[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <Y9o52aAC33YlRueI@hirez.programming.kicks-ass.net>
Date: Wed, 1 Feb 2023 11:07:21 +0100
From: Peter Zijlstra <peterz@...radead.org>
To: Song Liu <song@...nel.org>
Cc: linux-modules@...r.kernel.org, linux-kernel@...r.kernel.org,
hch@....de, kernel-team@...a.com,
Luis Chamberlain <mcgrof@...nel.org>,
Thomas Gleixner <tglx@...utronix.de>,
Guenter Roeck <linux@...ck-us.net>,
Christophe Leroy <christophe.leroy@...roup.eu>
Subject: Re: [PATCH v5] module: replace module_layout with module_memory
On Tue, Jan 31, 2023 at 10:47:20PM -0800, Song Liu wrote:
> diff --git a/include/linux/module.h b/include/linux/module.h
> index 8c5909c0076c..3429d354fec0 100644
> --- a/include/linux/module.h
> +++ b/include/linux/module.h
> @@ -320,17 +320,50 @@ struct mod_tree_node {
> struct latch_tree_node node;
> };
>
> +enum mod_mem_type {
MOD_TEXT = 0,
(just paranoia, and you explicitly rely on that below)
> + MOD_DATA,
> + MOD_RODATA,
> + MOD_RO_AFTER_INIT,
> + MOD_INIT_TEXT,
> + MOD_INIT_DATA,
> + MOD_INIT_RODATA,
> +
> + MOD_MEM_NUM_TYPES,
> + MOD_INVALID = -1,
> +};
> +
> +#define mod_mem_type_is_core_data(type) \
> + ((type) == MOD_DATA || \
> + (type) == MOD_RODATA || \
> + (type) == MOD_RO_AFTER_INIT)
> +
> +#define mod_mem_type_is_core(type) \
> + ((type) == MOD_TEXT || \
> + mod_mem_type_is_core_data(type))
> +
> +#define mod_mem_type_is_init(type) \
> + ((type) == MOD_INIT_TEXT || \
> + (type) == MOD_INIT_DATA || \
> + (type) == MOD_INIT_RODATA)
Note that, per definition:
core := !init
data := !text
(and vice-versa ofcourse, so pick the smallest set) also ISTR you
explicitly needing is_text somewhere.... ah yes, module_enable_nx().
That is; I'd write something like:
#define mod_mem_type_is_core(type) !mod_mem_type_is_init(type)
#define mod_mem_type_is_text(type) \
((type) == MOD_TEXT || \
(type) == MOD_INIT_TEXT)
#define mod_mem_type_is_data(type) !mod_mem_type_is_text(type)
and then possibly additional helpers like is_core_data etc.. where
needed.
#define mod_mem_type_is_core_data(type) \
(mod_mem_type_is_core(type) && \
mod_mem_type_is_data(type))
> +#define for_each_mod_mem_type(type) \
> + for (enum mod_mem_type (type) = 0; \
> + (type) < MOD_MEM_NUM_TYPES; (type)++)
So how about instead of this ...
> +#define for_core_mod_mem_type(type) \
> + for (enum mod_mem_type (type) = 0; \
> + (type) < MOD_MEM_NUM_TYPES; (type)++) \
> + if (mod_mem_type_is_core(type))
> +
> +#define for_init_mod_mem_type(type) \
> + for (enum mod_mem_type (type) = 0; \
> + (type) < MOD_MEM_NUM_TYPES; (type)++) \
> + if (mod_mem_type_is_init(type))
... you write something like:
#define for_class_mod_mem_type(type, class) \
for_each_mod_mem_type(type) \
if (mod_mem_type_is_##class(type))
Then we can write things like:
for_class_mod_mem_type(type, init)
for_class_mod_mem_type(type, data)
and
for_class_mod_mem_type(type, core_data)
(this last could be used in show_datasize() for example).
Does that make sense?
Powered by blists - more mailing lists