Skip to content
  • Mauro Carvalho Chehab's avatar
    [media] au8522: Avoid memory leak for device config data · aa37763f
    Mauro Carvalho Chehab authored
    
    
    As reported by kmemleak:
    
    	unreferenced object 0xffff880321e1da40 (size 32):
    	  comm "modprobe", pid 3309, jiffies 4295019569 (age 2359.636s)
    	  hex dump (first 32 bytes):
    	    47 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  G...............
    	    00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
    	  backtrace:
    	    [<ffffffff82278c8e>] kmemleak_alloc+0x4e/0xb0
    	    [<ffffffff8153c08c>] kmem_cache_alloc_trace+0x1ec/0x280
    	    [<ffffffffa13a896a>] au8522_probe+0x19a/0xa30 [au8522_decoder]
    	    [<ffffffff81de0032>] i2c_device_probe+0x2b2/0x490
    	    [<ffffffff81ca7004>] driver_probe_device+0x454/0xd90
    	    [<ffffffff81ca7c1b>] __device_attach_driver+0x17b/0x230
    	    [<ffffffff81ca15da>] bus_for_each_drv+0x11a/0x1b0
    	    [<ffffffff81ca6a4d>] __device_attach+0x1cd/0x2c0
    	    [<ffffffff81ca7d43>] device_initial_probe+0x13/0x20
    	    [<ffffffff81ca451f>] bus_probe_device+0x1af/0x250
    	    [<ffffffff81c9e0f3>] device_add+0x943/0x13b0
    	    [<ffffffff81c9eb7a>] device_register+0x1a/0x20
    	    [<ffffffff81de8626>] i2c_new_device+0x5d6/0x8f0
    	    [<ffffffffa0d88ea4>] v4l2_i2c_new_subdev_board+0x1e4/0x250 [v4l2_common]
    	    [<ffffffffa0d88fe7>] v4l2_i2c_new_subdev+0xd7/0x110 [v4l2_common]
    	    [<ffffffffa13b2f76>] au0828_card_analog_fe_setup+0x2e6/0x3f0 [au0828]
    
    Checking where the error happens:
    	(gdb) list *au8522_probe+0x19a
    	0x99a is in au8522_probe (drivers/media/dvb-frontends/au8522_decoder.c:761).
    	756			printk(KERN_INFO "au8522_decoder attach existing instance.\n");
    	757			break;
    	758		}
    	759
    	760		demod_config = kzalloc(sizeof(struct au8522_config), GFP_KERNEL);
    	761		if (demod_config == NULL) {
    	762			if (instance == 1)
    	763				kfree(state);
    	764			return -ENOMEM;
    	765		}
    
    Shows that the error path is not being handled properly.
    
    The are actually several issues here:
    
    1) config free should have been calling hybrid_tuner_release_state()
    function, by calling au8522_release_state();
    
    2) config is only allocated at the digital part. On the analog one,
    it is received from the caller.
    
    A complex logic could be added to address it, however, it is simpler
    to just embeed config inside the state.
    
    Signed-off-by: default avatarMauro Carvalho Chehab <mchehab@osg.samsung.com>
    aa37763f