linux/include
Yafang Shao 66048f8b3c net/cls_cgroup: Fix task_get_classid() during qdisc run
During recent testing with the netem qdisc to inject delays into TCP
traffic, we observed that our CLS BPF program failed to function correctly
due to incorrect classid retrieval from task_get_classid(). The issue
manifests in the following call stack:

        bpf_get_cgroup_classid+5
        cls_bpf_classify+507
        __tcf_classify+90
        tcf_classify+217
        __dev_queue_xmit+798
        bond_dev_queue_xmit+43
        __bond_start_xmit+211
        bond_start_xmit+70
        dev_hard_start_xmit+142
        sch_direct_xmit+161
        __qdisc_run+102             <<<<< Issue location
        __dev_xmit_skb+1015
        __dev_queue_xmit+637
        neigh_hh_output+159
        ip_finish_output2+461
        __ip_finish_output+183
        ip_finish_output+41
        ip_output+120
        ip_local_out+94
        __ip_queue_xmit+394
        ip_queue_xmit+21
        __tcp_transmit_skb+2169
        tcp_write_xmit+959
        __tcp_push_pending_frames+55
        tcp_push+264
        tcp_sendmsg_locked+661
        tcp_sendmsg+45
        inet_sendmsg+67
        sock_sendmsg+98
        sock_write_iter+147
        vfs_write+786
        ksys_write+181
        __x64_sys_write+25
        do_syscall_64+56
        entry_SYSCALL_64_after_hwframe+100

The problem occurs when multiple tasks share a single qdisc. In such cases,
__qdisc_run() may transmit skbs created by different tasks. Consequently,
task_get_classid() retrieves an incorrect classid since it references the
current task's context rather than the skb's originating task.

Given that dev_queue_xmit() always executes with bh disabled, we can use
softirq_count() instead to obtain the correct classid.

The simple steps to reproduce this issue:
1. Add network delay to the network interface:
  such as: tc qdisc add dev bond0 root netem delay 1.5ms
2. Build two distinct net_cls cgroups, each with a network-intensive task
3. Initiate parallel TCP streams from both tasks to external servers.

Under this specific condition, the issue reliably occurs. The kernel
eventually dequeues an SKB that originated from Task-A while executing in
the context of Task-B.

It is worth noting that it will change the established behavior for a
slightly different scenario:

  <sock S is created by task A>
  <class ID for task A is changed>
  <skb is created by sock S xmit and classified>

prior to this patch the skb will be classified with the 'new' task A
classid, now with the old/original one. The bpf_get_cgroup_classid_curr()
function is a more appropriate choice for this case.

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Thomas Graf <tgraf@suug.ch>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Nikolay Aleksandrov <razor@blackwall.org>
Link: https://patch.msgid.link/20250902062933.30087-1-laoar.shao@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
History 2025-09-14 11:55:04 -07:00
..
acpi mailbox/pcc: support mailbox management of the shared buffer 2025-08-07 23:49:56 -05:00
asm-generic Deferred unwind changes for 6.17 2025-08-01 09:46:24 -07:00
clocksource KVM/arm64 updates for 6.14 2025-01-28 09:01:36 -08:00
crypto This push fixes a regression that breaks hmac(sha3-224-s390). 2025-08-09 07:26:19 +03:00
cxl cxl/events: Trace Memory Sparing Event Record 2025-07-18 08:19:56 -07:00
drm drm/gpuvm: fix various typos in .c and .h gpuvm file 2025-08-25 21:48:50 +02:00
dt-bindings Pin control changes for v6.17 2025-08-02 12:07:09 -07:00
hyperv hyperv-next for v6.16 2025-06-03 08:39:20 -07:00
keys rxrpc: Add YFS RxGK (GSSAPI) security class 2025-04-14 17:36:41 -07:00
kunit linux_kselftest-kunit-6.17-rc1 2025-07-29 12:43:10 -07:00
kvm KVM/arm64 changes for 6.17, round #1 2025-07-29 12:27:40 -04:00
linux net: use NUMA drop counters for softnet_data.dropped 2025-09-14 11:35:17 -07:00
math-emu powerpc/math-emu: Remove -w build flag and fix warnings 2022-09-05 17:28:25 +10:00
media media: core: export v4l2_translate_cmd 2025-07-11 19:27:28 +02:00
memory memory: renesas-rpc-if: Add RZ/G3E xSPI support 2025-05-01 12:22:25 +02:00
misc cxl: Remove driver 2025-03-16 22:04:27 +11:00
net net/cls_cgroup: Fix task_get_classid() during qdisc run 2025-09-14 11:55:04 -07:00
pcmcia pcmcia: remove PCCARD_IODYN 2025-08-16 15:37:47 +02:00
ras tracing, AER: Hide PCIe AER event when PCIEAER is not configured 2025-07-21 16:40:56 -04:00
rdma IB: Extend UVERBS_METHOD_REG_MR to get DMAH 2025-07-23 01:42:11 -04:00
rv rv: Retry when da monitor detects race conditions 2025-07-28 16:47:34 -04:00
scsi Merge patch series "libsas cleanups" 2025-07-25 09:05:23 -04:00
soc This is the usual collection of primarily clk driver updates. The big part of 2025-07-31 13:36:27 -07:00
sound ASoC: Fixes for v6.17 2025-08-21 09:02:28 +02:00
target scsi: target: core: Constify enabled() in struct target_opcode_descriptor 2025-05-20 22:11:03 -04:00
trace ipv4: Convert ->flowi4_tos to dscp_t. 2025-08-26 17:34:31 -07:00
uapi netfilter pull request nf-next-25-09-11 2025-09-12 17:06:25 -07:00
ufs Merge patch series "ufs: ufs-qcom: Align programming sequence as per HW spec" 2025-07-14 21:00:41 -04:00
vdso vdso/gettimeofday: Add support for auxiliary clocks 2025-07-18 14:09:39 +02:00
video video: Make global edid_info depend on CONFIG_FIRMWARE_EDID 2025-06-16 11:00:29 +02:00
xen Significant patch series in this pull request: 2025-08-03 16:23:09 -07:00
Kbuild drm: ensure drm headers are self-contained and pass kernel-doc 2025-02-12 10:44:43 +02:00