Re: X86: build failure

[Jan Beulich <jbeulich@suse.com>]

On 17.12.2025 00:54, Andrew Cooper wrote:
> On 16/12/2025 11:35 pm, Grygorii Strashko wrote:
>> Hi All,
>>
>> I experience build failure with following Kconfig combination:
>>  CONFIG_DEBUG=n and CONFIG_XEN_IBT=n
>>
>> with gcc:
>>  gcc --version
>>  gcc (Ubuntu 11.4.0-1ubuntu1~22.04.2) 11.4.0
>>  Copyright (C) 2021 Free Software Foundation, Inc.
>>  This is free software; see the source for copying conditions.  There
>> is NO
>>  warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR
>> PURPOSE.
>>
>> Is this known issue?
>>
>> ====== log ===============
>>   CC      arch/x86/x86_emulate.o
>> In file included from arch/x86/x86_emulate.c:27:
>> arch/x86/x86_emulate/x86_emulate.c: In function ‘x86_emulate’:
>> arch/x86/x86_emulate/x86_emulate.c:70:36: error: writing 1 byte into a
>> region of size 0 [-Werror=stringop-overflow=]
>>    70 |     gcc11_wrap(ptr)[0 - PFX_BYTES] = ext < ext_8f08 ? 0xc4 :
>> 0x8f; \
>> arch/x86/x86_emulate/x86_emulate.c:78:9: note: in expansion of macro
>> ‘copy_VEX’
>>    78 |         copy_VEX(ptr, vex); \
>>       |         ^~~~~~~~
>> arch/x86/x86_emulate/x86_emulate.c:8140:13: note: in expansion of
>> macro ‘copy_REX_VEX’
>>  8140 |             copy_REX_VEX(opc, rex_prefix, vex);
>>       |             ^~~~~~~~~~~~
>> In file included from arch/x86/x86_emulate/x86_emulate.c:11,
>>                  from arch/x86/x86_emulate.c:27:
>> arch/x86/x86_emulate/private.h:691:17: note: at offset [0, 4095] into
>> destination object of size [0, 9223372036854775807] allocated by
>> ‘map_domain_page’
>>   691 |     (stb).ptr = map_domain_page(_mfn(this_cpu(stubs.mfn))) + \
>>       |                 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>> arch/x86/x86_emulate/x86_emulate.c:60:21: note: in expansion of macro
>> ‘get_stub’
>>    60 |     uint8_t *buf_ = get_stub(stub); \
>>       |                     ^~~~~~~~
>> arch/x86/x86_emulate/x86_emulate.c:3354:15: note: in expansion of
>> macro ‘init_prefixes’
>>  3354 |         opc = init_prefixes(stub);
>>       |               ^~~~~~~~~~~~~
>> arch/x86/x86_emulate/private.h:691:17: note: at offset [1, 4096] into
>> destination object of size [0, 9223372036854775807] allocated by
>> ‘map_domain_page’
>>   691 |     (stb).ptr = map_domain_page(_mfn(this_cpu(stubs.mfn))) + \
>>       |                 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>> arch/x86/x86_emulate/x86_emulate.c:94:21: note: in expansion of macro
>> ‘get_stub’
>>    94 |     uint8_t *buf_ = get_stub(stub); \
>>       |                     ^~~~~~~~
>> arch/x86/x86_emulate/x86_emulate.c:3402:15: note: in expansion of
>> macro ‘init_evex’
>>  3402 |         opc = init_evex(stub);
>> [...]
>>
> 
> Hmm - that looks like more of the GCC 11 bounds bugs, but that's what
> the gcc11_wrap() visible at the top is supposed to be fixing.
> 
> Neither of those CONFIG options ought to be relevant.

CONFIG_DEBUG=n may well matter, due to the different -O<n> settings.
CONFIG_XEN_IBT=n otoh is indeed curious.

May want experimenting some starting from the pre-processed file. Grygorii,
could you generate that in your build env and then make it available
somewhere (together with the exact options passed to gcc when you observe
the build failure)?

Jan