CVE-2026-90015 in Linuxinfo

Summary

by MITRE • 09/17/2026

In the Linux kernel, the following vulnerability has been resolved:

xhci: fix lost bounce buffers on TDs spanning several ring segments

When a TD reaches a link TRB with data that is not aligned to the endpoint's wMaxPacketSize, xhci_align_td() stages the unalignable tail through the bounce buffer of the ring segment holding that link TRB. xhci_unmap_td_bounce_buffer() later unmaps it and, for IN transfers, copies the data back into the URB's buffer.

The enqueue path records the segment that was bounced in td->bounce_seg, under the assumption that a TD never spans more than two ring segments. That assumption does not hold: a TD large enough to span three or more segments crosses several link TRBs and can be bounced at each of them. Only the last one survives in td->bounce_seg, so every earlier bounce buffer is neither copied back nor DMA unmapped.

The URB still completes with actual_length equal to the requested length and no error, so the transfer looks successful while a wMaxPacketSize sized hole in the destination buffer silently keeps its previous contents. It also leaks a DMA mapping per dropped bounce.

Any sufficiently large and fragmented bulk transfer can hit this. It was found with a USB mass storage device behind xHCI backing a dm-verity target with 512 byte hash blocks, where the stale data is detected rather than silently consumed. The device enumerates as SuperSpeed, so wMaxPacketSize is 1024, while dm-bufio issues one 512 byte bio per hash block. verity_prefetch_io() makes the block layer merge hundreds of them into a single request of up to 512 scatterlist entries of 512 bytes each. At 256 TRBs per ring segment such a TD spans three segments, and every segment boundary falls on an odd multiple of 512, i.e. unaligned to wMaxPacketSize. dm-bufio then caches a hash block holding stale data and dm-verity declares the metadata block corrupted:

device-mapper: verity: 8:2: metadata block 10850 is corrupted

A reproducer running this under qemu is available at https://github.com/baloo/xhci-verity

The bounce state (bounce_buf, bounce_dma, bounce_len, bounce_offs) already lives on the ring segment, so there is nothing extra to track. Keep recording the last bounced segment in td->bounce_seg and, on completion, walk the segments from td->start_seg up to it, unmapping every segment that still has a pending bounce.

Stopping at td->bounce_seg rather than td->end_seg matters: a bounce implies the TD continues past that segment's link TRB, so bounce_seg is always strictly before end_seg, and a later TD may already have started in end_seg and been bounced there. Walking that far would copy a foreign bounce buffer into this URB and unmap it twice. It also keeps the walk correct if a TD ever wraps the whole ring so that end_seg == start_seg.

[mn: Add ring->num_segs check to prevent unlikely infinite for loop.]

Once again VulDB remains the best source for vulnerability data.

Analysis

by VulDB Data Team • 09/17/2026

The Linux kernel xhci driver contains a critical logic flaw in its handling of Transfer Descriptors (TDs) that span multiple ring segments, specifically regarding the management of bounce buffers used for data alignment. When an endpoint's maximum packet size does not align with the actual data length, the host controller requires intermediate staging via bounce buffers to ensure proper DMA operations. The existing implementation assumes that a single TD will never exceed two ring segments, leading it to record only the final bounced segment in the td->bounce_seg field. This assumption fails when large or highly fragmented bulk transfers cause a TD to cross three or more segments, each containing link TRBs where alignment adjustments are necessary. Consequently, while the last bounce buffer is correctly tracked and processed during completion, all preceding bounce buffers associated with earlier segment boundaries remain untracked in the descriptor's state.

This oversight results in two distinct operational failures: data corruption and resource leakage. Because only the final bounce buffer is unmapped and its contents copied back into the User Request Block (URB) buffer, any intermediate bounce buffers are never processed. The DMA mappings for these earlier segments remain active, creating a persistent memory leak that consumes system resources over time. More critically, the data staged in those unprocessed bounce buffers is never transferred to the destination buffer. This leaves silent holes of uninitialized or stale data within the URB's payload area. Although the transfer completes successfully with no error code and reports the full requested length as actual_length, the application receives corrupted data that retains previous memory contents rather than the intended transmission from the USB device.

The vulnerability was identified in scenarios involving SuperSpeed USB mass storage devices operating behind a dm-verity target with 512-byte hash blocks. In this configuration, the block layer merges numerous small bios into large requests containing hundreds of scatterlist entries. When these requests are submitted to an xHCI controller with a wMaxPacketSize of 1024 bytes, the resulting TDs frequently span three or more ring segments because each segment boundary falls on odd multiples of 512 bytes, which are unaligned relative to the packet size. The dm-bufio driver subsequently caches hash blocks containing this stale data, causing dm-verity to detect and report metadata corruption errors such as device-mapper: verity: 8:2: metadata block corrupted. This demonstrates that while some systems may fail safely by detecting the inconsistency, others relying on silent success will suffer from subtle data integrity violations without any immediate indication of failure.

The resolution involves modifying the completion path to iterate through all ring segments between the start segment and the recorded bounce segment rather than processing only the final one. By walking this range during TD completion, the driver ensures that every pending bounce buffer is correctly unmapped and its contents copied back into the URB buffer. The logic carefully stops at td->bounce_seg instead of td->end_seg to prevent double-processing or copying foreign data from subsequent TDs that may have started in the same segment. Additionally, a check for ring->num_segs was introduced to guard against potential infinite loops if a TD wraps around the entire ring structure. This fix aligns with CWE-401 (Missing Release of Memory after Effective Lifetime) and CWE-200 (Exposure of Sensitive Information to an Unauthorized Actor), addressing both the resource leak and the data integrity issue inherent in the original implementation.

Responsible

Linux

Reservation

09/11/2026

Disclosure

09/17/2026

Moderation

accepted

CPE

ready

EPSS

0.00000

KEV

no

Activities

very low

Sources

Do you want to use VulDB in your project?

Use the official API to access entries easily!