net/sched: cls_api: remove block_cb from driver_list before freeing
authorVlad Buslov <vladbu@nvidia.com>
Wed, 26 Apr 2023 12:31:11 +0000 (14:31 +0200)
committerDavid S. Miller <davem@davemloft.net>
Fri, 28 Apr 2023 08:44:19 +0000 (09:44 +0100)
commitda94a7781fc3c92e7df7832bc2746f4d39bc624e
tree10ec3cfaea118e09b3dbd525cb438c99aea1e5b0
parent7e692df3933628d974acb9f5b334d2b3e885e2a6
net/sched: cls_api: remove block_cb from driver_list before freeing

Error handler of tcf_block_bind() frees the whole bo->cb_list on error.
However, by that time the flow_block_cb instances are already in the driver
list because driver ndo_setup_tc() callback is called before that up the
call chain in tcf_block_offload_cmd(). This leaves dangling pointers to
freed objects in the list and causes use-after-free[0]. Fix it by also
removing flow_block_cb instances from driver_list before deallocating them.

[0]:
[  279.868433] ==================================================================
[  279.869964] BUG: KASAN: slab-use-after-free in flow_block_cb_setup_simple+0x631/0x7c0
[  279.871527] Read of size 8 at addr ffff888147e2bf20 by task tc/2963

[  279.873151] CPU: 6 PID: 2963 Comm: tc Not tainted 6.3.0-rc6+ #4
[  279.874273] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014
[  279.876295] Call Trace:
[  279.876882]  <TASK>
[  279.877413]  dump_stack_lvl+0x33/0x50
[  279.878198]  print_report+0xc2/0x610
[  279.878987]  ? flow_block_cb_setup_simple+0x631/0x7c0
[  279.879994]  kasan_report+0xae/0xe0
[  279.880750]  ? flow_block_cb_setup_simple+0x631/0x7c0
[  279.881744]  ? mlx5e_tc_reoffload_flows_work+0x240/0x240 [mlx5_core]
[  279.883047]  flow_block_cb_setup_simple+0x631/0x7c0
[  279.884027]  tcf_block_offload_cmd.isra.0+0x189/0x2d0
[  279.885037]  ? tcf_block_setup+0x6b0/0x6b0
[  279.885901]  ? mutex_lock+0x7d/0xd0
[  279.886669]  ? __mutex_unlock_slowpath.constprop.0+0x2d0/0x2d0
[  279.887844]  ? ingress_init+0x1c0/0x1c0 [sch_ingress]
[  279.888846]  tcf_block_get_ext+0x61c/0x1200
[  279.889711]  ingress_init+0x112/0x1c0 [sch_ingress]
[  279.890682]  ? clsact_init+0x2b0/0x2b0 [sch_ingress]
[  279.891701]  qdisc_create+0x401/0xea0
[  279.892485]  ? qdisc_tree_reduce_backlog+0x470/0x470
[  279.893473]  tc_modify_qdisc+0x6f7/0x16d0
[  279.894344]  ? tc_get_qdisc+0xac0/0xac0
[  279.895213]  ? mutex_lock+0x7d/0xd0
[  279.896005]  ? __mutex_lock_slowpath+0x10/0x10
[  279.896910]  rtnetlink_rcv_msg+0x5fe/0x9d0
[  279.897770]  ? rtnl_calcit.isra.0+0x2b0/0x2b0
[  279.898672]  ? __sys_sendmsg+0xb5/0x140
[  279.899494]  ? do_syscall_64+0x3d/0x90
[  279.900302]  ? entry_SYSCALL_64_after_hwframe+0x46/0xb0
[  279.901337]  ? kasan_save_stack+0x2e/0x40
[  279.902177]  ? kasan_save_stack+0x1e/0x40
[  279.903058]  ? kasan_set_track+0x21/0x30
[  279.903913]  ? kasan_save_free_info+0x2a/0x40
[  279.904836]  ? ____kasan_slab_free+0x11a/0x1b0
[  279.905741]  ? kmem_cache_free+0x179/0x400
[  279.906599]  netlink_rcv_skb+0x12c/0x360
[  279.907450]  ? rtnl_calcit.isra.0+0x2b0/0x2b0
[  279.908360]  ? netlink_ack+0x1550/0x1550
[  279.909192]  ? rhashtable_walk_peek+0x170/0x170
[  279.910135]  ? kmem_cache_alloc_node+0x1af/0x390
[  279.911086]  ? _copy_from_iter+0x3d6/0xc70
[  279.912031]  netlink_unicast+0x553/0x790
[  279.912864]  ? netlink_attachskb+0x6a0/0x6a0
[  279.913763]  ? netlink_recvmsg+0x416/0xb50
[  279.914627]  netlink_sendmsg+0x7a1/0xcb0
[  279.915473]  ? netlink_unicast+0x790/0x790
[  279.916334]  ? iovec_from_user.part.0+0x4d/0x220
[  279.917293]  ? netlink_unicast+0x790/0x790
[  279.918159]  sock_sendmsg+0xc5/0x190
[  279.918938]  ____sys_sendmsg+0x535/0x6b0
[  279.919813]  ? import_iovec+0x7/0x10
[  279.920601]  ? kernel_sendmsg+0x30/0x30
[  279.921423]  ? __copy_msghdr+0x3c0/0x3c0
[  279.922254]  ? import_iovec+0x7/0x10
[  279.923041]  ___sys_sendmsg+0xeb/0x170
[  279.923854]  ? copy_msghdr_from_user+0x110/0x110
[  279.924797]  ? ___sys_recvmsg+0xd9/0x130
[  279.925630]  ? __perf_event_task_sched_in+0x183/0x470
[  279.926656]  ? ___sys_sendmsg+0x170/0x170
[  279.927529]  ? ctx_sched_in+0x530/0x530
[  279.928369]  ? update_curr+0x283/0x4f0
[  279.929185]  ? perf_event_update_userpage+0x570/0x570
[  279.930201]  ? __fget_light+0x57/0x520
[  279.931023]  ? __switch_to+0x53d/0xe70
[  279.931846]  ? sockfd_lookup_light+0x1a/0x140
[  279.932761]  __sys_sendmsg+0xb5/0x140
[  279.933560]  ? __sys_sendmsg_sock+0x20/0x20
[  279.934436]  ? fpregs_assert_state_consistent+0x1d/0xa0
[  279.935490]  do_syscall_64+0x3d/0x90
[  279.936300]  entry_SYSCALL_64_after_hwframe+0x46/0xb0
[  279.937311] RIP: 0033:0x7f21c814f887
[  279.938085] Code: 0a 00 f7 d8 64 89 02 48 c7 c0 ff ff ff ff eb b9 0f 1f 00 f3 0f 1e fa 64 8b 04 25 18 00 00 00 85 c0 75 10 b8 2e 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 51 c3 48 83 ec 28 89 54 24 1c 48 89 74 24 10
[  279.941448] RSP: 002b:00007fff11efd478 EFLAGS: 00000246 ORIG_RAX: 000000000000002e
[  279.942964] RAX: ffffffffffffffda RBX: 0000000064401979 RCX: 00007f21c814f887
[  279.944337] RDX: 0000000000000000 RSI: 00007fff11efd4e0 RDI: 0000000000000003
[  279.945660] RBP: 0000000000000000 R08: 0000000000000001 R09: 0000000000000000
[  279.947003] R10: 00007f21c8008708 R11: 0000000000000246 R12: 0000000000000001
[  279.948345] R13: 0000000000409980 R14: 000000000047e538 R15: 0000000000485400
[  279.949690]  </TASK>

[  279.950706] Allocated by task 2960:
[  279.951471]  kasan_save_stack+0x1e/0x40
[  279.952338]  kasan_set_track+0x21/0x30
[  279.953165]  __kasan_kmalloc+0x77/0x90
[  279.954006]  flow_block_cb_setup_simple+0x3dd/0x7c0
[  279.955001]  tcf_block_offload_cmd.isra.0+0x189/0x2d0
[  279.956020]  tcf_block_get_ext+0x61c/0x1200
[  279.956881]  ingress_init+0x112/0x1c0 [sch_ingress]
[  279.957873]  qdisc_create+0x401/0xea0
[  279.958656]  tc_modify_qdisc+0x6f7/0x16d0
[  279.959506]  rtnetlink_rcv_msg+0x5fe/0x9d0
[  279.960392]  netlink_rcv_skb+0x12c/0x360
[  279.961216]  netlink_unicast+0x553/0x790
[  279.962044]  netlink_sendmsg+0x7a1/0xcb0
[  279.962906]  sock_sendmsg+0xc5/0x190
[  279.963702]  ____sys_sendmsg+0x535/0x6b0
[  279.964534]  ___sys_sendmsg+0xeb/0x170
[  279.965343]  __sys_sendmsg+0xb5/0x140
[  279.966132]  do_syscall_64+0x3d/0x90
[  279.966908]  entry_SYSCALL_64_after_hwframe+0x46/0xb0

[  279.968407] Freed by task 2960:
[  279.969114]  kasan_save_stack+0x1e/0x40
[  279.969929]  kasan_set_track+0x21/0x30
[  279.970729]  kasan_save_free_info+0x2a/0x40
[  279.971603]  ____kasan_slab_free+0x11a/0x1b0
[  279.972483]  __kmem_cache_free+0x14d/0x280
[  279.973337]  tcf_block_setup+0x29d/0x6b0
[  279.974173]  tcf_block_offload_cmd.isra.0+0x226/0x2d0
[  279.975186]  tcf_block_get_ext+0x61c/0x1200
[  279.976080]  ingress_init+0x112/0x1c0 [sch_ingress]
[  279.977065]  qdisc_create+0x401/0xea0
[  279.977857]  tc_modify_qdisc+0x6f7/0x16d0
[  279.978695]  rtnetlink_rcv_msg+0x5fe/0x9d0
[  279.979562]  netlink_rcv_skb+0x12c/0x360
[  279.980388]  netlink_unicast+0x553/0x790
[  279.981214]  netlink_sendmsg+0x7a1/0xcb0
[  279.982043]  sock_sendmsg+0xc5/0x190
[  279.982827]  ____sys_sendmsg+0x535/0x6b0
[  279.983703]  ___sys_sendmsg+0xeb/0x170
[  279.984510]  __sys_sendmsg+0xb5/0x140
[  279.985298]  do_syscall_64+0x3d/0x90
[  279.986076]  entry_SYSCALL_64_after_hwframe+0x46/0xb0

[  279.987532] The buggy address belongs to the object at ffff888147e2bf00
                which belongs to the cache kmalloc-192 of size 192
[  279.989747] The buggy address is located 32 bytes inside of
                freed 192-byte region [ffff888147e2bf00ffff888147e2bfc0)

[  279.992367] The buggy address belongs to the physical page:
[  279.993430] page:00000000550f405c refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x147e2a
[  279.995182] head:00000000550f405c order:1 entire_mapcount:0 nr_pages_mapped:0 pincount:0
[  279.996713] anon flags: 0x200000000010200(slab|head|node=0|zone=2)
[  279.997878] raw: 0200000000010200 ffff888100042a00 0000000000000000 dead000000000001
[  279.999384] raw: 0000000000000000 0000000000200020 00000001ffffffff 0000000000000000
[  280.000894] page dumped because: kasan: bad access detected

[  280.002386] Memory state around the buggy address:
[  280.003338]  ffff888147e2be00: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[  280.004781]  ffff888147e2be80: fb fb fb fb fb fb fb fb fc fc fc fc fc fc fc fc
[  280.006224] >ffff888147e2bf00: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[  280.007700]                                ^
[  280.008592]  ffff888147e2bf80: fb fb fb fb fb fb fb fb fc fc fc fc fc fc fc fc
[  280.010035]  ffff888147e2c000: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[  280.011564] ==================================================================

Fixes: 59094b1e5094 ("net: sched: use flow block API")
Signed-off-by: Vlad Buslov <vladbu@nvidia.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/sched/cls_api.c