Re: [PATCH] fw_cfg: use __ATTR_RO_MODE to define rev sysfs

From: Michael S. Tsirkin
Date: Wed Feb 27 2019 - 08:51:17 EST


On Wed, Feb 27, 2019 at 01:33:19PM +0800, Wei Yang wrote:
> On Tue, Feb 26, 2019 at 11:10:06AM -0500, Michael S. Tsirkin wrote:
> >On Tue, Feb 26, 2019 at 03:31:59PM +0800, Wei Yang wrote:
> >> Leverage __ATTR_RO_MODE to define rev sysfs instead of using open code
> >> to define the attribute.
> >>
> >> Signed-off-by: Wei Yang <richardw.yang@xxxxxxxxxxxxxxx>
> >> ---
> >> drivers/firmware/qemu_fw_cfg.c | 13 ++++---------
> >> 1 file changed, 4 insertions(+), 9 deletions(-)
> >>
> >> diff --git a/drivers/firmware/qemu_fw_cfg.c b/drivers/firmware/qemu_fw_cfg.c
> >> index 039e0f91dba8..a1293cbd7adb 100644
> >> --- a/drivers/firmware/qemu_fw_cfg.c
> >> +++ b/drivers/firmware/qemu_fw_cfg.c
> >> @@ -296,18 +296,13 @@ static int fw_cfg_do_platform_probe(struct platform_device *pdev)
> >> return 0;
> >> }
> >>
> >> -static ssize_t fw_cfg_showrev(struct kobject *k, struct attribute *a, char *buf)
> >> +static ssize_t fw_cfg_rev_show(struct kobject *k, struct kobj_attribute *a,
> >> + char *buf)
> >> {
> >> return sprintf(buf, "%u\n", fw_cfg_rev);
> >> }
> >> -
> >> -static const struct {
> >> - struct attribute attr;
> >> - ssize_t (*show)(struct kobject *k, struct attribute *a, char *buf);
> >> -} fw_cfg_rev_attr = {
> >> - .attr = { .name = "rev", .mode = S_IRUSR },
> >> - .show = fw_cfg_showrev,
> >> -};
> >> +static const struct kobj_attribute fw_cfg_rev_attr =
> >> + __ATTR_RO_MODE(fw_cfg_rev, 0400);
> >>
> >> /* fw_cfg_sysfs_entry type */
> >> struct fw_cfg_sysfs_entry {
> >
> >
> >Looks like this will change the name from "rev" to "fw_cfg_rev".
> >That's a userspace visible change which we should not do lightly.
>
> You are right, I should keep the interface untouched.
>
> To keep it user un-visible, we could change like below:
>
> - __ATTR_RO(fw_cfg_rev);
> + __ATTR_RO(rev);
>
> Is this better for you?


Also why use 0400 and not S_IRUSR?

> >> --
> >> 2.19.1
>
> --
> Wei Yang
> Help you, Help me