> On May 7, 2024, at 17:52, Kees Cook <keesc...@chromium.org> wrote: > > On Tue, May 07, 2024 at 06:34:19PM +0000, Qing Zhao wrote: >> On May 7, 2024, at 13:57, Sebastian Huber >> <sebastian.hu...@embedded-brains.de> wrote: >>> On 07.05.24 16:26, Qing Zhao wrote: >>>> Hi, Sebastian, >>>> Thanks for your explanation. >>>> Our goal is to deprecate the GCC extension on structure >>>> containing a flexible array member not at the end of another >>>> structure. In order to achieve this goal, we provided the warning option >>>> -Wflex-array-member-not-at-end for the users to locate all such >>>> cases in their source code and update the source code to eliminate >>>> such cases. >>> >>> What is the benefit of deprecating this GCC extension? If GCC >>> extensions are removed, then it would be nice to enable the associated >>> warnings by default. > > The goal of all of the recent array bounds and flexible array work is to > make sizing information unambiguous (e.g. via __builtin_object_size(), > __builtin_dynamic_object_size(), and the array-bounds sanitizer). For > the compiler to be able to deterministically report size information > on arrays, we needed to deprecate this case even though it had been > supported in the past. (Though we also _added_ extensions to support > for other things, like flexible arrays in unions, and the coming > __counted_by attribute.) > > For example: > > struct flex { int length; char data[]; }; > struct mid_flex { int m; struct flex flex_data; int n; int o; }; > > #define SZ(p) __builtin_dynamic_object_size(p, 1) > > void foo(struct flex *f, struct mid_flex *m) > { > printf("%zu\n", SZ(f)); > printf("%zu\n", SZ(m->flex_data)); > } > > int main(void) > { > struct mid_flex m = { .flex_data.length = 8 }; > foo(&m->flex_data, &m); > return 0; > } > > This is printing the size of the same object. But the desired results > are ambiguous. Does m->flex_data have an unknown size (i.e. SIZE_MAX) > because it's a flex array, or does it contain 8 bytes, since it overlaps > with the other structure's trailing 2 ints? > > The answer from GCC 13 was neither: > > 18446744073709551615 > 4 > > It considered flex_data to be only the size of it's non-flex-array > members, but only when there was semantic context that it was part of > another structure. (Yet more ambiguity.) > > In GCC 14, this is "resolved" to be unknown since it is a flex array > which has no sizing info, and context doesn't matter: > > 18446744073709551615 > 18446744073709551615 > > But this paves the way for the coming 'counted_by' attribute which will > allow for struct flex above to be defined as: > > struct flex { int length; char data[] __attribute__((counted_by(length))); }; > > At which point GCC can deterministically report the object size. > > Hopefully I've captured this all correctly -- Qing can correct me. :) > >>> >>>> We had a long discussion before deciding to deprecating this GCC >>>> extension. Please see details here: >>>> >>>> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101832 >>>> >>>> Yes, we do plan to enable this warning by default before final >>>> deprecation. (Might consider to enable this warning by default in >>>> GCC15… and then deprecate it in the next release) >>>> >>>> Right now, there is an ongoing work in Linux kernel to get rid of >>>> all such cases. Kees might have more information on this. >>>> >>>> >>>> The static initialization of structures with flexible array members >>>> will still work as long as the flexible array members are at the end of >>>> the structures. >>> >>> Removing the support for flexible array members in the middle of >>> compounds will make the static initialization practically infeasible. >> >> If the flexible array member is moved to the end of the compounds, >> the static initialization still work. What’s the issue here? >> >>>> My question: is it possible to update your source code to move >>>> the structure with flexible array member to the end of the containing >>>> structure? >>>> >>>> i.e, in your example, in the struct Thread_Configured_control, >>>> move the field “Thread_Control Control” to the end of the structure? >>> >>> If we move the Thread_Control to the end, how would I add a >>> configuration defined number of elements at the end? >> >> Don’t understand this, why moving the Thread_Control Control” to >> the end of the containing structure will make this a problem? >> Could you please explain this with a simplified example? > > I found your example at [2] and tried to trim/summarize it here: > > > struct _Thread_Control { > Objects_Control Object; > ... > void *extensions[]; > }; > typedef struct _Thread_Control Thread_Control; > > struct Thread_Configured_control { > Thread_Control Control; > > #if CONFIGURE_MAXIMUM_USER_EXTENSIONS > 0 > void *extensions[ CONFIGURE_MAXIMUM_USER_EXTENSIONS + 1 ]; > #endif > Configuration_Scheduler_node Scheduler_nodes[ _CONFIGURE_SCHEDULER_COUNT ]; > RTEMS_API_Control API_RTEMS; > #ifdef RTEMS_POSIX_API > POSIX_API_Control API_POSIX; > #endif > #if CONFIGURE_MAXIMUM_THREAD_NAME_SIZE > 1 > char name[ CONFIGURE_MAXIMUM_THREAD_NAME_SIZE ]; > #endif > #if defined(_CONFIGURE_ENABLE_NEWLIB_REENTRANCY) && \ > !defined(_REENT_THREAD_LOCAL) > struct _reent Newlib; > #endif > }; > > #define THREAD_INFORMATION_DEFINE( name, api, cls, max ) \ > ... > static ... \ > Thread_Configured_control \ > name##_Objects[ _Objects_Maximum_per_allocation( max ) ]; \ > ... > > > I don't see any static initialization of struct _Thread_Control::extensions > nor any member initialization of the name##_Objects, and even then that > is all legal in any arrangement:
Thanks for the simplified code portion. From the above, I guess: For the structure: struct Thread_Configured_control { Thread_Control Control; #if CONFIGURE_MAXIMUM_USER_EXTENSIONS > 0 void *extensions[ CONFIGURE_MAXIMUM_USER_EXTENSIONS + 1 ]; #endif …. } The 2nd field “void *extensions[CONFIGURE_MAXIMUM_USER_EXTENSIONS + 1 ] ” is following the 1st field “Thread_Control Control”, whose type is a structure with trailing flexible array member field struct _Thread_Control { Objects_Control Object; ... void *extensions[]; }; So, the Thread_Configured_control.Control.extensions[] is overlapped with Thread_Configured_control.extensions[CONFIGURE_MAXIMUM_USER_EXTENSIONS + 1]. And the source code depends on the above overlapping relationship. (The initialization to the field Thread_Configured_control.extensions[CONFIGURE_MAXIMUM_USER_EXTENSIONS + 1] will be an initialization to the overlapped flexible array member field). If so, moving the field “Thread_Control Control” to the last field of the containing structure will break the overlapping relationship. Therefore resulting incorrect code. Not sure whether the above a correct understanding or not. If my understanding is correct, I do feel that such coding style is really not good, not easily to be understood or to be debugged. Not sure whether the new GCC extension: permitting flexible array member in union, could be used to replace such code? Qing > > truct flex { int length; char data[]; }; > struct mid_flex { int m; struct flex flex_data; int n; int o; }; > struct end_flex { int m; int n; struct flex flex_data; }; > > struct flex f = { .length = 2 }; > struct mid_flex m = { .m = 5 }; > struct end_flex e = { .m = 5 }; > > struct flex fa[4] = { { .length = 2 } }; > struct mid_flex ma[4] = { { .m = 5 } }; > struct end_flex ea[4] = { { .m = 5 } }; > > These all work. > > > But yes, I see why you can't move Thread_Control trivially to the end. It > looks like you're depending on the implicit overlapping memory locations > between struct _Thread_Control and things that include it as the first > struct member, like struct Thread_Configured_control above: > > cpukit/score/src/threaditerate.c: the_thread = (Thread_Control *) > information->local_table[ index ]; > > (In the Linux kernel we found this kind of open casting to be very > fragile and instead use a horrific wrapper called "container_of"[3] that > does the pointer math (possibly to an offset of 0 for a direct cast) to > find the member.) > > Anyway, for avoiding the warning, you can just keep using the extension > and add -Wno-... if it ever ends up in -Wall, or you can redefine struct > _Thread_Control to avoid having the "extensions" member at all. This is > what we've done in several cases in Linux. For example if we had this > again, but made to look more like Thread_Control: > > struct flex { int foo; int bar; char data[]; }; > struct mid_flex { struct flex hdr; int n; int o; }; > > It could be changed to: > > struct flex_hdr { int foo; int bar; }; > struct flex { struct flex_hdr hdr; char data[]; }; > struct mid_flex { struct flex_hdr hdr; int n; int o; }; > > This has some collateral changes needed to reference the struct flex_hdr > members from struct flex now (f->hdr.foo instead of f->foo). Sometimes > this can be avoided by using a union, as I did in a recent refactoring > in Linux: [4] > > For more complex cases in Linux we've handled this by using our > "struct_group"[5] macro, which allows for a union and tagged struct to > be constructed: > > struct flex { > __struct_group(flex_hdr, hdr,, > int foo; > int bar; > ); > char data[]; > }; > struct mid_flex { struct flex_hdr hdr; int n; int o; }; > > Then struct flex member names don't have to change, but if anything is > trying to get at struct flex::data through struct mid_flex::hdr, that'll > need casting. But it _shouldn't_ since it has "n" and "o". > > -Kees > > [1] https://gcc.gnu.org/pipermail/gcc-patches/2023-May/620122.html > [2] https://github.com/RTEMS/rtems > [3] > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/include/linux/container_of.h#n10 > [4] https://git.kernel.org/linus/896880ff30866f386ebed14ab81ce1ad3710cfc4 > [5] > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/include/uapi/linux/stddef.h?h=v6.8#n11 > > -- > Kees Cook