[lvc-project] [PATCH] comedi: drivers: do not detach device if driv->attach() fails

[Ian Abbott]

On 22/10/2025 11:45, Ian Abbott wrote:
> On 21/10/2025 14:16, Nikita Zhandarovich wrote:
>> Syzbot identified an issue [1] in comedi_device_attach() that occurs
>> when kernel attempts to detach a comedi device via
>> comedi_device_detach() even if a driver-specific attach() method
>> already failed. Attempts to follow through with detaching the
>> device and unregistering the driver trigger a warning.
>>
>> Fix this by rearranging cleanup calls so that comedi_device_detach()
>> runs only if the device in question has been successfully attached.
>>
>> Original idea for this patch belongs to Hillf Danton
>> <hdanton at sina.com>.
>>
>> [1] Syzbot crash:
>> Unexpected driver unregister!
>> WARNING: CPU: 0 PID: 5970 at drivers/base/driver.c:273 
>> driver_unregister drivers/base/driver.c:273 [inline]
>> WARNING: CPU: 0 PID: 5970 at drivers/base/driver.c:273 
>> driver_unregister+0x90/0xb0 drivers/base/driver.c:270
>> ...
>> Call Trace:
>>   <TASK>
>>   comedi_device_detach_locked+0x12f/0xa50 drivers/comedi/drivers.c:207
>>   comedi_device_detach+0x67/0xb0 drivers/comedi/drivers.c:215
>>   comedi_device_attach+0x43d/0x900 drivers/comedi/drivers.c:1011
>>   do_devconfig_ioctl+0x1b1/0x710 drivers/comedi/comedi_fops.c:872
>>   comedi_unlocked_ioctl+0x165d/0x2f00 drivers/comedi/comedi_fops.c:2178
>>   vfs_ioctl fs/ioctl.c:51 [inline]
>>   __do_sys_ioctl fs/ioctl.c:597 [inline]
>> ...
>>
>> Reported-by: syzbot+6616bba359cec7a1def1 at syzkaller.appspotmail.com
>> Closes: https://syzkaller.appspot.com/bug?extid=6616bba359cec7a1def1
>> Suggested-by: Hillf Danton <hdanton at sina.com>
>> Fixes: 74ece108f9e5 ("staging: comedi: move detach out of post-config")
>> Cc: stable at vger.kernel.org
>> Signed-off-by: Nikita Zhandarovich <n.zhandarovich at fintech.ru>
>> ---
>>   drivers/comedi/drivers.c | 9 ++++++---
>>   1 file changed, 6 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/comedi/drivers.c b/drivers/comedi/drivers.c
>> index c9ebaadc5e82..001083f96138 100644
>> --- a/drivers/comedi/drivers.c
>> +++ b/drivers/comedi/drivers.c
>> @@ -1005,10 +1005,13 @@ int comedi_device_attach(struct comedi_device 
>> *dev, struct comedi_devconfig *it)
>>       dev->board_name = dev->board_ptr ? *(const char **)dev->board_ptr
>>                        : dev->driver->driver_name;
>>       ret = driv->attach(dev, it);
>> -    if (ret >= 0)
>> +    if (ret >= 0) {
>>           ret = comedi_device_postconfig(dev);
>> -    if (ret < 0) {
>> -        comedi_device_detach(dev);
>> +        if (ret < 0) {
>> +            comedi_device_detach(dev);
>> +            module_put(driv->module);
>> +        }
>> +    } else {
>>           module_put(driv->module);
>>       }
>>       /* On success, the driver module count has been incremented. */
> 
> Unfortunately, the low-level drivers expect the `->detach()` handler to 
> be called to clean up even if the `->attach()` handler returns an error. 
>   So this won't work.
> 

The problem seems to be the "c6xdigio" driver 
("drivers/comedi/drivers/c6digio.c"). Its comedi `->attach()` handler 
`c6digio_attach()` can return an error before the call to 
`pnp_register_driver()`.  Also, it does not check the return value from 
`pnp_register_driver()`.  On error, comedi will call the `->detach()` 
handler `c6xdigio_detach()` which calls `pnp_unregister_driver()` 
unconditionally, leading to the warning reported by Syzbot.

-- 
-=( Ian Abbott <abbotti at mev.co.uk> || MEV Ltd. is a company  )=-
-=( registered in England & Wales.  Regd. number: 02862268.  )=-
-=( Regd. addr.: S11 & 12 Building 67, Europa Business Park, )=-
-=( Bird Hall Lane, STOCKPORT, SK3 0XA, UK. || www.mev.co.uk )=-