Re: DEV_MODULE doesn't support dynamic major numbers any longer?
Poul-Henning Kamp <[EMAIL PROTECTED]> writes:
> The cmaj and bmaj in DEV_MODULE are only used for ordering the drivers,
> and otherwise with no significance.
Yeah, I realized that later.
> The previous code was a hack and inflicted problems. The right
> solution, (until DEVFS of course) is to add two new functions:
> alloc_cmaj() and alloc_bmaj() and use those.
Like the following code? And should they be used by cdevsw_add when
d_maj and d_bmaj are some magical value such as -1?
/assar
Index: kern/kern_conf.c
===
RCS file: /src/fbsd-repository/src/sys/kern/kern_conf.c,v
retrieving revision 1.55
diff -u -w -u -w -r1.55 kern_conf.c
--- kern_conf.c 1999/08/08 18:42:47 1.55
+++ kern_conf.c 1999/08/12 13:24:04
@@ -88,6 +88,36 @@
}
/*
+ * Return an unused cdev major or -1 if there are none free.
+ */
+
+int
+alloc_cmaj (void)
+{
+int i;
+
+for (i = cdevsw_ALLOCSTART; i < NUMCDEVSW; ++i)
+ if (cdevsw[i] == NULL)
+ return i;
+return -1;
+}
+
+/*
+ * Return an unused bdev major or -1 if there are none free.
+ */
+
+int
+alloc_bmaj (void)
+{
+int i;
+
+for (i = cdevsw_ALLOCSTART; i < NUMCDEVSW; ++i)
+ if (bmaj2cmaj[i] == 254)
+ return i;
+return -1;
+}
+
+/*
* Add a cdevsw entry
*/
Index: sys/conf.h
===
RCS file: /src/fbsd-repository/src/sys/sys/conf.h,v
retrieving revision 1.69
diff -u -w -u -w -r1.69 conf.h
--- conf.h 1999/08/09 18:45:20 1.69
+++ conf.h 1999/08/12 13:21:23
@@ -248,6 +248,8 @@
DECLARE_MODULE(name, name##_mod, SI_SUB_DRIVERS, SI_ORDER_MIDDLE+cmaj*256+bmaj)
+intalloc_bmaj __P((void));
+intalloc_cmaj __P((void));
struct cdevsw *bdevsw __P((dev_t dev));
intcdevsw_add __P((struct cdevsw *new));
intcdevsw_remove __P((struct cdevsw *old));
Re: DEV_MODULE doesn't support dynamic major numbers any longer?
In message <[EMAIL PROTECTED]>, Assar Westerlund writes: >It seems to be the case that the possibility of specifying a major >number of NOMAJ in DEV_MODULE has vanished. The cmaj and bmaj in DEV_MODULE are only used for ordering the drivers, and otherwise with no significance. You should store your majors in your cdevsw structure. I don't particular like the concept of DEV_MODULE registrating the cdevsw by magic. >cdevsw_add() doesn't have any code for handling NOMAJ any longer. >(The only mention I can find of NOMAJ in a -current kernel tree from >19990811 are these: >which makes me wonder how the promcons works...). Fine, see above. >Is this intentional that there isn't there any support for dynamically >assigning major device numbers or is it an accident that I should go >and rectify? The previous code was a hack and inflicted problems. The right solution, (until DEVFS of course) is to add two new functions: alloc_cmaj() and alloc_bmaj() and use those. -- Poul-Henning Kamp FreeBSD coreteam member [EMAIL PROTECTED] "Real hackers run -current on their laptop." FreeBSD -- It will take a long time before progress goes too far! To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
DEV_MODULE doesn't support dynamic major numbers any longer?
It seems to be the case that the possibility of specifying a major number of NOMAJ in DEV_MODULE has vanished. cdevsw_add() doesn't have any code for handling NOMAJ any longer. (The only mention I can find of NOMAJ in a -current kernel tree from 19990811 are these: ./alpha/alpha/promcons.c:284:DEV_MODULE(prom, CDEV_MAJOR, NOMAJ, prom_cdevsw, 0, 0); ./sys/param.h:128:#define NOMAJ 256 /* non-existent device */ which makes me wonder how the promcons works...). Is this intentional that there isn't there any support for dynamically assigning major device numbers or is it an accident that I should go and rectify? /assar To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
