CVE-2026-99802: io_uring/register: enforce per-task restrictions in io_uring_register_blind()
In the Linux kernel, the following vulnerability has been resolved:
iouring/register: enforce per-task restrictions in iouringregisterblind()
When iouringregister(2) is invoked with fd == -1, it dispatches directly to iouringregisterblind() without checking current->iouringrestrict.
As a result, a task with per-task register restrictions installed via IORINGREGISTERRESTRICTIONS can still invoke blind registration opcodes such as IORINGREGISTERSENDMSGRING even when those opcodes are not in the task's allowed registerop bitmap.
Check current->iouringrestrict->registerop in iouringregisterblind() before dispatching blind registration opcodes.
Affected Software
Remediation
Recommended actions to resolve this vulnerability, in priority order.
- Upgrade
Upgrade
linux/Kernelto a version that resolves this vulnerability.Fixed in 7.3.2 - Upgrade
Upgrade
debian/linuxto a version that resolves this vulnerability.Fixed in 7.3.3-1 - Upgrade
Upgrade
debian/linuxto a version that resolves this vulnerability.Fixed in 6.1.176-1Fixed in 6.1.192-2Fixed in 6.12.107-1Fixed in 6.12.117-1Fixed in 7.3.4-1Fixed in 7.3.5-1
Frequently Asked Questions
Which systems are affected by this bypass?
The issue applies to Linux kernel environments that use io_uring per-task registration restrictions installed through IORING_REGISTER_RESTRICTIONS. It specifically affects blind registration requests made through io_uring_register(2) with fd set to -1.
What does an attacker need to exploit it?
An attacker needs to be able to invoke io_uring_register(2) in a task that has per-task register restrictions configured. They can then use a blind registration opcode, such as IORING_REGISTER_SEND_MSG_RING, that is absent from that task's allowed register_op bitmap.
Is the protection bypass limited to a particular registration path?
Yes. The bypass occurs because the fd == -1 path dispatches directly to io_uring_register_blind() without checking the task's register_op restrictions. Other behavior is not described in the available data.
How can administrators determine whether their kernel includes the fix?
The fixed behavior checks current->io_uring_restrict->register_op in io_uring_register_blind() before dispatching blind registration opcodes. Administrators can inspect their kernel source or vendor patch information for that check.