[RFC PATCH 5/8] module: add function to map address to containing module

[RFC PATCH 5/8] module: add function to map address to containing module

[Peter Oberparleiter]

From: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>

module_address() maps an address to the module containing it in either
text or data section. Required by the gcov profiling infrastructure to
associate profiling data structures with modules.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
---
 include/linux/module.h |    1 +
 kernel/module.c        |   23 ++++++++++++++++++-----
 2 files changed, 19 insertions(+), 5 deletions(-)

Index: linux-2.6.25/include/linux/module.h
===================================================================
--- linux-2.6.25.orig/include/linux/module.h
+++ linux-2.6.25/include/linux/module.h
@@ -380,6 +380,7 @@ static inline int module_is_live(struct 
 /* Is this address in a module? (second is with no locks, for oops) */
 struct module *module_text_address(unsigned long addr);
 struct module *__module_text_address(unsigned long addr);
+struct module *module_address(unsigned long addr);
 int is_module_address(unsigned long addr);
 
 /* Returns 0 and fills in value, defined and namebuf, or -ERANGE if
Index: linux-2.6.25/kernel/module.c
===================================================================
--- linux-2.6.25.orig/kernel/module.c
+++ linux-2.6.25/kernel/module.c
@@ -2527,10 +2527,16 @@ const struct exception_table_entry *sear
 	return e;
 }
 
-/*
- * Is this a valid module address?
+/**
+ * module_address - map an address to the containing module
+ * @addr: address for which the containing module should be searched
+ *
+ * Return the module which contains the address in either text or data
+ * section. Return NULL if no matching module was found. Callers should
+ * be aware that no reference counting is being performed for the returned
+ * module reference.
  */
-int is_module_address(unsigned long addr)
+struct module *module_address(unsigned long addr)
 {
 	struct module *mod;
 
@@ -2539,15 +2545,22 @@ int is_module_address(unsigned long addr
 	list_for_each_entry(mod, &modules, list) {
 		if (within(addr, mod->module_core, mod->core_size)) {
 			preempt_enable();
-			return 1;
+			return mod;
 		}
 	}
 
 	preempt_enable();
 
-	return 0;
+	return NULL;
 }
 
+/*
+ * Is this a valid module address?
+ */
+int is_module_address(unsigned long addr)
+{
+	return module_address(addr) ? 1 : 0;
+}
 
 /* Is this a valid kernel address? */
 struct module *__module_text_address(unsigned long addr)

Re: [RFC PATCH 5/8] module: add function to map address to containing module

[Rusty Russell]

On Monday 21 April 2008 22:34:39 Peter Oberparleiter wrote:
> From: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
>
> module_address() maps an address to the module containing it in either
> text or data section. Required by the gcov profiling infrastructure to
> associate profiling data structures with modules.

Locking problem; this isn't safe.  Note that we block preemption to traverse 
the module list.  You could grab a reference, and fix up all the callers to 
put it once they're done?

Thanks,
Rusty.

Re: [Ltp-coverage] [RFC PATCH 5/8] module: add function to map address to containing module

[Peter Oberparleiter]

Rusty Russell wrote:
> On Monday 21 April 2008 22:34:39 Peter Oberparleiter wrote:
>> From: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
>>
>> module_address() maps an address to the module containing it in either
>> text or data section. Required by the gcov profiling infrastructure to
>> associate profiling data structures with modules.
> 
> Locking problem; this isn't safe.  Note that we block preemption to traverse 
> the module list.  You could grab a reference, and fix up all the callers to 
> put it once they're done?

Hm, problem: in this patchset, module_address() is called when the module is going (mod->state = MODULE_STATE_GOING) and try_module_get() would most likely fail and/or make no sense since we're already past reference counting with this module anyway.

On the other hand, what I actually need is a way to find out if a given module contains a certain address so I could replace this function with something like module_contains(mod, addr). 
I'll go that route if there are no objections.


Regards,
  Peter