[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH 02/10] x86/ucode: Delete the microcode_init() initcall


  • To: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
  • From: Jan Beulich <jbeulich@xxxxxxxx>
  • Date: Tue, 29 Oct 2024 08:48:14 +0100
  • Autocrypt: addr=jbeulich@xxxxxxxx; keydata= xsDiBFk3nEQRBADAEaSw6zC/EJkiwGPXbWtPxl2xCdSoeepS07jW8UgcHNurfHvUzogEq5xk hu507c3BarVjyWCJOylMNR98Yd8VqD9UfmX0Hb8/BrA+Hl6/DB/eqGptrf4BSRwcZQM32aZK 7Pj2XbGWIUrZrd70x1eAP9QE3P79Y2oLrsCgbZJfEwCgvz9JjGmQqQkRiTVzlZVCJYcyGGsD /0tbFCzD2h20ahe8rC1gbb3K3qk+LpBtvjBu1RY9drYk0NymiGbJWZgab6t1jM7sk2vuf0Py O9Hf9XBmK0uE9IgMaiCpc32XV9oASz6UJebwkX+zF2jG5I1BfnO9g7KlotcA/v5ClMjgo6Gl MDY4HxoSRu3i1cqqSDtVlt+AOVBJBACrZcnHAUSuCXBPy0jOlBhxPqRWv6ND4c9PH1xjQ3NP nxJuMBS8rnNg22uyfAgmBKNLpLgAGVRMZGaGoJObGf72s6TeIqKJo/LtggAS9qAUiuKVnygo 3wjfkS9A3DRO+SpU7JqWdsveeIQyeyEJ/8PTowmSQLakF+3fote9ybzd880fSmFuIEJldWxp Y2ggPGpiZXVsaWNoQHN1c2UuY29tPsJgBBMRAgAgBQJZN5xEAhsDBgsJCAcDAgQVAggDBBYC AwECHgECF4AACgkQoDSui/t3IH4J+wCfQ5jHdEjCRHj23O/5ttg9r9OIruwAn3103WUITZee e7Sbg12UgcQ5lv7SzsFNBFk3nEQQCACCuTjCjFOUdi5Nm244F+78kLghRcin/awv+IrTcIWF hUpSs1Y91iQQ7KItirz5uwCPlwejSJDQJLIS+QtJHaXDXeV6NI0Uef1hP20+y8qydDiVkv6l IreXjTb7DvksRgJNvCkWtYnlS3mYvQ9NzS9PhyALWbXnH6sIJd2O9lKS1Mrfq+y0IXCP10eS FFGg+Av3IQeFatkJAyju0PPthyTqxSI4lZYuJVPknzgaeuJv/2NccrPvmeDg6Coe7ZIeQ8Yj t0ARxu2xytAkkLCel1Lz1WLmwLstV30g80nkgZf/wr+/BXJW/oIvRlonUkxv+IbBM3dX2OV8 AmRv1ySWPTP7AAMFB/9PQK/VtlNUJvg8GXj9ootzrteGfVZVVT4XBJkfwBcpC/XcPzldjv+3 HYudvpdNK3lLujXeA5fLOH+Z/G9WBc5pFVSMocI71I8bT8lIAzreg0WvkWg5V2WZsUMlnDL9 mpwIGFhlbM3gfDMs7MPMu8YQRFVdUvtSpaAs8OFfGQ0ia3LGZcjA6Ik2+xcqscEJzNH+qh8V m5jjp28yZgaqTaRbg3M/+MTbMpicpZuqF4rnB0AQD12/3BNWDR6bmh+EkYSMcEIpQmBM51qM EKYTQGybRCjpnKHGOxG0rfFY1085mBDZCH5Kx0cl0HVJuQKC+dV2ZY5AqjcKwAxpE75MLFkr wkkEGBECAAkFAlk3nEQCGwwACgkQoDSui/t3IH7nnwCfcJWUDUFKdCsBH/E5d+0ZnMQi+G0A nAuWpQkjM1ASeQwSHEeAWPgskBQL
  • Cc: Roger Pau Monné <roger.pau@xxxxxxxxxx>, "Daniel P . Smith" <dpsmith@xxxxxxxxxxxxxxxxxxxx>, Xen-devel <xen-devel@xxxxxxxxxxxxxxxxxxxx>
  • Delivery-date: Tue, 29 Oct 2024 07:48:22 +0000
  • List-id: Xen developer discussion <xen-devel.lists.xenproject.org>

On 28.10.2024 18:12, Andrew Cooper wrote:
> On 28/10/2024 1:38 pm, Jan Beulich wrote:
>> On 28.10.2024 10:18, Andrew Cooper wrote:
>>> The comment highlights just how bogus this really is.  Being an initcall, 
>>> the
>>> boot allocator is long gone, and bootstrap_unmap() is a no-op.
>> How's the boot allocator coming into the picture here? This is all about
>> (un)mapping, not allocating.
>>
>>> The fact there is nothing to do should be a giant red flag about the 
>>> validity
>>> of the mappings "being freed".  Indeed, they both constitute 
>>> use-after-frees.
>> I can't spot any use-after-free; the pointers in question ...
>>
>>> --- a/xen/arch/x86/cpu/microcode/core.c
>>> +++ b/xen/arch/x86/cpu/microcode/core.c
>>> @@ -758,28 +758,6 @@ int microcode_update(XEN_GUEST_HANDLE(const_void) buf,
>>>      return continue_hypercall_on_cpu(0, microcode_update_helper, buffer);
>>>  }
>>>  
>>> -static int __init cf_check microcode_init(void)
>>> -{
>>> -    /*
>>> -     * At this point, all CPUs should have updated their microcode
>>> -     * via the early_microcode_* paths so free the microcode blob.
>>> -     */
>>> -    if ( ucode_blob.size )
>>> -    {
>>> -        bootstrap_unmap();
>>> -        ucode_blob.size = 0;
>>> -        ucode_blob.data = NULL;
>>> -    }
>>> -    else if ( ucode_mod.mod_end )
>>> -    {
>>> -        bootstrap_unmap();
>>> -        ucode_mod.mod_end = 0;
>>> -    }
>>> -
>>> -    return 0;
>>> -}
>>> -__initcall(microcode_init);
>> ... aren't used anywhere. bootstrap_unmap() is "just in case" (perhaps indeed
>> a no-op at least nowadays), and the rest is field clobbering. I'm okay with 
>> the
>> code change, so
>> Acked-by: Jan Beulich <jbeulich@xxxxxxxx>
>> yet I'd like to ask for the description to be "softened" some.
> 
> As I said, this could be folded into patch 9, given this particular
> arrangement of the series.

I certainly don't mind the folding, but then I'd still like to see the
description of the resulting patch before giving a (hopefully) unconditional
ack.

> The UAFs are apparent *because* the comment demonstrates a false line of
> reasoning.
> 
> ucode_mod literally is used after free.  ucode=$n is genuinely buggy
> today, because its a stash of a physical pointer across move_xen().

Maybe my problem is that the UAF is elsewhere, not in the code you delete?
If so, it might help to simply point out where the actual bad use is. As it
stands, microcode_init() doesn't have any uses (reads), only writes. What I
agree with is that the comment there is at best misleading.

Jan

> ucode_blob stashes a virtual pointer.  This was even noticed in
> dc380df12acf ("x86/ucode: load microcode earlier on boot CPU")
> 
> ---
> It needs to rescan the modules in order to find the new virtual address
> of the ucode blob because it changes during the boot process, e.g.
> from 0x00000000010802fc to 0xffff83204dac52fc.
> ---
> 
> which highlighted the problem but duct-taped over it.
> 
> ~Andrew




 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.