tcx: Fix splat in ingress_destroy upon tcx_entry_free
Subsystems (4)
Daniel Borkmann
maintainerPATCH 1/1 tcx: Fix splat in ingress_destroy upon tcx_entry_free
On qdisc destruction, the ingress_destroy() needs to update the correct entry, that is, tcx_entry_update must NULL the dev->tcx_ingress pointer. Therefore, fix the typo.
--- net/sched/sch_ingress.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
net/sched/sch_ingress.c +1 -1
@@ -123,7 +123,7 @@ static void ingress_destroy(struct Qdisc *sch)if (entry) {tcx_miniq_set_active (entry, false);if (!tcx_entry_is_active (entry)) { -tcx_entry_update (dev,NULL , false); +tcx_entry_update (dev,NULL , true);tcx_entry_free (entry); } }
Signature · 1 line
2.34.1
Daniel Borkmann
maintainer Tested-by [email protected] Tested-by [email protected] Tested-by [email protected] on v1 1/1Quoted 9 lines · Daniel Borkmann
On qdisc destruction, the
ingress_destroy()needs to update the correct entry, that is,tcx_entry_updatemust NULL thedev->tcx_ingresspointer. Therefore, fix the typo.Fixes e420bed02507 ("bpf: Add fd-based tcx multi-prog infra with link support")Reported-by [email protected]Reported-by [email protected]Reported-by [email protected]Signed-off-by Daniel Borkmann <[email protected]>
Additionally, syzbot also tested that all 3 reports are fixed with this:
Quoted 9 lines · Daniel Borkmann
On qdisc destruction, the
ingress_destroy()needs to update the correct entry, that is,tcx_entry_updatemust NULL thedev->tcx_ingresspointer. Therefore, fix the typo.Fixes e420bed02507 ("bpf: Add fd-based tcx multi-prog infra with link support")Reported-by [email protected]Reported-by [email protected]Reported-by [email protected]Signed-off-by Daniel Borkmann <[email protected]>
Thanks!
This patch was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:
Hello:
This patch was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:
Quoted 11 lines · On Sat, 22 Jul 2023 01:33:30 +0200 you wrote:
On qdisc destruction, the
ingress_destroy()needs to update the correct entry, that is,tcx_entry_updatemust NULL thedev->tcx_ingresspointer. Therefore, fix the typo.Fixes e420bed02507 ("bpf: Add fd-based tcx multi-prog infra with link support")Reported-by [email protected]Reported-by [email protected]Reported-by [email protected]Signed-off-by Daniel Borkmann <[email protected]>[...]
Here is the summary with links:
- [net-next] tcx: Fix splat in ingress_destroy upon tcx_entry_free
https://git.kernel.org/netdev/net-next/c/dc644b540a2d
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html
Quoted 9 lines · Daniel Borkmann
On qdisc destruction, the
ingress_destroy()needs to update the correct entry, that is,tcx_entry_updatemust NULL thedev->tcx_ingresspointer. Therefore, fix the typo.Fixes e420bed02507 ("bpf: Add fd-based tcx multi-prog infra with link support")Reported-by [email protected]Reported-by [email protected]Reported-by [email protected]Signed-off-by Daniel Borkmann <[email protected]>
Hi Daniel,
Our nightly regression testing picked up new memory leaks which were bisected to this commit.
Unfortunately, I do not know the exact repro steps to trigger it, maybe the attached kmemeleak logs can help?
unreferenced object 0xffff88811ce37b80 (size 224):
comm "kworker/14:1", pid 7451, jiffies 4295350041 (age 64.444s) hex dump (first 32 bytes):
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
00 e0 69 29 81 88 ff ff 00 3a 39 0d 81 88 ff ff ..i).....:9.....
backtrace:
[<00000000a0f098fe>] __alloc_skb+0x1f4/0x2b0
[<000000000dabee54>] alloc_skb_with_frags+0x7a/0x6c0
[<00000000e681c78a>] sock_alloc_send_pskb+0x63f/0x7d0
[<00000000a4774143>] mld_newpack.isra.0+0x1ad/0x800
[<0000000060e32100>] add_grhead+0x271/0x320
[<00000000040e7099>] add_grec+0xc8b/0x1120
[<000000009853483c>] mld_ifc_work+0x387/0xae0
[<0000000079d8299d>] process_one_work+0x86a/0x1430
[<000000001968010b>] worker_thread+0x5b0/0xf00
[<0000000090c285b0>] kthread+0x2dd/0x3b0
[<000000001f322d79>] ret_from_fork+0x2d/0x70
[<000000008ad6bd7b>] ret_from_fork_asm+0x11/0x20
unreferenced object 0xffff888153058640 (size 224):
comm "softirq", pid 0, jiffies 4295350849 (age 61.212s)
hex dump (first 32 bytes):
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
00 e0 69 29 81 88 ff ff c0 32 39 0d 81 88 ff ff ..i).....29.....
backtrace:
[<00000000a0f098fe>] __alloc_skb+0x1f4/0x2b0
[<00000000bb2ddb4c>] ndisc_alloc_skb+0x133/0x340
[<0000000009614816>] ndisc_send_rs+0x1e0/0x4b0
[<000000004bc1b8be>] addrconf_rs_timer+0x25a/0x720
[<000000004d021706>] call_timer_fn+0x167/0x3d0
[<0000000088aa76a3>] __run_timers.part.0+0x546/0x8b0
[<0000000066f62ff3>] run_timer_softirq+0x6a/0x100
[<000000003732ddfb>] __do_softirq+0x264/0x80c
unreferenced object 0xffff888155d0a500 (size 224):
comm "softirq", pid 0, jiffies 4295352832 (age 53.328s)
hex dump (first 32 bytes):
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
00 e0 69 29 81 88 ff ff c0 32 39 0d 81 88 ff ff ..i).....29.....
backtrace:
[<00000000a0f098fe>] __alloc_skb+0x1f4/0x2b0
[<00000000bb2ddb4c>] ndisc_alloc_skb+0x133/0x340
[<0000000009614816>] ndisc_send_rs+0x1e0/0x4b0
[<000000004bc1b8be>] addrconf_rs_timer+0x25a/0x720
[<000000004d021706>] call_timer_fn+0x167/0x3d0
[<0000000088aa76a3>] __run_timers.part.0+0x546/0x8b0
[<0000000066f62ff3>] run_timer_softirq+0x6a/0x100
[<000000003732ddfb>] __do_softirq+0x264/0x80c
unreferenced object 0xffff88814e3f6040 (size 576):
comm "softirq", pid 0, jiffies 4295352832 (age 53.328s)
hex dump (first 32 bytes):
00 00 33 33 00 00 00 02 e8 eb d3 98 21 bc 86 dd ..33........!...
60 00 00 00 00 10 3a ff fe 80 00 00 00 00 00 00 `.....:.........
backtrace:
[<00000000525ad98b>] kmalloc_reserve+0x118/0x1f0
[<000000008d146183>] __alloc_skb+0x105/0x2b0
[<00000000bb2ddb4c>] ndisc_alloc_skb+0x133/0x340
[<0000000009614816>] ndisc_send_rs+0x1e0/0x4b0
[<000000004bc1b8be>] addrconf_rs_timer+0x25a/0x720
[<000000004d021706>] call_timer_fn+0x167/0x3d0
[<0000000088aa76a3>] __run_timers.part.0+0x546/0x8b0
[<0000000066f62ff3>] run_timer_softirq+0x6a/0x100
[<000000003732ddfb>] __do_softirq+0x264/0x80c
unreferenced object 0xffff88812acdebc0 (size 16):
comm "umount.nfs", pid 11626, jiffies 4295354796 (age 45.472s)
hex dump (first 16 bytes):
73 65 72 76 65 72 2d 32 00 eb cd 2a 81 88 ff ff server-2...*....
backtrace:
[<0000000010fb5130>] __kmalloc_node_track_caller+0x4c/0x170
[<00000000b866a733>] kvasprintf+0xb0/0x130
[<00000000b3564fca>] kasprintf+0xa6/0xd0
[<00000000f01d6cb3>] nfs_sysfs_move_sb_to_server+0x49/0xd0
[<000000009608708f>] nfs_kill_super+0x5f/0x90
[<0000000090d4108b>] deactivate_locked_super+0x80/0x130
[<000000000856aeb1>] cleanup_mnt+0x258/0x370
[<0000000040582e39>] task_work_run+0x12c/0x210
[<00000000378ea041>] exit_to_user_mode_prepare+0x1a0/0x1b0
[<00000000025e63dd>] syscall_exit_to_user_mode+0x19/0x50
[<00000000f34ad3ee>] do_syscall_64+0x4a/0x90
[<000000009d3e2403>] entry_SYSCALL_64_after_hwframe+0x46/0xb0Martin KaFai Lau
reviewerQuoted 11 lines · Gal Pressman
Quoted 9 lines · Daniel Borkmann
On qdisc destruction, the
ingress_destroy()needs to update the correct entry, that is,tcx_entry_updatemust NULL thedev->tcx_ingresspointer. Therefore, fix the typo.Fixes e420bed02507 ("bpf: Add fd-based tcx multi-prog infra with link support")Reported-by [email protected]Reported-by [email protected]Reported-by [email protected]Signed-off-by Daniel Borkmann <[email protected]>
Quoted 4 lines
Hi Daniel,
Our nightly regression testing picked up new memory leaks which were bisected to this commit.
Thanks for the report. Can you help to check if it can be reproduced with the latest net-next which has a tcx fix in commit 079082c60aff ("tcx: Fix splat during dev unregister")?
Quoted 13 lines · Martin KaFai Lau
Quoted 12 lines · Gal Pressman
Quoted 10 lines · Daniel Borkmann
On qdisc destruction, the
ingress_destroy()needs to update the correct entry, that is,tcx_entry_updatemust NULL thedev->tcx_ingresspointer. Therefore, fix the typo.Fixes e420bed02507 ("bpf: Add fd-based tcx multi-prog infra withlink support")
Reported-by [email protected]Reported-by [email protected]Reported-by [email protected]Signed-off-by Daniel Borkmann <[email protected]>
Quoted 9 lines
Quoted 4 lines
Hi Daniel,
Our nightly regression testing picked up new memory leaks which were bisected to this commit.
Thanks for the report. Can you help to check if it can be reproduced with the latest net-next which has a tcx fix in commit
079082c60aff("tcx: Fix splat during dev unregister")?
I think the issue is resolved after this patch, but we're still verifying. Thanks!
Daniel Borkmann
maintainerHi Gal,
On 8/3/23 1:10 PM, Gal Pressman wrote:
[...]
Quoted 4 lines
Our nightly regression testing picked up new memory leaks which were bisected to this commit.
Unfortunately, I do not know the exact repro steps to trigger it, maybe the attached kmemeleak logs can help?
Is this on latest net-next? Do you have some more info on what the test is doing? Does it trigger on qdisc cleanup only? Also, is there a way to run the regression suite?
Thanks,
Daniel
Quoted 79 lines
unreferenced object
0xffff88811ce37b80(size 224):comm "kworker/14:1", pid 7451, jiffies 4295350041 (age 64.444s) hex dump (first 32 bytes): 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 00 e0 69 29 81 88 ff ff 00 3a 39 0d 81 88 ff ff ..i).....:9..... backtrace: [<00000000a0f098fe>] __alloc_skb+0x1f4/0x2b0 [<000000000dabee54>] alloc_skb_with_frags+0x7a/0x6c0 [<00000000e681c78a>] sock_alloc_send_pskb+0x63f/0x7d0 [<00000000a4774143>] mld_newpack.isra.0+0x1ad/0x800 [<0000000060e32100>] add_grhead+0x271/0x320 [<00000000040e7099>] add_grec+0xc8b/0x1120 [<000000009853483c>] mld_ifc_work+0x387/0xae0 [<0000000079d8299d>] process_one_work+0x86a/0x1430 [<000000001968010b>] worker_thread+0x5b0/0xf00 [<0000000090c285b0>] kthread+0x2dd/0x3b0 [<000000001f322d79>] ret_from_fork+0x2d/0x70 [<000000008ad6bd7b>] ret_from_fork_asm+0x11/0x20 unreferenced object 0xffff888153058640 (size 224): comm "softirq", pid 0, jiffies 4295350849 (age 61.212s) hex dump (first 32 bytes): 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 00 e0 69 29 81 88 ff ff c0 32 39 0d 81 88 ff ff ..i).....29..... backtrace: [<00000000a0f098fe>] __alloc_skb+0x1f4/0x2b0 [<00000000bb2ddb4c>] ndisc_alloc_skb+0x133/0x340 [<0000000009614816>] ndisc_send_rs+0x1e0/0x4b0 [<000000004bc1b8be>] addrconf_rs_timer+0x25a/0x720 [<000000004d021706>] call_timer_fn+0x167/0x3d0 [<0000000088aa76a3>] __run_timers.part.0+0x546/0x8b0 [<0000000066f62ff3>] run_timer_softirq+0x6a/0x100 [<000000003732ddfb>] __do_softirq+0x264/0x80c unreferenced object 0xffff888155d0a500 (size 224): comm "softirq", pid 0, jiffies 4295352832 (age 53.328s) hex dump (first 32 bytes): 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 00 e0 69 29 81 88 ff ff c0 32 39 0d 81 88 ff ff ..i).....29..... backtrace: [<00000000a0f098fe>] __alloc_skb+0x1f4/0x2b0 [<00000000bb2ddb4c>] ndisc_alloc_skb+0x133/0x340 [<0000000009614816>] ndisc_send_rs+0x1e0/0x4b0 [<000000004bc1b8be>] addrconf_rs_timer+0x25a/0x720 [<000000004d021706>] call_timer_fn+0x167/0x3d0 [<0000000088aa76a3>] __run_timers.part.0+0x546/0x8b0 [<0000000066f62ff3>] run_timer_softirq+0x6a/0x100 [<000000003732ddfb>] __do_softirq+0x264/0x80c unreferenced object 0xffff88814e3f6040 (size 576): comm "softirq", pid 0, jiffies 4295352832 (age 53.328s) hex dump (first 32 bytes): 00 00 33 33 00 00 00 02 e8 eb d3 98 21 bc 86 dd ..33........!... 60 00 00 00 00 10 3a ff fe 80 00 00 00 00 00 00 `.....:......... backtrace: [<00000000525ad98b>] kmalloc_reserve+0x118/0x1f0 [<000000008d146183>] __alloc_skb+0x105/0x2b0 [<00000000bb2ddb4c>] ndisc_alloc_skb+0x133/0x340 [<0000000009614816>] ndisc_send_rs+0x1e0/0x4b0 [<000000004bc1b8be>] addrconf_rs_timer+0x25a/0x720 [<000000004d021706>] call_timer_fn+0x167/0x3d0 [<0000000088aa76a3>] __run_timers.part.0+0x546/0x8b0 [<0000000066f62ff3>] run_timer_softirq+0x6a/0x100 [<000000003732ddfb>] __do_softirq+0x264/0x80c unreferenced object 0xffff88812acdebc0 (size 16): comm "umount.nfs", pid 11626, jiffies 4295354796 (age 45.472s) hex dump (first 16 bytes): 73 65 72 76 65 72 2d 32 00 eb cd 2a 81 88 ff ff server-2...*.... backtrace: [<0000000010fb5130>] __kmalloc_node_track_caller+0x4c/0x170 [<00000000b866a733>] kvasprintf+0xb0/0x130 [<00000000b3564fca>] kasprintf+0xa6/0xd0 [<00000000f01d6cb3>] nfs_sysfs_move_sb_to_server+0x49/0xd0 [<000000009608708f>] nfs_kill_super+0x5f/0x90 [<0000000090d4108b>] deactivate_locked_super+0x80/0x130 [<000000000856aeb1>] cleanup_mnt+0x258/0x370 [<0000000040582e39>] task_work_run+0x12c/0x210 [<00000000378ea041>] exit_to_user_mode_prepare+0x1a0/0x1b0 [<00000000025e63dd>] syscall_exit_to_user_mode+0x19/0x50 [<00000000f34ad3ee>] do_syscall_64+0x4a/0x90 [<000000009d3e2403>] entry_SYSCALL_64_after_hwframe+0x46/0xb0
Quoted 4 lines · Daniel Borkmann
Hi Gal,
On 8/3/23 1:10 PM, Gal Pressman wrote:
[...]
Quoted 8 lines
Quoted 4 lines
Our nightly regression testing picked up new memory leaks which were bisected to this commit.
Unfortunately, I do not know the exact repro steps to trigger it, maybe the attached kmemeleak logs can help?Is this on latest net-next? Do you have some more info on what the test is doing? Does it trigger on qdisc cleanup only? Also, is there a way to run the regression suite?
I don't have more information ATM, I'll try to figure it out if the other fix doesn't resolve it.
Quoted 4 lines · Gal Pressman
Our nightly regression testing picked up new memory leaks which were bisected to this commit.
Unfortunately, I do not know the exact repro steps to trigger it, maybe the attached kmemeleak logs can help?
[...]
Quoted 17 lines
unreferenced object
0xffff88812acdebc0(size 16):
comm "umount.nfs", pid 11626, jiffies 4295354796 (age 45.472s) hex dump (first 16 bytes):73 65 72 76 65 72 2d 32 00 eb cd 2a 81 88 ff ff server-2...*.... backtrace: [<0000000010fb5130>] __kmalloc_node_track_caller+0x4c/0x170 [<00000000b866a733>] kvasprintf+0xb0/0x130 [<00000000b3564fca>] kasprintf+0xa6/0xd0 [<00000000f01d6cb3>] nfs_sysfs_move_sb_to_server+0x49/0xd0 [<000000009608708f>] nfs_kill_super+0x5f/0x90 [<0000000090d4108b>] deactivate_locked_super+0x80/0x130 [<000000000856aeb1>] cleanup_mnt+0x258/0x370 [<0000000040582e39>] task_work_run+0x12c/0x210 [<00000000378ea041>] exit_to_user_mode_prepare+0x1a0/0x1b0 [<00000000025e63dd>] syscall_exit_to_user_mode+0x19/0x50 [<00000000f34ad3ee>] do_syscall_64+0x4a/0x90 [<000000009d3e2403>] entry_SYSCALL_64_after_hwframe+0x46/0xb0
This one is caused by commit 1c7251187dc0 ("NFS: add superblock sysfs entries") and fixed by [1], so I'm not sure the bisection result is reliable.
Quoted 26 lines · Ido Schimmel
Quoted 4 lines · Gal Pressman
Our nightly regression testing picked up new memory leaks which were bisected to this commit.
Unfortunately, I do not know the exact repro steps to trigger it, maybe the attached kmemeleak logs can help?[...]
Quoted 17 lines
unreferenced object
0xffff88812acdebc0(size 16):
comm "umount.nfs", pid 11626, jiffies 4295354796 (age 45.472s) hex dump (first 16 bytes):73 65 72 76 65 72 2d 32 00 eb cd 2a 81 88 ff ff server-2...*.... backtrace: [<0000000010fb5130>] __kmalloc_node_track_caller+0x4c/0x170 [<00000000b866a733>] kvasprintf+0xb0/0x130 [<00000000b3564fca>] kasprintf+0xa6/0xd0 [<00000000f01d6cb3>] nfs_sysfs_move_sb_to_server+0x49/0xd0 [<000000009608708f>] nfs_kill_super+0x5f/0x90 [<0000000090d4108b>] deactivate_locked_super+0x80/0x130 [<000000000856aeb1>] cleanup_mnt+0x258/0x370 [<0000000040582e39>] task_work_run+0x12c/0x210 [<00000000378ea041>] exit_to_user_mode_prepare+0x1a0/0x1b0 [<00000000025e63dd>] syscall_exit_to_user_mode+0x19/0x50 [<00000000f34ad3ee>] do_syscall_64+0x4a/0x90 [<000000009d3e2403>] entry_SYSCALL_64_after_hwframe+0x46/0xb0
Quoted 5 lines
This one is caused by commit
1c7251187dc0("NFS: add superblock sysfs entries") and fixed by [1], so I'm not sure the bisection result is reliable.
Thanks, maybe there is more than one issue lurking.
We ran the bisect a few times, it always came back to this commit.