6.12-stable review patch. If anyone has any objections, please let me know.
------------------
From: Stefano Stabellini stefano.stabellini@amd.com
commit 7f9bbc1140ff8796230bc2634055763e271fd692 upstream.
dm_op hypercalls might come from userspace and pass memory addresses as parameters. The memory addresses typically correspond to buffers allocated in userspace to hold extra hypercall parameters.
On ARM, when CONFIG_ARM64_SW_TTBR0_PAN is enabled, they might not be accessible by Xen, as a result ioreq hypercalls might fail. See the existing comment in arch/arm64/xen/hypercall.S regarding privcmd_call for reference.
For privcmd_call, Linux calls uaccess_ttbr0_enable before issuing the hypercall thanks to commit 9cf09d68b89a. We need to do the same for dm_op. This resolves the problem.
Cc: stable@kernel.org Fixes: 9cf09d68b89a ("arm64: xen: Enable user access before a privcmd hvc call") Signed-off-by: Stefano Stabellini stefano.stabellini@amd.com Reviewed-by: Juergen Gross jgross@suse.com Message-ID: alpine.DEB.2.22.394.2505121446370.8380@ubuntu-linux-20-04-desktop Signed-off-by: Juergen Gross jgross@suse.com Signed-off-by: Greg Kroah-Hartman gregkh@linuxfoundation.org --- arch/arm64/xen/hypercall.S | 21 ++++++++++++++++++++- 1 file changed, 20 insertions(+), 1 deletion(-)
--- a/arch/arm64/xen/hypercall.S +++ b/arch/arm64/xen/hypercall.S @@ -83,7 +83,26 @@ HYPERCALL3(vcpu_op); HYPERCALL1(platform_op_raw); HYPERCALL2(multicall); HYPERCALL2(vm_assist); -HYPERCALL3(dm_op); + +SYM_FUNC_START(HYPERVISOR_dm_op) + mov x16, #__HYPERVISOR_dm_op; \ + /* + * dm_op hypercalls are issued by the userspace. The kernel needs to + * enable access to TTBR0_EL1 as the hypervisor would issue stage 1 + * translations to user memory via AT instructions. Since AT + * instructions are not affected by the PAN bit (ARMv8.1), we only + * need the explicit uaccess_enable/disable if the TTBR0 PAN emulation + * is enabled (it implies that hardware UAO and PAN disabled). + */ + uaccess_ttbr0_enable x6, x7, x8 + hvc XEN_IMM + + /* + * Disable userspace access from kernel once the hyp call completed. + */ + uaccess_ttbr0_disable x6, x7 + ret +SYM_FUNC_END(HYPERVISOR_dm_op);
SYM_FUNC_START(privcmd_call) mov x16, x0