On Tue Apr 28, 2026 at 10:27 PM CEST, Rosen Penev wrote: > On Tue, Apr 28, 2026 at 2:30 AM Luca Ceresoli <[email protected]> > wrote: >> >> Hello Thorsten, >> >> +Cc Romain, author of the involved code >> +Cc Rosen, author of a similar patch elsewhere in i2c-atr [0] >> >> [0] https://lore.kernel.org/linux-i2c/[email protected]/ >> >> On Mon Apr 27, 2026 at 7:28 PM CEST, Thorsten Blum wrote: >> > Add the __counted_by_ptr() compiler attribute to ->aliases to improve >> > bounds checking via CONFIG_UBSAN_BOUNDS and CONFIG_FORTIFY_SOURCE. >> > >> > Signed-off-by: Thorsten Blum <[email protected]> >> > --- >> > include/linux/i2c-atr.h | 2 +- >> > 1 file changed, 1 insertion(+), 1 deletion(-) >> > >> > diff --git a/include/linux/i2c-atr.h b/include/linux/i2c-atr.h >> > index 2bb54dc87c8e..b52a7b9ec536 100644 >> > --- a/include/linux/i2c-atr.h >> > +++ b/include/linux/i2c-atr.h >> > @@ -71,7 +71,7 @@ struct i2c_atr_adap_desc { >> > struct device *parent; >> > struct fwnode_handle *bus_handle; >> > size_t num_aliases; >> > - u16 *aliases; >> > + u16 *aliases __counted_by_ptr(num_aliases); >> >> I'm not fully proficient about all counted_by things, but the patch looks >> OK to me, and definitely @num_aliases is the number of items @aliases must >> have, so: > I prefer my patch. > > __counted_by is supported by more versions of GCC/Clang than __counted_by_ptr.
At a quick grep check, they seem to be different things: __counted_by for arrays embedded in structs, __counted_by_ptr for pointers in structs for dynamically allocated arrays. I wish this were documented, though, especially as their implementation is currently identical. :-/ Based on this, I think both Rosen's and Thorsten's changes are OK, using different macros for different members. Luca -- Luca Ceresoli, Bootlin Embedded Linux and Kernel engineering https://bootlin.com
