ghsa-vr5f-5953-gfj5
Vulnerability from github
In the Linux kernel, the following vulnerability has been resolved:
net/sched: taprio: avoid disabling offload when it was never enabled
In an incredibly strange API design decision, qdisc->destroy() gets called even if qdisc->init() never succeeded, not exclusively since commit 87b60cfacf9f ("net_sched: fix error recovery at qdisc creation"), but apparently also earlier (in the case of qdisc_create_dflt()).
The taprio qdisc does not fully acknowledge this when it attempts full offload, because it starts off with q->flags = TAPRIO_FLAGS_INVALID in taprio_init(), then it replaces q->flags with TCA_TAPRIO_ATTR_FLAGS parsed from netlink (in taprio_change(), tail called from taprio_init()).
But in taprio_destroy(), we call taprio_disable_offload(), and this determines what to do based on FULL_OFFLOAD_IS_ENABLED(q->flags).
But looking at the implementation of FULL_OFFLOAD_IS_ENABLED() (a bitwise check of bit 1 in q->flags), it is invalid to call this macro on q->flags when it contains TAPRIO_FLAGS_INVALID, because that is set to U32_MAX, and therefore FULL_OFFLOAD_IS_ENABLED() will return true on an invalid set of flags.
As a result, it is possible to crash the kernel if user space forces an error between setting q->flags = TAPRIO_FLAGS_INVALID, and the calling of taprio_enable_offload(). This is because drivers do not expect the offload to be disabled when it was never enabled.
The error that we force here is to attach taprio as a non-root qdisc, but instead as child of an mqprio root qdisc:
$ tc qdisc add dev swp0 root handle 1: \ mqprio num_tc 8 map 0 1 2 3 4 5 6 7 \ queues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 hw 0 $ tc qdisc replace dev swp0 parent 1:1 \ taprio num_tc 8 map 0 1 2 3 4 5 6 7 \ queues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 base-time 0 \ sched-entry S 0x7f 990000 sched-entry S 0x80 100000 \ flags 0x0 clockid CLOCK_TAI Unable to handle kernel paging request at virtual address fffffffffffffff8 [fffffffffffffff8] pgd=0000000000000000, p4d=0000000000000000 Internal error: Oops: 96000004 [#1] PREEMPT SMP Call trace: taprio_dump+0x27c/0x310 vsc9959_port_setup_tc+0x1f4/0x460 felix_port_setup_tc+0x24/0x3c dsa_slave_setup_tc+0x54/0x27c taprio_disable_offload.isra.0+0x58/0xe0 taprio_destroy+0x80/0x104 qdisc_create+0x240/0x470 tc_modify_qdisc+0x1fc/0x6b0 rtnetlink_rcv_msg+0x12c/0x390 netlink_rcv_skb+0x5c/0x130 rtnetlink_rcv+0x1c/0x2c
Fix this by keeping track of the operations we made, and undo the offload only if we actually did it.
I've added "bool offloaded" inside a 4 byte hole between "int clockid" and "atomic64_t picos_per_byte". Now the first cache line looks like below:
$ pahole -C taprio_sched net/sched/sch_taprio.o struct taprio_sched { struct Qdisc * * qdiscs; / 0 8 / struct Qdisc * root; / 8 8 / u32 flags; / 16 4 / enum tk_offsets tk_offset; / 20 4 / int clockid; / 24 4 / bool offloaded; / 28 1 /
/* XXX 3 bytes hole, try to pack */
atomic64_t picos_per_byte; /* 32 0 */
/* XXX 8 bytes hole, try to pack */
spinlock_t current_entry_lock; /* 40 0 */
/* XXX 8 bytes hole, try to pack */
struct sched_entry * current_entry; /* 48 8 */
struct sched_gate_list * oper_sched; /* 56 8 */
/* --- cacheline 1 boundary (64 bytes) --- */
{
"affected": [],
"aliases": [
"CVE-2022-48644"
],
"database_specific": {
"cwe_ids": [],
"github_reviewed": false,
"github_reviewed_at": null,
"nvd_published_at": "2024-04-28T13:15:07Z",
"severity": null
},
"details": "In the Linux kernel, the following vulnerability has been resolved:\n\nnet/sched: taprio: avoid disabling offload when it was never enabled\n\nIn an incredibly strange API design decision, qdisc-\u003edestroy() gets\ncalled even if qdisc-\u003einit() never succeeded, not exclusively since\ncommit 87b60cfacf9f (\"net_sched: fix error recovery at qdisc creation\"),\nbut apparently also earlier (in the case of qdisc_create_dflt()).\n\nThe taprio qdisc does not fully acknowledge this when it attempts full\noffload, because it starts off with q-\u003eflags = TAPRIO_FLAGS_INVALID in\ntaprio_init(), then it replaces q-\u003eflags with TCA_TAPRIO_ATTR_FLAGS\nparsed from netlink (in taprio_change(), tail called from taprio_init()).\n\nBut in taprio_destroy(), we call taprio_disable_offload(), and this\ndetermines what to do based on FULL_OFFLOAD_IS_ENABLED(q-\u003eflags).\n\nBut looking at the implementation of FULL_OFFLOAD_IS_ENABLED()\n(a bitwise check of bit 1 in q-\u003eflags), it is invalid to call this macro\non q-\u003eflags when it contains TAPRIO_FLAGS_INVALID, because that is set\nto U32_MAX, and therefore FULL_OFFLOAD_IS_ENABLED() will return true on\nan invalid set of flags.\n\nAs a result, it is possible to crash the kernel if user space forces an\nerror between setting q-\u003eflags = TAPRIO_FLAGS_INVALID, and the calling\nof taprio_enable_offload(). This is because drivers do not expect the\noffload to be disabled when it was never enabled.\n\nThe error that we force here is to attach taprio as a non-root qdisc,\nbut instead as child of an mqprio root qdisc:\n\n$ tc qdisc add dev swp0 root handle 1: \\\n\tmqprio num_tc 8 map 0 1 2 3 4 5 6 7 \\\n\tqueues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 hw 0\n$ tc qdisc replace dev swp0 parent 1:1 \\\n\ttaprio num_tc 8 map 0 1 2 3 4 5 6 7 \\\n\tqueues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 base-time 0 \\\n\tsched-entry S 0x7f 990000 sched-entry S 0x80 100000 \\\n\tflags 0x0 clockid CLOCK_TAI\nUnable to handle kernel paging request at virtual address fffffffffffffff8\n[fffffffffffffff8] pgd=0000000000000000, p4d=0000000000000000\nInternal error: Oops: 96000004 [#1] PREEMPT SMP\nCall trace:\n taprio_dump+0x27c/0x310\n vsc9959_port_setup_tc+0x1f4/0x460\n felix_port_setup_tc+0x24/0x3c\n dsa_slave_setup_tc+0x54/0x27c\n taprio_disable_offload.isra.0+0x58/0xe0\n taprio_destroy+0x80/0x104\n qdisc_create+0x240/0x470\n tc_modify_qdisc+0x1fc/0x6b0\n rtnetlink_rcv_msg+0x12c/0x390\n netlink_rcv_skb+0x5c/0x130\n rtnetlink_rcv+0x1c/0x2c\n\nFix this by keeping track of the operations we made, and undo the\noffload only if we actually did it.\n\nI\u0027ve added \"bool offloaded\" inside a 4 byte hole between \"int clockid\"\nand \"atomic64_t picos_per_byte\". Now the first cache line looks like\nbelow:\n\n$ pahole -C taprio_sched net/sched/sch_taprio.o\nstruct taprio_sched {\n struct Qdisc * * qdiscs; /* 0 8 */\n struct Qdisc * root; /* 8 8 */\n u32 flags; /* 16 4 */\n enum tk_offsets tk_offset; /* 20 4 */\n int clockid; /* 24 4 */\n bool offloaded; /* 28 1 */\n\n /* XXX 3 bytes hole, try to pack */\n\n atomic64_t picos_per_byte; /* 32 0 */\n\n /* XXX 8 bytes hole, try to pack */\n\n spinlock_t current_entry_lock; /* 40 0 */\n\n /* XXX 8 bytes hole, try to pack */\n\n struct sched_entry * current_entry; /* 48 8 */\n struct sched_gate_list * oper_sched; /* 56 8 */\n /* --- cacheline 1 boundary (64 bytes) --- */",
"id": "GHSA-vr5f-5953-gfj5",
"modified": "2024-04-28T15:30:29Z",
"published": "2024-04-28T15:30:29Z",
"references": [
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2022-48644"
},
{
"type": "WEB",
"url": "https://git.kernel.org/stable/c/586def6ebed195f3594a4884f7c5334d0e1ad1bb"
},
{
"type": "WEB",
"url": "https://git.kernel.org/stable/c/c7c9c7eb305ab8b4e93e4e4e1b78d8cfcbc26323"
},
{
"type": "WEB",
"url": "https://git.kernel.org/stable/c/d12a1eb07003e597077329767c6aa86a7e972c76"
},
{
"type": "WEB",
"url": "https://git.kernel.org/stable/c/db46e3a88a09c5cf7e505664d01da7238cd56c92"
},
{
"type": "WEB",
"url": "https://git.kernel.org/stable/c/f58e43184226e5e9662088ccf1389e424a3a4cbd"
}
],
"schema_version": "1.4.0",
"severity": []
}