Skip to contentSkip to conversation

hid: fix bug destroying hidraw device files after parent

PATCH Applied Fernando Luis Vázquez Cao 5 participants 8 messages Feb 26, 2014 – Mar 18, 2014 1 message not in the archive HID

lore
Share & tools
lore
#1 lore raw
In reply to Fernando Luis Vázquez Cao

I forgot to mention that if the fix below is acceptable it should be queued for 3.13-stable (older kernels are not affected).

Quoted 59 lines · Fernando Luis Vázquez Cao

I noticed that after hot unplugging a Logitech unifying receiver (drivers/hid/hid-logitech-dj.c) the kernel would occasionally spew a stack trace similar to this:

usb 1-1.1.2: USB disconnect, device number 7

WARNING: CPU: 0 PID: 2865 at fs/sysfs/group.c:216 device_del+0x40/0x1b0()
sysfs group ffffffff8187fa20 not found for kobject 'hidraw0'
[...]
CPU: 0 PID: 2865 Comm: upowerd Tainted: G        W 3.14.0-rc4 #7
Hardware name: LENOVO 7783PN4/        , BIOS 9HKT43AUS 07/11/2011
  0000000000000009 ffffffff814cd684 ffff880427ccfdf8 ffffffff810616e7
  ffff88041ec61800 ffff880427ccfe48 ffff88041e444d80 ffff880426fab8e8
  ffff880429359960 ffffffff8106174c ffffffff81714b98 0000000000000028
Call Trace:
  [<ffffffff814cd684>] ? dump_stack+0x41/0x51
  [<ffffffff810616e7>] ? warn_slowpath_common+0x77/0x90
  [<ffffffff8106174c>] ? warn_slowpath_fmt+0x4c/0x50
  [<ffffffff81374fd0>] ? device_del+0x40/0x1b0
  [<ffffffff8137516f>] ? device_unregister+0x2f/0x50
  [<ffffffff813751fa>] ? device_destroy+0x3a/0x40
  [<ffffffffa03ca245>] ? drop_ref+0x55/0x120 [hid]
  [<ffffffffa03ca3e6>] ? hidraw_release+0x96/0xb0 [hid]
  [<ffffffff811929da>] ? __fput+0xca/0x210
  [<ffffffff8107fe17>] ? task_work_run+0x97/0xd0
  [<ffffffff810139a9>] ? do_notify_resume+0x69/0xa0
  [<ffffffff814dbd22>] ? int_signal+0x12/0x17
---[ end trace 63f4a46f6566d737 ]---

During device removal hid_disconnect() is called via hid_hw_stop() to stop the device and free all its resources, including the sysfs files. The problem is that if a user space process, such as upowerd, holds a reference to a hidraw file the corresponding sysfs files will be kept around (drop_ref() does not call device_destroy() if the open counter is not 0) and it will be usb_disconnect() who, by calling device_del() for the USB device, will indirectly remove the sysfs files of the hidraw device (sysfs_remove_dir() is recursive these days). Because of this, by the time user space releases the last reference to the hidraw file and drop_ref() tries to destroy the device the sysfs files are already gone and the kernel will print the warning above.

Fix this by calling device_destroy() at USB disconnect time.

---

diff -urNp linux-3.14-rc4-orig/drivers/hid/hidraw.c linux-3.14-rc4/drivers/hid/hidraw.c

drivers/hid/hidraw.c +2 -0
@@ -320,13 +320,13 @@ static void drop_ref(struct hidraw *hidr
  			hid_hw_close(hidraw->hid);
  			wake_up_interruptible(&hidraw->wait);
  		}
+		device_destroy(hidraw_class,
+			       MKDEV(hidraw_major, hidraw->minor));

drivers/hid/hidraw.c static void drop_ref(struct hidraw *hidr · line 332

Quoted 10 lines
drivers/hid/hidraw.c +0 -2
  	} else {
  		--hidraw->open;
  	}
  	if (!hidraw->open) {
  		if (!hidraw->exist) {
-			device_destroy(hidraw_class,
-					MKDEV(hidraw_major, hidraw->minor));
  			hidraw_table[hidraw->minor] = NULL;
  			kfree(hidraw);
  		} else {

--
To unsubscribe from this list: send the line "unsubscribe linux-input" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#0 lore raw

PATCH 1/1 hid: fix bug destroying hidraw device files after parent

I noticed that after hot unplugging a Logitech unifying receiver (drivers/hid/hid-logitech-dj.c) the kernel would occasionally spew a stack trace similar to this:

usb 1-1.1.2: USB disconnect, device number 7

WARNING: CPU: 0 PID: 2865 at fs/sysfs/group.c:216 device_del+0x40/0x1b0()
sysfs group ffffffff8187fa20 not found for kobject 'hidraw0'
[...]
CPU: 0 PID: 2865 Comm: upowerd Tainted: G        W 3.14.0-rc4 #7
Hardware name: LENOVO 7783PN4/        , BIOS 9HKT43AUS 07/11/2011
 0000000000000009 ffffffff814cd684 ffff880427ccfdf8 ffffffff810616e7
 ffff88041ec61800 ffff880427ccfe48 ffff88041e444d80 ffff880426fab8e8
 ffff880429359960 ffffffff8106174c ffffffff81714b98 0000000000000028
Call Trace:
 [<ffffffff814cd684>] ? dump_stack+0x41/0x51
 [<ffffffff810616e7>] ? warn_slowpath_common+0x77/0x90
 [<ffffffff8106174c>] ? warn_slowpath_fmt+0x4c/0x50
 [<ffffffff81374fd0>] ? device_del+0x40/0x1b0
 [<ffffffff8137516f>] ? device_unregister+0x2f/0x50
 [<ffffffff813751fa>] ? device_destroy+0x3a/0x40
 [<ffffffffa03ca245>] ? drop_ref+0x55/0x120 [hid]
 [<ffffffffa03ca3e6>] ? hidraw_release+0x96/0xb0 [hid]
 [<ffffffff811929da>] ? __fput+0xca/0x210
 [<ffffffff8107fe17>] ? task_work_run+0x97/0xd0
 [<ffffffff810139a9>] ? do_notify_resume+0x69/0xa0
 [<ffffffff814dbd22>] ? int_signal+0x12/0x17
---[ end trace 63f4a46f6566d737 ]---

During device removal hid_disconnect() is called via hid_hw_stop() to stop the device and free all its resources, including the sysfs files. The problem is that if a user space process, such as upowerd, holds a reference to a hidraw file the corresponding sysfs files will be kept around (drop_ref() does not call device_destroy() if the open counter is not 0) and it will be usb_disconnect() who, by calling device_del() for the USB device, will indirectly remove the sysfs files of the hidraw device (sysfs_remove_dir() is recursive these days). Because of this, by the time user space releases the last reference to the hidraw file and drop_ref() tries to destroy the device the sysfs files are already gone and the kernel will print the warning above.

Fix this by calling device_destroy() at USB disconnect time.

---

drivers/hid/hidraw.c +2 -2
diff -urNp linux-3.14-rc4-orig/drivers/hid/hidraw.c linux-3.14-rc4/drivers/hid/hidraw.c
@@ -320,13 +320,13 @@ static void drop_ref(struct hidraw *hidr
 			hid_hw_close(hidraw->hid);
 			wake_up_interruptible(&hidraw->wait);
 		}
+		device_destroy(hidraw_class,
+			       MKDEV(hidraw_major, hidraw->minor));
 	} else {
 		--hidraw->open;
 	}
 	if (!hidraw->open) {
 		if (!hidraw->exist) {
-			device_destroy(hidraw_class,
-					MKDEV(hidraw_major, hidraw->minor));
 			hidraw_table[hidraw->minor] = NULL;
 			kfree(hidraw);
 		} else {
#2 lore raw
In reply to Fernando Luis Vázquez Cao

Hi Fernando,

Why are older kernels not affected ?

-nestor

Quoted 72 lines · Fernando Luis Vázquez Cao

I forgot to mention that if the fix below is acceptable it should be queued for 3.13-stable (older kernels are not affected).

Quoted 66 lines · Fernando Luis Vázquez Cao

I noticed that after hot unplugging a Logitech unifying receiver (drivers/hid/hid-logitech-dj.c) the kernel would occasionally spew a stack trace similar to this:

usb 1-1.1.2: USB disconnect, device number 7

WARNING: CPU: 0 PID: 2865 at fs/sysfs/group.c:216 device_del+0x40/0x1b0()
sysfs group ffffffff8187fa20 not found for kobject 'hidraw0'
[...]
CPU: 0 PID: 2865 Comm: upowerd Tainted: G        W 3.14.0-rc4 #7
Hardware name: LENOVO 7783PN4/        , BIOS 9HKT43AUS 07/11/2011
  0000000000000009 ffffffff814cd684 ffff880427ccfdf8 ffffffff810616e7
  ffff88041ec61800 ffff880427ccfe48 ffff88041e444d80 ffff880426fab8e8
  ffff880429359960 ffffffff8106174c ffffffff81714b98 0000000000000028
Call Trace:
  [<ffffffff814cd684>] ? dump_stack+0x41/0x51
  [<ffffffff810616e7>] ? warn_slowpath_common+0x77/0x90
  [<ffffffff8106174c>] ? warn_slowpath_fmt+0x4c/0x50
  [<ffffffff81374fd0>] ? device_del+0x40/0x1b0
  [<ffffffff8137516f>] ? device_unregister+0x2f/0x50
  [<ffffffff813751fa>] ? device_destroy+0x3a/0x40
  [<ffffffffa03ca245>] ? drop_ref+0x55/0x120 [hid]
  [<ffffffffa03ca3e6>] ? hidraw_release+0x96/0xb0 [hid]
  [<ffffffff811929da>] ? __fput+0xca/0x210
  [<ffffffff8107fe17>] ? task_work_run+0x97/0xd0
  [<ffffffff810139a9>] ? do_notify_resume+0x69/0xa0
  [<ffffffff814dbd22>] ? int_signal+0x12/0x17
---[ end trace 63f4a46f6566d737 ]---

During device removal hid_disconnect() is called via hid_hw_stop() to stop the device and free all its resources, including the sysfs files. The problem is that if a user space process, such as upowerd, holds a reference to a hidraw file the corresponding sysfs files will be kept around (drop_ref() does not call device_destroy() if the open counter is not 0) and it will be usb_disconnect() who, by calling device_del() for the USB device, will indirectly remove the sysfs files of the hidraw device (sysfs_remove_dir() is recursive these days). Because of this, by the time user space releases the last reference to the hidraw file and drop_ref() tries to destroy the device the sysfs files are already gone and the kernel will print the warning above.

Fix this by calling device_destroy() at USB disconnect time.

---

diff -urNp linux-3.14-rc4-orig/drivers/hid/hidraw.c linux-3.14-rc4/drivers/hid/hidraw.c
--- linux-3.14-rc4-orig/drivers/hid/hidraw.c 2014-02-26
14:21:48.622980475 +0900
+++ linux-3.14-rc4/drivers/hid/hidraw.c 2014-02-26 14:22:17.990979556 +0900

drivers/hid/hidraw.c +2 -1
@@ -320,13 +320,13 @@ static void drop_ref(struct hidraw *hidr
                        hid_hw_close(hidraw->hid);
                        wake_up_interruptible(&hidraw->wait);
                }
+               device_destroy(hidraw_class,
+                              MKDEV(hidraw_major, hidraw->minor));
        } else {
                --hidraw->open;
        }
        if (!hidraw->open) {
                if (!hidraw->exist) {
-                       device_destroy(hidraw_class,
Quoted 9 lines
Quoted 8 lines
drivers/hid/hidraw.c +0 -1
-                                       MKDEV(hidraw_major,
 hidraw->minor));
                        hidraw_table[hidraw->minor] = NULL;
                        kfree(hidraw);
                } else {

--
To unsubscribe from this list: send the line "unsubscribe linux-input" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

Reviewed-by on v1 1/1
#3 lore raw
In reply to Fernando Luis Vázquez Cao

Hi

Quoted 41 lines · Fernando Luis Vázquez Cao

I noticed that after hot unplugging a Logitech unifying receiver (drivers/hid/hid-logitech-dj.c) the kernel would occasionally spew a stack trace similar to this:

usb 1-1.1.2: USB disconnect, device number 7

WARNING: CPU: 0 PID: 2865 at fs/sysfs/group.c:216 device_del+0x40/0x1b0()
sysfs group ffffffff8187fa20 not found for kobject 'hidraw0'
[...]
CPU: 0 PID: 2865 Comm: upowerd Tainted: G        W 3.14.0-rc4 #7
Hardware name: LENOVO 7783PN4/        , BIOS 9HKT43AUS 07/11/2011
 0000000000000009 ffffffff814cd684 ffff880427ccfdf8 ffffffff810616e7
 ffff88041ec61800 ffff880427ccfe48 ffff88041e444d80 ffff880426fab8e8
 ffff880429359960 ffffffff8106174c ffffffff81714b98 0000000000000028
Call Trace:
 [<ffffffff814cd684>] ? dump_stack+0x41/0x51
 [<ffffffff810616e7>] ? warn_slowpath_common+0x77/0x90
 [<ffffffff8106174c>] ? warn_slowpath_fmt+0x4c/0x50
 [<ffffffff81374fd0>] ? device_del+0x40/0x1b0
 [<ffffffff8137516f>] ? device_unregister+0x2f/0x50
 [<ffffffff813751fa>] ? device_destroy+0x3a/0x40
 [<ffffffffa03ca245>] ? drop_ref+0x55/0x120 [hid]
 [<ffffffffa03ca3e6>] ? hidraw_release+0x96/0xb0 [hid]
 [<ffffffff811929da>] ? __fput+0xca/0x210
 [<ffffffff8107fe17>] ? task_work_run+0x97/0xd0
 [<ffffffff810139a9>] ? do_notify_resume+0x69/0xa0
 [<ffffffff814dbd22>] ? int_signal+0x12/0x17
---[ end trace 63f4a46f6566d737 ]---

During device removal hid_disconnect() is called via hid_hw_stop() to stop the device and free all its resources, including the sysfs files. The problem is that if a user space process, such as upowerd, holds a reference to a hidraw file the corresponding sysfs files will be kept around (drop_ref() does not call device_destroy() if the open counter is not 0) and it will be usb_disconnect() who, by calling device_del() for the USB device, will indirectly remove the sysfs files of the hidraw device (sysfs_remove_dir() is recursive these days). Because of this, by the time user space releases the last reference to the hidraw file and drop_ref() tries to destroy the device the sysfs files are already gone and the kernel will print the warning above.

Quoted 1 line

Fix this by calling device_destroy() at USB disconnect time.

This actually also fixes the issue that the device-node is kept even though the device is dead. By destroying the "struct device", we make sure to send out a remove-uevent so udev can drop the node. We correctly check ->exists during open(), but it is still nice to see the node go away once that bit is set.

Thanks
David

Quoted 26 lines

---

diff -urNp linux-3.14-rc4-orig/drivers/hid/hidraw.c linux-3.14-rc4/drivers/hid/hidraw.c

drivers/hid/hidraw.c 2014-02-26 14:22:17.990979556 +0900 renamed from drivers/hid/hidraw.c 2014-02-26 14:21:48.622980475 +0900 +2 -3
@@ -320,13 +320,13 @@ static void drop_ref(struct hidraw *hidr
                        hid_hw_close(hidraw->hid);
                        wake_up_interruptible(&hidraw->wait);
                }
+               device_destroy(hidraw_class,
+                              MKDEV(hidraw_major, hidraw->minor));
        } else {
                --hidraw->open;
        }
        if (!hidraw->open) {
                if (!hidraw->exist) {
-                       device_destroy(hidraw_class,
-                                       MKDEV(hidraw_major, hidraw->minor));
                        hidraw_table[hidraw->minor] = NULL;
                        kfree(hidraw);
                } else {
 
 
--
Quoted 3 lines

To unsubscribe from this list: send the line "unsubscribe linux-input" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

--
To unsubscribe from this list: send the line "unsubscribe linux-input" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

Applied
#4 lore raw
In reply to Fernando Luis Vázquez Cao
Quoted 18 lines · Fernando Luis Vázquez Cao

I noticed that after hot unplugging a Logitech unifying receiver (drivers/hid/hid-logitech-dj.c) the kernel would occasionally spew a stack trace similar to this:

usb 1-1.1.2: USB disconnect, device number 7

WARNING: CPU: 0 PID: 2865 at fs/sysfs/group.c:216 device_del+0x40/0x1b0()
sysfs group ffffffff8187fa20 not found for kobject 'hidraw0'
[...]
CPU: 0 PID: 2865 Comm: upowerd Tainted: G        W 3.14.0-rc4 #7
Hardware name: LENOVO 7783PN4/        , BIOS 9HKT43AUS 07/11/2011
 0000000000000009 ffffffff814cd684 ffff880427ccfdf8 ffffffff810616e7
 ffff88041ec61800 ffff880427ccfe48 ffff88041e444d80 ffff880426fab8e8
 ffff880429359960 ffffffff8106174c ffffffff81714b98 0000000000000028
Call Trace:
 [<ffffffff814cd684>] ? dump_stack+0x41/0x51
 [<ffffffff810616e7>] ? warn_slowpath_common+0x77/0x90
 [<ffffffff8106174c>] ? warn_slowpath_fmt+0x4c/0x50
 [<ffffffff81374fd0>] ? device_del+0x40/0x1b0
Quoted 8 lines
 [<ffffffff8137516f>] ? device_unregister+0x2f/0x50
 [<ffffffff813751fa>] ? device_destroy+0x3a/0x40
 [<ffffffffa03ca245>] ? drop_ref+0x55/0x120 [hid]
 [<ffffffffa03ca3e6>] ? hidraw_release+0x96/0xb0 [hid]
 [<ffffffff811929da>] ? __fput+0xca/0x210
 [<ffffffff8107fe17>] ? task_work_run+0x97/0xd0
 [<ffffffff810139a9>] ? do_notify_resume+0x69/0xa0
 [<ffffffff814dbd22>] ? int_signal+0x12/0x17

Applied, thanks.

I have slightly modified the patch title to make sure that it's obvious that what it fixes is actually a WARN_ON() splat.

Signature · 6 lines
Jiri Kosina
SUSE Labs
--
To unsubscribe from this list: send the line "unsubscribe linux-input" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
#6 lore raw
In reply to a message not in the archive

Hi Jiri,

On 02/26/2014 07:02 PM, Jiri Kosina wrote:
[...]

Quoted 4 lines

Applied, thanks.

I have slightly modified the patch title to make sure that it's obvious that what it fixes is actually a WARN_ON() splat.

Thank you.

Any chance we can get this to Linus before 3.14 comes out? According to Linus, rc7 may be the last rc.

- Fernando
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html

#7 lore raw
Quoted 7 lines · Fernando Luis Vázquez Cao
Quoted 2 lines

I have slightly modified the patch title to make sure that it's obvious that what it fixes is actually a WARN_ON() splat.

Thank you.

Any chance we can get this to Linus before 3.14 comes out? According to Linus, rc7 may be the last rc.

It will be in Linus' 3.14 final -- it went in post-rc7.

Thanks for taking care,

Signature · 6 lines
Jiri Kosina
SUSE Labs
--
To unsubscribe from this list: send the line "unsubscribe linux-input" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
#8 lore raw
Quoted 11 lines · Jiri Kosina
Quoted 6 lines · Fernando Luis Vázquez Cao
Quoted 2 lines

I have slightly modified the patch title to make sure that it's obvious that what it fixes is actually a WARN_ON() splat.

Thank you.

Any chance we can get this to Linus before 3.14 comes out? According to Linus, rc7 may be the last rc.

It will be in Linus' 3.14 final -- it went in post-rc7.

Thanks for taking care,

Thanks!

- Fernando
--
To unsubscribe from this list: send the line "unsubscribe linux-input" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html