Skip to contentSkip to conversation

udevng: Add OFONO_PATHNAME property to set modem dbus path name

PATCH In review Pau Espin Pedrol 4 participants 19 messages Feb 13, 2018 – Feb 20, 2018

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 udevng: Add OFONO_PATHNAME property to set modem dbus path name

The current udevng.c implementation sets dbus path names for modems based on type and a number incremented seuqntially for each new modem found. As a result, the dbus path for a given device is non deterministic, since it depends on the devices available during ofono startup.

Furthermore, if a modem crashes and reboots while in operation, then udev will trigger a remove event followed by a create event, and the same modem will now be given a different name (as the sequence number is bigger).

This is non suitable for systems handling several modems which want to identify them easily based on its path.

This patch introduces a way to be able to set persistent names for specific devices while still permitting previous dynamic naming methodology.

One can set a persistent name using udev rules for the target device which set the OFONO_PATHNAME env property. If ofono finds this property set, it will use its value as the dbus path name for the modem.

Example:

$ cat /etc/udev/rules.d/90-local.rules
SUBSYSTEMS=="usb", DEVPATH=="/devices/pci0000:00/0000:00:1a.0/usb3/3-1/3-1.1/3-1.1.4/3-1.1.4.1/3-1.1.4.1.1", ENV{OFONO_PATHNAME}="foo"
$ udevadm info -p /devices/pci0000:00/0000:00:1a.0/usb3/3-1/3-1.1/3-1.1.4/3-1.1.4.1/3-1.1.4.1.1
...
E: OFONO_PATHNAME=foo
...
$ mdbus2 -s org.ofono
/
/bluetooth
/bluetooth/profile
/bluetooth/profile/dun_gw
/bluetooth/profile/hfp_ag
/bluetooth/profile/hfp_hf
/foo
/mbm_0
---
 plugins/udevng.c | 22 +++++++++++++++-------
 1 file changed, 15 insertions(+), 7 deletions(-)
plugins/udevng.c +15 -7
index 9b78ab47..e9f27977 100644
@@ -46,6 +46,7 @@ enum modem_type {
 struct modem_info {
 	char *syspath;
 	char *devname;
+	char *pathname;
 	char *driver;
 	char *vendor;
 	char *model;
@@ -1302,6 +1303,7 @@ static void destroy_modem(gpointer data)
 
 	g_free(modem->syspath);
 	g_free(modem->devname);
+	g_free(modem->pathname);
 	g_free(modem->driver);
 	g_free(modem->vendor);
 	g_free(modem->model);
@@ -1382,7 +1384,7 @@ static struct udev_device* get_serial_modem_device(struct udev_device *dev)
  */
 static void add_serial_device(struct udev_device *dev)
 {
-	const char *syspath, *devpath, *devname, *devnode;
+	const char *syspath, *devpath, *devname, *devnode, *pathname;
 	struct modem_info *modem;
 	struct serial_device_info *info;
 	const char *subsystem;
@@ -1396,6 +1398,7 @@ static void add_serial_device(struct udev_device *dev)
 	}
 
 	driver = udev_device_get_property_value(mdev, "OFONO_DRIVER");
+	pathname = udev_device_get_property_value(mdev, "OFONO_PATHNAME");
 
 	syspath = udev_device_get_syspath(mdev);
 	devname = udev_device_get_devnode(mdev);
@@ -1415,6 +1418,7 @@ static void add_serial_device(struct udev_device *dev)
 		modem->type = MODEM_TYPE_SERIAL;
 		modem->syspath = g_strdup(syspath);
 		modem->devname = g_strdup(devname);
+		modem->pathname = g_strdup(pathname);
 		modem->driver = g_strdup(driver);
 
 		g_hash_table_replace(modem_list, modem->syspath, modem);
@@ -1440,7 +1444,8 @@ static void add_serial_device(struct udev_device *dev)
 
 static void add_device(const char *syspath, const char *devname,
 			const char *driver, const char *vendor,
-			const char *model, struct udev_device *device)
+			const char *model, const char *pathname,
+			struct udev_device *device)
 {
 	struct udev_device *usb_interface;
 	const char *devpath, *devnode, *interface, *number;
@@ -1477,6 +1482,7 @@ static void add_device(const char *syspath, const char *devname,
 		modem->driver = g_strdup(driver);
 		modem->vendor = g_strdup(vendor);
 		modem->model = g_strdup(model);
+		modem->pathname = g_strdup(pathname);
 
 		modem->sysattr = get_sysattr(driver);
 
@@ -1512,8 +1518,8 @@ static void add_device(const char *syspath, const char *devname,
 
 	DBG("%s", syspath);
 	DBG("%s", devpath);
-	DBG("%s (%s) %s [%s] ==> %s %s", devnode, driver,
-					interface, number, label, sysattr);
+	DBG("%s (%s) %s [%s] ==> %s %s %s",
+		devnode, driver, interface, number, label, sysattr, pathname);
 
 	info = g_try_new0(struct device_info, 1);
 	if (info == NULL)
@@ -1611,7 +1617,7 @@ static struct {
 static void check_usb_device(struct udev_device *device)
 {
 	struct udev_device *usb_device;
-	const char *syspath, *devname, *driver;
+	const char *syspath, *devname, *driver, *pathname;
 	const char *vendor = NULL, *model = NULL;
 
 	usb_device = udev_device_get_parent_with_subsystem_devtype(device,
@@ -1630,6 +1636,8 @@ static void check_usb_device(struct udev_device *device)
 	vendor = udev_device_get_property_value(usb_device, "ID_VENDOR_ID");
 	model = udev_device_get_property_value(usb_device, "ID_MODEL_ID");
 
+	pathname = udev_device_get_property_value(usb_device, "OFONO_PATHNAME");
+
 	driver = udev_device_get_property_value(usb_device, "OFONO_DRIVER");
 	if (!driver) {
 		struct udev_device *usb_interface =
@@ -1685,7 +1693,7 @@ static void check_usb_device(struct udev_device *device)
 			return;
 	}
 
-	add_device(syspath, devname, driver, vendor, model, device);
+	add_device(syspath, devname, driver, vendor, model, pathname, device);
 }
 
 static void check_device(struct udev_device *device)
@@ -1723,7 +1731,7 @@ static gboolean create_modem(gpointer key, gpointer value, gpointer user_data)
 
 	DBG("driver=%s", modem->driver);
 
-	modem->modem = ofono_modem_create(NULL, modem->driver);
+	modem->modem = ofono_modem_create(modem->pathname, modem->driver);
 	if (modem->modem == NULL)
 		return TRUE;
 
Signature · 1 line
2.16.1
#1 lore raw

Hi Pau,

Quoted 5 lines · Pau Espin Pedrol

The current udevng.c implementation sets dbus path names for modems based on type and a number incremented seuqntially for each new modem found. As a result, the dbus path for a given device is non deterministic, since it depends on the devices available during ofono startup.

Why should it be deterministic?

Quoted 5 lines

Furthermore, if a modem crashes and reboots while in operation, then udev will trigger a remove event followed by a create event, and the same modem will now be given a different name (as the sequence number is bigger).

Yes, and?

Quoted 4 lines

This is non suitable for systems handling several modems which want to identify them easily based on its path.

This sounds like a very specific non-real world usecase, right? I mean why don't you simply compare the IMEI if you really care?

Quoted 7 lines

This patch introduces a way to be able to set persistent names for specific devices while still permitting previous dynamic naming methodology.

One can set a persistent name using udev rules for the target device which set the OFONO_PATHNAME env property. If ofono finds this property set, it will use its value as the dbus path name for the modem.

I'm hesitant to take this patch as udev logic is already complex and this doesn't seem to be relevant for anyone else besides your test system.

Quoted 10 lines

Example:

$ cat /etc/udev/rules.d/90-local.rules
SUBSYSTEMS=="usb", DEVPATH=="/devices/pci0000:00/0000:00:1a.0/usb3/3-1/3-1.1/3-1.1.4/3-1.1.4.1/3-1.1.4.1.1", ENV{OFONO_PATHNAME}="foo"
$ udevadm info -p /devices/pci0000:00/0000:00:1a.0/usb3/3-1/3-1.1/3-1.1.4/3-1.1.4.1/3-1.1.4.1.1
...
E: OFONO_PATHNAME=foo
...
Quoted 8 lines
$ mdbus2 -s org.ofono
/
/bluetooth
/bluetooth/profile
/bluetooth/profile/dun_gw
/bluetooth/profile/hfp_ag
/bluetooth/profile/hfp_hf
/foo

How are you making sure that dbus path is valid? ofono_modem_create
can fail after all...

Quoted 5 lines

/mbm_0

---
  plugins/udevng.c | 22 +++++++++++++++-------
  1 file changed, 15 insertions(+), 7 deletions(-)

Regards,
-Denis

#2 lore raw

Hi Denis,

Quoted 3 lines · Denis Kenzior

This sounds like a very specific non-real world usecase, right?  I mean why don't you simply compare the IMEI if you really care?

First of all, I like thinking that I'm part of the real world!

As far as I understand the IMEI is available in org.ofono.Modem.GetProperties "Serial" property. To be able to access this information, I need to previously power on the modem (SetProperty Powered true), otherwise it doesn't show up.
Which means that for an ofono user, it is a requirement to power on the modem and thus use it before being able to identify it.

Simple case: Several modems attached to a PC, in order to find the one I'm interested in I need to power on all of them and find which one matches a given IMEI. Really? Do I need this if I know where is my device always physically attached? Are you going through all your network interfaces and check against its MAC address to find the one you are interested in? Or you configure your network manager based on interface names? Would you enjoy it if your enp0s25 interface was suddenly named enp1s67? Specially if you have iptables rules, I bet that's going to be funny.

More complex case (my scenario): Again, I repeat: To identify a modem, as of now, you seem to require to power it on (use it, change its state) before identifying it. Now think of a system which has a config file with modems available on the system (with its dbus paths) and a resources database of modems (and other types of resources) currently in use. Now imagine several processes reserving any specific subset of modems with specific capabilities (in parallel) against the resources database, then using them and among the use cases they power off and power them on. Even if you have locking mechanism in the database, it's a nightmare, because again, you must use a modem before identifying it, meaning you either race condition with other processes using the modem you are identifying (eg. by powering it on, or leaving it powered on after identifying it, or while powering it off after identifying while other process is starting to use it).

Then the easy way: Each modem is assigned a unique persistent path, and that's the one always being presented and used. No need to identify it, I already know which modem is which. If I have several system setups, I can assign same names in all systems for easy use, etc.

So the question should be better: Why should it NOT be deterministic?

Quoted 3 lines

How are you making sure that dbus path is valid?   ofono_modem_create can fail after all...

Yes, it can fail while validating the path (dbus_validate_path) which is expected if a bad OFONO_PATHMNAME is supplied. If it's fine for you, I can add an error message there stating that dbus_validate_path failed and printing the pathname. Agree?

Signature · 7 lines
- Pau Espin Pedrol <[email protected]>         http://www.sysmocom.de/
=======================================================================
* sysmocom - systems for mobile communications GmbH
* Alt-Moabit 93
* 10559 Berlin, Germany
* Sitz / Registered office: Berlin, HRB 134158 B
* Geschaeftsfuehrer / Managing Director: Harald Welte
#3 lore raw

Hi Pau,

Quoted 6 lines

As far as I understand the IMEI is available in org.ofono.Modem.GetProperties "Serial" property. To be able to access this information, I need to previously power on the modem (SetProperty Powered true), otherwise it doesn't show up.
Which means that for an ofono user, it is a requirement to power on the modem and thus use it before being able to identify it.

Correct, but why is this a problem? Under normal circumstances the modems are auto-powered by connman anyway. That's the only reason that oFono doesn't do this automagically.

Quoted 11 lines

Simple case: Several modems attached to a PC, in order to find the one I'm interested in I need to power on all of them and find which one matches a given IMEI. Really? Do I need this if I know where is my device always physically attached? Are you going through all your network interfaces and check against its MAC address to find the one you are interested in? Or you configure your network manager based on interface names? Would you enjoy it if your enp0s25 interface was suddenly named enp1s67? Specially if you have iptables rules, I bet that's going to be funny.

Quoted 14 lines

More complex case (my scenario): Again, I repeat: To identify a modem, as of now, you seem to require to power it on (use it, change its state) before identifying it. Now think of a system which has a config file with modems available on the system (with its dbus paths) and a resources database of modems (and other types of resources) currently in use. Now imagine several processes reserving any specific subset of modems with specific capabilities (in parallel) against the resources database, then using them and among the use cases they power off and power them on. Even if you have locking mechanism in the database, it's a nightmare, because again, you must use a modem before identifying it, meaning you either race condition with other processes using the modem you are identifying (eg. by powering it on, or leaving it powered on after identifying it, or while powering it off after identifying while other process is starting to use it).

oFono doesn't expose hardware details, only hardware capabilities. So I understand how this is helpful for your usecase, but I'm not convinced this is relevant for what oFono was designed to do.

Quoted 7 lines

Then the easy way: Each modem is assigned a unique persistent path, and that's the one always being presented and used. No need to identify it, I already know which modem is which. If I have several system setups, I can assign same names in all systems for easy use, etc. >
So the question should be better: Why should it NOT be deterministic?

In general oFono does not use deterministic paths, the Manager APIs should be used to query the list of objects. In fact the docs even specifically state that the modem object path might have a variable prefix (e.g. a random number) and that variable prefix can change across oFono boots. This is mainly to encourage UI/client writers to use the APIs properly.

We have used hardcoded paths in one or two cases for synthetic test cases (e.g. stktest) but those are highly special and not user-visible anyhow.

Quoted 9 lines
Quoted 4 lines

How are you making sure that dbus path is valid?   ofono_modem_create can fail after all...

Yes, it can fail while validating the path (dbus_validate_path) which is expected if a bad OFONO_PATHMNAME is supplied. If it's fine for you, I can add an error message there stating that dbus_validate_path failed and printing the pathname. Agree?

Go ahead and send a V2, I'll let it sit for a week or so to see if anyone else gives any feedback.

Regards,
-Denis

#4 lore raw

Hi Pau, hi Denis,

Quoted 68 lines · Denis Kenzior

Hi Pau,

Quoted 6 lines

As far as I understand the IMEI is available in org.ofono.Modem.GetProperties "Serial" property. To be able to access this information, I need to previously power on the modem (SetProperty Powered true), otherwise it doesn't show up. Which means that for an ofono user, it is a requirement to power on the modem and thus use it before being able to identify it.

Correct, but why is this a problem?  Under normal circumstances the modems are auto-powered by connman anyway.  That's the only reason that oFono doesn't do this automagically.

Quoted 11 lines

Simple case: Several modems attached to a PC, in order to find the one I'm interested in I need to power on all of them and find which one matches a given IMEI. Really? Do I need this if I know where is my device always physically attached? Are you going through all your network interfaces and check against its MAC address to find the one you are interested in? Or you configure your network manager based on interface names? Would you enjoy it if your enp0s25 interface was suddenly named enp1s67? Specially if you have iptables rules, I bet that's going to be funny.

Quoted 15 lines

More complex case (my scenario): Again, I repeat: To identify a modem, as of now, you seem to require to power it on (use it, change its state) before identifying it. Now think of a system which has a config file with modems available on the system (with its dbus paths) and a resources database of modems (and other types of resources) currently in use. Now imagine several processes reserving any specific subset of modems with specific capabilities (in parallel) against the resources database, then using them and among the use cases they power off and power them on. Even if you have locking mechanism in the database, it's a nightmare, because again, you must use a modem before identifying it, meaning you either race condition with other processes using the modem you are identifying (eg. by powering it on, or leaving it powered on after identifying it, or while powering it off after identifying while other process is starting to use it).

oFono doesn't expose hardware details, only hardware capabilities.  So I understand how this is helpful for your usecase, but I'm not convinced this is relevant for what oFono was designed to do.

Quoted 8 lines

Then the easy way: Each modem is assigned a unique persistent path, and that's the one always being presented and used. No need to identify it, I already know which modem is which. If I have several system setups, I can assign same names in all systems for easy use, etc. >
So the question should be better: Why should it NOT be deterministic?

In general oFono does not use deterministic paths, the Manager APIs should be used to query the list of objects.  In fact the docs even specifically state that the modem object path might have a variable prefix (e.g. a random number) and that variable prefix can change across oFono boots.  This is mainly to encourage UI/client writers to use the APIs properly.

We have used hardcoded paths in one or two cases for synthetic test cases (e.g. stktest) but those are highly special and not user-visible anyhow.

Quoted 5 lines
Quoted 4 lines

How are you making sure that dbus path is valid? ofono_modem_create can fail after all...

Quoted 7 lines
Quoted 4 lines

Yes, it can fail while validating the path (dbus_validate_path) which is expected if a bad OFONO_PATHMNAME is supplied. If it's fine for you, I can add an error message there stating that dbus_validate_path failed and printing the pathname. Agree?

Go ahead and send a V2, I'll let it sit for a week or so to see if anyone else gives any feedback.

I have the same use case on my boards: several modems and I need to know where they physically are to be able to power the one I want and to physically reset a none working modem (using gpios corresponding to its hardware position). So I have a patch in Ofono to do that.
I have added a ModemSyspath property in org.ofono.Modem displaying dbus path. The patch is good enough for me but maybe not for upstream. As information comes from udevng plugin, this property might be in a specific interface added in udevng plugin and not in org.ofono.Modem.

This is just another way to have needed information (without changing modem name). I'll send my patch.

Christophe

#5 lore raw

PATCH 1/1 add syspath property to modem properties

---
 plugins/udevng.c |  2 ++
 src/modem.c      | 74 ++++++++++++++++++++++++++++++++------------------------
 2 files changed, 45 insertions(+), 31 deletions(-)
plugins/udevng.c +2 -0
index 2279bbe9..d5139586 100644
@@ -1330,6 +1330,8 @@ static gboolean create_modem(gpointer key, gpointer value, gpointer user_data)
 			continue;
 
 		if (driver_list[i].setup(modem) == TRUE) {
+			ofono_modem_set_string(modem->modem, "Syspath",
+								syspath);
 			ofono_modem_register(modem->modem);
 			return FALSE;
 		}
src/modem.c +43 -31

98 lines not on this page · raw message

Signature · 1 line
2.11.0
#6 lore raw

Re: [PATCH] add syspath property to modem properties

Hi Christophe,

Thanks for providing feedback on the topic. As I expected, I'm not alone and other people also have this requirement of being able to easily identify modems.

The patch you provide is enough for my needs, as I can easily identify a modem from the list without having to change its state (powering them on). I tested it and seems to be working fine:

mdbus2 -s org.ofono / org.ofono.Manager.GetModems
([('/mbm_1', {'Online': <false>, 'Powered': <false>, 'Lockdown':
<false>, 'Emergency': <false>, 'ModemSyspath':
<'/sys/devices/pci0000:00/0000:00:1d.0/usb4/4-1/4-1.4'>, 'Interfaces':
<@as []>, 'Features': <@as []>, 'Type': <'hardware'>}), ('/gobi_0', {'Online': <false>, 'Powered': <false>, 'Lockdown': <false>, 'Emergency': <false>, 'ModemSyspath':
<'/sys/devices/pci0000:00/0000:00:1a.0/usb1/1-1/1-1.2'>, 'Interfaces':
<@as []>, 'Features': <@as []>, 'Type': <'hardware'>})],)

So I'd like seeing merged either something Christophe patch or mine (or both). I'll send a v2 of my patch as suggested and let you deliver on the best approach. If not good solution is provided, I'd need to keep it in a separate fork, which imho would be really sad as it seems other people may potentially use this kind of feature.

Regards,

Signature · 7 lines
- Pau Espin Pedrol <[email protected]>         http://www.sysmocom.de/
=======================================================================
* sysmocom - systems for mobile communications GmbH
* Alt-Moabit 93
* 10559 Berlin, Germany
* Sitz / Registered office: Berlin, HRB 134158 B
* Geschaeftsfuehrer / Managing Director: Harald Welte
#7 lore raw

Re: [PATCH] add syspath property to modem properties

In reply to Christophe Ronco

Hi Christophe,

Quoted 10 lines · Christophe Ronco
---
  plugins/udevng.c |  2 ++
  src/modem.c      | 74 ++++++++++++++++++++++++++++++++------------------------
  2 files changed, 45 insertions(+), 31 deletions(-)
plugins/udevng.c +0 -0
index 2279bbe9..d5139586 100644

plugins/udevng.c static gboolean create_modem(gpointer key, gpointer value, gpointer user_data) · line 1333

Quoted 4 lines
plugins/udevng.c +1 -0
@@ -1330,6 +1330,8 @@ static gboolean create_modem(gpointer key, gpointer value, gpointer user_data)
  			continue;
 
  		if (driver_list[i].setup(modem) == TRUE) {
+			ofono_modem_set_string(modem->modem, "Syspath",

Use camel case, so "SysPath" or "SystemPath"

Quoted 59 lines
plugins/udevng.c +1 -0
+								syspath);
  			ofono_modem_register(modem->modem);
  			return FALSE;
  		}
src/modem.c +36 -0
index b1e8d3e2..7759b567 100644
@@ -773,6 +773,41 @@ ofono_bool_t ofono_modem_get_online(struct ofono_modem *modem)
  	return modem->online;
  }
 
+static gboolean get_modem_property(struct ofono_modem *modem, const char *name,
+					enum property_type type,
+					void *value)
+{
+	struct modem_property *property;
+
+	DBG("modem %p property %s", modem, name);
+
+	property = g_hash_table_lookup(modem->properties, name);
+
+	if (property == NULL) {
+		DBG("property %s does not exist", name);
+		return FALSE;
+	}
+
+	if (property->type != type) {
+		DBG("property %s wrong type", name);
+		return FALSE;
+	}
+
+	switch (property->type) {
+	case PROPERTY_TYPE_STRING:
+		*((const char **) value) = property->value;
+		return TRUE;
+	case PROPERTY_TYPE_INTEGER:
+		memcpy(value, property->value, sizeof(int));
+		return TRUE;
+	case PROPERTY_TYPE_BOOLEAN:
+		memcpy(value, property->value, sizeof(ofono_bool_t));
+		return TRUE;
+	default:
+		return FALSE;
+	}
+}
+
  void __ofono_modem_append_properties(struct ofono_modem *modem,
  						DBusMessageIter *dict)
  {
@@ -783,6 +818,7 @@ void __ofono_modem_append_properties(struct ofono_modem *modem,
  	struct ofono_devinfo *info;
  	dbus_bool_t emergency = ofono_modem_get_emergency_mode(modem);
  	const char *strtype;
+	char *value;
 
  	ofono_dbus_dict_append(dict, "Online", DBUS_TYPE_BOOLEAN,
  				&modem->online);

src/modem.c void __ofono_modem_append_properties(struct ofono_modem *modem, · line 868

Quoted 10 lines
src/modem.c +7 -0
@@ -823,6 +859,13 @@ void __ofono_modem_append_properties(struct ofono_modem *modem,
  						&info->svn);
  	}
 
+	if (get_modem_property(modem, "Syspath", PROPERTY_TYPE_STRING,
+							&value) == TRUE) {
+		ofono_dbus_dict_append(dict, "ModemSyspath",
+						DBUS_TYPE_STRING,
+						&value);
+	}
+

There's no need for {}. And why don't you make things much easier on yourself and simply use ofono_modem_get_string()

Quoted 32 lines
src/modem.c +0 -24
  	interfaces = g_new0(char *, g_slist_length(modem->interface_list) + 1);
  	for (i = 0, l = modem->interface_list; l; l = l->next, i++)
  		interfaces[i] = l->data;
@@ -1731,37 +1774,6 @@ static int set_modem_property(struct ofono_modem *modem, const char *name,
  	return 0;
  }
 
-static gboolean get_modem_property(struct ofono_modem *modem, const char *name,
-					enum property_type type,
-					void *value)
-{
-	struct modem_property *property;
-
-	DBG("modem %p property %s", modem, name);
-
-	property = g_hash_table_lookup(modem->properties, name);
-
-	if (property == NULL)
-		return FALSE;
-
-	if (property->type != type)
-		return FALSE;
-
-	switch (property->type) {
-	case PROPERTY_TYPE_STRING:
-		*((const char **) value) = property->value;
-		return TRUE;
-	case PROPERTY_TYPE_INTEGER:
-		memcpy(value, property->value, sizeof(int));
-		return TRUE;
-	case PROPERTY_TYPE_BOOLEAN:

src/modem.c static int set_modem_property(struct ofono_modem *modem, const char *name, · line 1779

Quoted 10 lines
src/modem.c +0 -7
-		memcpy(value, property->value, sizeof(ofono_bool_t));
-		return TRUE;
-	default:
-		return FALSE;
-	}
-}
-
  int ofono_modem_set_string(struct ofono_modem *modem,
  				const char *key, const char *value)
  {

Regards,
-Denis

#8 lore raw

Re: [PATCH] add syspath property to modem properties

Hi Denis,

I didn't think this patch could be something used on Upstream Ofono. If it is, I also have remarks on my own patch:
 - shouldn't it be two patches (one for plugin directory, another for
src directory)?
 - what is an acceptable name for the new DBUS property (currently it
is ModemSyspath):
   - ModemSysPath?
   - ModemSystemPath?
   - SystemPath?
 - what about modems that don't come from udevng plugin? Property is
not set for these modems, is it ok? Do we want a generic property name to be able to also modify these plugins to give an information about hardware identification (something like: HardwareId, HardwarePath)? Or maybe SystemPath is also ok for these modems.

Let me know what you think about that and I will rework my patch to integrate your remarks,

Christophe

Quoted 143 lines · Denis Kenzior

Hi Christophe,

Quoted 11 lines · Christophe Ronco

---
  plugins/udevng.c |  2 ++
  src/modem.c      | 74
++++++++++++++++++++++++++++++++------------------------

  2 files changed, 45 insertions(+), 31 deletions(-)
plugins/udevng.c +0 -0
index 2279bbe9..d5139586 100644

plugins/udevng.c static gboolean create_modem(gpointer key, · line 1333

Quoted 4 lines
plugins/udevng.c +1 -0
@@ -1330,6 +1330,8 @@ static gboolean create_modem(gpointer key, 
 gpointer value, gpointer user_data)
               continue;
             if (driver_list[i].setup(modem) == TRUE) {
+            ofono_modem_set_string(modem->modem, "Syspath",

Use camel case, so "SysPath" or "SystemPath"

Quoted 60 lines
plugins/udevng.c +1 -0
+                                syspath);
               ofono_modem_register(modem->modem);
               return FALSE;
           }
src/modem.c +34 -0
index b1e8d3e2..7759b567 100644
@@ -773,6 +773,41 @@ ofono_bool_t ofono_modem_get_online(struct 
 ofono_modem *modem)
       return modem->online;
   }
   +static gboolean get_modem_property(struct ofono_modem *modem, 
 const char *name,
+                    enum property_type type,
+                    void *value)
+{
+    struct modem_property *property;
+
+    DBG("modem %p property %s", modem, name);
+
+    property = g_hash_table_lookup(modem->properties, name);
+
+    if (property == NULL) {
+        DBG("property %s does not exist", name);
+        return FALSE;
+    }
+
+    if (property->type != type) {
+        DBG("property %s wrong type", name);
+        return FALSE;
+    }
+
+    switch (property->type) {
+    case PROPERTY_TYPE_STRING:
+        *((const char **) value) = property->value;
+        return TRUE;
+    case PROPERTY_TYPE_INTEGER:
+        memcpy(value, property->value, sizeof(int));
+        return TRUE;
+    case PROPERTY_TYPE_BOOLEAN:
+        memcpy(value, property->value, sizeof(ofono_bool_t));
+        return TRUE;
+    default:
+        return FALSE;
+    }
+}
+
   void __ofono_modem_append_properties(struct ofono_modem *modem,

                          DBusMessageIter *dict)
  {

src/modem.c +1 -0
@@ -783,6 +818,7 @@ void __ofono_modem_append_properties(struct 
 ofono_modem *modem,
       struct ofono_devinfo *info;
       dbus_bool_t emergency = ofono_modem_get_emergency_mode(modem);
       const char *strtype;
+    char *value;
         ofono_dbus_dict_append(dict, "Online", DBUS_TYPE_BOOLEAN,
                   &modem->online);

src/modem.c void __ofono_modem_append_properties(struct · line 868

Quoted 10 lines
src/modem.c +6 -0
@@ -823,6 +859,13 @@ void __ofono_modem_append_properties(struct 
 ofono_modem *modem,
                           &info->svn);
       }
   +    if (get_modem_property(modem, "Syspath", PROPERTY_TYPE_STRING,
+                            &value) == TRUE) {
+        ofono_dbus_dict_append(dict, "ModemSyspath",
+                        DBUS_TYPE_STRING,
+                        &value);
+    }
+

There's no need for {}.  And why don't you make things much easier on yourself and simply use ofono_modem_get_string()

Quoted 41 lines
src/modem.c +0 -0
       interfaces = g_new0(char *, 
 g_slist_length(modem->interface_list) + 1);

      for (i = 0, l = modem->interface_list; l; l = l->next, i++) interfaces[i] = l->data;

src/modem.c +0 -30
@@ -1731,37 +1774,6 @@ static int set_modem_property(struct 
 ofono_modem *modem, const char *name,
       return 0;
   }
   -static gboolean get_modem_property(struct ofono_modem *modem, 
 const char *name,
-                    enum property_type type,
-                    void *value)
-{
-    struct modem_property *property;
-
-    DBG("modem %p property %s", modem, name);
-
-    property = g_hash_table_lookup(modem->properties, name);
-
-    if (property == NULL)
-        return FALSE;
-
-    if (property->type != type)
-        return FALSE;
-
-    switch (property->type) {
-    case PROPERTY_TYPE_STRING:
-        *((const char **) value) = property->value;
-        return TRUE;
-    case PROPERTY_TYPE_INTEGER:
-        memcpy(value, property->value, sizeof(int));
-        return TRUE;
-    case PROPERTY_TYPE_BOOLEAN:
-        memcpy(value, property->value, sizeof(ofono_bool_t));
-        return TRUE;
-    default:
-        return FALSE;
-    }
-}
-
   int ofono_modem_set_string(struct ofono_modem *modem,
Quoted 3 lines

                  const char *key, const char *value) {

Regards,
-Denis

#9 lore raw

Re: [PATCH] add syspath property to modem properties

Hi Christophe,

No top posting please.

Quoted 6 lines · Christophe Ronco

Hi Denis,

I didn't think this patch could be something used on Upstream Ofono. If it is, I also have remarks on my own patch:
 - shouldn't it be two patches (one for plugin directory, another for
src directory)?

yes, all the normal patch submission guidelines apply. You should also include a doc patch.

Quoted 5 lines

 - what is an acceptable name for the new DBUS property (currently it
is ModemSyspath):
   - ModemSysPath?
   - ModemSystemPath?
   - SystemPath?

Since you want to put it on the .Modem interface the 'Modem' prefix is redundant. SystemPath should be okay, or maybe SysfsPath

Quoted 5 lines

 - what about modems that don't come from udevng plugin? Property is
not set for these modems, is it ok? Do we want a generic property name to be able to also modify these plugins to give an information about hardware identification (something like: HardwareId, HardwarePath)? Or maybe SystemPath is also ok for these modems.

Or simply mark this property [optional] and be done with it.

Go ahead and resubmit and lets see what feedback we get.

Regards,
-Denis

#10 lore raw

cover Add SystemPath property to modem interface

Here is the patch serie as discussed in this thread of mails.

Christophe Ronco (3):
udevng: Add modem string SystemPath
modem: Add SystemPath dbus property
doc: Add SystemPath to Modem interface

 doc/modem-api.txt | 8 ++++++++
 plugins/udevng.c  | 2 ++
 src/modem.c       | 6 ++++++
 3 files changed, 16 insertions(+)
Signature · 1 line
2.7.4
#11 lore raw

PATCH 1/3 udevng: Add modem string SystemPath

---
 plugins/udevng.c | 2 ++
 1 file changed, 2 insertions(+)
plugins/udevng.c +2 -0
index 9b78ab4..3c7d99e 100644
@@ -1732,6 +1732,8 @@ static gboolean create_modem(gpointer key, gpointer value, gpointer user_data)
 			continue;
 
 		if (driver_list[i].setup(modem) == TRUE) {
+			ofono_modem_set_string(modem->modem, "SystemPath",
+								syspath);
 			ofono_modem_register(modem->modem);
 			return FALSE;
 		}
Signature · 1 line
2.7.4
#12 lore raw

PATCH 2/3 modem: Add SystemPath dbus property

In reply to Christophe Ronco
---
 src/modem.c | 6 ++++++
 1 file changed, 6 insertions(+)
src/modem.c +6 -0
index d5fda7c..0cee861 100644
@@ -792,6 +792,7 @@ void __ofono_modem_append_properties(struct ofono_modem *modem,
 	struct ofono_devinfo *info;
 	dbus_bool_t emergency = ofono_modem_get_emergency_mode(modem);
 	const char *strtype;
+	const char *system_path;
 
 	ofono_dbus_dict_append(dict, "Online", DBUS_TYPE_BOOLEAN,
 				&modem->online);
@@ -832,6 +833,11 @@ void __ofono_modem_append_properties(struct ofono_modem *modem,
 						&info->svn);
 	}
 
+	system_path = ofono_modem_get_string(modem, "SystemPath");
+	if (system_path)
+		ofono_dbus_dict_append(dict, "SystemPath", DBUS_TYPE_STRING,
+					&system_path);
+
 	interfaces = g_new0(char *, g_slist_length(modem->interface_list) + 1);
 	for (i = 0, l = modem->interface_list; l; l = l->next, i++)
 		interfaces[i] = l->data;
Signature · 1 line
2.7.4
#13 lore raw

PATCH 3/3 doc: Add SystemPath to Modem interface

In reply to Christophe Ronco
---
 doc/modem-api.txt | 8 ++++++++
 1 file changed, 8 insertions(+)
doc/modem-api.txt +8 -0
index 89e6191..e03c64a 100644
@@ -95,6 +95,14 @@ Properties	boolean Powered [readwrite]
 			String representing the software version number of the
 			modem device.
 
+		string SystemPath [readonly, optional]
+
+			String giving a consistent hardware path for the modem
+			device. It will not change in case of reboot of modem
+			or ofono.
+			For modems detected by udev events, gives modem sysfs
+			path.
+
 		array{string} Features [readonly]
 
 			List of currently enabled features. It uses simple
Signature · 1 line
2.7.4
#14 lore raw

Re: [PATCH 3/3] doc: Add SystemPath to Modem interface

Quoted 9 lines · Christophe Ronco
---
  doc/modem-api.txt | 8 ++++++++
  1 file changed, 8 insertions(+)
doc/modem-api.txt +0 -0
index 89e6191..e03c64a 100644

doc/modem-api.txt Properties boolean Powered [readwrite] · line 102

Quoted 8 lines
doc/modem-api.txt +5 -0
@@ -95,6 +95,14 @@ Properties	boolean Powered [readwrite]
  			String representing the software version number of the
  			modem device.
 
+		string SystemPath [readonly, optional]
+
+			String giving a consistent hardware path for the modem
+			device. It will not change in case of reboot of modem
+			or ofono.

Given that many modems are USB devices, I don't think you can actually promise this. I can envision a (slightly pathological) hardware design whereby the presence or absence of an _external_ device at startup may change the device enumeration order and thereby the syspath...

/Jonas

Quoted 7 lines
doc/modem-api.txt +3 -0
+			For modems detected by udev events, gives modem sysfs
+			path.
+
  		array{string} Features [readonly]
 
  			List of currently enabled features. It uses simple
#15 lore raw

Re: [PATCH 3/3] doc: Add SystemPath to Modem interface

Hi,

Quoted 11 lines
Quoted 5 lines

+        string SystemPath [readonly, optional]
+
+            String giving a consistent hardware path for the modem +            device. It will not change in case of reboot of modem +            or ofono.

Given that many modems are USB devices, I don't think you can actually promise this.  I can envision a (slightly pathological) hardware design whereby the presence or absence of an _external_ device at startup may change the device enumeration order and thereby the syspath...

Right, this can depend on the kernel version also. I wouldn't claim any guarantees at the D-Bus API level as it is outside our control.

If someone chooses to use it as a consistent path, then they do it at their own risk.

Regards,
-Denis

#16 lore raw

PATCH 1/1 doc: Add SystemPath to Modem interface

---
 doc/modem-api.txt | 7 +++++++
 1 file changed, 7 insertions(+)
doc/modem-api.txt +7 -0
index 89e6191..ff9facd 100644
@@ -95,6 +95,13 @@ Properties	boolean Powered [readwrite]
 			String representing the software version number of the
 			modem device.
 
+		string SystemPath [readonly, optional]
+
+			String representing the system path for the modem
+			device.
+			For modems detected by udev events, this corresponds to
+			the modem sysfs path.
+
 		array{string} Features [readonly]
 
 			List of currently enabled features. It uses simple
Signature · 1 line
2.7.4
#17 lore raw

Re: [PATCH 0/3] Add SystemPath property to modem interface

In reply to Christophe Ronco

Hi Christophe,

Quoted 7 lines · Christophe Ronco

Here is the patch serie as discussed in this thread of mails.

Christophe Ronco (3):
udevng: Add modem string SystemPath
modem: Add SystemPath dbus property
doc: Add SystemPath to Modem interface

Quoted 5 lines
  doc/modem-api.txt | 8 ++++++++
  plugins/udevng.c  | 2 ++
  src/modem.c       | 6 ++++++
  3 files changed, 16 insertions(+)

I went ahead and applied all these (with the updated doc 3/3 patch).

Regards,
-Denis

#18 lore raw

PATCH 1/1 modem: ofono_modem_create: log invalid paths

In reply to Pau Espin Pedrol
---
 src/modem.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
src/modem.c +3 -1
index d5fda7ca..698a7ca8 100644
@@ -1876,8 +1876,10 @@ struct ofono_modem *ofono_modem_create(const char *name, const char *type)
 	else
 		snprintf(path, sizeof(path), "/%s", name);
 
-	if (!dbus_validate_path(path, NULL))
+	if (!dbus_validate_path(path, NULL)) {
+		DBG("Invalid dbus path for modem: %s", path);
 		return NULL;
+	}
 
 	modem = g_try_new0(struct ofono_modem, 1);
 
Signature · 1 line
2.16.1