CVE-2026-89986 PUBLISHED

mm/mempolicy: fix sleeping allocation in alloc_pages_bulk_weighted_interleave()

Assigner: Linux
Reserved: 11.09.2026 Published: 16.09.2026 Updated: 16.09.2026

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

mm/mempolicy: fix sleeping allocation in alloc_pages_bulk_weighted_interleave()

syzbot reported a sleeping function called from invalid context splat in bucket_table_alloc().

When rhashtable_insert_slow() rehashes the table under rcu_read_lock(), it calls bucket_table_alloc(..., GFP_ATOMIC | __GFP_NOWARN). If the bucket table allocation uses vmalloc, __vmalloc_node_range_noprof() invokes vm_area_alloc_pages() -> alloc_pages_bulk_mempolicy_noprof() with the passed GFP_ATOMIC flags.

If the current task has an MPOL_WEIGHTED_INTERLEAVE mempolicy, alloc_pages_bulk_weighted_interleave() is called and currently hardcodes GFP_KERNEL when allocating the temporary weights array, triggering a might_alloc() splat in atomic/RCU contexts.

Pass the gfp flags (masked with GFP_RECLAIM_MASK to strip page-allocator zone modifiers like __GFP_HIGHMEM) received by alloc_pages_bulk_weighted_interleave() to kmalloc() instead of hardcoding GFP_KERNEL. Since the weights buffer is immediately initialized in full, kmalloc() is sufficient.

Metrics

CVSS Vector: CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H
CVSS Score: 7.8

AV:L - Triggering requires set_mempolicy(MPOL_WEIGHTED_INTERLEAVE) on the attacker's task and a local syscall that grows an rhashtable until GFP_ATOMIC kvmalloc falls back to vmalloc (nftables hash sets, tc flower, BPF rhashtable maps); softirq/packet paths use default_policy via in_interrupt() and cannot apply the attacker's mempolicy. AC:L - The attacker fully controls setup: set_mempolicy on their own task, insert enough rhashtable elements for a >PAGE_SIZE bucket table, and apply memory pressure so GFP_ATOMIC kmalloc fails and kvmalloc uses vmalloc; CONFIG_NUMA is standard on x86_64/arm64 distro and cloud kernels, and no attacker-uncontrollable race is required. PR:L - set_mempolicy() has no capability check, and the rhashtable rehash path is reachable by an unprivileged user via nftables or tc in a user+net namespace where CAP_NET_ADMIN is namespace-local; init-namespace root is not required. UI:N - The attacker sets their own mempolicy and drives the rhashtable inserts themselves; no victim action such as mounting a filesystem or opening a file is required. S:U - The illegal GFP_KERNEL allocation and any resulting RCU use-after-free of the rhashtable bucket table stay inside the host kernel's memory-management authority; this is standard in-kernel impact, not a VM, IOMMU, or sandbox boundary crossing. C:H - Sleeping under rcu_read_lock() in rhashtable_insert_slow() on non-PREEMPT_RCU kernels can let the grace period complete and free the old bucket table before rhashtable_rehash_attach() uses it, yielding a use-after-free read of kernel heap. I:H - The same RCU-break UAF lets rhashtable_rehash_attach() cmpxchg old_tbl->future_tbl on a freed table, which is an attacker-sprayable kernel write primitive via heap reuse rather than a clean failure. A:H - GFP_KERNEL under rcu_read_lock produces a sleeping-from-invalid-context BUG/WARN (panic with panic_on_warn), can RCU-stall or deadlock if reclaim actually sleeps, and the bucket-table UAF can oops the kernel.

Product Status

Vendor Linux
Product Linux
Versions Default: unaffected
  • affected from fa3bea4e1f8202d787709b7e3654eb0a99aed758 to bcb3d0c867ee40dc48e9c085bf328fbc679b6656 (excl.)
  • affected from fa3bea4e1f8202d787709b7e3654eb0a99aed758 to 0ceda28f371df9e0bbdaa29214f71fe8298f23d8 (excl.)
  • affected from fa3bea4e1f8202d787709b7e3654eb0a99aed758 to 2943f1f4b7f2816177060eb9f551f2e6d8b629ba (excl.)
  • affected from fa3bea4e1f8202d787709b7e3654eb0a99aed758 to 540e583b66d6402bf556fde5e53c817a54c1afe5 (excl.)
Vendor Linux
Product Linux
Versions Default: affected
  • Version 6.9 is affected
  • unaffected from 0 to 6.9 (excl.)
  • unaffected from 6.12.110 to 6.12.* (incl.)
  • unaffected from 6.18.51 to 6.18.* (incl.)
  • unaffected from 7.2.5 to 7.2.* (incl.)
  • unaffected from 7.3-rc2 to * (incl.)

References