fs: move inode fields used during fast path lookup closer together
authorMateusz Guzik <mjguzik@gmail.com>
Sun, 9 Nov 2025 12:19:31 +0000 (13:19 +0100)
committerChristian Brauner <brauner@kernel.org>
Tue, 11 Nov 2025 09:49:54 +0000 (10:49 +0100)
commitdca3aa666fbd71118905d88bb1c353881002b647
treec0287393a35f1d1536d5ed4c4e542994f7815ae9
parent5b8ed52866e3d19e02860c7cf1d6bbbd70b619e9
fs: move inode fields used during fast path lookup closer together

This should avoid *some* cache misses.

Successful path lookup is guaranteed to load at least ->i_mode,
->i_opflags and ->i_acl. At the same time the common case will avoid
looking at more fields.

struct inode is not guaranteed to have any particular alignment, notably
ext4 has it only aligned to 8 bytes meaning nearby fields might happen
to be on the same or only adjacent cache lines depending on luck (or no
luck).

According to pahole:
        umode_t                    i_mode;               /*     0     2 */
        short unsigned int         i_opflags;            /*     2     2 */
        kuid_t                     i_uid;                /*     4     4 */
        kgid_t                     i_gid;                /*     8     4 */
        unsigned int               i_flags;              /*    12     4 */
        struct posix_acl *         i_acl;                /*    16     8 */
        struct posix_acl *         i_default_acl;        /*    24     8 */

->i_acl is unnecessarily separated by 8 bytes from the other fields.
With struct inode being offset 48 bytes into the cacheline this means an
avoidable miss. Note it will still be there for the 56 byte case.

New layout:
        umode_t                    i_mode;               /*     0     2 */
        short unsigned int         i_opflags;            /*     2     2 */
        unsigned int               i_flags;              /*     4     4 */
        struct posix_acl *         i_acl;                /*     8     8 */
        struct posix_acl *         i_default_acl;        /*    16     8 */
        kuid_t                     i_uid;                /*    24     4 */
        kgid_t                     i_gid;                /*    28     4 */

I verified with pahole there are no size or hole changes.

This is stopgap until someone(tm) sanitizes the layout in the first
place, allocation methods aside.

Signed-off-by: Mateusz Guzik <mjguzik@gmail.com>
Link: https://patch.msgid.link/20251109121931.1285366-1-mjguzik@gmail.com
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>
include/linux/fs.h