On Wed, Sep 03, 2025 at 10:00:22AM -0400, Yazen Ghannam wrote:
> But any reason to use u32? Why not u8? Alignment or something?
Struct padding:
$ pahole --header elf64_hdr vmlinux
...
struct mce_amd_cpu_data {
mce_banks_t thr_intr_banks; /* 0 8 */
mce_banks_t dfr_intr_banks; /* 8 8 */
u8 thr_intr_en:1; /* 16: 0 1 */
u8 dfr_intr_en:1; /* 16: 1 1 */
u8 __resv:6; /* 16: 2 1 */
/* size: 24, cachelines: 1, members: 5 */
/* padding: 7 */
^^^^^^^^^^^
/* last cacheline: 24 bytes */
};
vs
struct mce_amd_cpu_data {
mce_banks_t thr_intr_banks; /* 0 8 */
mce_banks_t dfr_intr_banks; /* 8 8 */
u32 thr_intr_en:1; /* 16: 0 4 */
u32 dfr_intr_en:1; /* 16: 1 4 */
u32 __resv:30; /* 16: 2 4 */
/* size: 24, cachelines: 1, members: 5 */
/* padding: 4 */
^^^^^^^^^^
/* last cacheline: 24 bytes */
};
The end result is the same unless your do __packed.
But you might as well use the u32 in the distant case you need more flags
- you never know with our RAS folks.
:-P
--
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquette
