Re: [PATCH] btf: Use BIN_ATTR_SIMPLE_RO() to define vmlinux attribute

["Thomas Weißschuh" <linux@weissschuh.net>]

On 2024-11-26 17:52:29-0800, Alexei Starovoitov wrote:
> On Fri, Nov 22, 2024 at 4:57 AM Thomas Weißschuh <linux@weissschuh.net> wrote:
> >
> > The usage of the macro allows to remove the custom handler function,
> > saving some memory. Additionally the code is easier to read.
> >
> > Signed-off-by: Thomas Weißschuh <linux@weissschuh.net>
> > ---
> > Something similar can be done to btf_module_read() in kernel/bpf/btf.c.
> > But doing it here and now would lead to some conflicts with some other
> > sysfs refactorings I'm doing. It will be part of a future series.
> > ---
> >  kernel/bpf/sysfs_btf.c | 21 +++++----------------
> >  1 file changed, 5 insertions(+), 16 deletions(-)
> >
> > diff --git a/kernel/bpf/sysfs_btf.c b/kernel/bpf/sysfs_btf.c
> > index fedb54c94cdb830a4890d33677dcc5a6e236c13f..a24381f933d0b80b11116d05463c35e9fa66acb1 100644
> > --- a/kernel/bpf/sysfs_btf.c
> > +++ b/kernel/bpf/sysfs_btf.c
> > @@ -12,34 +12,23 @@
> >  extern char __start_BTF[];
> >  extern char __stop_BTF[];
> >
> > -static ssize_t
> > -btf_vmlinux_read(struct file *file, struct kobject *kobj,
> > -                struct bin_attribute *bin_attr,
> > -                char *buf, loff_t off, size_t len)
> > -{
> > -       memcpy(buf, __start_BTF + off, len);
> > -       return len;
> > -}
> > -
> > -static struct bin_attribute bin_attr_btf_vmlinux __ro_after_init = {
> > -       .attr = { .name = "vmlinux", .mode = 0444, },
> > -       .read = btf_vmlinux_read,
> > -};
> > +static __ro_after_init BIN_ATTR_SIMPLE_RO(vmlinux);
> 
> To be honest I really don't like when code is hidden by macros like this.
> Looks like you guys already managed to sprinkle it in a few places.
> 
> btf_vmlinux_read() can be replaced with sysfs_bin_attr_simple_read().
> This part is fine, but macro pls dont.
> It doesn't help readability.
> imo mode = 0444 vs mode = 0400 is easier to understand
> instead of _RO vs _ADMIN_RO suffix.

I'm fine with either solution.

My patch is motivated by my current effort to constify 'struct
bin_attribute' throughout the kernel.
With the macro I only have to touch this location once,
without it twice.

If we go with a plain sysfs_bin_attr_simple_read() please let me do the
patch in another series I have prepared, to be submitted after 6.13-rc1.

> __ro_after_init should be a part of it, at least.

I see where you are coming from, it would break the pattern with all the
other attribute macros however.

> I'd like to hear what other maintainers think about
> such obfuscation.

Ack.