On 5/14/2026 7:46 AM, Philippe Mathieu-Daudé wrote:
> On 14/5/26 16:16, Stefan Hajnoczi wrote:
>> On Tue, May 12, 2026 at 2:32 PM Pierrick Bouvier
>> <[email protected]> wrote:
>>>
>>> The following changes since commit
>>> 5e61afe211e82a9af15a8794a0bd29bb574e953b:
>>>
>>>    Merge tag 'ui-pull-request' of https://gitlab.com/
>>> marcandre.lureau/qemu into staging (2026-05-11 10:49:53 -0400)
>>>
>>> are available in the Git repository at:
>>>
>>>    https://gitlab.com/p-b-o/qemu tags/pbouvier/pr/target_info-20260512
>>>
>>> for you to fetch changes up to 0fc2022c6331c5d5bdd9060c73d067abba3f5515:
>>>
>>>    target-info: replace target_info() in system-mode (2026-05-12
>>> 11:29:18 -0700)
>>>
>>> ----------------------------------------------------------------
>>> Changes:
>>> - [PATCH v6 0/5] single-binary: deduplicate target_info() (Pierrick
>>> Bouvier <[email protected]>)
>>>    Link: https://lore.kernel.org/qemu-devel/20260512150208.1167711-1-
>>> [email protected]
>>>
>>> ----------------------------------------------------------------
>>> Pierrick Bouvier (5):
>>>        qom/object: register OBJECT and INTERFACE QOM types before main
>>>        qom/object: initialize type_table in static ctor with
>>> fundamental QOM types
>>>        target-info: introduce TargetInfo in QOM
>>>        target-info-qom: detect target from QOM
>>>        target-info: replace target_info() in system-mode
>>
>> Please take a look at this CI failure:
>>
>> Testing ./build-oss-fuzz/DEST_DIR/qemu-fuzz-i386-target-generic-fuzz-
>> cirrus-vga
>> ...
>> AddressSanitizer:DEADLYSIGNAL
>> =================================================================
>> ==5334==ERROR: AddressSanitizer: SEGV on unknown address
>> 0x000000000020 (pc 0x5bd0beb39ef4 bp 0x7ffe0ad764d0 sp 0x7ffe0ad764c0
>> T0)
>> ==5334==The signal is caused by a READ memory access.
>> ==5334==Hint: address points to the zero page.
>> #0 0x5bd0beb39ef4 in target_endian_mode
>> /builds/qemu-project/qemu/build-oss-fuzz/../target-info.c:42:27
>> #1 0x5bd0beb39ef4 in target_big_endian
>> /builds/qemu-project/qemu/build-oss-fuzz/../target-info.c:47:12
>> #2 0x5bd0bf4f1da1 in qtest_process_command
>> /builds/qemu-project/qemu/build-oss-fuzz/../system/qtest.c:705:13
>> #3 0x5bd0bf4f1da1 in qtest_process_inbuf
>> /builds/qemu-project/qemu/build-oss-fuzz/../system/qtest.c:778:9
>> #4 0x5bd0bf4eef54 in qtest_server_inproc_recv
>> /builds/qemu-project/qemu/build-oss-fuzz/../system/qtest.c:910:9
>> #5 0x5bd0bfe18f85 in qtest_sendf
>> /builds/qemu-project/qemu/build-oss-fuzz/../tests/qtest/libqtest.c:715:5
>> #6 0x5bd0bfe15294 in qtest_query_target_endianness
>> /builds/qemu-project/qemu/build-oss-fuzz/../tests/qtest/libqtest.c:804:5
>> #7 0x5bd0bfe1f2f3 in qtest_inproc_init
>> /builds/qemu-project/qemu/build-oss-fuzz/../tests/qtest/
>> libqtest.c:2043:23
>> #8 0x5bd0bfe71778 in qtest_setup
>> /builds/qemu-project/qemu/build-oss-fuzz/../tests/qtest/fuzz/fuzz.c:64:12
>> #9 0x5bd0bfe71778 in LLVMFuzzerInitialize
>> /builds/qemu-project/qemu/build-oss-fuzz/../tests/qtest/fuzz/
>> fuzz.c:206:16
>> #10 0x5bd0be9b3b56 in fuzzer::FuzzerDriver(int*, char***, int
>> (*)(unsigned char const*, unsigned long)) fuzzer.o
>> #11 0x5bd0be9e2166 in main
> 
> 
> IIUC the fuzzer mode doesn't call main() -> qemu_init() but a distinct
> one via LLVMFuzzerInitialize().
> 
> We'd need to squash something like:
> 
> -- >8 --
> diff --git a/tests/qtest/fuzz/fuzz.c b/tests/qtest/fuzz/fuzz.c
> index ca248a51a6c..90f07dcda3d 100644
> --- a/tests/qtest/fuzz/fuzz.c
> +++ b/tests/qtest/fuzz/fuzz.c
> @@ -172,6 +172,7 @@ int LLVMFuzzerInitialize(int *argc, char ***argv,
> char ***envp)
>      /* Initialize qgraph and modules */
>      qos_graph_init();
>      module_call_init(MODULE_INIT_FUZZ_TARGET);
> +    module_call_init(MODULE_INIT_TARGET_INFO);
>      module_call_init(MODULE_INIT_QOM);
>      module_call_init(MODULE_INIT_LIBQOS);
>

Yes, this looks like the issue.
I'll reproduce, send a v7, and send PR once all tests are completed.

> ---
> 
>> (/builds/qemu-project/qemu/build-oss-fuzz/DEST_DIR/qemu-fuzz-i386-
>> target-generic-fuzz-cirrus-vga+0x46166)
>> (BuildId: 845b9265a3ae1439c62fb86e3bc4973f7ea7c960)
>> #12 0x7f7fef9a65b4 in __libc_start_call_main (/lib64/libc.so.6+0x35b4)
>> (BuildId: abd2a4d896fed122b3f7da571746f193feeb86a1)
>> #13 0x7f7fef9a6667 in __libc_start_main@GLIBC_2.2.5
>> (/lib64/libc.so.6+0x3667) (BuildId:
>> abd2a4d896fed122b3f7da571746f193feeb86a1)
>> #14 0x5bd0be9a8f34 in _start
>> (/builds/qemu-project/qemu/build-oss-fuzz/DEST_DIR/qemu-fuzz-i386-
>> target-generic-fuzz-cirrus-vga+0xcf34)
>> (BuildId: 845b9265a3ae1439c62fb86e3bc4973f7ea7c960)
>>
>> https://gitlab.com/qemu-project/qemu/-/jobs/14371480130
>>
>> If this issue is caused by this pull request, please send a new
>> revision that solves the issue.
>>
>> If this issue is not caused by this pull request, please let me know.
>> Thanks!
>>

Seems like it's definitely related to this PR, I'll fix the series.

>> Stefan
>>
>>>
>>>   configs/targets/aarch64-softmmu.c |  6 ++---
>>>   configs/targets/arm-softmmu.c     |  6 ++---
>>>   include/qemu/module.h             |  1 +
>>>   include/qemu/target-info-init.h   | 53 ++++++++++++++++++++++++++++
>>> +++++++++++
>>>   include/qemu/target-info-qom.h    | 30 ++++++++++++++++++++++
>>>   qom/object.c                      | 22 +++++-----------
>>>   system/vl.c                       |  4 +++
>>>   target-info-qom.c                 | 45 ++++++++++++++++++++++++++++
>>> +++++
>>>   target-info-stub.c                |  6 ++---
>>>   9 files changed, 145 insertions(+), 28 deletions(-)
>>>   create mode 100644 include/qemu/target-info-init.h
>>>   create mode 100644 include/qemu/target-info-qom.h
>>>
>>
> 

Thanks,
Pierrick

Reply via email to