In the Linux kernel, the following vulnerability has been resolved:<\/p>\n
net: cap tx_queue_len at S16_MAX to prevent oversized ring allocations<\/p>\n
Several subsystems allocate ring buffers sized by dev->tx_queue_len
\nwith no upper bound. An unprivileged user (via unshare -Urn) can set a
\nhuge tx_queue_len and exhaust global memory with ring allocations:<\/p>\n
– pfifo_fast: pfifo_fast_init() and pfifo_fast_change_tx_queue_len()
\n allocate 3 skb_array rings of tx_queue_len entries each.
\n– tun: tun_queue_resize() and the queue-attach path resize ptr_rings
\n to tx_queue_len on the NETDEV_CHANGE_TX_QUEUE_LEN notifier.
\n– tap (macvtap\/ipvtap): tap_queue_resize() and tap_init() resize\/init
\n ptr_rings to tx_queue_len on the same notifier.<\/p>\n
netif_change_tx_queue_len() is the single entry point for IFLA_TXQLEN,
\nsysfs, and the SIOCSIFTXQLEN ioctl. Cap new_len at S16_MAX (32767)
\nthere so the oversized value is rejected at set time. This takes
\neffect whether the device is up or down, before dev->tx_queue_len is
\nwritten, before any notifier fires, and before any ring is allocated.
\nThe “> S16_MAX” check also subsumes the previous unsigned-long
\ntruncation test, and a negative ifr_qlen from the ioctl lands far
\nabove the cap after conversion, so both old failure modes are covered
\nby the one comparison.<\/p>\n
tx_queue_len is ambigious: both a per-ring sizing multiplier and a
\ndefault queue-length\/limit knob for consumers that allocate
\nnothing at set time (pfifo\/bfifo\/gred\/plug\/sfb limits, htb
\ndirect_qlen, qfq max_classes, teql). 32767 is chosen as the largest
\nvalue NLA_POLICY_FULL_RANGE can express for the u32 IFLA_TXQLEN
\npolicy in patch 2\/3 while staying a legitimate queue length on
\nhigh-BDP paths; the ring-memory trade-off of a shared knob is
\ndisclosed below.<\/p>\n
Conditions to recreate the bug:
\n– CONFIG_NET_SCHED=y, CONFIG_VETH=y, CONFIG_USER_NS=y, CONFIG_NET_NS=y.
\n– Unprivileged user in a fresh user+net namespace (unshare -Urn).
\n– pfifo_fast: create veth pairs, set tx_queue_len to 500000, attach
\n mq+pfifo_fast. ~28 iterations OOMs a 2GB guest.
\n– tun: create 50 tun devices with IFF_MULTI_QUEUE, set tx_queue_len to
\n 500000, open 8 queues each. ~1.6GB of ptr_ring allocations OOMs a
\n 512MB guest.
\n– tap: same as tun with IFF_TAP. ~960MB OOMs a 512MB guest.
\n– On the fixed kernel the oversized tx_queue_len is rejected with
\n -ERANGE at set time (all four paths: RTM_SETLINK, RTM_NEWLINK
\n create, sysfs, ioctl – the latter two via this check, the former
\n two via this check and the 2\/3 parse policy respectively).<\/p>\n","protected":false},"excerpt":{"rendered":"
In the Linux kernel, the following vulnerability has been resolved:<\/p>\n
net: cap tx_queue_len at S16_MAX to prevent oversized ring allocations<\/p>\n
Several subsystems allocate ring buffers sized by dev->tx_q<\/p>\n","protected":false},"template":"","meta":[],"categories":[],"tags":[],"class_list":["post-29177","vuln","type-vuln","status-publish","hentry"],"_links":{"self":[{"href":"https:\/\/www.54iter.cn\/index.php?rest_route=\/wp\/v2\/vuln\/29177","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.54iter.cn\/index.php?rest_route=\/wp\/v2\/vuln"}],"about":[{"href":"https:\/\/www.54iter.cn\/index.php?rest_route=\/wp\/v2\/types\/vuln"}],"wp:attachment":[{"href":"https:\/\/www.54iter.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=29177"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.54iter.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=29177"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.54iter.cn\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=29177"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}