Re: [PATCH 1/4] ACPI: EC: Add ec_get_handle()

[Thomas Renninger <trenn@suse.de>]

On Friday, December 16, 2011 01:33:33 AM Matthew Garrett wrote:
> On Fri, Dec 16, 2011 at 01:22:35AM +0100, Thomas Renninger wrote:
> 
> > I think best is to move the thinkpad implementation of getting
> > ACPI handles based on HIDs to osl.c and make it global.
> > I'll send patches.
> > Please review them carefully, they are only compile tested.
> 
> The ec driver is already finding the hardware on the basis of the HID - 
> is there any reason to do this twice rather than just exporting the 
> information the ec driver already has?

I don't object to export ec_handle, but thinkpad_acpi.c should get
adjusted as well.

I had a closer look and can come up with a replacement for acpi_get_devices
using bus.c matching function bus_find_device() avoiding the namespace
walk.

What do you think about this one?
I found these occurences of acpi_get_devices, which I would
adjust in follow up patches if it's worth it:
drivers/char/agp/hp-agp.c:      acpi_get_devices("HWP0003", zx1_gart_probe, "HWP0003", NULL);
drivers/char/agp/hp-agp.c:      acpi_get_devices("HWP0007", zx1_gart_probe, "HWP0007", NULL);
drivers/acpi/processor_core.c:  acpi_get_devices("ACPI0007", early_init_pdc, NULL, NULL);
drivers/platform/x86/eeepc-wmi.c:       status = acpi_get_devices(EEEPC_ACPI_HID, eeepc_wmi_parse_device,
drivers/platform/x86/thinkpad_acpi.c:   status = acpi_get_devices(hid, tpacpi_acpi_handle_locate_callback,


Argh, these are called before ACPI objects are initialized
(or parsed at all) and won't work:
drivers/acpi/ec.c:      status = acpi_get_devices(ec_device_ids[0].id, ec_parse_device,
arch/ia64/kernel/acpi.c:        acpi_get_devices(NULL, acpi_map_iosapic, NULL, NULL);
drivers/pnp/pnpacpi/core.c:     acpi_get_devices(NULL, pnpacpi_add_device_handler, NULL, NULL);
arch/x86/pci/mmconfig-shared.c: acpi_get_devices("PNP0C01", find_mboard_resource, &mcfg_res, NULL);
arch/x86/pci/mmconfig-shared.c:         acpi_get_devices("PNP0C02", find_mboard_resource, &mcfg_res,
arch/ia64/sn/kernel/io_acpi_init.c:     acpi_get_devices("SGIHUB", sn_acpi_hubdev_init, NULL, NULL);
arch/ia64/sn/kernel/io_acpi_init.c:     acpi_get_devices("SGITIO", sn_acpi_hubdev_init, NULL, NULL);

Looks like it's not worth it.
I paste my patch anyway, if there is any use-case,
it's already tested and works...:

---
ACPI: Provide an acpi device search on acpi bus registered devices

These should replace unnecessary acpi_get_devices invokations
which ends in a namespace walk.

Signed-off-by: Thomas Renninger <trenn@suse.de>
CC: Matthew Garrett <mjg@redhat.com>
CC: Len Brown <lenb@kernel.org>
CC: linux-acpi@vger.kernel.org
CC: seth.forshee@canonical.com
CC: Azael Avalos <coproscefalo@gmail.com>
CC: Henrique de Moraes Holschuh <hmh@hmh.eng.br>

---
 drivers/acpi/scan.c  |   66 ++++++++++++++++++++++++++++++++++++++++++++++++++
 include/linux/acpi.h |    4 +++
 2 files changed, 70 insertions(+), 0 deletions(-)

diff --git a/drivers/acpi/scan.c b/drivers/acpi/scan.c
index 8ab80ba..efd97ea 100644
--- a/drivers/acpi/scan.c
+++ b/drivers/acpi/scan.c
@@ -1606,3 +1606,69 @@ int __init acpi_scan_init(void)
 
 	return result;
 }
+
+static int acpi_bus_find_cb(struct device *dev, void *data)
+{
+	struct acpi_device_id *ids = data;
+
+	if (acpi_match_device_ids(to_acpi_device(dev), ids) == 0)
+		return 1;
+	return 0;
+}
+
+/**
+ * acpi_find_devices - find acpi devices of a given HID and execute callback
+ * @ids: The Hardware IDs for matching devices
+ * @match: A callback function which is executed for matching HIDs
+ *
+ * This is the counterpart of acpi_get_devices from ACPICA sources which
+ * should not get executed because it performs a namespace walk.
+ * This function instead iterates only over all ACPI devices found by initially
+ * parsing the ACPI namespace (and thus got added to the ACPI bus).
+ * One difference to acpi_get_devices is that it only finds active devices
+ * which were successfully added to the acpi bus and show up in
+ * /sys/devices/LNXSYSTM:00 and for which an ACPI driver's .add function would
+ * get called.
+ *
+ * ACPI devices with a matching HID or CID are recognized.
+ *
+ * If match returns zero, no further, possibly matching devices are processed.
+ */
+int acpi_find_devices(struct acpi_device_id *ids, void *data,
+		      int (*match) (struct acpi_device *dev, void *data))
+{
+	struct device *hit_dev = NULL;
+
+	while(1) {
+		hit_dev = bus_find_device(&acpi_bus_type, hit_dev, ids,
+					  acpi_bus_find_cb);
+		if (!hit_dev)
+			break;
+		if (!match(to_acpi_device(hit_dev), data))
+			break;
+	}
+	return 0;		
+}
+EXPORT_SYMBOL_GPL(acpi_find_devices);
+
+/**
+ * acpi_find_device - find an acpi device of a given HID
+ * @ids: The Hardware IDs for matching devices
+ *
+ * Same as acpi_find_devices, but the search will stop on the first found
+ * device and its structure is returned.
+ *
+ * Use this one if you are sure that only 1 ACPI device of a given HID can
+ * exist in ACPI namespace.
+ *
+ * Returns NULL if no device is found.
+ */
+struct acpi_device *acpi_find_device(struct acpi_device_id *ids)
+{
+	struct device *dev = bus_find_device(&acpi_bus_type, NULL, ids,
+					     acpi_bus_find_cb);
+	if (dev)
+		return to_acpi_device(dev);
+	return NULL;
+}
+EXPORT_SYMBOL_GPL(acpi_find_device);
diff --git a/include/linux/acpi.h b/include/linux/acpi.h
index 6001b4da..2f9e09e 100644
--- a/include/linux/acpi.h
+++ b/include/linux/acpi.h
@@ -146,6 +146,10 @@ struct acpi_pci_driver {
 int acpi_pci_register_driver(struct acpi_pci_driver *driver);
 void acpi_pci_unregister_driver(struct acpi_pci_driver *driver);
 
+extern int acpi_find_devices(struct acpi_device_id *ids, void *data,
+			     int (*match) (struct acpi_device *dev, void *data));
+extern struct acpi_device *acpi_find_device(struct acpi_device_id *ids);
+
 extern int ec_read(u8 addr, u8 *val);
 extern int ec_write(u8 addr, u8 val);
 extern int ec_transaction(u8 command,