issue处理注意事项:
1. 提交正常修复分支的修复PR时,必须关联当前issue,否则无法关闭当前issue;
2. 模板内容需要填写完整, 无论是受影响或者不受影响都需要填写完整内容;
3. 以下为模板中需要填写完整的内容, 请复制到评论区回复;
注: 内容的关键词(影响性分析说明, openEuler评分, 受影响版本排查(受影响/不受影响), 修复是否涉及abi变化(是/否), 原因说明)不能省略,省略后cve-manager将无法正常解析填写内容.
影响性分析说明:
openEuler评分: (评分和向量)
受影响版本排查(受影响/不受影响):
1.master(6.18.40):
2.openEuler-20.03-LTS-SP4(4.19.90):
3.openEuler-22.03-LTS-SP4(5.10.0):
4.openEuler-24.03-LTS-Next(6.6.0):
5.openEuler-24.03-LTS-SP1(6.6.0):
6.openEuler-24.03-LTS-SP3(6.6.0):
7.openEuler-24.03-LTS-SP4(6.6.0):
修复是否涉及abi变化(是/否)：
1.master(6.18.40):
2.openEuler-20.03-LTS-SP4(4.19.90):
3.openEuler-22.03-LTS-SP4(5.10.0):
4.openEuler-24.03-LTS-Next(6.6.0):
5.openEuler-24.03-LTS-SP1(6.6.0):
6.openEuler-24.03-LTS-SP3(6.6.0):
7.openEuler-24.03-LTS-SP4(6.6.0):
原因说明：
1.master(6.18.40):
2.openEuler-20.03-LTS-SP4(4.19.90):
3.openEuler-22.03-LTS-SP4(5.10.0):
4.openEuler-24.03-LTS-Next(6.6.0):
5.openEuler-24.03-LTS-SP1(6.6.0):
6.openEuler-24.03-LTS-SP3(6.6.0):
7.openEuler-24.03-LTS-SP4(6.6.0):
原因说明填写请参考下方表格(注意：版本是否受影响和版本的原因说明必须对应，例如master版本分支受影响，那原因说明只能是受影响对应的原因之一!)：
分支状态 |
原因说明 | 使用场景 |
|---|---|---|
| 受影响 | 正常修复 | 受影响且需要修复（包含升级版本修复）的漏洞； 受影响且已经修复的漏洞（历史修复PR也需要关联issue）； 若因特殊原因无法修复，应修改原因说明为【不修复-特殊原因】，并在安委会备案相关情况。 |
| 受影响 | 漏洞仍在分析中 | 已关注到相关漏洞，正在处理，未明确漏洞影响和修复方案。 |
| 受影响 | 暂不修复-暂无解决方案或补丁 | 当前没有可用的修复或补救措施。【影响性分析】中应包含有关为什么没有修复或补救措施的详细说明、上游相关PR等。 |
| 受影响 | 不修复-超出修复范围 | 没有漏洞的修复计划。当版本停维、软件包宣布生命周期终止或弃用使用。 |
| 受影响 | 不修复-特殊原因导致不再修复 | 如存在其他特殊情况不修复相关漏洞，或评估后无法升级修复，应在openEuler社区安全委员会例会进行说明备案。 【影响性分析】中应包含不发布修复的详细说明、特殊情况还应有安委会会议纪要。 |
| 不受影响 | 不受影响-组件不存在 | 软件不受影响，因为易受攻击的组件不在产品中。 |
| 不受影响 | 不受影响-已有内置的内联控制或缓解措施 | 内置的内联控制或缓解措施可防止攻击者利用漏洞 |
| 不受影响 | 不受影响-漏洞代码不能被攻击者触发 | 易受攻击的组件存在，并且该组件包含易受攻击的代码。但是，易受攻击的代码的使用方式使得攻击者无法进行任何预期的攻击。 |
| 不受影响 | 不受影响-漏洞代码不在执行路径 | 易受影响的代码在执行过程中不可访问，包括产品的非预期状态。产品不使用也不执行的组件。 |
| 不受影响 | 不受影响-漏洞代码不存在 | 产品不受影响，因为漏洞背后的代码在产品中不存在。与component_not_present不同的是，有问题的组件存在，但由于某种原因（例如安全的编译器选项）使漏洞的特定代码不存在于组件中。 |
issue处理具体操作请参考:
https://atomgit.com/openeuler/cve-manager/blob/master/cve-vulner-manager/doc/md/manual.md
pr关联issue具体操作请参考:
https://docs.atomgit.com/docs/help/home/org_project/pullrequests/pr-related-issue
检测到新建 CVE Issue，CVE-ID: CVE-2026-74691。
你可以通过以下两种方式使用 CVE 修复服务：
-
分支影响分析
- 在本 Issue 下评论：
/analysis_branches - 系统将自动分析各个分支是否受影响，并在本 Issue 下给出表格形式的分析结果；
- 在本 Issue 下评论：
-
在指定分支上创建 PR（修复提交）
- 在本 Issue 下按如下格式评论（
branch、name、email为必填字段，每个字段单独一行）：
- 在本 Issue 下按如下格式评论（
/create_pr
branch: OLK-6.6 OLK-5.10
name: dev
email: dev@devstation.com
backport-engine: opencode
- 说明：
branch: 需要创建 PR 的目标分支列表，例如OLK-6.6 OLK-5.10；name: PR 签名人姓名（如责任人或安全负责人）；email: PR 签名人邮箱；backport-engine（可选）：可选portgpt、mystique或opencode；不填时默认使用portgpt，建议使用opencode；
系统会根据上述信息自动在指定分支上创建 PR，并在本 Issue 下反馈结果。
provided by DevStation.
说明：补丁链接仅供初步排查参考，实际可用性请人工再次确认，补丁下载验证可使用CVE补丁工具。
若补丁不准确，烦请在此issue下评论 '/report-patch 参考网址 补丁链接1,补丁链接2' 反馈正确信息，便于我们不断优化工具，不胜感激。
如 /report-patch https://security-tracker.debian.org/tracker/CVE-2021-3997 https://github.com/systemd/systemd/commit/5b1cf7a9be37e20133c0208005274ce4a5b5c6a1
Welcome To openEuler Community
Hey @openeuler-ci-bot , thanks for your contribution to the community.
Bot Usage Manual
I'm the Bot here serving you. You can find the instructions on how to interact with me at Here . That means you can comment below every pull request or issue to trigger Bot Commands.
Contact Guide
If you have any questions, please contact the SIG: Kernel ,
and any of the maintainers: @hanjunguo, @oekernel, @sanglipeng, @wkfxxx, @zeng_zhaorong ,
and any of the committers: @CTC-XiboWang, @Frank_Sae, @GoGo_phytium, @GongLei-, @LiuYongQiang0816, @SuperSix173, @Tankll2021, @TrueAI, @YiweiZ, @allen-shi, @baratta, @bibo_mao, @caixu-blue, @chen-jun-hw, @chenjiesong, @chenjunxin1992, @chenke2026, @chiqijun, @chriszjh, @duanqiangwen, @eingesch, @fangfeng123, @fanghaiqinghw, @gang_he, @gaojuxin09, @gouhao2022, @guohaocs2c, @guzitao, @hanjunguo, @hanliyang, @hellotcc, @henryze, @hewanhan, @hjx_gitff, @hongwu-wang, @htforge, @hu-chunzhi, @hunan4222, @jackknight, @jerry_lilijun, @jiayi0118, @junlong-zheng, @juntianlinux, @kailiu42, @kaitiandu, @kazero00, @kevinzhu1, @kile2009, @klmengkd, @koishimind, @kongzizaixian, @kylin-mayukun, @leoliu-oc, @li-huisong, @linan888, @linyunsheng, @liulongfang, @liyihang0226, @lostway1, @lujialin2, @mao-hongbo, @markyuan4ta21, @mawupeng, @mingqian218472, @mingrui-liu, @mufengyan, @pigalsofine, @robinorg, @rock_hw, @sanglipeng, @shu-shengming, @shuaijiakun, @sming56_admin, @stavewu, @stkid, @sun_nanyong, @wangboe2022, @wanghang73, @wenzhiwei11, @whoisxxx, @wkfxxx, @woqidaideshi, @wsoydl, @xingmz1, @xukuohai, @yeweihua999, @ygn-ndwd-official, @yonghu_4dc5, @young-sun, @yubo-liu1, @yuehaibing_planb, @yuzenghui1, @zhang-changzhong, @zhangyi089, @zhujianwei001, @zichengqu, @zouyipeng, @zqiao216 .
CVE-2026-74691
影响性分析说明：
In the Linux kernel, the following vulnerability has been resolved:
net: thunderbolt: Tear down DMA paths before stopping the rings
tbnet_tear_down() stops both rings and frees their frame buffers before
calling tb_xdomain_disable_paths(). tb_ring_stop() zeroes the ring's
descriptor base and tbnet_free_buffers() unmaps and frees the pages the
frames sit in, so by the time __tb_path_deactivate_hop() polls the hop's
'pending' bit, anything still in flight has nowhere to drain to.
The teardown sequence has been in this order since the driver was added.
The setup path has not: commit ff7cd07f3064 ("net: thunderbolt: Enable
DMA paths only after rings are enabled") moved the path enable to the end
of tbnet_connected_work() and documented why:
/* Both logins successful so enable the rings, high-speed DMA
* paths and start the network device queue.
*
* Note we enable the DMA paths last to make sure we have primed
* the Rx ring before any incoming packets are allowed to
* arrive.
*/
Teardown was never updated to match, so the rings and the paths now come
down in the same order they go up instead of in reverse.
On an ASMedia ASM4242 host router the 'pending' bit then never clears:
every teardown burns the full 500 ms timeout and
__tb_path_deactivate_hop() returns -ETIMEDOUT. Raising the timeout to
5 s does not help, so the hop is not slow to drain, it never drains
at all.
The failure is invisible above the thunderbolt core.
__tb_path_deactivate_hops() is void and only calls tb_port_warn();
tb_path_deactivate(), tb_tunnel_deactivate() and
__tb_disconnect_xdomain_paths() are void as well, and
tb_disconnect_xdomain_paths() ends in an unconditional "return 0". So
tb_xdomain_disable_paths() reports success and the netdev_warn() below
it never fires. Repeated teardowns eventually take the XDomain control
channel down, after which the peer node is gone and only a power cycle
brings the controller back.
Deactivating the paths first fixes it. Measured with kretprobes on a
stock v6.17 tree with no other patches applied, on a link that was up
and had just carried traffic:
before: __tb_path_deactivate_hop() returns 0 for the first hop, then
-ETIMEDOUT for the second 500335 us later
after: 0 for both, 525 us apart
Alternating the two orderings ABBA over three load levels, four
teardowns per arm: every teardown failed before the change (21 of 21
that ran), none failed after (0 of 24). The before arms ran short
because the link died partway through. The same split shows up when
the interface is enslaved to a bond instead of just brought down, which
is how I ran into this in the first place. Throughput and latency after
the change are unchanged.
Hosts whose routers drain the hop despite the stale descriptor base see
no functional difference, since the paths end up deactivated either way.
The Linux kernel CVE team has assigned CVE-2026-74691 to this issue.
openEuler评分:(评分和向量)
3.9
CVSS:3.1/AV:L/AC:H/PR:H/UI:N/S:U/C:L/I:L/A:L
受影响版本排查(受影响/不受影响):
1.master:不受影响
2.openEuler-20.03-LTS-SP4:受影响
3.openEuler-22.03-LTS-SP3:受影响
4.openEuler-22.03-LTS-SP4:受影响
5.openEuler-24.03-LTS:受影响
6.openEuler-24.03-LTS-Next:不受影响
7.openEuler-24.03-LTS-SP1:受影响
8.openEuler-24.03-LTS-SP2:受影响
9.openEuler-24.03-LTS-SP3:受影响
10.openEuler-24.03-LTS-SP4:受影响
修复是否涉及abi变化(是/否)：
1.master:否
2.openEuler-20.03-LTS-SP4:否
3.openEuler-22.03-LTS-SP3:否
4.openEuler-22.03-LTS-SP4:否
5.openEuler-24.03-LTS:否
6.openEuler-24.03-LTS-Next:否
7.openEuler-24.03-LTS-SP1:否
8.openEuler-24.03-LTS-SP2:否
9.openEuler-24.03-LTS-SP3:否
10.openEuler-24.03-LTS-SP4:否
原因说明：
1.master:不受影响-漏洞代码不能被攻击者触发
2.openEuler-20.03-LTS-SP4:暂不修复-漏洞仍在分析中
3.openEuler-22.03-LTS-SP3:不修复-超出修复范围
4.openEuler-22.03-LTS-SP4:暂不修复-漏洞仍在分析中
5.openEuler-24.03-LTS:不修复-超出修复范围
6.openEuler-24.03-LTS-Next:不受影响-漏洞代码不能被攻击者触发
7.openEuler-24.03-LTS-SP1:暂不修复-漏洞仍在分析中
8.openEuler-24.03-LTS-SP2:不修复-超出修复范围
9.openEuler-24.03-LTS-SP3:暂不修复-漏洞仍在分析中
10.openEuler-24.03-LTS-SP4:暂不修复-漏洞仍在分析中
@LiuYongQiang0816 经过 cve-manager 解析, 已分析的内容如下表所示:
| 状态 | 分析项目 | 内容 |
|---|---|---|
| 已分析 | 1.影响性分析说明 | In the Linux kernel, the following vulnerability has been resolved:net: thunderbolt: Tear down DMA paths before stopping the ringstbnet_tear_down() stops both rings and frees their frame buffers beforecalling tb_xdomain_disable_paths(). tb_ring_stop() zeroes the ring'sdescriptor base and tbnet_free_buffers() unmaps and frees the pages theframes sit in, so by the time __tb_path_deactivate_hop() polls the hop's'pending' bit, anything still in flight has nowhere to drain to.The teardown sequence has been in this order since the driver was added.The setup path has not: commit ff7cd07f3064 ("net: thunderbolt: EnableDMA paths only after rings are enabled") moved the path enable to the endof tbnet_connected_work() and documented why:/* Both logins successful so enable the rings, high-speed DMA * paths and start the network device queue. * * Note we enable the DMA paths last to make sure we have primed * the Rx ring before any incoming packets are allowed to * arrive. */Teardown was never updated to match, so the rings and the paths now comedown in the same order they go up instead of in reverse.On an ASMedia ASM4242 host router the 'pending' bit then never clears:every teardown burns the full 500 ms timeout and__tb_path_deactivate_hop() returns -ETIMEDOUT. Raising the timeout to5 s does not help, so the hop is not slow to drain, it never drainsat all.The failure is invisible above the thunderbolt core.__tb_path_deactivate_hops() is void and only calls tb_port_warn();tb_path_deactivate(), tb_tunnel_deactivate() and__tb_disconnect_xdomain_paths() are void as well, andtb_disconnect_xdomain_paths() ends in an unconditional "return 0". Sotb_xdomain_disable_paths() reports success and the netdev_warn() belowit never fires. Repeated teardowns eventually take the XDomain controlchannel down, after which the peer node is gone and only a power cyclebrings the controller back.Deactivating the paths first fixes it. Measured with kretprobes on astock v6.17 tree with no other patches applied, on a link that was upand had just carried traffic: before: __tb_path_deactivate_hop() returns 0 for the first hop, then -ETIMEDOUT for the second 500335 us later after: 0 for both, 525 us apartAlternating the two orderings ABBA over three load levels, fourteardowns per arm: every teardown failed before the change (21 of 21that ran), none failed after (0 of 24). The before arms ran shortbecause the link died partway through. The same split shows up whenthe interface is enslaved to a bond instead of just brought down, whichis how I ran into this in the first place. Throughput and latency afterthe change are unchanged.Hosts whose routers drain the hop despite the stale descriptor base seeno functional difference, since the paths end up deactivated either way.The Linux kernel CVE team has assigned CVE-2026-74691 to this issue. |
| 已分析 | 2.openEulerScore | 3.9 |
| 已分析 | 3.openEulerVector | AV:L/AC:H/PR:H/UI:N/S:U/C:L/I:L/A:L |
| 已分析 | 4.受影响版本排查 | master:不受影响,openEuler-20.03-LTS-SP4:受影响,openEuler-22.03-LTS-SP4:受影响,openEuler-24.03-LTS-Next:不受影响,openEuler-24.03-LTS-SP1:受影响,openEuler-24.03-LTS-SP3:受影响,openEuler-24.03-LTS-SP4:受影响,openEuler-22.03-LTS-SP3:受影响,openEuler-24.03-LTS:受影响,openEuler-24.03-LTS-SP2:受影响 |
| 已分析 | 5.是否涉及abi变化 | master:否,openEuler-20.03-LTS-SP4:否,openEuler-22.03-LTS-SP4:否,openEuler-24.03-LTS-Next:否,openEuler-24.03-LTS-SP1:否,openEuler-24.03-LTS-SP3:否,openEuler-24.03-LTS-SP4:否,openEuler-22.03-LTS-SP3:否,openEuler-24.03-LTS:否,openEuler-24.03-LTS-SP2:否 |
| 已分析 | 6.原因说明 | master:不受影响-漏洞代码不能被攻击者触发,openEuler-20.03-LTS-SP4:漏洞仍在分析中,openEuler-22.03-LTS-SP4:漏洞仍在分析中,openEuler-24.03-LTS-Next:不受影响-漏洞代码不能被攻击者触发,openEuler-24.03-LTS-SP1:漏洞仍在分析中,openEuler-24.03-LTS-SP3:漏洞仍在分析中,openEuler-24.03-LTS-SP4:漏洞仍在分析中,openEuler-22.03-LTS-SP3:不修复-超出修复范围,openEuler-24.03-LTS:不修复-超出修复范围,openEuler-24.03-LTS-SP2:不修复-超出修复范围 |
请确认分析内容的准确性, 确认无误后, 您可以进行后续步骤, 否则您可以继续分析.
openEuler评分:(评分和向量)
8.8
CVSS:3.1/AV:A/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
@LiuYongQiang0816 经过 cve-manager 解析, 已分析的内容如下表所示:
| 状态 | 分析项目 | 内容 |
|---|---|---|
| 已分析 | 1.影响性分析说明 | In the Linux kernel, the following vulnerability has been resolved:net: thunderbolt: Tear down DMA paths before stopping the ringstbnet_tear_down() stops both rings and frees their frame buffers beforecalling tb_xdomain_disable_paths(). tb_ring_stop() zeroes the ring'sdescriptor base and tbnet_free_buffers() unmaps and frees the pages theframes sit in, so by the time __tb_path_deactivate_hop() polls the hop's'pending' bit, anything still in flight has nowhere to drain to.The teardown sequence has been in this order since the driver was added.The setup path has not: commit ff7cd07f3064 ("net: thunderbolt: EnableDMA paths only after rings are enabled") moved the path enable to the endof tbnet_connected_work() and documented why:/* Both logins successful so enable the rings, high-speed DMA * paths and start the network device queue. * * Note we enable the DMA paths last to make sure we have primed * the Rx ring before any incoming packets are allowed to * arrive. */Teardown was never updated to match, so the rings and the paths now comedown in the same order they go up instead of in reverse.On an ASMedia ASM4242 host router the 'pending' bit then never clears:every teardown burns the full 500 ms timeout and__tb_path_deactivate_hop() returns -ETIMEDOUT. Raising the timeout to5 s does not help, so the hop is not slow to drain, it never drainsat all.The failure is invisible above the thunderbolt core.__tb_path_deactivate_hops() is void and only calls tb_port_warn();tb_path_deactivate(), tb_tunnel_deactivate() and__tb_disconnect_xdomain_paths() are void as well, andtb_disconnect_xdomain_paths() ends in an unconditional "return 0". Sotb_xdomain_disable_paths() reports success and the netdev_warn() belowit never fires. Repeated teardowns eventually take the XDomain controlchannel down, after which the peer node is gone and only a power cyclebrings the controller back.Deactivating the paths first fixes it. Measured with kretprobes on astock v6.17 tree with no other patches applied, on a link that was upand had just carried traffic: before: __tb_path_deactivate_hop() returns 0 for the first hop, then -ETIMEDOUT for the second 500335 us later after: 0 for both, 525 us apartAlternating the two orderings ABBA over three load levels, fourteardowns per arm: every teardown failed before the change (21 of 21that ran), none failed after (0 of 24). The before arms ran shortbecause the link died partway through. The same split shows up whenthe interface is enslaved to a bond instead of just brought down, whichis how I ran into this in the first place. Throughput and latency afterthe change are unchanged.Hosts whose routers drain the hop despite the stale descriptor base seeno functional difference, since the paths end up deactivated either way.The Linux kernel CVE team has assigned CVE-2026-74691 to this issue. |
| 已分析 | 2.openEulerScore | 8.8 |
| 已分析 | 3.openEulerVector | AV:A/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H |
| 已分析 | 4.受影响版本排查 | master:不受影响,openEuler-20.03-LTS-SP4:受影响,openEuler-22.03-LTS-SP4:受影响,openEuler-24.03-LTS-Next:不受影响,openEuler-24.03-LTS-SP1:受影响,openEuler-24.03-LTS-SP3:受影响,openEuler-24.03-LTS-SP4:受影响,openEuler-22.03-LTS-SP3:受影响,openEuler-24.03-LTS:受影响,openEuler-24.03-LTS-SP2:受影响 |
| 已分析 | 5.是否涉及abi变化 | master:否,openEuler-20.03-LTS-SP4:否,openEuler-22.03-LTS-SP4:否,openEuler-24.03-LTS-Next:否,openEuler-24.03-LTS-SP1:否,openEuler-24.03-LTS-SP3:否,openEuler-24.03-LTS-SP4:否,openEuler-22.03-LTS-SP3:否,openEuler-24.03-LTS:否,openEuler-24.03-LTS-SP2:否 |
| 已分析 | 6.原因说明 | master:不受影响-漏洞代码不能被攻击者触发,openEuler-20.03-LTS-SP4:漏洞仍在分析中,openEuler-22.03-LTS-SP4:漏洞仍在分析中,openEuler-24.03-LTS-Next:不受影响-漏洞代码不能被攻击者触发,openEuler-24.03-LTS-SP1:漏洞仍在分析中,openEuler-24.03-LTS-SP3:漏洞仍在分析中,openEuler-24.03-LTS-SP4:漏洞仍在分析中,openEuler-22.03-LTS-SP3:不修复-超出修复范围,openEuler-24.03-LTS:不修复-超出修复范围,openEuler-24.03-LTS-SP2:不修复-超出修复范围 |
请确认分析内容的准确性, 确认无误后, 您可以进行后续步骤, 否则您可以继续分析.
/create_pr
branch: OLK-5.10
name: Xi Qi
email: xiqi2@huawei.com
已触发 CVE 修复服务，CVE-ID: CVE-2026-74691。
provided by DevStation.
已完成 CVE 修复全流程汇总，CVE-ID: CVE-2026-74691。
关联修复提交信息：net: thunderbolt: Tear down DMA paths before stopping the rings
| 目标分支 | 是否受影响 | 适配状态 | 是否存在冲突 | PR 链接 |
|---|---|---|---|---|
| OLK-6.6 | Affected | success | No | - |
| OLK-5.10 | Affected | success | Yes | 查看PR |
| openEuler-1.0-LTS | Affected | success | Yes | - |
provided by DevStation.
/create_pr
branch: OLK-6.6
name: Xi Qi
email: xiqi2@huawei.com
已触发 CVE 修复服务，CVE-ID: CVE-2026-74691。
provided by DevStation.
已完成 CVE 修复全流程汇总，CVE-ID: CVE-2026-74691。
关联修复提交信息：net: thunderbolt: Tear down DMA paths before stopping the rings
| 目标分支 | 是否受影响 | 适配状态 | 是否存在冲突 | PR 链接 |
|---|---|---|---|---|
| OLK-6.6 | Affected | success | No | - |
| OLK-5.10 | Affected | success | Yes | 查看PR |
| openEuler-1.0-LTS | Affected | success | Yes | - |
provided by DevStation.
/create_pr
branch: OLK-6.6
name: Xi Qi
email: xiqi2@huawei.com
已触发 CVE 修复服务，CVE-ID: CVE-2026-74691。
provided by DevStation.
CVE-2026-74691
影响性分析说明：
In the Linux kernel, the following vulnerability has been resolved:
net: thunderbolt: Tear down DMA paths before stopping the rings
tbnet_tear_down() stops both rings and frees their frame buffers before
calling tb_xdomain_disable_paths(). tb_ring_stop() zeroes the ring's
descriptor base and tbnet_free_buffers() unmaps and frees the pages the
frames sit in, so by the time __tb_path_deactivate_hop() polls the hop's
'pending' bit, anything still in flight has nowhere to drain to.
The teardown sequence has been in this order since the driver was added.
The setup path has not: commit ff7cd07f3064 ("net: thunderbolt: Enable
DMA paths only after rings are enabled") moved the path enable to the end
of tbnet_connected_work() and documented why:
/* Both logins successful so enable the rings, high-speed DMA
* paths and start the network device queue.
*
* Note we enable the DMA paths last to make sure we have primed
* the Rx ring before any incoming packets are allowed to
* arrive.
*/
Teardown was never updated to match, so the rings and the paths now come
down in the same order they go up instead of in reverse.
On an ASMedia ASM4242 host router the 'pending' bit then never clears:
every teardown burns the full 500 ms timeout and
__tb_path_deactivate_hop() returns -ETIMEDOUT. Raising the timeout to
5 s does not help, so the hop is not slow to drain, it never drains
at all.
The failure is invisible above the thunderbolt core.
__tb_path_deactivate_hops() is void and only calls tb_port_warn();
tb_path_deactivate(), tb_tunnel_deactivate() and
__tb_disconnect_xdomain_paths() are void as well, and
tb_disconnect_xdomain_paths() ends in an unconditional "return 0". So
tb_xdomain_disable_paths() reports success and the netdev_warn() below
it never fires. Repeated teardowns eventually take the XDomain control
channel down, after which the peer node is gone and only a power cycle
brings the controller back.
Deactivating the paths first fixes it. Measured with kretprobes on a
stock v6.17 tree with no other patches applied, on a link that was up
and had just carried traffic:
before: __tb_path_deactivate_hop() returns 0 for the first hop, then
-ETIMEDOUT for the second 500335 us later
after: 0 for both, 525 us apart
Alternating the two orderings ABBA over three load levels, four
teardowns per arm: every teardown failed before the change (21 of 21
that ran), none failed after (0 of 24). The before arms ran short
because the link died partway through. The same split shows up when
the interface is enslaved to a bond instead of just brought down, which
is how I ran into this in the first place. Throughput and latency after
the change are unchanged.
Hosts whose routers drain the hop despite the stale descriptor base see
no functional difference, since the paths end up deactivated either way.
The Linux kernel CVE team has assigned CVE-2026-74691 to this issue.
openEuler评分:(评分和向量)
8.8
CVSS:3.1/AV:A/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
受影响版本排查(受影响/不受影响):
1.master:不受影响
2.openEuler-20.03-LTS-SP4:受影响
3.openEuler-22.03-LTS-SP3:受影响
4.openEuler-22.03-LTS-SP4:受影响
5.openEuler-24.03-LTS:受影响
6.openEuler-24.03-LTS-Next:不受影响
7.openEuler-24.03-LTS-SP1:受影响
8.openEuler-24.03-LTS-SP2:受影响
9.openEuler-24.03-LTS-SP3:受影响
10.openEuler-24.03-LTS-SP4:受影响
修复是否涉及abi变化(是/否)：
1.master:否
2.openEuler-20.03-LTS-SP4:否
3.openEuler-22.03-LTS-SP3:否
4.openEuler-22.03-LTS-SP4:否
5.openEuler-24.03-LTS:否
6.openEuler-24.03-LTS-Next:否
7.openEuler-24.03-LTS-SP1:否
8.openEuler-24.03-LTS-SP2:否
9.openEuler-24.03-LTS-SP3:否
10.openEuler-24.03-LTS-SP4:否
原因说明：
1.master:不受影响-漏洞代码不能被攻击者触发
2.openEuler-20.03-LTS-SP4:暂不修复-暂无解决方案或补丁
3.openEuler-22.03-LTS-SP3:不修复-超出修复范围
4.openEuler-22.03-LTS-SP4:暂不修复-漏洞仍在分析中
5.openEuler-24.03-LTS:不修复-超出修复范围
6.openEuler-24.03-LTS-Next:不受影响-漏洞代码不能被攻击者触发
7.openEuler-24.03-LTS-SP1:正常修复
8.openEuler-24.03-LTS-SP2:不修复-超出修复范围
9.openEuler-24.03-LTS-SP3:正常修复
10.openEuler-24.03-LTS-SP4:正常修复
@LiuYongQiang0816 经过 cve-manager 解析, 已分析的内容如下表所示:
| 状态 | 分析项目 | 内容 |
|---|---|---|
| 已分析 | 1.影响性分析说明 | In the Linux kernel, the following vulnerability has been resolved:net: thunderbolt: Tear down DMA paths before stopping the ringstbnet_tear_down() stops both rings and frees their frame buffers beforecalling tb_xdomain_disable_paths(). tb_ring_stop() zeroes the ring'sdescriptor base and tbnet_free_buffers() unmaps and frees the pages theframes sit in, so by the time __tb_path_deactivate_hop() polls the hop's'pending' bit, anything still in flight has nowhere to drain to.The teardown sequence has been in this order since the driver was added.The setup path has not: commit ff7cd07f3064 ("net: thunderbolt: EnableDMA paths only after rings are enabled") moved the path enable to the endof tbnet_connected_work() and documented why:/* Both logins successful so enable the rings, high-speed DMA * paths and start the network device queue. * * Note we enable the DMA paths last to make sure we have primed * the Rx ring before any incoming packets are allowed to * arrive. */Teardown was never updated to match, so the rings and the paths now comedown in the same order they go up instead of in reverse.On an ASMedia ASM4242 host router the 'pending' bit then never clears:every teardown burns the full 500 ms timeout and__tb_path_deactivate_hop() returns -ETIMEDOUT. Raising the timeout to5 s does not help, so the hop is not slow to drain, it never drainsat all.The failure is invisible above the thunderbolt core.__tb_path_deactivate_hops() is void and only calls tb_port_warn();tb_path_deactivate(), tb_tunnel_deactivate() and__tb_disconnect_xdomain_paths() are void as well, andtb_disconnect_xdomain_paths() ends in an unconditional "return 0". Sotb_xdomain_disable_paths() reports success and the netdev_warn() belowit never fires. Repeated teardowns eventually take the XDomain controlchannel down, after which the peer node is gone and only a power cyclebrings the controller back.Deactivating the paths first fixes it. Measured with kretprobes on astock v6.17 tree with no other patches applied, on a link that was upand had just carried traffic: before: __tb_path_deactivate_hop() returns 0 for the first hop, then -ETIMEDOUT for the second 500335 us later after: 0 for both, 525 us apartAlternating the two orderings ABBA over three load levels, fourteardowns per arm: every teardown failed before the change (21 of 21that ran), none failed after (0 of 24). The before arms ran shortbecause the link died partway through. The same split shows up whenthe interface is enslaved to a bond instead of just brought down, whichis how I ran into this in the first place. Throughput and latency afterthe change are unchanged.Hosts whose routers drain the hop despite the stale descriptor base seeno functional difference, since the paths end up deactivated either way.The Linux kernel CVE team has assigned CVE-2026-74691 to this issue. |
| 已分析 | 2.openEulerScore | 8.8 |
| 已分析 | 3.openEulerVector | AV:A/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H |
| 已分析 | 4.受影响版本排查 | master:不受影响,openEuler-20.03-LTS-SP4:受影响,openEuler-22.03-LTS-SP4:受影响,openEuler-24.03-LTS-Next:不受影响,openEuler-24.03-LTS-SP1:受影响,openEuler-24.03-LTS-SP3:受影响,openEuler-24.03-LTS-SP4:受影响,openEuler-22.03-LTS-SP3:受影响,openEuler-24.03-LTS:受影响,openEuler-24.03-LTS-SP2:受影响 |
| 已分析 | 5.是否涉及abi变化 | master:否,openEuler-20.03-LTS-SP4:否,openEuler-22.03-LTS-SP4:否,openEuler-24.03-LTS-Next:否,openEuler-24.03-LTS-SP1:否,openEuler-24.03-LTS-SP3:否,openEuler-24.03-LTS-SP4:否,openEuler-22.03-LTS-SP3:否,openEuler-24.03-LTS:否,openEuler-24.03-LTS-SP2:否 |
| 已分析 | 6.原因说明 | master:不受影响-漏洞代码不能被攻击者触发,openEuler-20.03-LTS-SP4:暂不修复-暂无解决方案或补丁,openEuler-22.03-LTS-SP4:漏洞仍在分析中,openEuler-24.03-LTS-Next:不受影响-漏洞代码不能被攻击者触发,openEuler-24.03-LTS-SP1:正常修复,openEuler-24.03-LTS-SP3:正常修复,openEuler-24.03-LTS-SP4:正常修复,openEuler-22.03-LTS-SP3:不修复-超出修复范围,openEuler-24.03-LTS:不修复-超出修复范围,openEuler-24.03-LTS-SP2:不修复-超出修复范围 |
请确认分析内容的准确性, 确认无误后, 您可以进行后续步骤, 否则您可以继续分析.
经过cve-manager解析，部分分支PR未合入，如红色字体所示：
原因说明：
1.master:不受影响-漏洞代码不能被攻击者触发
2.openEuler-20.03-LTS-SP4:暂不修复-暂无解决方案或补丁
3.openEuler-22.03-LTS-SP4:漏洞仍在分析中
4.openEuler-24.03-LTS-Next:不受影响-漏洞代码不能被攻击者触发
5.openEuler-24.03-LTS-SP1:正常修复(PR未合入)
6.openEuler-24.03-LTS-SP3:正常修复(PR未合入)
7.openEuler-24.03-LTS-SP4:正常修复(PR未合入)
8.openEuler-22.03-LTS-SP3:不修复-超出修复范围
9.openEuler-24.03-LTS:不修复-超出修复范围
10.openEuler-24.03-LTS-SP2:不修复-超出修复范围
CVE-2026-74691
影响性分析说明：
In the Linux kernel, the following vulnerability has been resolved:
net: thunderbolt: Tear down DMA paths before stopping the rings
tbnet_tear_down() stops both rings and frees their frame buffers before
calling tb_xdomain_disable_paths(). tb_ring_stop() zeroes the ring's
descriptor base and tbnet_free_buffers() unmaps and frees the pages the
frames sit in, so by the time __tb_path_deactivate_hop() polls the hop's
'pending' bit, anything still in flight has nowhere to drain to.
The teardown sequence has been in this order since the driver was added.
The setup path has not: commit ff7cd07f3064 ("net: thunderbolt: Enable
DMA paths only after rings are enabled") moved the path enable to the end
of tbnet_connected_work() and documented why:
/* Both logins successful so enable the rings, high-speed DMA
* paths and start the network device queue.
*
* Note we enable the DMA paths last to make sure we have primed
* the Rx ring before any incoming packets are allowed to
* arrive.
*/
Teardown was never updated to match, so the rings and the paths now come
down in the same order they go up instead of in reverse.
On an ASMedia ASM4242 host router the 'pending' bit then never clears:
every teardown burns the full 500 ms timeout and
__tb_path_deactivate_hop() returns -ETIMEDOUT. Raising the timeout to
5 s does not help, so the hop is not slow to drain, it never drains
at all.
The failure is invisible above the thunderbolt core.
__tb_path_deactivate_hops() is void and only calls tb_port_warn();
tb_path_deactivate(), tb_tunnel_deactivate() and
__tb_disconnect_xdomain_paths() are void as well, and
tb_disconnect_xdomain_paths() ends in an unconditional "return 0". So
tb_xdomain_disable_paths() reports success and the netdev_warn() below
it never fires. Repeated teardowns eventually take the XDomain control
channel down, after which the peer node is gone and only a power cycle
brings the controller back.
Deactivating the paths first fixes it. Measured with kretprobes on a
stock v6.17 tree with no other patches applied, on a link that was up
and had just carried traffic:
before: __tb_path_deactivate_hop() returns 0 for the first hop, then
-ETIMEDOUT for the second 500335 us later
after: 0 for both, 525 us apart
Alternating the two orderings ABBA over three load levels, four
teardowns per arm: every teardown failed before the change (21 of 21
that ran), none failed after (0 of 24). The before arms ran short
because the link died partway through. The same split shows up when
the interface is enslaved to a bond instead of just brought down, which
is how I ran into this in the first place. Throughput and latency after
the change are unchanged.
Hosts whose routers drain the hop despite the stale descriptor base see
no functional difference, since the paths end up deactivated either way.
The Linux kernel CVE team has assigned CVE-2026-74691 to this issue.
openEuler评分:(评分和向量)
8.8
CVSS:3.1/AV:A/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
受影响版本排查(受影响/不受影响):
1.master:不受影响
2.openEuler-20.03-LTS-SP4:受影响
3.openEuler-22.03-LTS-SP3:受影响
4.openEuler-22.03-LTS-SP4:受影响
5.openEuler-24.03-LTS:受影响
6.openEuler-24.03-LTS-Next:不受影响
7.openEuler-24.03-LTS-SP1:受影响
8.openEuler-24.03-LTS-SP2:受影响
9.openEuler-24.03-LTS-SP3:受影响
10.openEuler-24.03-LTS-SP4:受影响
修复是否涉及abi变化(是/否)：
1.master:否
2.openEuler-20.03-LTS-SP4:否
3.openEuler-22.03-LTS-SP3:否
4.openEuler-22.03-LTS-SP4:否
5.openEuler-24.03-LTS:否
6.openEuler-24.03-LTS-Next:否
7.openEuler-24.03-LTS-SP1:否
8.openEuler-24.03-LTS-SP2:否
9.openEuler-24.03-LTS-SP3:否
10.openEuler-24.03-LTS-SP4:否
原因说明：
1.master:不受影响-漏洞代码不能被攻击者触发
2.openEuler-20.03-LTS-SP4:暂不修复-暂无解决方案或补丁
3.openEuler-22.03-LTS-SP3:不修复-超出修复范围
4.openEuler-22.03-LTS-SP4:正常修复
5.openEuler-24.03-LTS:不修复-超出修复范围
6.openEuler-24.03-LTS-Next:不受影响-漏洞代码不能被攻击者触发
7.openEuler-24.03-LTS-SP1:正常修复
8.openEuler-24.03-LTS-SP2:不修复-超出修复范围
9.openEuler-24.03-LTS-SP3:正常修复
10.openEuler-24.03-LTS-SP4:正常修复
@LiuYongQiang0816 经过 cve-manager 解析, 已分析的内容如下表所示:
| 状态 | 分析项目 | 内容 |
|---|---|---|
| 已分析 | 1.影响性分析说明 | In the Linux kernel, the following vulnerability has been resolved:net: thunderbolt: Tear down DMA paths before stopping the ringstbnet_tear_down() stops both rings and frees their frame buffers beforecalling tb_xdomain_disable_paths(). tb_ring_stop() zeroes the ring'sdescriptor base and tbnet_free_buffers() unmaps and frees the pages theframes sit in, so by the time __tb_path_deactivate_hop() polls the hop's'pending' bit, anything still in flight has nowhere to drain to.The teardown sequence has been in this order since the driver was added.The setup path has not: commit ff7cd07f3064 ("net: thunderbolt: EnableDMA paths only after rings are enabled") moved the path enable to the endof tbnet_connected_work() and documented why:/* Both logins successful so enable the rings, high-speed DMA * paths and start the network device queue. * * Note we enable the DMA paths last to make sure we have primed * the Rx ring before any incoming packets are allowed to * arrive. */Teardown was never updated to match, so the rings and the paths now comedown in the same order they go up instead of in reverse.On an ASMedia ASM4242 host router the 'pending' bit then never clears:every teardown burns the full 500 ms timeout and__tb_path_deactivate_hop() returns -ETIMEDOUT. Raising the timeout to5 s does not help, so the hop is not slow to drain, it never drainsat all.The failure is invisible above the thunderbolt core.__tb_path_deactivate_hops() is void and only calls tb_port_warn();tb_path_deactivate(), tb_tunnel_deactivate() and__tb_disconnect_xdomain_paths() are void as well, andtb_disconnect_xdomain_paths() ends in an unconditional "return 0". Sotb_xdomain_disable_paths() reports success and the netdev_warn() belowit never fires. Repeated teardowns eventually take the XDomain controlchannel down, after which the peer node is gone and only a power cyclebrings the controller back.Deactivating the paths first fixes it. Measured with kretprobes on astock v6.17 tree with no other patches applied, on a link that was upand had just carried traffic: before: __tb_path_deactivate_hop() returns 0 for the first hop, then -ETIMEDOUT for the second 500335 us later after: 0 for both, 525 us apartAlternating the two orderings ABBA over three load levels, fourteardowns per arm: every teardown failed before the change (21 of 21that ran), none failed after (0 of 24). The before arms ran shortbecause the link died partway through. The same split shows up whenthe interface is enslaved to a bond instead of just brought down, whichis how I ran into this in the first place. Throughput and latency afterthe change are unchanged.Hosts whose routers drain the hop despite the stale descriptor base seeno functional difference, since the paths end up deactivated either way.The Linux kernel CVE team has assigned CVE-2026-74691 to this issue. |
| 已分析 | 2.openEulerScore | 8.8 |
| 已分析 | 3.openEulerVector | AV:A/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H |
| 已分析 | 4.受影响版本排查 | master:不受影响,openEuler-20.03-LTS-SP4:受影响,openEuler-22.03-LTS-SP4:受影响,openEuler-24.03-LTS-Next:不受影响,openEuler-24.03-LTS-SP1:受影响,openEuler-24.03-LTS-SP3:受影响,openEuler-24.03-LTS-SP4:受影响,openEuler-22.03-LTS-SP3:受影响,openEuler-24.03-LTS:受影响,openEuler-24.03-LTS-SP2:受影响 |
| 已分析 | 5.是否涉及abi变化 | master:否,openEuler-20.03-LTS-SP4:否,openEuler-22.03-LTS-SP4:否,openEuler-24.03-LTS-Next:否,openEuler-24.03-LTS-SP1:否,openEuler-24.03-LTS-SP3:否,openEuler-24.03-LTS-SP4:否,openEuler-22.03-LTS-SP3:否,openEuler-24.03-LTS:否,openEuler-24.03-LTS-SP2:否 |
| 已分析 | 6.原因说明 | master:不受影响-漏洞代码不能被攻击者触发,openEuler-20.03-LTS-SP4:暂不修复-暂无解决方案或补丁,openEuler-22.03-LTS-SP4:正常修复,openEuler-24.03-LTS-Next:不受影响-漏洞代码不能被攻击者触发,openEuler-24.03-LTS-SP1:正常修复,openEuler-24.03-LTS-SP3:正常修复,openEuler-24.03-LTS-SP4:正常修复,openEuler-22.03-LTS-SP3:不修复-超出修复范围,openEuler-24.03-LTS:不修复-超出修复范围,openEuler-24.03-LTS-SP2:不修复-超出修复范围 |
请确认分析内容的准确性, 确认无误后, 您可以进行后续步骤, 否则您可以继续分析.
经过cve-manager解析，部分分支PR未合入，如红色字体所示：
原因说明：
1.master:不受影响-漏洞代码不能被攻击者触发
2.openEuler-20.03-LTS-SP4:暂不修复-暂无解决方案或补丁
3.openEuler-22.03-LTS-SP4:正常修复(PR未合入)
4.openEuler-24.03-LTS-Next:不受影响-漏洞代码不能被攻击者触发
5.openEuler-24.03-LTS-SP1:正常修复(PR未合入)
6.openEuler-24.03-LTS-SP3:正常修复(PR未合入)
7.openEuler-24.03-LTS-SP4:正常修复(PR未合入)
8.openEuler-22.03-LTS-SP3:不修复-超出修复范围
9.openEuler-24.03-LTS:不修复-超出修复范围
10.openEuler-24.03-LTS-SP2:不修复-超出修复范围
一、漏洞信息
漏洞编号：CVE-2026-74691
漏洞归属组件：kernel
漏洞归属的版本：4.19.140,4.19.194,4.19.90,5.10.0,6.1.19,6.12.33,6.18.18,6.4.0,6.6.0
CVSS评分：
BaseScore：8.8 High
Vector：CVSS:3.1/AV:A/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
漏洞简述：
In the Linux kernel, the following vulnerability has been resolved:net: thunderbolt: Tear down DMA paths before stopping the ringstbnet_tear_down() stops both rings and frees their frame buffers beforecalling tb_xdomain_disable_paths(). tb_ring_stop() zeroes the ring sdescriptor base and tbnet_free_buffers() unmaps and frees the pages theframes sit in, so by the time __tb_path_deactivate_hop() polls the hop s pending bit, anything still in flight has nowhere to drain to.The teardown sequence has been in this order since the driver was added.The setup path has not: commit ff7cd07f3064 ( net: thunderbolt: EnableDMA paths only after rings are enabled ) moved the path enable to the endof tbnet_connected_work() and documented why: /* Both logins successful so enable the rings, high-speed DMA * paths and start the network device queue. * * Note we enable the DMA paths last to make sure we have primed * the Rx ring before any incoming packets are allowed to * arrive. */Teardown was never updated to match, so the rings and the paths now comedown in the same order they go up instead of in reverse.On an ASMedia ASM4242 host router the pending bit then never clears:every teardown burns the full 500 ms timeout and__tb_path_deactivate_hop() returns -ETIMEDOUT. Raising the timeout to5 s does not help, so the hop is not slow to drain, it never drainsat all.The failure is invisible above the thunderbolt core.__tb_path_deactivate_hops() is void and only calls tb_port_warn();tb_path_deactivate(), tb_tunnel_deactivate() and__tb_disconnect_xdomain_paths() are void as well, andtb_disconnect_xdomain_paths() ends in an unconditional return 0 . Sotb_xdomain_disable_paths() reports success and the netdev_warn() belowit never fires. Repeated teardowns eventually take the XDomain controlchannel down, after which the peer node is gone and only a power cyclebrings the controller back.Deactivating the paths first fixes it. Measured with kretprobes on astock v6.17 tree with no other patches applied, on a link that was upand had just carried traffic: before: __tb_path_deactivate_hop() returns 0 for the first hop, then -ETIMEDOUT for the second 500335 us later after: 0 for both, 525 us apartAlternating the two orderings ABBA over three load levels, fourteardowns per arm: every teardown failed before the change (21 of 21that ran), none failed after (0 of 24). The before arms ran shortbecause the link died partway through. The same split shows up whenthe interface is enslaved to a bond instead of just brought down, whichis how I ran into this in the first place. Throughput and latency afterthe change are unchanged.Hosts whose routers drain the hop despite the stale descriptor base seeno functional difference, since the paths end up deactivated either way.
漏洞公开时间：2026-08-23 00:16:43
漏洞创建时间：2026-08-23 00:24:26
漏洞详情参考链接：
https://nvd.nist.gov/vuln/detail/CVE-2026-74691
更多参考(点击展开)
漏洞分析指导链接：
https://atomgit.com/openeuler/cve-manager/blob/master/cve-vulner-manager/doc/md/manual.md
漏洞数据来源:
七彩瞬析开源风险感知平台
漏洞补丁信息：
详情(点击展开)
二、漏洞分析结构反馈
影响性分析说明：
In the Linux kernel, the following vulnerability has been resolved:net: thunderbolt: Tear down DMA paths before stopping the ringstbnet_tear_down() stops both rings and frees their frame buffers beforecalling tb_xdomain_disable_paths(). tb_ring_stop() zeroes the ring'sdescriptor base and tbnet_free_buffers() unmaps and frees the pages theframes sit in, so by the time __tb_path_deactivate_hop() polls the hop's'pending' bit, anything still in flight has nowhere to drain to.The teardown sequence has been in this order since the driver was added.The setup path has not: commit ff7cd07f3064 ("net: thunderbolt: EnableDMA paths only after rings are enabled") moved the path enable to the endof tbnet_connected_work() and documented why:/* Both logins successful so enable the rings, high-speed DMA * paths and start the network device queue. * * Note we enable the DMA paths last to make sure we have primed * the Rx ring before any incoming packets are allowed to * arrive. */Teardown was never updated to match, so the rings and the paths now comedown in the same order they go up instead of in reverse.On an ASMedia ASM4242 host router the 'pending' bit then never clears:every teardown burns the full 500 ms timeout and__tb_path_deactivate_hop() returns -ETIMEDOUT. Raising the timeout to5 s does not help, so the hop is not slow to drain, it never drainsat all.The failure is invisible above the thunderbolt core.__tb_path_deactivate_hops() is void and only calls tb_port_warn();tb_path_deactivate(), tb_tunnel_deactivate() and__tb_disconnect_xdomain_paths() are void as well, andtb_disconnect_xdomain_paths() ends in an unconditional "return 0". Sotb_xdomain_disable_paths() reports success and the netdev_warn() belowit never fires. Repeated teardowns eventually take the XDomain controlchannel down, after which the peer node is gone and only a power cyclebrings the controller back.Deactivating the paths first fixes it. Measured with kretprobes on astock v6.17 tree with no other patches applied, on a link that was upand had just carried traffic: before: __tb_path_deactivate_hop() returns 0 for the first hop, then -ETIMEDOUT for the second 500335 us later after: 0 for both, 525 us apartAlternating the two orderings ABBA over three load levels, fourteardowns per arm: every teardown failed before the change (21 of 21that ran), none failed after (0 of 24). The before arms ran shortbecause the link died partway through. The same split shows up whenthe interface is enslaved to a bond instead of just brought down, whichis how I ran into this in the first place. Throughput and latency afterthe change are unchanged.Hosts whose routers drain the hop despite the stale descriptor base seeno functional difference, since the paths end up deactivated either way.The Linux kernel CVE team has assigned CVE-2026-74691 to this issue.
openEuler评分：
8.8
Vector：CVSS:3.1/AV:A/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
受影响版本排查(受影响/不受影响)：
1.master(6.18.40):不受影响
2.openEuler-20.03-LTS-SP4(4.19.90):受影响
3.openEuler-22.03-LTS-SP4(5.10.0):受影响
4.openEuler-24.03-LTS-Next(6.6.0):不受影响
5.openEuler-24.03-LTS-SP1(6.6.0):受影响
6.openEuler-24.03-LTS-SP3(6.6.0):受影响
7.openEuler-24.03-LTS-SP4(6.6.0):受影响
8.openEuler-22.03-LTS-SP3(5.10.0):受影响
9.openEuler-24.03-LTS(6.6.0):受影响
10.openEuler-24.03-LTS-SP2(6.6.0):受影响
修复是否涉及abi变化(是/否)：
1.master(6.18.40):否
2.openEuler-20.03-LTS-SP4(4.19.90):否
3.openEuler-22.03-LTS-SP4(5.10.0):否
4.openEuler-24.03-LTS-Next(6.6.0):否
5.openEuler-24.03-LTS-SP1(6.6.0):否
6.openEuler-24.03-LTS-SP3(6.6.0):否
7.openEuler-24.03-LTS-SP4(6.6.0):否
8.openEuler-22.03-LTS-SP3(5.10.0):否
9.openEuler-24.03-LTS(6.6.0):否
10.openEuler-24.03-LTS-SP2(6.6.0):否
原因说明：
1.master(6.18.40):不受影响-漏洞代码不能被攻击者触发
2.openEuler-20.03-LTS-SP4(4.19.90):暂不修复-暂无解决方案或补丁
3.openEuler-22.03-LTS-SP4(5.10.0):正常修复
4.openEuler-24.03-LTS-Next(6.6.0):不受影响-漏洞代码不能被攻击者触发
5.openEuler-24.03-LTS-SP1(6.6.0):正常修复
6.openEuler-24.03-LTS-SP3(6.6.0):正常修复
7.openEuler-24.03-LTS-SP4(6.6.0):正常修复
8.openEuler-22.03-LTS-SP3(5.10.0):不修复-超出修复范围
9.openEuler-24.03-LTS(6.6.0):不修复-超出修复范围
10.openEuler-24.03-LTS-SP2(6.6.0):不修复-超出修复范围