CVE-2026-31787

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> xen/privcmd: fix double free via VMA splitting<br /> <br /> privcmd_vm_ops defines .close (privcmd_close), but neither .may_split<br /> nor .open. When userspace does a partial munmap() on a privcmd mapping,<br /> the kernel splits the VMA via __split_vma(). Since may_split is NULL,<br /> the split is allowed. vm_area_dup() copies vm_private_data (a pages<br /> array allocated in alloc_empty_pages()) into the new VMA without any<br /> fixup, because there is no .open callback.<br /> <br /> Both VMAs now point to the same pages array. When the unmapped portion<br /> is closed, privcmd_close() calls:<br /> - xen_unmap_domain_gfn_range()<br /> - xen_free_unpopulated_pages()<br /> - kvfree(pages)<br /> <br /> The surviving VMA still holds the dangling pointer. When it is later<br /> destroyed, the same sequence runs again, which leads to a double free.<br /> <br /> Fix this issue by adding a .may_split callback denying the VMA split.<br /> <br /> This is XSA-487 / CVE-2026-31787