CVE-2026-99590: ceph: clamp the inline data length in ceph_fill_inline_data()
In the Linux kernel, the following vulnerability has been resolved:
ceph: clamp the inline data length in cephfillinlinedata()
The MDS decides how much inline data to attach to a reply, and the client parses inlinelen without any upper bound: cephdecodeneed() only verifies that the message actually carries that many bytes. cephfillinlinedata() then copies the data into a single folio with no length check, so a malicious or buggy MDS returning more inline data than the folio can hold writes past its end.
Clamp the length to foliosize() so both callers, handlecapgrant() and fillinode(), are covered.
Affected Software
Frequently Asked Questions
Who is exposed to this issue?
Linux kernel Ceph clients that process inline data supplied in MDS replies are exposed. The issue can be triggered by a malicious or buggy MDS that returns more inline data than a destination folio can hold.
What must an attacker or faulty component do to trigger it?
The MDS must provide an inline_len value and corresponding inline-data payload that exceed the size of the folio used by ceph_fill_inline_data(). The existing message-length validation only confirms that the reply contains the claimed bytes; it does not ensure they fit in the folio.
What is the impact of receiving an oversized inline-data reply?
The client can copy inline data past the end of a single folio, causing an out-of-bounds write. Both handle_cap_grant() and fill_inode() can reach the affected copy path.
What should be done if patching cannot happen immediately?
The provided information identifies the MDS reply as the source of the oversized data. Limiting clients to trusted, correctly functioning MDS servers is the available mitigation described by the issue; no client-side configuration workaround is provided.