Pull vfs fixes from Christian Brauner: "The kernfs rbtree is keyed by (hash, ns, name) where the hash is seeded with the raw namespace pointer via init_name_hash(ns). The resulting hash values are exposed to userspace through readdir seek positions, and the pointer-based ordering in kernfs_name_compare() is observable through entry order. Switch from raw pointers to ns_common::ns_id for both hashing and comparison. A preparatory commit first replaces all const void * namespace parameters with const struct ns_common * throughout kernfs, sysfs, and kobject so the code can access ns->ns_id. Also compare the ns_id when hashes match in the rbtree to handle crafted collisions. Also fix eventpoll RCU grace period issue and a cachefiles refcount problem" * tag 'vfs-7.0-rc8.fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs: kernfs: make directory seek namespace-aware kernfs: use namespace id instead of pointer for hashing and comparison kernfs: pass struct ns_common instead of const void * for namespace tags eventpoll: defer struct eventpoll free to RCU grace period cachefiles: fix incorrect dentry refcount in cachefiles_cull()
kernfs has historically used const void * to pass around namespace tags used for directory-level namespace filtering. The only current user of this is sysfs network namespace tagging where struct net pointers are cast to void *. Replace all const void * namespace parameters with const struct ns_common * throughout the kernfs, sysfs, and kobject namespace layers. This includes the kobj_ns_type_operations callbacks, kobject_namespace(), and all sysfs/kernfs APIs that accept or return namespace tags. Passing struct ns_common is needed because various codepaths require access to the underlying namespace. A struct ns_common can always be converted back to the concrete namespace type (e.g., struct net) via container_of() or to_ns_common() in the reverse direction. This is a preparatory change for switching to ns_id-based directory iteration to prevent a KASLR pointer leak through the current use of raw namespace pointers as hash seeds and comparison keys. Signed-off-by: Christian Brauner <brauner@kernel.org>
Sharing a global hash table among all queues is tempting, but
it can cause crash:
BUG: KASAN: slab-use-after-free in nfqnl_recv_verdict+0x11ac/0x15e0 [nfnetlink_queue]
[..]
nfqnl_recv_verdict+0x11ac/0x15e0 [nfnetlink_queue]
nfnetlink_rcv_msg+0x46a/0x930
kmem_cache_alloc_node_noprof+0x11e/0x450
struct nf_queue_entry is freed via kfree, but parallel cpu can still
encounter such an nf_queue_entry when walking the list.
Alternative fix is to free the nf_queue_entry via kfree_rcu() instead,
but as we have to alloc/free for each skb this will cause more mem
pressure.
Cc: Scott Mitchell <scott.k.mitch1@gmail.com>
Fixes: e19079adcd26 ("netfilter: nfnetlink_queue: optimize verdict lookup with hash table")
Signed-off-by: Florian Westphal <fw@strlen.de>
nft_ct_timeout_obj_destroy() frees the timeout object with kfree()
immediately after nf_ct_untimeout(), without waiting for an RCU grace
period. Concurrent packet processing on other CPUs may still hold
RCU-protected references to the timeout object obtained via
rcu_dereference() in nf_ct_timeout_data().
Add an rcu_head to struct nf_ct_timeout and use kfree_rcu() to defer
freeing until after an RCU grace period, matching the approach already
used in nfnetlink_cttimeout.c.
KASAN report:
BUG: KASAN: slab-use-after-free in nf_conntrack_tcp_packet+0x1381/0x29d0
Read of size 4 at addr ffff8881035fe19c by task exploit/80
Call Trace:
nf_conntrack_tcp_packet+0x1381/0x29d0
nf_conntrack_in+0x612/0x8b0
nf_hook_slow+0x70/0x100
__ip_local_out+0x1b2/0x210
tcp_sendmsg_locked+0x722/0x1580
__sys_sendto+0x2d8/0x320
Allocated by task 75:
nft_ct_timeout_obj_init+0xf6/0x290
nft_obj_init+0x107/0x1b0
nf_tables_newobj+0x680/0x9c0
nfnetlink_rcv_batch+0xc29/0xe00
Freed by task 26:
nft_obj_destroy+0x3f/0xa0
nf_tables_trans_destroy_work+0x51c/0x5c0
process_one_work+0x2c4/0x5a0
Fixes: 7e0b2b57f01d ("netfilter: nft_ct: add ct timeout support")
Cc: stable@vger.kernel.org
Signed-off-by: Tuan Do <tuan@calif.io>
Signed-off-by: Florian Westphal <fw@strlen.de>
Currently xp_assign_dev_shared() is missing XDP_USE_SG being propagated
to flags so set it in order to preserve mtu check that is supposed to be
done only when no multi-buffer setup is in picture.
Also, this flag has the same value as XDP_UMEM_TX_SW_CSUM so we could
get unexpected SG setups for software Tx checksums. Since csum flag is
UAPI, modify value of XDP_UMEM_SG_FLAG.
Fixes: d609f3d228a8 ("xsk: add multi-buffer support for sockets sharing umem")
Reviewed-by: Björn Töpel <bjorn@kernel.org>
Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Link: https://patch.msgid.link/20260402154958.562179-4-maciej.fijalkowski@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Multi-buffer XDP stores information about frags in skb_shared_info that
sits at the tailroom of a packet. The storage space is reserved via
xdp_data_hard_end():
((xdp)->data_hard_start + (xdp)->frame_sz - \
SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
and then we refer to it via macro below:
static inline struct skb_shared_info *
xdp_get_shared_info_from_buff(const struct xdp_buff *xdp)
{
return (struct skb_shared_info *)xdp_data_hard_end(xdp);
}
Currently we do not respect this tailroom space in multi-buffer AF_XDP
ZC scenario. To address this, introduce xsk_pool_get_tailroom() and use
it within xsk_pool_get_rx_frame_size() which is used in ZC drivers to
configure length of HW Rx buffer.
Typically drivers on Rx Hw buffers side work on 128 byte alignment so
let us align the value returned by xsk_pool_get_rx_frame_size() in order
to avoid addressing this on driver's side. This addresses the fact that
idpf uses mentioned function *before* pool->dev being set so we were at
risk that after subtracting tailroom we would not provide 128-byte
aligned value to HW.
Since xsk_pool_get_rx_frame_size() is actively used in xsk_rcv_check()
and __xsk_rcv(), add a variant of this routine that will not include 128
byte alignment and therefore old behavior is preserved.
Reviewed-by: Björn Töpel <bjorn@kernel.org>
Acked-by: Stanislav Fomichev <sdf@fomichev.me>
Fixes: 24ea50127ecf ("xsk: support mbuf on ZC RX")
Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Link: https://patch.msgid.link/20260402154958.562179-3-maciej.fijalkowski@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In configurations with multiple tunnel layers and MPLS lwtunnel routing, a
single tunnel hop can increment the counter beyond this limit. This causes
packets to be dropped with the "Dead loop on virtual device" message even
when a routing loop doesn't exist.
Increase IP_TUNNEL_RECURSION_LIMIT from 4 to 5 to handle this use-case.
Fixes: 6f1a9140ecda ("net: add xmit recursion limit to tunnel xmit functions")
Link: https://lore.kernel.org/netdev/88deb91b-ef1b-403c-8eeb-0f971f27e34f@redhat.com/
Signed-off-by: Chris J Arges <carges@cloudflare.com>
Link: https://patch.msgid.link/20260402222401.3408368-1-carges@cloudflare.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The RCU-protected codepaths (mpls_forward, mpls_dump_routes) can have
an inconsistent view of platform_labels vs platform_label in case of a
concurrent resize (resize_platform_label_table, under
platform_mutex). This can lead to OOB accesses.
This patch adds a seqcount, so that we get a consistent snapshot.
Note that mpls_label_ok is also susceptible to this, so the check
against RTA_DST in rtm_to_route_config, done outside platform_mutex,
is not sufficient. This value gets passed to mpls_label_ok once more
in both mpls_route_add and mpls_route_del, so there is no issue, but
that additional check must not be removed.
Reported-by: Yuan Tan <tanyuan98@outlook.com>
Reported-by: Yifan Wu <yifanwucs@gmail.com>
Reported-by: Juefei Pu <tomapufckgml@gmail.com>
Reported-by: Xin Liu <bird@lzu.edu.cn>
Fixes: 7720c01f3f590 ("mpls: Add a sysctl to control the size of the mpls label table")
Fixes: dde1b38e873c ("mpls: Convert mpls_dump_routes() to RCU.")
Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Link: https://patch.msgid.link/cd8fca15e3eb7e212b094064cd83652e20fd9d31.1774284088.git.sd@queasysnail.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
__nf_ct_expect_find() and nf_ct_expect_find_get() are called under rcu_read_lock() but they dereference the master conntrack via exp->master. Since the expectation does not hold a reference on the master conntrack, this could be dying conntrack or different recycled conntrack than the real master due to SLAB_TYPESAFE_RCU. Store the netns, the master_tuple and the zone in struct nf_conntrack_expect as a safety measure. This patch is required by the follow up fix not to dump expectations that do not belong to this netns. Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Holding reference on the expectation is not sufficient, the master conntrack object can just go away, making exp->master invalid. To access exp->master safely: - Grab the nf_conntrack_expect_lock, this gets serialized with clean_from_lists() which also holds this lock when the master conntrack goes away. - Hold reference on master conntrack via nf_conntrack_find_get(). Not so easy since the master tuple to look up for the master conntrack is not available in the existing problematic paths. This patch goes for extending the nf_conntrack_expect_lock section to address this issue for simplicity, in the cases that are described below this is just slightly extending the lock section. The add expectation command already holds a reference to the master conntrack from ctnetlink_create_expect(). However, the delete expectation command needs to grab the spinlock before looking up for the expectation. Expand the existing spinlock section to address this to cover the expectation lookup. Note that, the nf_ct_expect_iterate_net() calls already grabs the spinlock while iterating over the expectation table, which is correct. The get expectation command needs to grab the spinlock to ensure master conntrack does not go away. This also expands the existing spinlock section to cover the expectation lookup too. I needed to move the netlink skb allocation out of the spinlock to keep it GFP_KERNEL. For the expectation events, the IPEXP_DESTROY event is already delivered under the spinlock, just move the delivery of IPEXP_NEW under the spinlock too because the master conntrack event cache is reached through exp->master. While at it, add lockdep notations to help identify what codepaths need to grab the spinlock. Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>