Re: usb/gadget: potential deadlock in gadgetfs_suspend
From: Alan Stern
Date: Mon Jun 12 2017 - 16:25:51 EST
On Mon, 12 Jun 2017, Andrey Konovalov wrote:
> Aha, got KASAN report:
>
> ==================================================================
> BUG: KASAN: use-after-free in __lock_acquire+0x3069/0x3690
> kernel/locking/lockdep.c:3246
> Read of size 8 at addr ffff88003a2bdaf8 by task kworker/3:1/903
>
> CPU: 3 PID: 903 Comm: kworker/3:1 Not tainted 4.12.0-rc4+ #35
> Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
> Workqueue: usb_hub_wq hub_event
> Call Trace:
> __dump_stack lib/dump_stack.c:16 [inline]
> dump_stack+0x292/0x395 lib/dump_stack.c:52
> print_address_description+0x78/0x280 mm/kasan/report.c:252
> kasan_report_error mm/kasan/report.c:351 [inline]
> kasan_report+0x230/0x340 mm/kasan/report.c:408
> __asan_report_load8_noabort+0x19/0x20 mm/kasan/report.c:429
> __lock_acquire+0x3069/0x3690 kernel/locking/lockdep.c:3246
> lock_acquire+0x22d/0x560 kernel/locking/lockdep.c:3855
> __raw_spin_lock include/linux/spinlock_api_smp.h:142 [inline]
> _raw_spin_lock+0x2f/0x40 kernel/locking/spinlock.c:151
> spin_lock include/linux/spinlock.h:299 [inline]
> gadgetfs_suspend+0x89/0x130 drivers/usb/gadget/legacy/inode.c:1682
> set_link_state+0x88e/0xae0 drivers/usb/gadget/udc/dummy_hcd.c:455
> dummy_hub_control+0xd7e/0x1fb0 drivers/usb/gadget/udc/dummy_hcd.c:2074
> rh_call_control drivers/usb/core/hcd.c:689 [inline]
> rh_urb_enqueue drivers/usb/core/hcd.c:846 [inline]
> usb_hcd_submit_urb+0x92f/0x20b0 drivers/usb/core/hcd.c:1650
> usb_submit_urb+0x8b2/0x12c0 drivers/usb/core/urb.c:542
> usb_start_wait_urb+0x148/0x5b0 drivers/usb/core/message.c:56
> usb_internal_control_msg drivers/usb/core/message.c:100 [inline]
> usb_control_msg+0x341/0x4d0 drivers/usb/core/message.c:151
> usb_clear_port_feature+0x74/0xa0 drivers/usb/core/hub.c:412
> hub_port_disable+0x123/0x510 drivers/usb/core/hub.c:4177
> hub_port_init+0x1ed/0x2940 drivers/usb/core/hub.c:4648
> hub_port_connect drivers/usb/core/hub.c:4826 [inline]
> hub_port_connect_change drivers/usb/core/hub.c:4999 [inline]
> port_event drivers/usb/core/hub.c:5105 [inline]
> hub_event+0x1ae1/0x3d40 drivers/usb/core/hub.c:5185
> process_one_work+0xc08/0x1bd0 kernel/workqueue.c:2097
> process_scheduled_works kernel/workqueue.c:2157 [inline]
> worker_thread+0xb2b/0x1860 kernel/workqueue.c:2233
> kthread+0x363/0x440 kernel/kthread.c:231
> ret_from_fork+0x2a/0x40 arch/x86/entry/entry_64.S:424
>
> Allocated by task 9958:
> save_stack_trace+0x1b/0x20 arch/x86/kernel/stacktrace.c:59
> save_stack+0x43/0xd0 mm/kasan/kasan.c:513
> set_track mm/kasan/kasan.c:525 [inline]
> kasan_kmalloc+0xad/0xe0 mm/kasan/kasan.c:617
> kmem_cache_alloc_trace+0x87/0x280 mm/slub.c:2745
> kmalloc include/linux/slab.h:492 [inline]
> kzalloc include/linux/slab.h:665 [inline]
> dev_new drivers/usb/gadget/legacy/inode.c:170 [inline]
> gadgetfs_fill_super+0x24f/0x540 drivers/usb/gadget/legacy/inode.c:1993
> mount_single+0xf6/0x160 fs/super.c:1192
> gadgetfs_mount+0x31/0x40 drivers/usb/gadget/legacy/inode.c:2019
> mount_fs+0x9c/0x2d0 fs/super.c:1223
> vfs_kern_mount.part.25+0xcb/0x490 fs/namespace.c:976
> vfs_kern_mount fs/namespace.c:2509 [inline]
> do_new_mount fs/namespace.c:2512 [inline]
> do_mount+0x41b/0x2d90 fs/namespace.c:2834
> SYSC_mount fs/namespace.c:3050 [inline]
> SyS_mount+0xb0/0x120 fs/namespace.c:3027
> entry_SYSCALL_64_fastpath+0x1f/0xbe
>
> Freed by task 9960:
> save_stack_trace+0x1b/0x20 arch/x86/kernel/stacktrace.c:59
> save_stack+0x43/0xd0 mm/kasan/kasan.c:513
> set_track mm/kasan/kasan.c:525 [inline]
> kasan_slab_free+0x72/0xc0 mm/kasan/kasan.c:590
> slab_free_hook mm/slub.c:1357 [inline]
> slab_free_freelist_hook mm/slub.c:1379 [inline]
> slab_free mm/slub.c:2961 [inline]
> kfree+0xed/0x2b0 mm/slub.c:3882
> put_dev+0x124/0x160 drivers/usb/gadget/legacy/inode.c:163
> gadgetfs_kill_sb+0x33/0x60 drivers/usb/gadget/legacy/inode.c:2027
> deactivate_locked_super+0x8d/0xd0 fs/super.c:309
> deactivate_super+0x21e/0x310 fs/super.c:340
> cleanup_mnt+0xb7/0x150 fs/namespace.c:1112
> __cleanup_mnt+0x1b/0x20 fs/namespace.c:1119
> task_work_run+0x1a0/0x280 kernel/task_work.c:116
> exit_task_work include/linux/task_work.h:21 [inline]
> do_exit+0x18a8/0x2820 kernel/exit.c:878
> do_group_exit+0x14e/0x420 kernel/exit.c:982
> get_signal+0x784/0x1780 kernel/signal.c:2318
> do_signal+0xd7/0x2130 arch/x86/kernel/signal.c:808
> exit_to_usermode_loop+0x1ac/0x240 arch/x86/entry/common.c:157
> prepare_exit_to_usermode arch/x86/entry/common.c:194 [inline]
> syscall_return_slowpath+0x3ba/0x410 arch/x86/entry/common.c:263
> entry_SYSCALL_64_fastpath+0xbc/0xbe
>
> The buggy address belongs to the object at ffff88003a2bdae0
> which belongs to the cache kmalloc-1024 of size 1024
> The buggy address is located 24 bytes inside of
> 1024-byte region [ffff88003a2bdae0, ffff88003a2bdee0)
> The buggy address belongs to the page:
> page:ffffea0000e8ae00 count:1 mapcount:0 mapping: (null)
> index:0x0 compound_mapcount: 0
> flags: 0x100000000008100(slab|head)
> raw: 0100000000008100 0000000000000000 0000000000000000 0000000100170017
> raw: ffffea0000ed3020 ffffea0000f5f820 ffff88003e80efc0 0000000000000000
> page dumped because: kasan: bad access detected
>
> Memory state around the buggy address:
> ffff88003a2bd980: fb fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
> ffff88003a2bda00: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
> >ffff88003a2bda80: fc fc fc fc fc fc fc fc fc fc fc fc fb fb fb fb
> ^
> ffff88003a2bdb00: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
> ffff88003a2bdb80: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
> ==================================================================
As you surmised, this was caused by a race. The race was between
dummy_udc_stop() and set_link_state(), both in dummy_hcd.c. A symptom
of this race is that the first routine clears dum->driver while the
second dereferences it, and neither access is protected by a lock.
It turns out this problem affects at least one other UDC driver
(net2280.c). Below is a patch that adds the missing lock (and removes
some unneeded unlocks) from these drivers. It turns out that the
disconnect, reset, suspend, and resume callbacks should all be invoked
while the UDC's lock is held, because they can race with gadget driver
unbinding.
Adding locked regions can lead to the possibility of deadlock or
lockdep violations. Please let me know if the patch below raises any
problems.
Alan Stern
Index: usb-4.x/drivers/usb/gadget/legacy/inode.c
===================================================================
--- usb-4.x.orig/drivers/usb/gadget/legacy/inode.c
+++ usb-4.x/drivers/usb/gadget/legacy/inode.c
@@ -1679,9 +1679,10 @@ static void
gadgetfs_suspend (struct usb_gadget *gadget)
{
struct dev_data *dev = get_gadget_data (gadget);
+ unsigned long flags;
INFO (dev, "suspended from state %d\n", dev->state);
- spin_lock (&dev->lock);
+ spin_lock_irqsave(&dev->lock, flags);
switch (dev->state) {
case STATE_DEV_SETUP: // VERY odd... host died??
case STATE_DEV_CONNECTED:
@@ -1692,7 +1693,7 @@ gadgetfs_suspend (struct usb_gadget *gad
default:
break;
}
- spin_unlock (&dev->lock);
+ spin_unlock_irqrestore(&dev->lock, flags);
}
static struct usb_gadget_driver gadgetfs_driver = {
Index: usb-4.x/drivers/usb/gadget/udc/dummy_hcd.c
===================================================================
--- usb-4.x.orig/drivers/usb/gadget/udc/dummy_hcd.c
+++ usb-4.x/drivers/usb/gadget/udc/dummy_hcd.c
@@ -442,23 +442,16 @@ static void set_link_state(struct dummy_
/* Report reset and disconnect events to the driver */
if (dum->driver && (disconnect || reset)) {
stop_activity(dum);
- spin_unlock(&dum->lock);
if (reset)
usb_gadget_udc_reset(&dum->gadget, dum->driver);
else
dum->driver->disconnect(&dum->gadget);
- spin_lock(&dum->lock);
}
} else if (dum_hcd->active != dum_hcd->old_active) {
- if (dum_hcd->old_active && dum->driver->suspend) {
- spin_unlock(&dum->lock);
+ if (dum_hcd->old_active && dum->driver->suspend)
dum->driver->suspend(&dum->gadget);
- spin_lock(&dum->lock);
- } else if (!dum_hcd->old_active && dum->driver->resume) {
- spin_unlock(&dum->lock);
+ else if (!dum_hcd->old_active && dum->driver->resume)
dum->driver->resume(&dum->gadget);
- spin_lock(&dum->lock);
- }
}
dum_hcd->old_status = dum_hcd->port_status;
@@ -983,7 +976,9 @@ static int dummy_udc_stop(struct usb_gad
struct dummy_hcd *dum_hcd = gadget_to_dummy_hcd(g);
struct dummy *dum = dum_hcd->dum;
+ spin_lock_irq(&dum->lock);
dum->driver = NULL;
+ spin_unlock_irq(&dum->lock);
return 0;
}
Index: usb-4.x/drivers/usb/gadget/udc/net2280.c
===================================================================
--- usb-4.x.orig/drivers/usb/gadget/udc/net2280.c
+++ usb-4.x/drivers/usb/gadget/udc/net2280.c
@@ -2470,11 +2470,8 @@ static void stop_activity(struct net2280
nuke(&dev->ep[i]);
/* report disconnect; the driver is already quiesced */
- if (driver) {
- spin_unlock(&dev->lock);
+ if (driver)
driver->disconnect(&dev->gadget);
- spin_lock(&dev->lock);
- }
usb_reinit(dev);
}
@@ -3348,8 +3345,6 @@ next_endpoints:
BIT(PCI_RETRY_ABORT_INTERRUPT))
static void handle_stat1_irqs(struct net2280 *dev, u32 stat)
-__releases(dev->lock)
-__acquires(dev->lock)
{
struct net2280_ep *ep;
u32 tmp, num, mask, scratch;
@@ -3390,14 +3385,12 @@ __acquires(dev->lock)
if (disconnect || reset) {
stop_activity(dev, dev->driver);
ep0_start(dev);
- spin_unlock(&dev->lock);
if (reset)
usb_gadget_udc_reset
(&dev->gadget, dev->driver);
else
(dev->driver->disconnect)
(&dev->gadget);
- spin_lock(&dev->lock);
return;
}
}