Re: how interesting are data->bss patches?

From: Keith Owens (kaos@ocs.com.au)
Date: Sun Sep 24 2000 - 08:43:16 EST


On Sun, 24 Sep 2000 08:31:12 -0500 (CDT),
Jeff Garzik <jgarzik@mandrakesoft.mandrakesoft.com> wrote:
>> On Sat, 23 Sep 2000 20:59:59 -0500 (CDT),
>> Peter Samuelson <peter@cadcamlab.org> wrote:
>> >(A related question: __initdata *does* have to be initialized, right?)
>>
>> If __initdata is not initialized then it ends up in the global .bss.
>> This would defeat the purpose of using __initdata.
>
>I am glad this was mentioned... It is a valid use of __initdata for
>static variables which you want to go away after boot. There might be
>some wasted space lurking here and there due to un-init'd __initdata
>vars.

Another case to look at. I discovered that

char __initdata *cmd[] = {
        "command1",
        "command2",
        "command3",
        NULL
        };

Stores the 4 pointers of cmd in .init.data but the strings are down in
.rodata, i.e. they are not discarded after init. I got around it by

static __initdata char str_command1[] = "command1";
static __initdata char str_command2[] = "command2";
static __initdata char str_command3[] = "command3";
char __initdata *cmd[] = {
        str_command1,
        str_command2,
        str_command3,
        NULL
        };

That put the strings in .init.data as well as the array. This simpler
version might work but I have not tried it.

char __initdata *cmd[] = {
        __initdata "command1",
        __initdata "command2",
        __initdata "command3",
        NULL
        };

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Sat Sep 30 2000 - 21:00:12 EST