xref: /titanic_52/usr/src/man/man9f/ddi_modopen.9f (revision 6a5408e613dc0b372f722907d6c0b18e99c182dd)
te
Copyright (c) 2008, Sun Microsystems, Inc., All Rights Reserved
The contents of this file are subject to the terms of the Common Development and Distribution License (the "License"). You may not use this file except in compliance with the License.
You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE or http://www.opensolaris.org/os/licensing. See the License for the specific language governing permissions and limitations under the License.
When distributing Covered Code, include this CDDL HEADER in each file and include the License file at usr/src/OPENSOLARIS.LICENSE. If applicable, add the following below this CDDL HEADER, with the fields enclosed by brackets "[]" replaced with your own identifying information: Portions Copyright [yyyy] [name of copyright owner]
ddi_modopen 9F "17 March 2008" "SunOS 5.11" "Kernel Functions for Drivers"
NAME
ddi_modopen, ddi_modsym, ddi_modclose - dynamically-loaded kernel module functions
SYNOPSIS

#include <sys/modctl.h>

ddi_modhandle_t ddi_modopen(const char*modname, int mode,
 int *errnop);

void *ddi_modsym(ddi_modhandle_t handle, const char *symname,
 int *errnop);

int ddi_modclose(ddi_modhandle_t handle);
INTERFACE LEVEL

Solaris DDI specific (Solaris DDI).

PARAMETERS

modname

The name of the dynamically-loaded kernel module (file) to be opened. The modname string is of the form:

"[namespace/[dirspace/]]modulename"
Each "namespace/" directory along the standard kernel moddir/module-path path (system(4)) is searched to locate the module. If "namespace/" is not specified, "misc/" is assumed. If "dirspace" is specified, then "namespace/" must be explicitly provided.

mode

Currently, KRTLD_MODE_FIRST.

errnop

Pointer to errno returned on error, if NULL then no additional error information is available.

handle

Opaque handle returned from ddi_modopen(), invalidated by ddi_modclose().

symname

Symbol's name as a character string.

DESCRIPTION

The function prototypes for ddi_modopen(), ddi_modsym(), and ddi_modclose() are modeled after the userland libdl(3LIB), dlopen(3C), dlsym(3C) , anddlclose(3C) interfaces, however not all userland features are available and the kernel symbol resolution is different. The dlerror(3C) interface is not appropriate for the kernel environment, so the new errnop return argument was added for ddi_modopen() and ddi_modsym().

The ddi_modopen()function makes a dynamically-loaded kernel module named by "modname" available to a running kernel. ddi_modopen() returns a handle that the caller can use on subsequent calls to ddi_modsym() and ddi_modclose(). The value of this handle should not be interpreted in any way by the caller.

The ddi_modopen() interface works best as a dynamic component/object plug-in mechanism when targeting kernel "misc" modules that contain a single "struct modlmisc" module linkage, however non-"misc" modules and modules with multiple linkage structures can also be targeted.

There are two different symbol resolution search orders associated with the ddi_modopen() function: one search order to resolve symbols during the load of the targeted module, another search order o resolve ddi_modsym() calls against the handle returned by ddi_modopen(). To resolve symbols during module load, the standard kernel module load search order is used; to resolve symbols during module "A" load, the order is as follows:

A -> A's _depends_on -> unix -> unix's _depends_on

A single-level, left-to-right search in _depends_on (or the "ld -N" alternative) modules occurs. For UNIX on Sparc, _depends_on is similar to "genunix misc/platmod cpu/SUNW,UltraSPARC-III+ dtracestubs" for Intel, it is "genunix dtracestubs". The ddi_modsym() search is limited to the module directly associated with the handle.

The ddi_modopen() function increments the reference count on the named kernel module. Upon the first load of a module, the_init(9E) initialization code in the module is called; ddi_modopen() does not return until _init completes.

The ddi_modsym() function allows a caller to obtain the address of a symbol that is defined within a module. The handle argument is a valid ddi_modhandle_t as returned by ddi_modopen(), the symname argument is the symbol's name as a character string. The special handle values supported by ddi_modsym(3C) are not supported.

The ddi_modclose() function decrements the reference count of the kernel module associated with the specified handle. After the ddi_modclose() function is called, all ddi_modsym() resolutions obtained (either directly or indirectly) using the now closed handle are invalid; further use of these resolutions can cause undefined behavior (that is, may lead to a panic). When the last ddi_modclose() of a module occurs, and there are no further references to the module, the module _fini(9E)entry point may be called. If _fini returns success then the module may be unloaded.

RETURN VALUES

The ddi_modopen() function returns a handle to the dynamically-loaded kernel module. The ddi_modopen() function returns NULL if the module cannot be found, the object cannot be relocated, or an error occurs during the process of resolving and relocating its symbolic references.

The ddi_modsym() function returns NULL if the symname symbol cannot be found directly within the module associated with the handle.

If the handle was not referenced, ddi_modclose() returns 0. If the handle is invalid, ddi_modclose() may return a non-zero value.

When either ddi_modopen() or ddi_modsym() return NULL, additional errno information related to the failure is returned in *errnop if it is not NULL.

CONTEXT

ddi_modopen() can be called from user context only.

EXAMPLES

Example 1 Coding a Dynamically Loaded Kernel Module

The following example shows code to dynamically load and call a "test" interface in a module called "dltest". The "test" interface then adds one to its integer argument.

ddi_modhandle_t modh;
int (*test)(int);
int i = 0;
int errno;
---%<---
/* dynamically load "dltest" kernel 'misc' module */
modh = ddi_modopen("dltest", KRTLD_MODE_FIRST, &errno);
if (modh == NULL)
 goto fail; /* failed to open dltest module */

test = (int (*)())ddi_modsym(modh, "test", &errno);
if (test == NULL) {
 (void) ddi_modclose(modh);
 goto fail; /* failed to find "test" interface */
}

/* invoke test interface and verify result */
i = (*test)(0);
ASSERT(i == 1);

(void) ddi_modclose(modh);
---%<---

The implementation of the "dltest" "misc" module is as follows:

#include <sys/modctl.h>
static dltest_add = 0;

/* define the module linkage */
static struct modlmisc modlmisc = {&mod_miscops, "dltest"};
static struct modlinkage modlinkage = {
 MODREV_1, (void *)&modmisc, NULL
};
int
_init(void)
{
 int i;

 dltest_add = 1; /* initialization */
 if ((i = mod_install(&modlinkage)) != 0)
 dltest_add = -1; /* un-initialization */
 return (i);
}
int
_fini()
{
 int i;

 if ((i = mod_remove(&modlinkage)) == 0)
 dltest_add = -1; /* un-initialization */
 return (i);
}
int
_info(struct modinfo *modinfop)
{
 return (mod_info(&modlinkage, modinfop));
}

/* "test" interface */
int
test(int i)
{
 return (i + dltest_add);
}

Example 2 Dynamically Accessing a Kernel Module within a Drive

The following example shows driver code to dynamically load into the kernel a module constructed via the elfwrap(1) utility and containing firmware intended for download to a device. The "start" and "end" pointers provide the addresses of the beginning of the data and first byte beyond the data.

ddi_modhandle_t modp;
char *data_startp, *data_endp;
size_t nbytes;
int rv;

modp = ddi_modopen("firmware-rev1.2a", KRTLD_MODE_FIRST, &rv);
data_startp = (char *)ddi_modsym(modp, "fw-rev1.2a_start", &rv);
data_endp = (char *)ddi_modsym(modp, "fw-rev1.2a_end", &rv);
nbytes = data_endp - data_startp;
rv = ddi_modclose(modp);
SEE ALSO

dlclose(3C), dlopen(3C), dlsym(3C), libdl(3LIB), boot(1M), elfwrap(1), modload(1M), system(4), _fini(9E), _info(9E), _init(9E)

Writing Device Drivers

WARNINGS

A system(4)forceload must be established for modules targeted by ddi_modopen() by code involved in the mount of root on "bootdev" during machine boot(1M).