On Sat, Feb 18, 2006 at 06:27:44PM +0100, Dag-Erling Sm?rgrav wrote: +> Pawel Jakub Dawidek <[EMAIL PROTECTED]> writes: +> > It doesn't change your API, just extends it. With this change I'll be +> > able to use this functions in geom(8) without wrappers around them. +> > And this is the whole point of having them, right? +> +> Please tell me what semantics you want instead of expecting me to read +> your mind.
There are some cases where you cannot decide which module to load,
because it is provided at run-time.
In geom(8) I've something like this currently:
static void
load_module(void)
{
char name1[64], name2[64];
snprintf(name1, sizeof(name1), "g_%s", class_name);
snprintf(name2, sizeof(name2), "geom_%s", class_name);
if (modfind(name1) < 0) {
/* Not present in kernel, try loading it. */
if (kldload(name2) < 0 || modfind(name1) < 0) {
if (errno != EEXIST) {
errx(EXIT_FAILURE,
"%s module not available!", name2);
}
}
}
}
Changing it to the version below doesn't really buy me anything:
static void
load_module(void)
{
char name1[64], name2[64];
snprintf(name1, sizeof(name1), "g_%s", class_name);
snprintf(name2, sizeof(name2), "geom_%s", class_name);
if (!kld_isloaded(name1) && kld_load(name2) == -1)
err(1, "failed to load %s module", name2);
}
But changing it to the code below, or removing load_module entirely looks
like improvement:
static void
load_module(void)
{
if (!kld_isloaded("g_%s", class_name) &&
kld_load("geom_%s", class_name) == -1) {
err(1, "failed to load geom_%s module", class_name);
}
}
Yet another option is to reimplement kld_load(3) to:
kld_load(const char *modname, const char *kldname);
(and eventually remove kld_isloaded(3)).
--
Pawel Jakub Dawidek http://www.wheel.pl
[EMAIL PROTECTED] http://www.FreeBSD.org
FreeBSD committer Am I Evil? Yes, I Am!
pgpLFPFcDSsYv.pgp
Description: PGP signature
