module scanning in kgdb 2.x

module scanning in kgdb 2.x

[Amit S. Kale]

Hi,

Here is code to scan modules in kgdb for 2.6 kernels. It's been contributed by 
TimeSys Corporation. 

It does following things:
1. Adds MODULE_STATE_GONE to indicate that a module was removed. This is 
differnent from MODULE_STATE_GOING. gdb needs to be notified of a module 
event _after_ a module has been removed. Or else it'll still find the module 
during a module list scan and will not remove it from its core.

2. Defines a structure mod_section which stores module section names and 
offsets preserved during loading of a module.

3. Adds a couple of fields to struct module to keep module section 
information.

4. Adds a few notifications for gdb to know module related events.

5. Saves module section names and offsets in load_module.

-Amit

Index: linux-2.6.3-kgdb/include/linux/module.h
===================================================================
--- linux-2.6.3-kgdb.orig/include/linux/module.h	2004-02-24 10:44:47.000000000 
+0530
+++ linux-2.6.3-kgdb/include/linux/module.h	2004-03-04 18:58:47.116645760 
+0530
@@ -186,8 +186,17 @@
 	MODULE_STATE_LIVE,
 	MODULE_STATE_COMING,
 	MODULE_STATE_GOING,
+	MODULE_STATE_GONE,
 };
 
+#ifdef CONFIG_KGDB
+#define MAX_SECTNAME 31
+struct mod_section {
+	void *address;
+	char name[MAX_SECTNAME + 1];
+};
+#endif
+
 struct module
 {
 	enum module_state state;
@@ -198,6 +207,13 @@
 	/* Unique handle for this module */
 	char name[MODULE_NAME_LEN];
 
+#ifdef CONFIG_KGDB
+	/* keep kgdb info at the begining so that gdb doesn't have a chance to
+	 * miss out any fields */
+	unsigned long num_sections;
+	struct mod_section *mod_sections;
+#endif
+
 	/* Exported symbols */
 	const struct kernel_symbol *syms;
 	unsigned int num_syms;
Index: linux-2.6.3-kgdb/kernel/module.c
===================================================================
--- linux-2.6.3-kgdb.orig/kernel/module.c	2004-02-24 10:44:56.000000000 +0530
+++ linux-2.6.3-kgdb/kernel/module.c	2004-03-04 18:55:59.136182672 +0530
@@ -727,6 +727,11 @@
 	mod->state = MODULE_STATE_GOING;
 	restart_refcounts();
 
+	down(&notify_mutex);
+	notifier_call_chain(&module_notify_list, MODULE_STATE_GOING,
+				mod);
+	up(&notify_mutex);
+
 	/* Never wait if forced. */
 	if (!forced && module_refcount(mod) != 0)
 		wait_for_zero_refcount(mod);
@@ -734,6 +739,10 @@
 	/* Final destruction now noone is using it. */
 	mod->exit();
 	free_module(mod);
+	down(&notify_mutex);
+	notifier_call_chain(&module_notify_list, MODULE_STATE_GONE,
+				NULL);
+	up(&notify_mutex);
 
  out:
 	up(&module_mutex);
@@ -1087,6 +1096,11 @@
 	/* Arch-specific cleanup. */
 	module_arch_cleanup(mod);
 
+#ifdef CONFIG_KGDB
+	/* kgdb info */
+	vfree(mod->mod_sections);
+#endif
+
 	/* Module unload stuff */
 	module_unload_free(mod);
 
@@ -1302,6 +1316,30 @@
 	return NULL;
 }
 
+#ifdef CONFIG_KGDB
+int add_modsects (struct module *mod, Elf_Ehdr *hdr, Elf_Shdr *sechdrs, const
+		char *secstrings)
+{
+	int i;
+
+	mod->num_sections = hdr->e_shnum - 1;
+	mod->mod_sections = vmalloc((hdr->e_shnum - 1)* sizeof (struct 
mod_section));
+
+	if (mod->mod_sections == NULL) {
+		return -ENOMEM;
+	}
+
+	for (i = 1; i < hdr->e_shnum; i++) {
+		mod->mod_sections[i - 1].address = (void *)sechdrs[i].sh_addr;
+		strncpy(mod->mod_sections[i - 1].name, secstrings +
+				sechdrs[i].sh_name, MAX_SECTNAME);
+		mod->mod_sections[i - 1].name[MAX_SECTNAME] = '\0';
+	}
+
+	return 0;
+}
+#endif
+
 #ifdef CONFIG_KALLSYMS
 int is_exported(const char *name, const struct module *mod)
 {
@@ -1650,6 +1688,12 @@
 	percpu_modcopy(mod->percpu, (void *)sechdrs[pcpuindex].sh_addr,
 		       sechdrs[pcpuindex].sh_size);
 
+#ifdef CONFIG_KGDB
+	if ((err = add_modsects(mod, hdr, sechdrs, secstrings)) < 0) {
+		goto nomodsectinfo;
+	}
+#endif
+
 	err = module_finalize(hdr, sechdrs, mod);
 	if (err < 0)
 		goto cleanup;
@@ -1688,6 +1732,11 @@
  arch_cleanup:
 	module_arch_cleanup(mod);
  cleanup:
+
+#ifdef CONFIG_KGDB
+nomodsectinfo:
+	vfree(mod->mod_sections);
+#endif
 	module_unload_free(mod);
 	module_free(mod, mod->module_init);
  free_core:
@@ -1758,7 +1807,12 @@
 	if (ret < 0) {
 		/* Init routine failed: abort.  Try to protect us from
                    buggy refcounters. */
+
 		mod->state = MODULE_STATE_GOING;
+		down(&notify_mutex);
+		notifier_call_chain(&module_notify_list, MODULE_STATE_GOING,
+					mod);
+		up(&notify_mutex);
 		synchronize_kernel();
 		if (mod->unsafe)
 			printk(KERN_ERR "%s: module is now stuck!\n",

Re: [Kgdb-bugreport] module scanning in kgdb 2.x

[Amit S. Kale]

The gdb code for this is completely different from the one that works with 2.4 
kernels. I'll post it at kgdb.sourceforge.net when I checkin this change.

-Amit

On Friday 12 Mar 2004 12:06 pm, Amit S. Kale wrote:
> Hi,
>
> Here is code to scan modules in kgdb for 2.6 kernels. It's been contributed
> by TimeSys Corporation.
>
> It does following things:
> 1. Adds MODULE_STATE_GONE to indicate that a module was removed. This is
> differnent from MODULE_STATE_GOING. gdb needs to be notified of a module
> event _after_ a module has been removed. Or else it'll still find the
> module during a module list scan and will not remove it from its core.
>
> 2. Defines a structure mod_section which stores module section names and
> offsets preserved during loading of a module.
>
> 3. Adds a couple of fields to struct module to keep module section
> information.
>
> 4. Adds a few notifications for gdb to know module related events.
>
> 5. Saves module section names and offsets in load_module.
>
> -Amit
>
> Index: linux-2.6.3-kgdb/include/linux/module.h
> ===================================================================
> --- linux-2.6.3-kgdb.orig/include/linux/module.h	2004-02-24
> 10:44:47.000000000 +0530
> +++ linux-2.6.3-kgdb/include/linux/module.h	2004-03-04 18:58:47.116645760
> +0530
> @@ -186,8 +186,17 @@
>  	MODULE_STATE_LIVE,
>  	MODULE_STATE_COMING,
>  	MODULE_STATE_GOING,
> +	MODULE_STATE_GONE,
>  };
>
> +#ifdef CONFIG_KGDB
> +#define MAX_SECTNAME 31
> +struct mod_section {
> +	void *address;
> +	char name[MAX_SECTNAME + 1];
> +};
> +#endif
> +
>  struct module
>  {
>  	enum module_state state;
> @@ -198,6 +207,13 @@
>  	/* Unique handle for this module */
>  	char name[MODULE_NAME_LEN];
>
> +#ifdef CONFIG_KGDB
> +	/* keep kgdb info at the begining so that gdb doesn't have a chance to
> +	 * miss out any fields */
> +	unsigned long num_sections;
> +	struct mod_section *mod_sections;
> +#endif
> +
>  	/* Exported symbols */
>  	const struct kernel_symbol *syms;
>  	unsigned int num_syms;
> Index: linux-2.6.3-kgdb/kernel/module.c
> ===================================================================
> --- linux-2.6.3-kgdb.orig/kernel/module.c	2004-02-24 10:44:56.000000000
> +0530 +++ linux-2.6.3-kgdb/kernel/module.c	2004-03-04 18:55:59.136182672
> +0530 @@ -727,6 +727,11 @@
>  	mod->state = MODULE_STATE_GOING;
>  	restart_refcounts();
>
> +	down(&notify_mutex);
> +	notifier_call_chain(&module_notify_list, MODULE_STATE_GOING,
> +				mod);
> +	up(&notify_mutex);
> +
>  	/* Never wait if forced. */
>  	if (!forced && module_refcount(mod) != 0)
>  		wait_for_zero_refcount(mod);
> @@ -734,6 +739,10 @@
>  	/* Final destruction now noone is using it. */
>  	mod->exit();
>  	free_module(mod);
> +	down(&notify_mutex);
> +	notifier_call_chain(&module_notify_list, MODULE_STATE_GONE,
> +				NULL);
> +	up(&notify_mutex);
>
>   out:
>  	up(&module_mutex);
> @@ -1087,6 +1096,11 @@
>  	/* Arch-specific cleanup. */
>  	module_arch_cleanup(mod);
>
> +#ifdef CONFIG_KGDB
> +	/* kgdb info */
> +	vfree(mod->mod_sections);
> +#endif
> +
>  	/* Module unload stuff */
>  	module_unload_free(mod);
>
> @@ -1302,6 +1316,30 @@
>  	return NULL;
>  }
>
> +#ifdef CONFIG_KGDB
> +int add_modsects (struct module *mod, Elf_Ehdr *hdr, Elf_Shdr *sechdrs,
> const +		char *secstrings)
> +{
> +	int i;
> +
> +	mod->num_sections = hdr->e_shnum - 1;
> +	mod->mod_sections = vmalloc((hdr->e_shnum - 1)* sizeof (struct
> mod_section));
> +
> +	if (mod->mod_sections == NULL) {
> +		return -ENOMEM;
> +	}
> +
> +	for (i = 1; i < hdr->e_shnum; i++) {
> +		mod->mod_sections[i - 1].address = (void *)sechdrs[i].sh_addr;
> +		strncpy(mod->mod_sections[i - 1].name, secstrings +
> +				sechdrs[i].sh_name, MAX_SECTNAME);
> +		mod->mod_sections[i - 1].name[MAX_SECTNAME] = '\0';
> +	}
> +
> +	return 0;
> +}
> +#endif
> +
>  #ifdef CONFIG_KALLSYMS
>  int is_exported(const char *name, const struct module *mod)
>  {
> @@ -1650,6 +1688,12 @@
>  	percpu_modcopy(mod->percpu, (void *)sechdrs[pcpuindex].sh_addr,
>  		       sechdrs[pcpuindex].sh_size);
>
> +#ifdef CONFIG_KGDB
> +	if ((err = add_modsects(mod, hdr, sechdrs, secstrings)) < 0) {
> +		goto nomodsectinfo;
> +	}
> +#endif
> +
>  	err = module_finalize(hdr, sechdrs, mod);
>  	if (err < 0)
>  		goto cleanup;
> @@ -1688,6 +1732,11 @@
>   arch_cleanup:
>  	module_arch_cleanup(mod);
>   cleanup:
> +
> +#ifdef CONFIG_KGDB
> +nomodsectinfo:
> +	vfree(mod->mod_sections);
> +#endif
>  	module_unload_free(mod);
>  	module_free(mod, mod->module_init);
>   free_core:
> @@ -1758,7 +1807,12 @@
>  	if (ret < 0) {
>  		/* Init routine failed: abort.  Try to protect us from
>                     buggy refcounters. */
> +
>  		mod->state = MODULE_STATE_GOING;
> +		down(&notify_mutex);
> +		notifier_call_chain(&module_notify_list, MODULE_STATE_GOING,
> +					mod);
> +		up(&notify_mutex);
>  		synchronize_kernel();
>  		if (mod->unsafe)
>  			printk(KERN_ERR "%s: module is now stuck!\n",
>
>
>
> -------------------------------------------------------
> This SF.Net email is sponsored by: IBM Linux Tutorials
> Free Linux tutorial presented by Daniel Robbins, President and CEO of
> GenToo technologies. Learn everything from fundamentals to system
> administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click
> _______________________________________________
> Kgdb-bugreport mailing list
> Kgdb-bugreport@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/kgdb-bugreport

Re: module scanning in kgdb 2.x

[Rusty Russell]

On Fri, 2004-03-12 at 17:36, Amit S. Kale wrote:
> Hi,

Hi Amit,

	FYI: you would have received a quicker response if you'd CC'd me.

> It does following things:
> 1. Adds MODULE_STATE_GONE to indicate that a module was removed. This is 
> differnent from MODULE_STATE_GOING. gdb needs to be notified of a module 
> event _after_ a module has been removed. Or else it'll still find the module 
> during a module list scan and will not remove it from its core.

Makes sense.

> 2. Defines a structure mod_section which stores module section names and 
> offsets preserved during loading of a module.
> 
> 3. Adds a couple of fields to struct module to keep module section 
> information.

Why not just set the section strings to SHF_ALLOC rather than copying
(and possibly truncating) the names into your struct mod_section? 
struct mod_section is then simply void *addr; char *name;

Cheers,
Rusty.
-- 
Anyone who quotes me in their signature is an idiot -- Rusty Russell

Re: [Kgdb-bugreport] module scanning in kgdb 2.x

[Amit S. Kale]

Hi,

I have checked in this kgdb change and uploaded the gdb to be used with it.
Both kgdb and gdb code for this feature has been contributed by TimeSys 
Corporation.

-Amit


On Friday 12 Mar 2004 12:30 pm, Amit S. Kale wrote:
> The gdb code for this is completely different from the one that works with
> 2.4 kernels. I'll post it at kgdb.sourceforge.net when I checkin this
> change.
>
> -Amit
>
> On Friday 12 Mar 2004 12:06 pm, Amit S. Kale wrote:
> > Hi,
> >
> > Here is code to scan modules in kgdb for 2.6 kernels. It's been
> > contributed by TimeSys Corporation.
> >
> > It does following things:
> > 1. Adds MODULE_STATE_GONE to indicate that a module was removed. This is
> > differnent from MODULE_STATE_GOING. gdb needs to be notified of a module
> > event _after_ a module has been removed. Or else it'll still find the
> > module during a module list scan and will not remove it from its core.
> >
> > 2. Defines a structure mod_section which stores module section names and
> > offsets preserved during loading of a module.
> >
> > 3. Adds a couple of fields to struct module to keep module section
> > information.
> >
> > 4. Adds a few notifications for gdb to know module related events.
> >
> > 5. Saves module section names and offsets in load_module.
> >
> > -Amit
> >
> > Index: linux-2.6.3-kgdb/include/linux/module.h
> > ===================================================================
> > --- linux-2.6.3-kgdb.orig/include/linux/module.h	2004-02-24
> > 10:44:47.000000000 +0530
> > +++ linux-2.6.3-kgdb/include/linux/module.h	2004-03-04 18:58:47.116645760
> > +0530
> > @@ -186,8 +186,17 @@
> >  	MODULE_STATE_LIVE,
> >  	MODULE_STATE_COMING,
> >  	MODULE_STATE_GOING,
> > +	MODULE_STATE_GONE,
> >  };
> >
> > +#ifdef CONFIG_KGDB
> > +#define MAX_SECTNAME 31
> > +struct mod_section {
> > +	void *address;
> > +	char name[MAX_SECTNAME + 1];
> > +};
> > +#endif
> > +
> >  struct module
> >  {
> >  	enum module_state state;
> > @@ -198,6 +207,13 @@
> >  	/* Unique handle for this module */
> >  	char name[MODULE_NAME_LEN];
> >
> > +#ifdef CONFIG_KGDB
> > +	/* keep kgdb info at the begining so that gdb doesn't have a chance to
> > +	 * miss out any fields */
> > +	unsigned long num_sections;
> > +	struct mod_section *mod_sections;
> > +#endif
> > +
> >  	/* Exported symbols */
> >  	const struct kernel_symbol *syms;
> >  	unsigned int num_syms;
> > Index: linux-2.6.3-kgdb/kernel/module.c
> > ===================================================================
> > --- linux-2.6.3-kgdb.orig/kernel/module.c	2004-02-24 10:44:56.000000000
> > +0530 +++ linux-2.6.3-kgdb/kernel/module.c	2004-03-04 18:55:59.136182672
> > +0530 @@ -727,6 +727,11 @@
> >  	mod->state = MODULE_STATE_GOING;
> >  	restart_refcounts();
> >
> > +	down(&notify_mutex);
> > +	notifier_call_chain(&module_notify_list, MODULE_STATE_GOING,
> > +				mod);
> > +	up(&notify_mutex);
> > +
> >  	/* Never wait if forced. */
> >  	if (!forced && module_refcount(mod) != 0)
> >  		wait_for_zero_refcount(mod);
> > @@ -734,6 +739,10 @@
> >  	/* Final destruction now noone is using it. */
> >  	mod->exit();
> >  	free_module(mod);
> > +	down(&notify_mutex);
> > +	notifier_call_chain(&module_notify_list, MODULE_STATE_GONE,
> > +				NULL);
> > +	up(&notify_mutex);
> >
> >   out:
> >  	up(&module_mutex);
> > @@ -1087,6 +1096,11 @@
> >  	/* Arch-specific cleanup. */
> >  	module_arch_cleanup(mod);
> >
> > +#ifdef CONFIG_KGDB
> > +	/* kgdb info */
> > +	vfree(mod->mod_sections);
> > +#endif
> > +
> >  	/* Module unload stuff */
> >  	module_unload_free(mod);
> >
> > @@ -1302,6 +1316,30 @@
> >  	return NULL;
> >  }
> >
> > +#ifdef CONFIG_KGDB
> > +int add_modsects (struct module *mod, Elf_Ehdr *hdr, Elf_Shdr *sechdrs,
> > const +		char *secstrings)
> > +{
> > +	int i;
> > +
> > +	mod->num_sections = hdr->e_shnum - 1;
> > +	mod->mod_sections = vmalloc((hdr->e_shnum - 1)* sizeof (struct
> > mod_section));
> > +
> > +	if (mod->mod_sections == NULL) {
> > +		return -ENOMEM;
> > +	}
> > +
> > +	for (i = 1; i < hdr->e_shnum; i++) {
> > +		mod->mod_sections[i - 1].address = (void *)sechdrs[i].sh_addr;
> > +		strncpy(mod->mod_sections[i - 1].name, secstrings +
> > +				sechdrs[i].sh_name, MAX_SECTNAME);
> > +		mod->mod_sections[i - 1].name[MAX_SECTNAME] = '\0';
> > +	}
> > +
> > +	return 0;
> > +}
> > +#endif
> > +
> >  #ifdef CONFIG_KALLSYMS
> >  int is_exported(const char *name, const struct module *mod)
> >  {
> > @@ -1650,6 +1688,12 @@
> >  	percpu_modcopy(mod->percpu, (void *)sechdrs[pcpuindex].sh_addr,
> >  		       sechdrs[pcpuindex].sh_size);
> >
> > +#ifdef CONFIG_KGDB
> > +	if ((err = add_modsects(mod, hdr, sechdrs, secstrings)) < 0) {
> > +		goto nomodsectinfo;
> > +	}
> > +#endif
> > +
> >  	err = module_finalize(hdr, sechdrs, mod);
> >  	if (err < 0)
> >  		goto cleanup;
> > @@ -1688,6 +1732,11 @@
> >   arch_cleanup:
> >  	module_arch_cleanup(mod);
> >   cleanup:
> > +
> > +#ifdef CONFIG_KGDB
> > +nomodsectinfo:
> > +	vfree(mod->mod_sections);
> > +#endif
> >  	module_unload_free(mod);
> >  	module_free(mod, mod->module_init);
> >   free_core:
> > @@ -1758,7 +1807,12 @@
> >  	if (ret < 0) {
> >  		/* Init routine failed: abort.  Try to protect us from
> >                     buggy refcounters. */
> > +
> >  		mod->state = MODULE_STATE_GOING;
> > +		down(&notify_mutex);
> > +		notifier_call_chain(&module_notify_list, MODULE_STATE_GOING,
> > +					mod);
> > +		up(&notify_mutex);
> >  		synchronize_kernel();
> >  		if (mod->unsafe)
> >  			printk(KERN_ERR "%s: module is now stuck!\n",
> >
> >
> >
> > -------------------------------------------------------
> > This SF.Net email is sponsored by: IBM Linux Tutorials
> > Free Linux tutorial presented by Daniel Robbins, President and CEO of
> > GenToo technologies. Learn everything from fundamentals to system
> > administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click
> > _______________________________________________
> > Kgdb-bugreport mailing list
> > Kgdb-bugreport@lists.sourceforge.net
> > https://lists.sourceforge.net/lists/listinfo/kgdb-bugreport

Re: module scanning in kgdb 2.x

[Amit S. Kale]

On Wednesday 17 Mar 2004 2:48 am, Rusty Russell wrote:
> On Fri, 2004-03-12 at 17:36, Amit S. Kale wrote:
> > Hi,
>
> Hi Amit,
>
> 	FYI: you would have received a quicker response if you'd CC'd me.

Hi Rusty,

Thanks.
Yep! I'll CC you on any further modules stuff.

>
> > It does following things:
> > 1. Adds MODULE_STATE_GONE to indicate that a module was removed. This is
> > differnent from MODULE_STATE_GOING. gdb needs to be notified of a module
> > event _after_ a module has been removed. Or else it'll still find the
> > module during a module list scan and will not remove it from its core.
>
> Makes sense.
>
> > 2. Defines a structure mod_section which stores module section names and
> > offsets preserved during loading of a module.
> >
> > 3. Adds a couple of fields to struct module to keep module section
> > information.
>
> Why not just set the section strings to SHF_ALLOC rather than copying
> (and possibly truncating) the names into your struct mod_section?
> struct mod_section is then simply void *addr; char *name;

How can I do that? Do I have to use objcopy on module files for this purpose?

-Amit

Re: module scanning in kgdb 2.x

[Rusty Russell]

On Sat, 2004-03-20 at 02:06, Amit S. Kale wrote:
> On Wednesday 17 Mar 2004 2:48 am, Rusty Russell wrote:
> > Why not just set the section strings to SHF_ALLOC rather than copying
> > (and possibly truncating) the names into your struct mod_section?
> > struct mod_section is then simply void *addr; char *name;
> 
> How can I do that? Do I have to use objcopy on module files for this purpose?

That'd be one way.  But I was thinking you should add it in the loading
code the way CONFIG_KALLSYMS does for some other sections.

Cheers,
Rusty.
-- 
Anyone who quotes me in their signature is an idiot -- Rusty Russell