Re: [PATCH net] infiniband: fix use-after-free when rename device name

[Leon Romanovsky <leon@kernel.org>]

On Tue, Mar 11, 2025 at 10:55:42AM +0800, Wang Liang wrote:
> 
> 在 2025/3/10 18:14, Leon Romanovsky 写道:
> > On Mon, Mar 10, 2025 at 02:45:16PM +0800, Wang Liang wrote:
> > > Syzbot reported a slab-use-after-free with the following call trace:
> > > 
> > > ==================================================================
> > > BUG: KASAN: slab-use-after-free in nla_put+0xd3/0x150 lib/nlattr.c:1099
> > > Read of size 5 at addr ffff888140ea1c60 by task syz.0.988/10025
> > > 
> > > CPU: 0 UID: 0 PID: 10025 Comm: syz.0.988 Not tainted 6.14.0-rc4-syzkaller-00859-gf77f12010f67 #0
> > > Hardware name: Google Compute Engine, BIOS Google 02/12/2025
> > > Call Trace:
> > >   <TASK>
> > >   __dump_stack lib/dump_stack.c:94 [inline]
> > >   dump_stack_lvl+0x241/0x360 lib/dump_stack.c:120
> > >   print_address_description mm/kasan/report.c:408 [inline]
> > >   print_report+0x16e/0x5b0 mm/kasan/report.c:521
> > >   kasan_report+0x143/0x180 mm/kasan/report.c:634
> > >   kasan_check_range+0x282/0x290 mm/kasan/generic.c:189
> > >   __asan_memcpy+0x29/0x70 mm/kasan/shadow.c:105
> > >   nla_put+0xd3/0x150 lib/nlattr.c:1099
> > >   nla_put_string include/net/netlink.h:1621 [inline]
> > >   fill_nldev_handle+0x16e/0x200 drivers/infiniband/core/nldev.c:265
> > >   rdma_nl_notify_event+0x561/0xef0 drivers/infiniband/core/nldev.c:2857
> > >   ib_device_notify_register+0x22/0x230 drivers/infiniband/core/device.c:1344
> > >   ib_register_device+0x1292/0x1460 drivers/infiniband/core/device.c:1460
> > >   rxe_register_device+0x233/0x350 drivers/infiniband/sw/rxe/rxe_verbs.c:1540
> > >   rxe_net_add+0x74/0xf0 drivers/infiniband/sw/rxe/rxe_net.c:550
> > >   rxe_newlink+0xde/0x1a0 drivers/infiniband/sw/rxe/rxe.c:212
> > >   nldev_newlink+0x5ea/0x680 drivers/infiniband/core/nldev.c:1795
> > >   rdma_nl_rcv_skb drivers/infiniband/core/netlink.c:239 [inline]
> > >   rdma_nl_rcv+0x6dd/0x9e0 drivers/infiniband/core/netlink.c:259
> > >   netlink_unicast_kernel net/netlink/af_netlink.c:1313 [inline]
> > >   netlink_unicast+0x7f6/0x990 net/netlink/af_netlink.c:1339
> > >   netlink_sendmsg+0x8de/0xcb0 net/netlink/af_netlink.c:1883
> > >   sock_sendmsg_nosec net/socket.c:709 [inline]
> > >   __sock_sendmsg+0x221/0x270 net/socket.c:724
> > >   ____sys_sendmsg+0x53a/0x860 net/socket.c:2564
> > >   ___sys_sendmsg net/socket.c:2618 [inline]
> > >   __sys_sendmsg+0x269/0x350 net/socket.c:2650
> > >   do_syscall_x64 arch/x86/entry/common.c:52 [inline]
> > >   do_syscall_64+0xf3/0x230 arch/x86/entry/common.c:83
> > >   entry_SYSCALL_64_after_hwframe+0x77/0x7f
> > > RIP: 0033:0x7f42d1b8d169
> > > Code: ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 48 89 f8 48 ...
> > > RSP: 002b:00007f42d2960038 EFLAGS: 00000246 ORIG_RAX: 000000000000002e
> > > RAX: ffffffffffffffda RBX: 00007f42d1da6320 RCX: 00007f42d1b8d169
> > > RDX: 0000000000000000 RSI: 00004000000002c0 RDI: 000000000000000c
> > > RBP: 00007f42d1c0e2a0 R08: 0000000000000000 R09: 0000000000000000
> > > R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
> > > R13: 0000000000000000 R14: 00007f42d1da6320 R15: 00007ffe399344a8
> > >   </TASK>
> > > 
> > > Allocated by task 10025:
> > >   kasan_save_stack mm/kasan/common.c:47 [inline]
> > >   kasan_save_track+0x3f/0x80 mm/kasan/common.c:68
> > >   poison_kmalloc_redzone mm/kasan/common.c:377 [inline]
> > >   __kasan_kmalloc+0x98/0xb0 mm/kasan/common.c:394
> > >   kasan_kmalloc include/linux/kasan.h:260 [inline]
> > >   __do_kmalloc_node mm/slub.c:4294 [inline]
> > >   __kmalloc_node_track_caller_noprof+0x28b/0x4c0 mm/slub.c:4313
> > >   __kmemdup_nul mm/util.c:61 [inline]
> > >   kstrdup+0x42/0x100 mm/util.c:81
> > >   kobject_set_name_vargs+0x61/0x120 lib/kobject.c:274
> > >   dev_set_name+0xd5/0x120 drivers/base/core.c:3468
> > >   assign_name drivers/infiniband/core/device.c:1202 [inline]
> > >   ib_register_device+0x178/0x1460 drivers/infiniband/core/device.c:1384
> > >   rxe_register_device+0x233/0x350 drivers/infiniband/sw/rxe/rxe_verbs.c:1540
> > >   rxe_net_add+0x74/0xf0 drivers/infiniband/sw/rxe/rxe_net.c:550
> > >   rxe_newlink+0xde/0x1a0 drivers/infiniband/sw/rxe/rxe.c:212
> > >   nldev_newlink+0x5ea/0x680 drivers/infiniband/core/nldev.c:1795
> > >   rdma_nl_rcv_skb drivers/infiniband/core/netlink.c:239 [inline]
> > >   rdma_nl_rcv+0x6dd/0x9e0 drivers/infiniband/core/netlink.c:259
> > >   netlink_unicast_kernel net/netlink/af_netlink.c:1313 [inline]
> > >   netlink_unicast+0x7f6/0x990 net/netlink/af_netlink.c:1339
> > >   netlink_sendmsg+0x8de/0xcb0 net/netlink/af_netlink.c:1883
> > >   sock_sendmsg_nosec net/socket.c:709 [inline]
> > >   __sock_sendmsg+0x221/0x270 net/socket.c:724
> > >   ____sys_sendmsg+0x53a/0x860 net/socket.c:2564
> > >   ___sys_sendmsg net/socket.c:2618 [inline]
> > >   __sys_sendmsg+0x269/0x350 net/socket.c:2650
> > >   do_syscall_x64 arch/x86/entry/common.c:52 [inline]
> > >   do_syscall_64+0xf3/0x230 arch/x86/entry/common.c:83
> > >   entry_SYSCALL_64_after_hwframe+0x77/0x7f
> > > 
> > > Freed by task 10035:
> > >   kasan_save_stack mm/kasan/common.c:47 [inline]
> > >   kasan_save_track+0x3f/0x80 mm/kasan/common.c:68
> > >   kasan_save_free_info+0x40/0x50 mm/kasan/generic.c:576
> > >   poison_slab_object mm/kasan/common.c:247 [inline]
> > >   __kasan_slab_free+0x59/0x70 mm/kasan/common.c:264
> > >   kasan_slab_free include/linux/kasan.h:233 [inline]
> > >   slab_free_hook mm/slub.c:2353 [inline]
> > >   slab_free mm/slub.c:4609 [inline]
> > >   kfree+0x196/0x430 mm/slub.c:4757
> > >   kobject_rename+0x38f/0x410 lib/kobject.c:524
> > >   device_rename+0x16a/0x200 drivers/base/core.c:4525
> > >   ib_device_rename+0x270/0x710 drivers/infiniband/core/device.c:402
> > >   nldev_set_doit+0x30e/0x4c0 drivers/infiniband/core/nldev.c:1146
> > >   rdma_nl_rcv_skb drivers/infiniband/core/netlink.c:239 [inline]
> > >   rdma_nl_rcv+0x6dd/0x9e0 drivers/infiniband/core/netlink.c:259
> > >   netlink_unicast_kernel net/netlink/af_netlink.c:1313 [inline]
> > >   netlink_unicast+0x7f6/0x990 net/netlink/af_netlink.c:1339
> > >   netlink_sendmsg+0x8de/0xcb0 net/netlink/af_netlink.c:1883
> > >   sock_sendmsg_nosec net/socket.c:709 [inline]
> > >   __sock_sendmsg+0x221/0x270 net/socket.c:724
> > >   ____sys_sendmsg+0x53a/0x860 net/socket.c:2564
> > >   ___sys_sendmsg net/socket.c:2618 [inline]
> > >   __sys_sendmsg+0x269/0x350 net/socket.c:2650
> > >   do_syscall_x64 arch/x86/entry/common.c:52 [inline]
> > >   do_syscall_64+0xf3/0x230 arch/x86/entry/common.c:83
> > >   entry_SYSCALL_64_after_hwframe+0x77/0x7f
> > > 
> > > This is because if rename device happens, the old name is freed in
> > > ib_device_rename() with lock, but fill_nldev_handle() may visit the dev
> > > name locklessly triggered by rxe_newlink().
> > > 
> > > Fix this by add lock around rdma_nl_notify_event() in
> > > ib_device_notify_register().
> > > 
> > > Reported-by: syzbot+f60349ba1f9f08df349f@syzkaller.appspotmail.com
> > > Closes: https://syzkaller.appspot.com/bug?extid=25bc6f0ed2b88b9eb9b8
> > > Fixes: 9cbed5aab5ae ("RDMA/nldev: Add support for RDMA monitoring")
> > > Signed-off-by: Wang Liang <wangliang74@huawei.com>
> > > ---
> > >   drivers/infiniband/core/device.c | 2 ++
> > >   1 file changed, 2 insertions(+)
> > > 
> > > diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c
> > > index 0ded91f056f3..4536621ada0d 100644
> > > --- a/drivers/infiniband/core/device.c
> > > +++ b/drivers/infiniband/core/device.c
> > > @@ -1341,7 +1341,9 @@ static void ib_device_notify_register(struct ib_device *device)
> > >   	u32 port;
> > >   	int ret;
> > > +	down_write(&devices_rwsem);
> > The analysis looks correct to me, however this should be down_read(&devices_rwsem)
> > together with comment about possible race with RDMA netlink, which can change
> > internals of struct ib_device.
> > 
> > I wonder if this read semaphore should be hold for whole
> > ib_device_notify_register() function and not only for RDMA_REGISTER_EVENT event.
> > 
> > Thanks
> Yes, you are right! The RDMA_NETDEV_ATTACH_EVENT event in function
> ib_device_notify_register() can also visit the dev name locklessly:
> 
> rxe_newlink
>     ...
>     ib_device_notify_register
>         rdma_nl_notify_event(RDMA_NETDEV_ATTACH_EVENT)
>             fill_mon_netdev_association
>                 // visit dev_name without devices_rwsem
> 
> What's more, the following trace may also trigger the issue when function
> nldev_set_doit() rename device:
> 
> (1)
> RDMA_NLDEV_CMD_GET
>     nldev_get_doit
>         fill_dev_info
>             fill_nldev_handle
>                 // visit dev_name without devices_rwsem
> 
> (2)
> RDMA_NLDEV_CMD_PORT_GET
>     nldev_port_get_doit
>         fill_port_info
>             fill_nldev_handle
>                 // visit dev_name without devices_rwsem
> 
> (3)
> RDMA_NLDEV_CMD_RES_GET
>     nldev_res_get_doit
>         fill_res_info
>             fill_nldev_handle
>                 // visit dev_name without devices_rwsem
> 
> And more...
> 
> However, some scenarios hold devices_rwsem, like function
> nldev_get_dumpit(),
> which was modified by 921eab1143aa("RDMA/devices: Re-organize device.c
> locking"):
> 
> RDMA_NLDEV_CMD_GET
>     nldev_get_dumpit
>         ib_enum_all_devs
>             down_read(&devices_rwsem); // hold devices_rwsem read
>                 _nldev_get_dumpit
>                     fill_dev_info
>                         fill_nldev_handle
>                             // visit dev_name with devices_rwsem
>             up_read(&devices_rwsem);
> 
> So only hold devices_rwsem for whole ib_device_notify_register() may be not
> enough.
> 
> How about add down_read(&devices_rwsem) around
> fill_dev_info()/fill_port_info()/
> fill_res_info()/ib_device_notify_register().

I don't think so, as all flows in nldev.c except rdma_nl_notify_event() are
single threaded and call to ib_device_rename() will make sure that no
other RDMA_NLDEV_CMD_* commands are executed.

In your case call to rdma_nl_notify_event() can happen in parallel to
device_rename only.

Thanks