Re: [PATCH 3/3] x86/alternative: make kernel ITS thunks read-only

["Jürgen Groß" <jgross@suse.com>]

[-- Attachment #1.1.1: Type: text/plain, Size: 2586 bytes --]

On 28.05.25 15:22, Peter Zijlstra wrote:
> On Wed, May 28, 2025 at 03:19:24PM +0200, Jürgen Groß wrote:
>> On 28.05.25 15:10, Peter Zijlstra wrote:
>>> On Wed, May 28, 2025 at 02:35:57PM +0200, Juergen Gross wrote:
>>>> When allocating memory pages for kernel ITS thunks, make them read-only
>>>> after having written the last thunk.
>>>>
>>>> This will be needed when X86_FEATURE_PSE isn't available, as the thunk
>>>> memory will have PAGE_KERNEL_EXEC protection, which is including the
>>>> write permission.
>>>>
>>>> Cc: <stable@vger.kernel.org>
>>>> Fixes: 5185e7f9f3bd ("x86/module: enable ROX caches for module text on 64 bit")
>>>> Signed-off-by: Juergen Gross <jgross@suse.com>
>>>> ---
>>>>    arch/x86/kernel/alternative.c | 16 ++++++++++++++++
>>>>    1 file changed, 16 insertions(+)
>>>>
>>>> diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
>>>> index ecfe7b497cad..bd974a0ac88a 100644
>>>> --- a/arch/x86/kernel/alternative.c
>>>> +++ b/arch/x86/kernel/alternative.c
>>>> @@ -217,6 +217,15 @@ static void *its_alloc(void)
>>>>    	return no_free_ptr(page);
>>>>    }
>>>> +static void its_set_kernel_ro(void *addr)
>>>> +{
>>>> +#ifdef CONFIG_MODULES
>>>> +	if (its_mod)
>>>> +		return;
>>>> +#endif
>>>> +	execmem_restore_rox(addr, PAGE_SIZE);
>>>> +}
>>>> +
>>>>    static void *its_allocate_thunk(int reg)
>>>>    {
>>>>    	int size = 3 + (reg / 8);
>>>> @@ -234,6 +243,8 @@ static void *its_allocate_thunk(int reg)
>>>>    #endif
>>>>    	if (!its_page || (its_offset + size - 1) >= PAGE_SIZE) {
>>>> +		if (its_page)
>>>> +			its_set_kernel_ro(its_page);
>>>>    		its_page = its_alloc();
>>>>    		if (!its_page) {
>>>>    			pr_err("ITS page allocation failed\n");
>>>> @@ -2338,6 +2349,11 @@ void __init alternative_instructions(void)
>>>>    	apply_retpolines(__retpoline_sites, __retpoline_sites_end);
>>>>    	apply_returns(__return_sites, __return_sites_end);
>>>> +	/* Make potential last thunk page read-only. */
>>>> +	if (its_page)
>>>> +		its_set_kernel_ro(its_page);
>>>> +	its_page = NULL;
>>>> +
>>>>    	/*
>>>>    	 * Adjust all CALL instructions to point to func()-10, including
>>>>    	 * those in .altinstr_replacement.
>>>
>>> No, this is all sorts of wrong. Execmem API should ensure this.
>>
>> You are aware that this patch is basically mirroring the work which is
>> already done for modules in alternative.c?
> 
> I am having trouble parsing that -- where does alternative.c do this to
> modules?

Have a look at its_fini_mod().


Juergen

[-- Attachment #1.1.2: OpenPGP public key --]
[-- Type: application/pgp-keys, Size: 3743 bytes --]

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 495 bytes --]