Re: [PATCH] drivers/base: Fix memory leak in error paths

[jouni.hogander@unikie.com (Jouni Högander)]

Greg Kroah-Hartman <gregkh@linuxfoundation.org> writes:

>> >> This memory leak was reported by Syzkaller:
>> >> 
>> >> BUG: memory leak unreferenced object 0xffff8880675ca008 (size 256):
>> >>   comm "netdev_register", pid 281, jiffies 4294696663 (age 6.808s)
>> >>   hex dump (first 32 bytes):
>> >>     00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
>> >>     00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
>> >>   backtrace:
>> >>     [<0000000058ca4711>] kmem_cache_alloc_trace+0x167/0x280
>> >>     [<000000002340019b>] device_add+0x882/0x1750
>> >>     [<000000001d588c3a>] netdev_register_kobject+0x128/0x380
>> >>     [<0000000011ef5535>] register_netdevice+0xa1b/0xf00
>> >>     [<000000007fcf1c99>] __tun_chr_ioctl+0x20d5/0x3dd0
>> >>     [<000000006a5b7b2b>] tun_chr_ioctl+0x2f/0x40
>> >>     [<00000000f30f834a>] do_vfs_ioctl+0x1c7/0x1510
>> >>     [<00000000fba062ea>] ksys_ioctl+0x99/0xb0
>> >>     [<00000000b1c1b8d2>] __x64_sys_ioctl+0x78/0xb0
>> >>     [<00000000984cabb9>] do_syscall_64+0x16f/0x580
>> >>     [<000000000bde033d>] entry_SYSCALL_64_after_hwframe+0x44/0xa9
>> >>     [<00000000e6ca2d9f>] 0xffffffffffffffff
>> >
>> > How is this a leak?  This is in device_add(), not removing the device.
>> > When the structure really is freed then it can be removed.
>> 
>> In net/core/net-sysfs.c:netdev_register_kobject device_add allocates
>> dev->p. Now if register_queue_kobjects fails the error path is calling
>> device_del and dev->p is never freed. Proper fix here could be to call
>> put_device after device_del?
>
> Hm, this sounds like you have a reference count leak here, as
> put_device() should be properly called already in this case.  You might
> want to look further to see where exactly the register_queue_kobjects()
> call fails in order to see if we grabbed a reference we forgot to put
> back on an error path.

Ok, did some more debugging on
this. net/core/net-sysfs.c:netdev_register_kobject is doing
device_initialize(dev). This is in
drivers/base/core.c:device_initialize:

 * NOTE: Use put_device() to give up your reference instead of freeing
 * @dev directly once you have called this function.

My understanding is that remaining reference on error path is taken by
device_initialize and as instructed in the note above it should be given
up using put_device? Tested this and it's fixing the memory leak I found
in my Syzkaller exercise. Addition to that it seems to be fixing also
this one:

https://syzkaller.appspot.com/bug?id=f5f4af9fb9ffb3112ad6e30f717f769decdccdfc

BR,

Jouni Högander