Skip to contentSkip to conversation

nubus: switch to dynamic root device

PATCH In review Johan Hovold 3 participants 7 messages Apr 24 – May 4 Nubus

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 nubus: switch to dynamic root device

Driver core expects devices to be dynamically allocated and will, for example, complain loudly if a device that lacks a release function is ever freed.

Use root_device_register() to allocate and register the root device instead of open coding using a static device.

---
 drivers/nubus/nubus.c | 19 ++++++++-----------
 1 file changed, 8 insertions(+), 11 deletions(-)
drivers/nubus/nubus.c +8 -11
index 559dce302d06..40ce4991c356 100644
@@ -41,9 +41,7 @@ module_param_named(populate_procfs, nubus_populate_procfs, bool, 0);
 
 LIST_HEAD(nubus_func_rsrcs);
 
-static struct device nubus_parent = {
-	.init_name	= "nubus",
-};
+static struct device *nubus_parent;
 
 /* Meaning of "bytelanes":
 
@@ -833,7 +831,7 @@ static void __init nubus_add_board(int slot, int bytelanes)
 		list_add_tail(&fres->list, &nubus_func_rsrcs);
 	}
 
-	if (nubus_device_register(&nubus_parent, board))
+	if (nubus_device_register(nubus_parent, board))
 		put_device(&board->dev);
 }
 
@@ -880,18 +878,17 @@ static void __init nubus_scan_bus(void)
 
 static int __init nubus_init(void)
 {
-	int err;
-
 	if (!MACH_IS_MAC)
 		return 0;
 
 	nubus_proc_init();
-	err = device_register(&nubus_parent);
-	if (err) {
-		put_device(&nubus_parent);
-		return err;
-	}
+
+	nubus_parent = root_device_register("nubus");
+	if (IS_ERR(nubus_parent))
+		return PTR_ERR(nubus_parent);
+
 	nubus_scan_bus();
+
 	return 0;
 }
 
Signature · 1 line
2.53.0
maintainer
#1 lore raw
Quoted 4 lines · Johan Hovold

Driver core expects devices to be dynamically allocated and will, for example, complain loudly if a device that lacks a release function is ever freed.

Yes, in drivers/base/core.c, there is a warning in device_release().

WARN(1, KERN_ERR "Device '%s' does not have a release()
function, it is broken and must be fixed. See Documentation/core-api/kobject.rst.\n",

But there's no way for the refcount for the nubus parent device to reach zero that I can see. Did I miss something?

Quoted 3 lines

Use root_device_register() to allocate and register the root device instead of open coding using a static device.

Well, dynamic allocation makes sense for busses that might be instantiated more than once. But I don't know of any hardware like that. The nubus parent device is a singleton.

I suppose I could see some benefit to converting CONFIG_NUBUS into a tristate. Maybe the module link in root_device_register() would become applicable if someone wanted to do that conversion.

#2 lore raw
Quoted 7 lines · Finn Thain
Quoted 4 lines · Johan Hovold

Driver core expects devices to be dynamically allocated and will, for example, complain loudly if a device that lacks a release function is ever freed.

Quoted 8 lines

Yes, in drivers/base/core.c, there is a warning in device_release().

WARN(1, KERN_ERR "Device '%s' does not have a release()
function, it is broken and must be fixed. See Documentation/core-api/kobject.rst.\n",

But there's no way for the refcount for the nubus parent device to reach zero that I can see. Did I miss something?

It will if registration ever fails (e.g. due to fault injection or name collision):

	err = device_register(&nubus_parent);
	if (err) {
		put_device(&nubus_parent);
		return err;
	}

Johan

maintainer Acked-by
#3 lore raw
Quoted 17 lines · Johan Hovold
Quoted 7 lines · Finn Thain
Quoted 4 lines · Johan Hovold

Driver core expects devices to be dynamically allocated and will, for example, complain loudly if a device that lacks a release function is ever freed.

Quoted 8 lines

Yes, in drivers/base/core.c, there is a warning in device_release().

WARN(1, KERN_ERR "Device '%s' does not have a release()
function, it is broken and must be fixed. See Documentation/core-api/kobject.rst.\n",

But there's no way for the refcount for the nubus parent device to reach zero that I can see. Did I miss something?

Quoted 9 lines

It will if registration ever fails (e.g. due to fault injection or name collision):

	err = device_register(&nubus_parent);
	if (err) {
		put_device(&nubus_parent);
		return err;
	}

In that situation the kernel error message would say the device "is broken and must be fixed" when it was deliberately broken by fault injection. I think the commit log should state that the aim of your patch is to avoid that error message for that use-case.
Aside from that quibble, the patch looks fine to me.

#4 lore raw
Quoted 27 lines · Finn Thain
Quoted 16 lines · Johan Hovold
Quoted 6 lines · Finn Thain
Quoted 3 lines · Johan Hovold

Driver core expects devices to be dynamically allocated and will, for example, complain loudly if a device that lacks a release function is ever freed.

Quoted 8 lines

Yes, in drivers/base/core.c, there is a warning in device_release().

WARN(1, KERN_ERR "Device '%s' does not have a release()
function, it is broken and must be fixed. See Documentation/core-api/kobject.rst.\n",

But there's no way for the refcount for the nubus parent device to reach zero that I can see. Did I miss something?

Quoted 9 lines

It will if registration ever fails (e.g. due to fault injection or name collision):

	err = device_register(&nubus_parent);
	if (err) {
		put_device(&nubus_parent);
		return err;
	}
Quoted 4 lines

In that situation the kernel error message would say the device "is broken and must be fixed" when it was deliberately broken by fault injection. I think the commit log should state that the aim of your patch is to avoid that error message for that use-case.

No, the aim is to git rid of statically allocated device structures. The error message is there to notify people that the driver model expects dynamically allocated objects (e.g. as documented in kobject.rst) even if you're unlikely to hit it with this particular module.

Quoted 3 lines

Aside from that quibble, the patch looks fine to me.

Thanks for taking a look.

Johan

maintainer
#5 lore raw
Quoted 3 lines · Johan Hovold

the aim is to git rid of statically allocated device structures

Indeed, and the aim of the commit log is to provide some rationale. When I reviewed your patch I found that it was missing. Like I said to you already, the nubus parent device is a singleton (just like the ISA parent device).

Reviewed-by on v1 1/1
#6 lore raw
In reply to Johan Hovold
Quoted 8 lines · Johan Hovold

Driver core expects devices to be dynamically allocated and will, for example, complain loudly if a device that lacks a release function is ever freed.

Use root_device_register() to allocate and register the root device instead of open coding using a static device.

i.e. will queue in the m68k tree for v7.2.

Gr{oetje,eeting}s,

Geert

Signature · 5 lines
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds