CVE-2026-74447drm/amdkfd: fix uint32_t overflow in EOP ring buffer size alignment

Official description

In the Linux kernel, the following vulnerability has been resolved: drm/amdkfd: fix uint32_t overflow in EOP ring buffer size alignment eop_ring_buffer_size in struct queue_properties is a u32. In kfd_queue_acquire_buffers() the expected EOP buffer size is computed as ALIGN(eop_ring_buffer_size, PAGE_SIZE); ALIGN uses typeof(x), so the addition is done in 32-bit. A user-supplied size of 0xFFFFF001 wraps to 0, causing kfd_queue_buffer_get() to skip its exact-size check (gated on size != 0) and accept any BO mapped at the address. On GFX8/GFX9 the MQD cp_hqd_eop_control is then programmed for an 8KB EOP ring backed by a 4KB BO, so CP EOP writes can land past the buffer and fault the GPU. Cast the operand to u64 so the alignment is computed in 64-bit; the size check in kfd_queue_buffer_get() then rejects the oversized request. (cherry picked from commit ae443117b742c357bfef3a7bddabf76fcf86e9ef)

Affected packages & fixed versions

Affected products (as reported)

References

• https://git.kernel.org/stable/c/273548eb997c6be85230c1236b18784b09f6203c
• https://git.kernel.org/stable/c/6dc0b4b39ed4f11ef70f76ecea8537e35f45342b
• https://git.kernel.org/stable/c/7c54bd225d83bc1bcb44430ed4b4d3a5c36b1961
• https://git.kernel.org/stable/c/83463a96ea3c7d8ae636a4d6a0ba63c9ce410724