Skip to contentSkip to conversation

usb_id segfaulting on 2.6.23

DISCUSSION Daniel Drake 5 participants 21 messages Oct 15, 2007 – Nov 6, 2007

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

Hi,

Since upgrading to 2.6.23, usb_id is segfaulting when dealing with my input devices. I'm using udev-116. I can get an instant segfault simply by replugging my keyboard.

I have dug around a bit. The segfault happens inside usb_id, see the "fallback" section.

It reaches here:

		dbg("No USB vendor information available\n");
		sprintf(vendor_str,"0000");

then a few lines later:

		set_str(vendor_str,usb_vendor, sizeof(vendor_str) - 1);

usb_vendor is NULL. set_str does various string manipulations with this NULL pointer which causes the segfault.

I suggest fixing this by replacing the sprintf line with:

	usb_vendor = "0000";

The same bug exists a few lines below with the USB model handling.

Now for what actually caused the change in behaviour: the devpath passed to usb_id has changed between 2.6.22 and 2.6.23.

In 2.6.22, it was /class/input/input4/event2 and I would get this output:

Quoted 9 lines
# /lib/udev/usb_id /class/input/input4/event2
Microsoft_Microsoft_Keyboard_with_Fingerprint_Reader_{3A1B506A-BB77-444C-851B-C424997B2763}
ID_VENDOR=Microsoft
ID_MODEL=Microsoft_Keyboard_with_Fingerprint_Reader
ID_REVISION00
ID_SERIAL=Microsoft_Microsoft_Keyboard_with_Fingerprint_Reader_{3A1B506A-BB77-444C-851B-C424997B2763}
ID_SERIAL_SHORT={3A1B506A-BB77-444C-851B-C424997B2763}
ID_TYPE=hid
ID_BUS=usb

In 2.6.23 the devpath now provided to usb_id is /devices/pci0000:00/0000:00:02.0/usb2/2-2/2-2:1.0/input/input13/event2

After applying my segfault fixes I get the following from usb_id:

Quoted 2 lines
# /lib/udev/usb_id /devices/pci0000:00/0000:00:02.0/usb2/2-2/2-21.0/input/input13/event2
0000_0000

Why has the devpath changed? Is this a kernel bug for breaking userspace compat? Or should usb_id just be adapted to be able to handle both types of devpath?

Thanks!
Daniel

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#1 lore raw
Quoted 13 lines · Daniel Drake

Since upgrading to 2.6.23, usb_id is segfaulting when dealing with my input devices. I'm using udev-116. I can get an instant segfault simply by replugging my keyboard.

I have dug around a bit. The segfault happens inside usb_id, see the "fallback" section.

It reaches here:

                dbg("No USB vendor information available\n");
                sprintf(vendor_str,"0000");

then a few lines later:

Quoted 7 lines
                set_str(vendor_str,usb_vendor, sizeof(vendor_str) - 1);

usb_vendor is NULL. set_str does various string manipulations with this NULL pointer which causes the segfault.

I suggest fixing this by replacing the sprintf line with:

        usb_vendor = "0000";

That's clearly a silly bug in usb_id. We should never return any 0000 values and just fail completely instead. No idea why such crap is in there, it needs to be removed.

Quoted 1 line

The same bug exists a few lines below with the USB model handling.

I'll go fix up the mess here. Thanks for debugging this.

Quoted 18 lines

Now for what actually caused the change in behaviour: the devpath passed to usb_id has changed between 2.6.22 and 2.6.23.

In 2.6.22, it was /class/input/input4/event2 and I would get this output:

Quoted 9 lines
# /lib/udev/usb_id /class/input/input4/event2
Microsoft_Microsoft_Keyboard_with_Fingerprint_Reader_{3A1B506A-BB77-444C-851B-C424997B2763}
ID_VENDOR=Microsoft
ID_MODEL=Microsoft_Keyboard_with_Fingerprint_Reader
ID_REVISION00
ID_SERIAL=Microsoft_Microsoft_Keyboard_with_Fingerprint_Reader_{3A1B506A-BB77-444C-851B-C424997B2763}
ID_SERIAL_SHORT={3A1B506A-BB77-444C-851B-C424997B2763}
ID_TYPE=hid
ID_BUS=usb

In 2.6.23 the devpath now provided to usb_id is /devices/pci0000:00/0000:00:02.0/usb2/2-2/2-2:1.0/input/input13/event2

Quoted 7 lines

After applying my segfault fixes I get the following from usb_id:

Quoted 2 lines
# /lib/udev/usb_id /devices/pci0000:00/0000:00:02.0/usb2/2-2/2-21.0/input/input13/event2
0000_0000

Why has the devpath changed? Is this a kernel bug for breaking userspace compat? Or should usb_id just be adapted to be able to handle both types of devpath?

It's fine, "input" uses the unified sysfs tree now. If you don't configure CONFIG_SYSFS_DEPRECATED, you will have all devices in /sys/devices instead of spreaded around in several places. Usb_id must be fixed, the kernel is fine.

Thanks,
Kay

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#2 lore raw
In reply to Daniel Drake
Quoted 22 lines · Kay Sievers
Quoted 13 lines · Daniel Drake

Since upgrading to 2.6.23, usb_id is segfaulting when dealing with my input devices. I'm using udev-116. I can get an instant segfault simply by replugging my keyboard.

I have dug around a bit. The segfault happens inside usb_id, see the "fallback" section.

It reaches here:

                dbg("No USB vendor information available\n");
                sprintf(vendor_str,"0000");

then a few lines later:

Quoted 7 lines
                set_str(vendor_str,usb_vendor, sizeof(vendor_str) - 1);

usb_vendor is NULL. set_str does various string manipulations with this NULL pointer which causes the segfault.

I suggest fixing this by replacing the sprintf line with:

        usb_vendor = "0000";
Quoted 29 lines

That's clearly a silly bug in usb_id. We should never return any 0000 values and just fail completely instead. No idea why such crap is in there, it needs to be removed.

Quoted 1 line

The same bug exists a few lines below with the USB model handling.

I'll go fix up the mess here. Thanks for debugging this.

Quoted 21 lines

Now for what actually caused the change in behaviour: the devpath passed to usb_id has changed between 2.6.22 and 2.6.23.

In 2.6.22, it was /class/input/input4/event2 and I would get this output:

Quoted 9 lines
# /lib/udev/usb_id /class/input/input4/event2
Microsoft_Microsoft_Keyboard_with_Fingerprint_Reader_{3A1B506A-BB77-444C-851B-C424997B2763}
ID_VENDOR=Microsoft
ID_MODEL=Microsoft_Keyboard_with_Fingerprint_Reader
ID_REVISION00
ID_SERIAL=Microsoft_Microsoft_Keyboard_with_Fingerprint_Reader_{3A1B506A-BB77-444C-851B-C424997B2763}
ID_SERIAL_SHORT={3A1B506A-BB77-444C-851B-C424997B2763}
ID_TYPE=hid
ID_BUS=usb

In 2.6.23 the devpath now provided to usb_id is /devices/pci0000:00/0000:00:02.0/usb2/2-2/2-2:1.0/input/input13/event2

After applying my segfault fixes I get the following from usb_id:

Quoted 2 lines
# /lib/udev/usb_id /devices/pci0000:00/0000:00:02.0/usb2/2-2/2-21.0/input/input13/event2
0000_0000

Hmm, works fine here with my USB keyboard. Can you provide the output of:
udevinfo -a -p <devpath>
for the failing setup?

Thanks,
Kay

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#3 lore raw
In reply to Daniel Drake
Quoted 3 lines · Kay Sievers

Hmm, works fine here with my USB keyboard. Can you provide the output of:
udevinfo -a -p <devpath>
for the failing setup?

# udevinfo -a -p 
/devices/pci0000:00/0000:00:02.0/usb2/2-2/2-2:1.0/input/input2/event2

Udevinfo starts with the device specified by the devpath and then walks up the chain of parent devices. It prints for every device found, all possible attributes in the udev rules key format. A rule to match, can be composed by the attributes of the device and the attributes from one single parent device.

looking at device
'/devices/pci0000:00/0000:00:02.0/usb2/2-2/2-2:1.0/input/input2/event2':

     KERNEL="event2"
     SUBSYSTEM="input"
     DRIVER=""
     ATTR{dev}="13:66"

looking at parent device
'/devices/pci0000:00/0000:00:02.0/usb2/2-2/2-2:1.0/input/input2':

     KERNELS="input2"
     SUBSYSTEMS="input"
     DRIVERS=""

ATTRS{modalias}="input:b0003v045Ep00BBe0111-e0,1,11,14,k71,72,73,74,75,77,79,7A,7B,7C,7D,7E,7F,80,81,82,83,84,85,86,87,88,89,8A,B7,B8,B9,BA,BB,BC,BD,BE,BF,C0,C1,C2,F0,raml0,1,2,8,sfw"

     ATTRS{uniq}="_3A1B506A-BB77-444C-851B-C424997B2763_"
     ATTRS{phys}="usb-0000:00:02.0-2/input0"

looking at parent device
'/devices/pci0000:00/0000:00:02.0/usb2/2-2/2-2:1.0/input':

     KERNELS="input"
     SUBSYSTEMS=""
     DRIVERS=""

looking at parent device
'/devices/pci0000:00/0000:00:02.0/usb2/2-2/2-2:1.0':

     KERNELS="2-2:1.0"
     SUBSYSTEMS="usb"
     DRIVERS="usbhid"
     ATTRS{modalias}="usb:v045Ep00BBd0100dc00dsc00dp00ic03isc01ip01"
     ATTRS{bInterfaceProtocol}="01"
     ATTRS{bInterfaceSubClass}="01"
     ATTRS{bInterfaceClass}="03"
     ATTRS{bNumEndpoints}="01"
     ATTRS{bAlternateSetting}=" 0"
     ATTRS{bInterfaceNumber}="00"

looking at parent device '/devices/pci0000:00/0000:00:02.0/usb2/2-2':

     KERNELS="2-2"
     SUBSYSTEMS="usb"
     DRIVERS="usb"
     ATTRS{dev}="189:129"

looking at parent device '/devices/pci0000:00/0000:00:02.0/usb2':

     KERNELS="usb2"
     SUBSYSTEMS="usb"
     DRIVERS="usb"
     ATTRS{dev}="189:128"

looking at parent device '/devices/pci0000:00/0000:00:02.0':

     KERNELS="0000:00:02.0"
     SUBSYSTEMS="pci"
     DRIVERS="ohci_hcd"
     ATTRS{msi_bus}=""
     ATTRS{broken_parity_status}="0"
     ATTRS{enable}="1"

ATTRS{modalias}="pci:v000010DEd0000036Csv0000147Bsd00001C25bc0Csc03i10"

     ATTRS{local_cpus}="ffffffff"
     ATTRS{irq}="23"
     ATTRS{class}="0x0c0310"
     ATTRS{subsystem_device}="0x1c25"
     ATTRS{subsystem_vendor}="0x147b"
     ATTRS{device}="0x036c"
     ATTRS{vendor}="0x10de"

looking at parent device '/devices/pci0000:00':

     KERNELS="pci0000:00"
     SUBSYSTEMS=""
     DRIVERS=""
     ATTRS{uevent}=""

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#4 lore raw
In reply to Daniel Drake
Quoted 79 lines · Daniel Drake
Quoted 3 lines · Kay Sievers

Hmm, works fine here with my USB keyboard. Can you provide the output of:
udevinfo -a -p <devpath>
for the failing setup?

# udevinfo -a -p
/devices/pci0000:00/0000:00:02.0/usb2/2-2/2-2:1.0/input/input2/event2

Udevinfo starts with the device specified by the devpath and then walks up the chain of parent devices. It prints for every device found, all possible attributes in the udev rules key format. A rule to match, can be composed by the attributes of the device and the attributes from one single parent device.

looking at device
'/devices/pci0000:00/0000:00:02.0/usb2/2-2/2-2:1.0/input/input2/event2':

     KERNEL="event2"
     SUBSYSTEM="input"
     DRIVER=""
     ATTR{dev}="13:66"

looking at parent device
'/devices/pci0000:00/0000:00:02.0/usb2/2-2/2-2:1.0/input/input2':

     KERNELS="input2"
     SUBSYSTEMS="input"
     DRIVERS=""

ATTRS{modalias}="input:b0003v045Ep00BBe0111-e0,1,11,14,k71,72,73,74,75,77,79,7A,7B,7C,7D,7E,7F,80,81,82,83,84,85,86,87,88,89,8A,B7,B8,B9,BA,BB,BC,BD,BE,BF,C0,C1,C2,F0,raml0,1,2,8,sfw"

     ATTRS{uniq}="_3A1B506A-BB77-444C-851B-C424997B2763_"
     ATTRS{phys}="usb-0000:00:02.0-2/input0"

looking at parent device
'/devices/pci0000:00/0000:00:02.0/usb2/2-2/2-2:1.0/input':

     KERNELS="input"
     SUBSYSTEMS=""
     DRIVERS=""

looking at parent device
'/devices/pci0000:00/0000:00:02.0/usb2/2-2/2-2:1.0':

     KERNELS="2-2:1.0"
     SUBSYSTEMS="usb"
     DRIVERS="usbhid"
     ATTRS{modalias}="usb:v045Ep00BBd0100dc00dsc00dp00ic03isc01ip01"
     ATTRS{bInterfaceProtocol}="01"
     ATTRS{bInterfaceSubClass}="01"
     ATTRS{bInterfaceClass}="03"
     ATTRS{bNumEndpoints}="01"
     ATTRS{bAlternateSetting}=" 0"
     ATTRS{bInterfaceNumber}="00"
   looking at parent device '/devices/pci0000:00/0000:00:02.0/usb2/2-2':
     KERNELS="2-2"
     SUBSYSTEMS="usb"
     DRIVERS="usb"
     ATTRS{dev}="189:129"

   looking at parent device '/devices/pci0000:00/0000:00:02.0/usb2':
     KERNELS="usb2"
     SUBSYSTEMS="usb"
     DRIVERS="usb"
     ATTRS{dev}="189:128"

   looking at parent device '/devices/pci0000:00/0000:00:02.0':
     KERNELS="0000:00:02.0"
     SUBSYSTEMS="pci"
     DRIVERS="ohci_hcd"
     ATTRS{msi_bus}=""
     ATTRS{broken_parity_status}="0"
     ATTRS{enable}="1"

ATTRS{modalias}="pci:v000010DEd0000036Csv0000147Bsd00001C25bc0Csc03i10"

     ATTRS{local_cpus}="ffffffff"
     ATTRS{irq}="23"
     ATTRS{class}="0x0c0310"
     ATTRS{subsystem_device}="0x1c25"
     ATTRS{subsystem_vendor}="0x147b"
     ATTRS{device}="0x036c"
     ATTRS{vendor}="0x10de"
Quoted 5 lines
   looking at parent device '/devices/pci0000:00':
     KERNELS="pci0000:00"
     SUBSYSTEMS=""
     DRIVERS=""
     ATTRS{uevent}=""

Strange, seems some attributes are not catched by udev. What does "udevinfo" print for the working setup? Do you see the missing attributes there?

Kay

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#5 lore raw
In reply to Daniel Drake
Quoted 16 lines · Kay Sievers
Quoted 15 lines · Daniel Drake

Since upgrading to 2.6.23, usb_id is segfaulting when dealing with my input devices. I'm using udev-116. I can get an instant segfault simply by replugging my keyboard.

I have dug around a bit. The segfault happens inside usb_id, see the "fallback" section.

It reaches here:

                dbg("No USB vendor information available\n");
                sprintf(vendor_str,"0000");

then a few lines later:

                set_str(vendor_str,usb_vendor, sizeof(vendor_str) - 1);
Quoted 8 lines
Quoted 5 lines

usb_vendor is NULL. set_str does various string manipulations with this NULL pointer which causes the segfault.

I suggest fixing this by replacing the sprintf line with:

        usb_vendor = "0000";

That's clearly a silly bug in usb_id. We should never return any 0000 values and just fail completely instead.

I've removed the 0000 logic entirely, and we fail if we are unable to retrieve the information we are looking for.

Now we only need to know what goes wrong with your keyboard. :)

Thanks again for the debugging,
Kay

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#6 lore raw
In reply to Daniel Drake

Sorry for the delayed response. I was away from this computer for a while, and I wanted to get a decent amount of info before we continue.

Quoted 3 lines · Kay Sievers

Strange, seems some attributes are not catched by udev. What does "udevinfo" print for the working setup? Do you see the missing attributes there?

You're referring to idVendor, idProduct, etc. Actually, those attributes don't even exist in sysfs(!!) which must be the root of the problem.

So, I bisected the kernel. Ready for a strange result? First bad commit is 296699de: Introduce CONFIG_SUSPEND for suspend-to-Ram and standby

On this system, I always disable suspend/hibernate/etc support, and continued to do so throughout the bisection. This lead me to the above result.

On a kernel tree with HEAD as 296699de, I see the bug (no idVendor, etc) when CONFIG_SUSPEND=n (my normal setup). However, if I set CONFIG_SUSPEND=y, the missing USB sysfs attributes show up!

On 2.6.24-rc1-git14, the same is true. CONFIG_SUSPEND=n, no idVendor nodes, udev usb_id crashes on boot. CONFIG_SUSPEND=y, idVendor and friends are back, udev behaves as normal.

I know you have experience working with sysfs/kobject internals, do you have any suggestions or thoughts before I take this to LKML?

Thanks!
Daniel

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#7 lore raw
In reply to Daniel Drake
Quoted 16 lines · Daniel Drake
Quoted 3 lines · Kay Sievers

Strange, seems some attributes are not catched by udev. What does "udevinfo" print for the working setup? Do you see the missing attributes there?

You're referring to idVendor, idProduct, etc. Actually, those attributes don't even exist in sysfs(!!) which must be the root of the problem.

So, I bisected the kernel. Ready for a strange result? First bad commit is 296699de: Introduce CONFIG_SUSPEND for suspend-to-Ram and standby

On this system, I always disable suspend/hibernate/etc support, and continued to do so throughout the bisection. This lead me to the above result.

Quoted 7 lines

On a kernel tree with HEAD as 296699de, I see the bug (no idVendor, etc) when CONFIG_SUSPEND=n (my normal setup). However, if I set CONFIG_SUSPEND=y, the missing USB sysfs attributes show up!

On 2.6.24-rc1-git14, the same is true. CONFIG_SUSPEND=n, no idVendor nodes, udev usb_id crashes on boot. CONFIG_SUSPEND=y, idVendor and friends are back, udev behaves as normal.

I fixes usb_id in the meantime not to crash, but that's a different story.

Quoted 2 lines

I know you have experience working with sysfs/kobject internals, do you have any suggestions or thoughts before I take this to LKML?

Oh, great, it was a complete mystery what was going wrong, maybe this is pointing in the right direction. We got more than one report for this bug.

Maybe Oliver or Alan have an idea, what could go wrong with the USB attributes depending on CONFIG_SUSPEND, I don't expect a driver core issue here.

Thanks Daniel for the debugging,
Kay

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#8 lore raw
In reply to Daniel Drake

Am Dienstag 06 November 2007 schrieb Kay Sievers:

Quoted 31 lines · Daniel Drake
Quoted 16 lines
Quoted 3 lines · Kay Sievers

Strange, seems some attributes are not catched by udev. What does "udevinfo" print for the working setup? Do you see the missing attributes there?

You're referring to idVendor, idProduct, etc. Actually, those attributes don't even exist in sysfs(!!) which must be the root of the problem.

So, I bisected the kernel. Ready for a strange result? First bad commit is 296699de: Introduce CONFIG_SUSPEND for suspend-to-Ram and standby

On this system, I always disable suspend/hibernate/etc support, and continued to do so throughout the bisection. This lead me to the above result.

Quoted 7 lines

On a kernel tree with HEAD as 296699de, I see the bug (no idVendor, etc) when CONFIG_SUSPEND=n (my normal setup). However, if I set CONFIG_SUSPEND=y, the missing USB sysfs attributes show up!

On 2.6.24-rc1-git14, the same is true. CONFIG_SUSPEND=n, no idVendor nodes, udev usb_id crashes on boot. CONFIG_SUSPEND=y, idVendor and friends are back, udev behaves as normal.

I fixes usb_id in the meantime not to crash, but that's a different story.

Quoted 2 lines

I know you have experience working with sysfs/kobject internals, do you have any suggestions or thoughts before I take this to LKML?

Quoted 7 lines

Oh, great, it was a complete mystery what was going wrong, maybe this is pointing in the right direction. We got more than one report for this bug.

Maybe Oliver or Alan have an idea, what could go wrong with the USB attributes depending on CONFIG_SUSPEND, I don't expect a driver core issue here.

Is bcdDevice present without CONFIG_SUSPEND? Which attributes are present?

Regards
Oliver

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#9 lore raw
In reply to Daniel Drake
Quoted 2 lines · Oliver Neukum

Is bcdDevice present without CONFIG_SUSPEND? Which attributes are present?

Without CONFIG_SUSPEND, the contents of my /sys/devices/pci0000:00/0000:00:02.0/usb2/2-10/ are as follows:

2-10:1.0
dev
driver
subsystem
uevent
usb_device

The missing attributes are:

authorized
bcdDevice
bConfigurationValue
bDeviceClass
bDeviceProtocol
bDeviceSubClass
bmAttributes
bMaxPacketSize0
bMaxPower
bNumConfigurations
bNumInterfaces
busnum
configuration
descriptors
devnum
ep_00
idProduct
idVendor
manufacturer
maxchild
power
product
quirks
speed
urbnum
usb_endpoint
version

Daniel

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#10 lore raw
In reply to Daniel Drake

Am Dienstag 06 November 2007 schrieb Daniel Drake:

Quoted 3 lines
Quoted 2 lines · Oliver Neukum

Is bcdDevice present without CONFIG_SUSPEND? Which attributes are present?

Looks like a failure in drivers/usb/core/sysfs.c::usb_create_sysfs_dev_files() Can you put printks there to show where it fails?
I am trying to replicate this, but the compilation will take some time.

Regards
Oliver

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#11 lore raw
In reply to Daniel Drake

Am Dienstag 06 November 2007 schrieb Daniel Drake:

Quoted 6 lines
Quoted 2 lines · Oliver Neukum

Is bcdDevice present without CONFIG_SUSPEND? Which attributes are present?

Without CONFIG_SUSPEND, the contents of my /sys/devices/pci0000:00/0000:00:02.0/usb2/2-10/ are as follows:

Can you please provide your full .config?

Regards
Oliver

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#12 lore raw
In reply to Daniel Drake
Quoted 7 lines · Oliver Neukum

Am Dienstag 06 November 2007 schrieb Daniel Drake:

Quoted 3 lines
Quoted 2 lines · Oliver Neukum

Is bcdDevice present without CONFIG_SUSPEND? Which attributes are present?

Looks like a failure in drivers/usb/core/sysfs.c::usb_create_sysfs_dev_files() Can you put printks there to show where it fails?

Done. It fails here:

	retval = add_power_attributes(dev);
	if (retval)
		goto error;

add_power_attributes returns -2.

Digging further, inside add_power_attributes(), the first call to sysfs_add_file_to_group() returns -2. The second one is therefore not attempted.

Daniel

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#15 lore raw
In reply to Daniel Drake
Quoted 35 lines · Daniel Drake
Quoted 2 lines · Oliver Neukum

Is bcdDevice present without CONFIG_SUSPEND? Which attributes are present?

Without CONFIG_SUSPEND, the contents of my /sys/devices/pci0000:00/0000:00:02.0/usb2/2-10/ are as follows:

2-10:1.0
dev
driver
subsystem
uevent
usb_device

The missing attributes are:

authorized
bcdDevice
bConfigurationValue
bDeviceClass
bDeviceProtocol
bDeviceSubClass
bmAttributes
bMaxPacketSize0
bMaxPower
bNumConfigurations
bNumInterfaces
busnum
configuration
descriptors
devnum
ep_00
idProduct
idVendor
manufacturer

Quoted 8 lines

maxchild
power
product
quirks
speed
urbnum
usb_endpoint
version

Can you post a dmesg log showing what happens when a USB device is plugged into a kernel with CONFIG_USB_DEBUG enabled and CONFIG_USB_SUSPEND disabled?

Also, you might want to experiment on your own to trace the flow of control. Here's what _should_ happen when a new USB device is detected:

In hub.c, usb_new_device() calls device_add(&udev->dev).

In generic.c, generic_probe() is called to bind to the new device.

generic_probe() calls usb_create_sysfs_dev_files() in sysfs.c.

usb_create_sysfs_dev_files() does a whole bunch of sysfs_create_group() and related function calls.

The rest is handled by the sysfs core.

Maybe you can find out at which stage things go wrong.

Alan Stern

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#13 lore raw
In reply to Daniel Drake
Quoted 15 lines · Daniel Drake
Quoted 7 lines · Oliver Neukum

Am Dienstag 06 November 2007 schrieb Daniel Drake:

Quoted 3 lines
Quoted 2 lines · Oliver Neukum

Is bcdDevice present without CONFIG_SUSPEND? Which attributes are present?

Looks like a failure in drivers/usb/core/sysfs.c::usb_create_sysfs_dev_files() Can you put printks there to show where it fails?

Done. It fails here:

	retval = add_power_attributes(dev);
	if (retval)
		goto error;
Quoted 5 lines

add_power_attributes returns -2.

Digging further, inside add_power_attributes(), the first call to sysfs_add_file_to_group() returns -2. The second one is therefore not attempted.

Ah, okay. This is a known problem, a configuration-related issue. The PM core doesn't create a device's /sys/.../power subdirectory unless CONFIG_PM_SLEEP is set. In fact it should be created whenever CONFIG_PM is set.

I haven't had time to write a patch for this yet. Do you want to try?

Alan Stern

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#14 lore raw
Quoted 5 lines · Alan Stern

Ah, okay. This is a known problem, a configuration-related issue. The PM core doesn't create a device's /sys/.../power subdirectory unless CONFIG_PM_SLEEP is set. In fact it should be created whenever
CONFIG_PM is set.

I haven't had time to write a patch for this yet. Do you want to try?

OK. Any comments on this before I sent it to the proper channels?

Thanks,
Daniel

From: Daniel Drake <[email protected]>

drivers/base/core.c +6 -1
Index: linux-2.6.24-rc1-git14/drivers/base/core.c
===================================================================
@@ -770,6 +770,9 @@ int device_add(struct device *dev)
 	error = device_add_attrs(dev);
 	if (error)
 		goto AttrsError;
+	error = dpm_sysfs_add(dev);
+	if (error)
+		goto PMsysfsError;
 	error = device_pm_add(dev);
 	if (error)
 		goto PMError;
@@ -797,11 +800,13 @@ int device_add(struct device *dev)
 	return error;
  BusError:
 	device_pm_remove(dev);
- PMError:
+ PMsysfsError:
 	if (dev->bus)
 		blocking_notifier_call_chain(&dev->bus->bus_notifier,
 					     BUS_NOTIFY_DEL_DEVICE, dev);
 	device_remove_attrs(dev);
+ PMError:
+	dpm_sysfs_remove(dev);
  AttrsError:
 	device_remove_class_symlinks(dev);
  SymlinkError:
drivers/base/power/Makefile +2 -1
Index: linux-2.6.24-rc1-git14/drivers/base/power/Makefile
===================================================================
@@ -1,5 +1,6 @@
 obj-y			:= shutdown.o
-obj-$(CONFIG_PM_SLEEP)	+= main.o sysfs.o
+obj-$(CONFIG_PM)	+= sysfs.o
+obj-$(CONFIG_PM_SLEEP)	+= main.o
 obj-$(CONFIG_PM_TRACE)	+= trace.o
 
 ifeq ($(CONFIG_DEBUG_DRIVER),y)
drivers/base/power/power.h +21 -5
Index: linux-2.6.24-rc1-git14/drivers/base/power/power.h
===================================================================
@@ -21,6 +21,22 @@ static inline struct device * to_device(
 extern int device_pm_add(struct device *);
 extern void device_pm_remove(struct device *);
 
+#else /* CONFIG_PM_SLEEP */
+
+
+static inline int device_pm_add(struct device * dev)
+{
+	return 0;
+}
+static inline void device_pm_remove(struct device * dev)
+{
+
+}
+
+#endif
+
+#ifdef CONFIG_PM
+
 /*
  * sysfs.c
  */
@@ -28,16 +44,16 @@ extern void device_pm_remove(struct devi
 extern int dpm_sysfs_add(struct device *);
 extern void dpm_sysfs_remove(struct device *);
 
-#else /* CONFIG_PM_SLEEP */
-
+#else /* CONFIG_PM */
 
-static inline int device_pm_add(struct device * dev)
+static inline int dpm_sysfs_add(struct device * dev)
 {
 	return 0;
 }
-static inline void device_pm_remove(struct device * dev)
+static inline void dpm_sysfs_remove(struct device * dev)
 {
 
 }
 
-#endif
+#endif /* CONFIG_PM */
+
#16 lore raw
In reply to Daniel Drake
Quoted 5 lines · Alan Stern

Ah, okay. This is a known problem, a configuration-related issue. The PM core doesn't create a device's /sys/.../power subdirectory unless CONFIG_PM_SLEEP is set. In fact it should be created whenever
CONFIG_PM is set.

I haven't had time to write a patch for this yet. Do you want to try?

OK. Any comments on this before I sent it to the proper channels?

Thanks,
Daniel

From: Daniel Drake <[email protected]>

drivers/base/core.c +6 -1
Index: linux-2.6.24-rc1-git14/drivers/base/core.c
===================================================================
@@ -770,6 +770,9 @@ int device_add(struct device *dev)
 	error = device_add_attrs(dev);
 	if (error)
 		goto AttrsError;
+	error = dpm_sysfs_add(dev);
+	if (error)
+		goto PMsysfsError;
 	error = device_pm_add(dev);
 	if (error)
 		goto PMError;
@@ -797,11 +800,13 @@ int device_add(struct device *dev)
 	return error;
  BusError:
 	device_pm_remove(dev);
- PMError:
+ PMsysfsError:
 	if (dev->bus)
 		blocking_notifier_call_chain(&dev->bus->bus_notifier,
 					     BUS_NOTIFY_DEL_DEVICE, dev);
 	device_remove_attrs(dev);
+ PMError:
+	dpm_sysfs_remove(dev);
  AttrsError:
 	device_remove_class_symlinks(dev);
  SymlinkError:
drivers/base/power/Makefile +2 -1
Index: linux-2.6.24-rc1-git14/drivers/base/power/Makefile
===================================================================
@@ -1,5 +1,6 @@
 obj-y			:= shutdown.o
-obj-$(CONFIG_PM_SLEEP)	+= main.o sysfs.o
+obj-$(CONFIG_PM)	+= sysfs.o
+obj-$(CONFIG_PM_SLEEP)	+= main.o
 obj-$(CONFIG_PM_TRACE)	+= trace.o
 
 ifeq ($(CONFIG_DEBUG_DRIVER),y)
drivers/base/power/power.h +21 -5
Index: linux-2.6.24-rc1-git14/drivers/base/power/power.h
===================================================================
@@ -21,6 +21,22 @@ static inline struct device * to_device(
 extern int device_pm_add(struct device *);
 extern void device_pm_remove(struct device *);
 
+#else /* CONFIG_PM_SLEEP */
+
+
+static inline int device_pm_add(struct device * dev)
+{
+	return 0;
+}
+static inline void device_pm_remove(struct device * dev)
+{
+
+}
+
+#endif
+
+#ifdef CONFIG_PM
+
 /*
  * sysfs.c
  */
@@ -28,16 +44,16 @@ extern void device_pm_remove(struct devi
 extern int dpm_sysfs_add(struct device *);
 extern void dpm_sysfs_remove(struct device *);
 
-#else /* CONFIG_PM_SLEEP */
-
+#else /* CONFIG_PM */
 
-static inline int device_pm_add(struct device * dev)
+static inline int dpm_sysfs_add(struct device * dev)
 {
 	return 0;
 }
-static inline void device_pm_remove(struct device * dev)
+static inline void dpm_sysfs_remove(struct device * dev)
 {
 
 }
 
-#endif
+#endif /* CONFIG_PM */
+

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/

_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#17 lore raw
Quoted 8 lines · Daniel Drake
Quoted 5 lines · Alan Stern

Ah, okay. This is a known problem, a configuration-related issue. The PM core doesn't create a device's /sys/.../power subdirectory unless CONFIG_PM_SLEEP is set. In fact it should be created whenever
CONFIG_PM is set.

I haven't had time to write a patch for this yet. Do you want to try?

OK. Any comments on this before I sent it to the proper channels?

Missed a hunk where I remove the existing dpm_sysfs_add() call from drivers/base/power/main.c but you get the idea.. although I'm increasingly unsure if this is the right approach.

Daniel

#18 lore raw
In reply to Daniel Drake
Quoted 9 lines · Daniel Drake
Quoted 6 lines · Alan Stern

Ah, okay. This is a known problem, a configuration-related issue. The PM core doesn't create a device's /sys/.../power subdirectory unless CONFIG_PM_SLEEP is set. In fact it should be created whenever CONFIG_PM is set.

I haven't had time to write a patch for this yet. Do you want to try?

OK. Any comments on this before I sent it to the proper channels?

Missed a hunk where I remove the existing dpm_sysfs_add() call from drivers/base/power/main.c but you get the idea.. although I'm increasingly unsure if this is the right approach.

Daniel

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#19 lore raw
Quoted 7 lines · Daniel Drake
Quoted 6 lines · Daniel Drake
Quoted 5 lines · Alan Stern

Ah, okay. This is a known problem, a configuration-related issue. The PM core doesn't create a device's /sys/.../power subdirectory unless CONFIG_PM_SLEEP is set. In fact it should be created whenever CONFIG_PM is set.

Quoted 6 lines
Quoted 2 lines
Quoted 1 line

I haven't had time to write a patch for this yet. Do you want to try?

OK. Any comments on this before I sent it to the proper channels?

Missed a hunk where I remove the existing dpm_sysfs_add() call from drivers/base/power/main.c but you get the idea.. although I'm increasingly unsure if this is the right approach.

It looks like a sane fix to me, as long as all of the different config combinations work properly.

thanks,

greg k-h

#20 lore raw
In reply to Daniel Drake
Quoted 7 lines · Daniel Drake
Quoted 6 lines · Daniel Drake
Quoted 5 lines · Alan Stern

Ah, okay. This is a known problem, a configuration-related issue. The PM core doesn't create a device's /sys/.../power subdirectory unless CONFIG_PM_SLEEP is set. In fact it should be created whenever CONFIG_PM is set.

Quoted 6 lines
Quoted 2 lines
Quoted 1 line

I haven't had time to write a patch for this yet. Do you want to try?

OK. Any comments on this before I sent it to the proper channels?

Missed a hunk where I remove the existing dpm_sysfs_add() call from drivers/base/power/main.c but you get the idea.. although I'm increasingly unsure if this is the right approach.

It looks like a sane fix to me, as long as all of the different config combinations work properly.

thanks,

greg k-h

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel