfidentiality impact, none
- The bug skips unmap/free of an XLT translation buffer; it does not corrupt memory, dereference freed objects, or provide an arbitrary kernel read primitive.
- I:NIntegrity impact, none
- No out-of-bounds write, use-after-free, or control-flow corruption occurs; the defect only leaks resources rather than enabling arbitrary data modification.
- A:HAvailability impact, high
- Each failure leaks a DMA mapping and buffer, and on the emergency-page path permanently locks xlt_emergency_page_mutex, deadlocking subsequent UMR XLT updates and disabling ODP/RDMA on the mlx5 device.