| Age | Commit message (Collapse) | Author |
|---|
|
Extend enic_sriov_configure() to handle V2 SR-IOV VFs. When the PF
detects V2 VF device IDs, the enable path allocates per-VF MBOX state,
initializes the MBOX protocol, opens the admin channel, and then calls
pci_enable_sriov(). The admin channel must be ready before VFs are
created so that VF drivers can immediately begin the MBOX capability
and registration handshake during their probe.
The enic_sriov_configure() dispatcher and its V2 helpers
(enic_sriov_v2_enable, enic_sriov_v2_disable) are defined here but
intentionally not yet wired into struct pci_driver via
.sriov_configure -- hence the __maybe_unused annotations. This
series introduces only the admin channel and MBOX infrastructure;
sysfs-driven V2 enable/disable will be activated in a follow-up
patch by adding ".sriov_configure = enic_sriov_configure," to
enic_driver.
Because .sriov_configure is not registered yet, enic_sriov_configure()
cannot run concurrently with the rtnl-protected reset paths
(enic_reset(), enic_tx_hang_reset()) in this series, so there is no
reachable locking race between SR-IOV enable/disable and reset. The
follow-up patch that wires the callback will add the necessary
serialization against those paths. Note that simply taking rtnl_lock()
around the enable path is not viable, because pci_enable_sriov()
triggers VF probe and register_netdev(), which themselves acquire rtnl;
the wiring patch therefore uses finer-grained serialization.
The disable path first clears ENIC_SRIOV_ENABLED and flushes the
link-notify work, so no further VF link-state broadcast can run, then
calls pci_disable_sriov() (VF drivers unregister via MBOX), closes the
admin channel, and frees per-VF state. Clearing the flag and flushing
the work before vf_state is freed closes a use-after-free window
against the link-notify path.
Notify registered VFs of PF link transitions: enic_link_check()
schedules link_notify_work on each carrier up/down edge, and the work
handler sends PF_LINK_STATE_NOTIF to the VFs from process context.
The broadcast cannot run directly in enic_link_check() because the
MBOX send path may sleep and link check runs in the notify timer/ISR
context.
On a V2 VF the admin-channel (PF) link-state notification is the sole
authority for carrier state, so enic_link_check() returns early for
such VFs. As a side effect the VF retains the firmware-provided static
Rx interrupt coalescing (config.intr_timer_usec) rather than PF-driven
speed-adaptive coalescing; this is intentional, as adaptive Rx
coalescing is a PF-only responsibility for V2 VFs.
Re-establish the admin/MBOX channel across a PF reset. enic_reset()
and enic_tx_hang_reset() fully close the admin channel before the
soft/hang reset (which wipes all hardware queues, including the admin
WQ/RQ), then reopen it and re-run enic_mbox_init() after the data path
is back up, and re-push the current link state to registered VFs.
Reject VF port profile requests when V2 SR-IOV is active
(enic_is_valid_pp_vf), since enic->pp is not reallocated for V2 VFs
and the V2 protocol uses MBOX instead of port profiles.
Update enic_remove() to run enic_dev_deinit() and vnic_dev_close()
after SR-IOV teardown, so the PF device remains functional while VFs
are being cleaned up. This ordering applies to both V1 and V2 SR-IOV
paths.
Restrict the probe-time SR-IOV auto-enable to the legacy VF types (V1
and usNIC). A V2-capable adapter whose firmware lacks V2 support is
downgraded to ENIC_VF_TYPE_NONE, and V2 VFs require the admin channel
which is only brought up via sysfs enic_sriov_configure(); neither
must be auto-enabled through the legacy pci_enable_sriov() path at
probe.
Signed-off-by: Satish Kharat <satishkh@cisco.com>
Link: https://patch.msgid.link/20260812-enic-sriov-v2-admin-channel-v2-v13-9-b3809e448aba@cisco.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Implement VF-side mailbox message processing for SR-IOV V2
admin channel communication.
VF receive handlers:
- VF_CAPABILITY_REPLY: store PF protocol version, signal
completion
- VF_REGISTER_REPLY: mark VF as registered, signal completion
- VF_UNREGISTER_REPLY: mark VF as unregistered, signal
completion
- PF_LINK_STATE_NOTIF: update carrier state via
netif_carrier_on/off, send ACK back to PF
VF initiation functions for the probe-time handshake:
- enic_mbox_vf_capability_check: send capability request,
wait for PF reply via completion
- enic_mbox_vf_register: send register request, wait for
PF confirmation via completion
- enic_mbox_vf_unregister: send unregister request, wait
for PF confirmation
The wait helper (enic_mbox_wait_reply) uses
wait_for_completion_timeout, signaled when the admin ISR and
CQ-poll/dispatch workqueue pipeline delivers the reply message.
mbox_expected_reply is written by the request thread and read by the
admin CQ poll/dispatch context that runs the receive handlers; annotate
those accesses with READ_ONCE()/WRITE_ONCE() under the
single-outstanding-reply invariant.
Signed-off-by: Satish Kharat <satishkh@cisco.com>
Link: https://patch.msgid.link/20260812-enic-sriov-v2-admin-channel-v2-v13-8-b3809e448aba@cisco.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Implement PF-side mailbox message processing for SR-IOV V2
admin channel communication.
When the PF receives messages from VFs, the dispatch routes
them to type-specific handlers:
- VF_CAPABILITY_REQUEST: reply with protocol version 1
- VF_REGISTER_REQUEST: send the register reply, mark the
VF registered on success, then send PF_LINK_STATE_NOTIF
reflecting the PF's current carrier state
- VF_UNREGISTER_REQUEST: mark VF unregistered, send reply
- PF_LINK_STATE_ACK: log errors from VF acknowledgment
Per-VF state (struct enic_vf_state) is tracked via enic->vf_state
which will be allocated when SRIOV V2 is enabled.
Remove the CONFIG_PCI_IOV guard from num_vfs in struct enic. The
PF handlers reference enic->num_vfs for VF ID bounds checking in
enic_mbox.c, which is compiled unconditionally. The field must be
visible regardless of CONFIG_PCI_IOV to avoid build failures.
Add enic_mbox_send_link_state() helper for PF-initiated link
state notifications.
Signed-off-by: Satish Kharat <satishkh@cisco.com>
Link: https://patch.msgid.link/20260812-enic-sriov-v2-admin-channel-v2-v13-7-b3809e448aba@cisco.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Implement the mailbox protocol engine used for PF-VF communication
over the admin channel.
The send path (enic_mbox_send_msg) builds a message with a common
header, DMA-maps it, posts a single WQ descriptor with the
destination vnic ID encoded in the VLAN tag field, and polls
the WQ CQ for completion.
The total message length is computed as a size_t, and the payload is
bounded before the send lock is taken: a payload larger than the admin
buffer minus the header is rejected with -EINVAL. This keeps the length
sum from wrapping and stops the on-the-wire u16 length from overflowing
or the DMA buffer from being overrun.
MBOX sends are gated by enic->mbox_send_disabled: enic_mbox_send_msg()
returns early while it is set. It is set at the very start of both
enic_admin_channel_open() and enic_admin_channel_close(), and is
cleared in enic_admin_channel_open() only once the admin WQ/RQ/CQ and
interrupt are fully allocated, programmed and enabled. Keeping it set
for the whole open sequence means an early failure that returns before
the channel is ready (as well as a not-yet-ready or torn-down channel)
leaves sends disabled, so a concurrent sender can never race an MBOX
send against a half-open or freed admin_wq.
The receive path (enic_mbox_recv_handler) is installed as the admin
RQ callback and validates incoming message headers. PF/VF-specific
dispatch will be added in subsequent commits.
Signed-off-by: Satish Kharat <satishkh@cisco.com>
Link: https://patch.msgid.link/20260812-enic-sriov-v2-admin-channel-v2-v13-6-b3809e448aba@cisco.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Define the mailbox protocol structures for PF-VF communication:
message header, generic reply, and per-message-type payloads for
capability negotiation, VF registration/unregistration, and link
state notification/acknowledgment.
Include linux/types.h and linux/bits.h for __le16/__le32/__le64
and BIT() used in the header.
Message types use an even=request / odd=reply convention. The
header carries source and destination VNIC IDs, a per-channel
message sequence number, and the total message length.
Signed-off-by: Satish Kharat <satishkh@cisco.com>
Link: https://patch.msgid.link/20260812-enic-sriov-v2-admin-channel-v2-v13-5-b3809e448aba@cisco.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Add completion queue (CQ) service for the admin channel work queue
(WQ) and receive queue (RQ), driven by a dedicated MSI-X interrupt
and a workqueue-based CQ poller.
The admin WQ CQ service advances the completion ring and returns the
number of descriptors consumed. The admin RQ CQ service does the
same for receive completions and copies each received message out of
its pre-posted DMA buffer into a dynamically allocated queue entry.
The pending queue is bounded to ENIC_ADMIN_MSG_MAX (256) entries so a
buggy or hostile VF cannot drive the host out of memory; messages are
enqueued for deferred dispatch by a separate work_struct so the CQ
poller stays short.
When the MSI-X interrupt fires, the ISR schedules the CQ poll work.
The work handler drains all pending completions, kicks message
dispatch if work was done, and returns credits to unmask the
interrupt. The admin vector is kept masked from the time the IRQ is
requested until the rings are initialised and filled during channel
open, so an early or spurious interrupt cannot run the poll handler
against uninitialised rings.
The poll handler snapshots the pending credit count before draining
the CQ so it acknowledges exactly what the hardware reported for this
interrupt; any credits that accrue during draining are serviced by the
next interrupt. The credit write also sets the mask bit to re-arm the
vector, and that unmask is applied independently of the credit count,
so the vector is re-armed even when zero credits are returned -- which
matters here because the admin channel is not re-polled like the NAPI
data path.
If an admin RQ buffer refill fails under transient memory pressure,
reschedule the CQ poll work itself after a short delay to retry the
refill and re-arm the RQ, so the admin channel cannot stall when the
ring would otherwise be left empty with no completion to drive the
next refill. The poll work is a delayed_work for this reason; routing
the retry through it keeps the admin RQ ring owned by a single context
so refills never run concurrently.
Signed-off-by: Satish Kharat <satishkh@cisco.com>
Link: https://patch.msgid.link/20260812-enic-sriov-v2-admin-channel-v2-v13-4-b3809e448aba@cisco.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The admin receive queue needs pre-posted DMA buffers for incoming
mailbox messages from VFs. Each buffer is a kzalloc'd region mapped
for DMA (2048 bytes, sufficient for any MBOX message). Zeroing on
allocation ensures that if a completion reports more bytes than
hardware actually DMA-wrote, the parser reads zero padding rather
than uninitialised heap contents.
Add enic_admin_rq_fill(gfp) to post buffers at open time, and
enic_admin_rq_drain() to unmap and free them at close time.
Wire both into the admin channel open/close paths. The gfp_t
parameter lets the caller pass the allocation context; both current
callers -- channel open and the CQ-poll work handler that refills
after draining (added in the next patch) -- run in process context
and use GFP_KERNEL.
Signed-off-by: Satish Kharat <satishkh@cisco.com>
Link: https://patch.msgid.link/20260812-enic-sriov-v2-admin-channel-v2-v13-3-b3809e448aba@cisco.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The V2 SR-IOV design uses a dedicated admin channel (WQ/RQ/CQ
resources plus an MSI-X interrupt) for PF-VF mailbox communication rather
than firmware-proxied devcmds.
Introduce enic_admin_channel_open() and enic_admin_channel_close().
Open allocates and initialises the admin WQ, RQ, and two CQs (one per
direction), then issues CMD_QP_TYPE_SET to tell firmware the queues are
admin-type. Close reverses the sequence.
enic_admin_wq_buf_clean() unmaps and frees any WQ buffers still held
at close time, fixing a DMA mapping leak when a send times out.
Add CMD_QP_TYPE_SET (97), QP_TYPE_ADMIN/DATA, and QP_ENABLE/QP_DISABLE
defines to vnic_devcmd.h. Add VNIC_CQ_* named constants to vnic_cq.h
so CQ initialisation parameters are self-documenting from their first
introduction.
Signed-off-by: Satish Kharat <satishkh@cisco.com>
Link: https://patch.msgid.link/20260812-enic-sriov-v2-admin-channel-v2-v13-2-b3809e448aba@cisco.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
During PF probe, query the firmware get-supported-feature interface
to verify that the running firmware supports V2 SR-IOV. Firmware
version 5.3(4.72) and later report VIC_FEATURE_SRIOV via
CMD_GET_SUPP_FEATURE_VER. If the firmware does not support the
feature, set vf_type to ENIC_VF_TYPE_NONE and log a warning so the
admin knows a firmware upgrade is needed.
The V2 admin-channel and MBOX bring-up added later in this series is
gated on ENIC_VF_TYPE_V2, so this downgrade keeps those paths from
running on firmware that does not support V2 SR-IOV.
VIC_FEATURE_SRIOV is assigned the explicit value 4 to match the
firmware ABI. Slot 3 (firmware's VIC_FEATURE_PTP) is reserved with
a comment rather than a placeholder enum entry, since PTP is not
used by the upstream driver.
Suggested-by: Breno Leitao <leitao@debian.org>
Signed-off-by: Satish Kharat <satishkh@cisco.com>
Reviewed-by: Breno Leitao <leitao@debian.org>
Link: https://patch.msgid.link/20260812-enic-sriov-v2-admin-channel-v2-v13-1-b3809e448aba@cisco.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Matthieu Baerts says:
====================
mptcp: misc. features for v7.3
This series contains a few independent new features, and small fixes for
net-next:
- Patch 1: Add WARN_ON_ONCE guards around extra_subflows to catch issues
with this counter, similar to what is done with other PM counters.
- Patches 2-3: Follow-up patches to remove data_ack field from struct
mptcp_ext -- now unused after recent fixes -- and makes a userspace PM
helper static.
- Patch 4: Honour tcp_rto_{min_us,max_ms} sysctls for MPTCP-level
retransmit timers like with DATA_FIN's and fallback timeout.
- Patches 5-6: Add per-event MIB counters for MPTCP_RST_EMPTCP resets to
help to spot such situations in production.
- Patches 7-9: Small pcap-related improvements in the selftests.
- Patch 10: Fix compiler warning in the selftests.
- Patch 11: Avoid a buffer overflow when misusing the mptcp_diag tool
from the selftests.
====================
Link: https://patch.msgid.link/20260812-net-next-mptcp-misc-feat-7-3-v1-0-1905a818f6cb@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
get_subflow_info() parses the subflow address string with:
char saddr[64], daddr[64];
ret = sscanf(subflow_addrs, "%[^:]:%d %[^:]:%d",
saddr, &sport, daddr, &dport);
The subflow_addrs buffer holds up to 1024 bytes and is taken directly
from the command line ("-c" argument). The "%[^:]" conversions have no
maximum field width, so if the address substring before the ':' exceeds
63 bytes, sscanf() writes past the end of the 64-byte saddr/daddr stack
buffers. This overflows the stack, corrupting adjacent stack data such
as the saved return address, and can crash the tool or lead to
out-of-bounds writes controlled by user-supplied input.
Bound both string conversions to the destination buffer size by adding
an explicit maximum field width of 63 (leaving room for the terminating
NUL), so at most 63 bytes are written into each 64-byte buffer:
ret = sscanf(subflow_addrs, "%63[^:]:%d %63[^:]:%d",
saddr, &sport, daddr, &dport);
The subflow address can be passed in argument, so fixing this is helpful
when the tool is manually used.
Reviewed-by: Geliang Tang <geliang@kernel.org>
Signed-off-by: Jiangshan Yi <yijiangshan@kylinos.cn>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20260812-net-next-mptcp-misc-feat-7-3-v1-11-1905a818f6cb@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
In mptcp_connect.c, strchr() returns a pointer to a character within
the input string, which is declared as const char *. Assigning this
return value to a non-const char * discards the const qualifier,
triggering compiler warnings:
make: Entering directory 'tools/testing/selftests/net/mptcp'
CC mptcp_connect
mptcp_connect.c: In function 'parse_cmsg_types':
mptcp_connect.c:1267:22: warning: initialization discards 'const'
qualifier from pointer target type [-Wdiscarded-qualifiers]
1267 | char *next = strchr(type, ',');
| ^~~~~~
mptcp_connect.c: In function 'parse_setsock_options':
mptcp_connect.c:1295:22: warning: initialization discards 'const'
qualifier from pointer target type [-Wdiscarded-qualifiers]
1295 | char *next = strchr(name, ',');
| ^~~~~~
make: Leaving directory 'tools/testing/selftests/net/mptcp'
Fix these warnings by declaring the 'next' variable as const char *,
as it is only used for read-only parsing.
Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
Reviewed-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20260812-net-next-mptcp-misc-feat-7-3-v1-10-1905a818f6cb@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Limit the size of each captured packet to 108B (IPv4 only) or 128B (a
mix of v4 and v6): this should drop most of the payload that is
generally not needed when debugging an issue.
8 bytes are left in this payload, to be able to inspect the beginning,
just in case.
Please also note that generally, this payload is usually mostly filled
with 0, except at the end. This reduces the .pcap sizes, and reduce IO
usage, which helps debugging issues.
Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20260812-net-next-mptcp-misc-feat-7-3-v1-9-1905a818f6cb@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
To be able to easily find out which pcap was produced by which test, the
selftest name is now added to the pcap file, similar to the other tests.
While at it, print the prefix name to be able to find which capture
files have been produced by which test after several runs. This prefix
was not printed anywhere before.
Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20260812-net-next-mptcp-misc-feat-7-3-v1-8-1905a818f6cb@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Even if the pcap prefix is printed in the test, it is clearer if this
prefix also include the test name: mptcp_connect.
With this, it is easily possible to find out which pcap was produced by
which test, and easily delete the right ones.
Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20260812-net-next-mptcp-misc-feat-7-3-v1-7-1905a818f6cb@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Add named env-var expectations for each per-event MPTCP_RST_EMPTCP
counter, matching the pattern used by the existing JOIN/RST checks.
Each defaults to 0 and is checked silently on success; a mismatch prints
a check line and fails the test. Counters absent from the running
kernel are skipped silently so older kernels do not false-fail.
The JOIN-related counters (MPJoinSynAckNoMPJoin, MPJoinAckNoMPJoin,
MPJoinAckNoCtx, MPJoinNotEstablished, MPJoinNoIdFound) are checked in
chk_join_nr() on fixed namespaces; the two remaining reset counters
(MD5SigReset, DssReset) stay in chk_rst_nr().
Add a test at the end of signal_address_tests that triggers
MPJoinSynAckNoMPJoin: ns1 signals an address that is already bound on
the client (ns2), where a TCP-only mptcp_connect listener is started.
The client's MP_JOIN routes locally to the TCP listener, which responds
with a plain SYN/ACK without the MP_JOIN option, and the new counter
increments on the client side.
Other per-event counters (MD5SigReset, MPJoinAckNoMPJoin, MPJoinAckNoCtx,
DssReset, MPJoinNotEstablished, MPJoinNoIdFound) are not currently
reachable from mptcp_join.sh; the env-var hooks are in place for future
tests to set expectations explicitly.
Signed-off-by: Shardul Bankar <shardul.b@mpiricsoftware.com>
Reviewed-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20260812-net-next-mptcp-misc-feat-7-3-v1-6-1905a818f6cb@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
MPTCP_RST_EMPTCP (reset reason 1) is used as a catch-all for several
distinct error conditions across subflow setup, authentication, and
data-path validation. The existing MPRstTx/MPRstRx counters only
track aggregate reset volume, making it difficult to diagnose which
code path is triggering subflow resets in production.
Add per-event MIB counters covering each MPTCP_RST_EMPTCP use site
that is not already covered by an existing counter, named after the
underlying event or condition rather than the reset action:
MD5SigReset MD5SIG enabled on listener (incompatible)
MPJoinSynAckNoMPJoin SYN/ACK missing MP_JOIN option
MPJoinAckNoMPJoin server-side ACK missing MP_JOIN option
(fallback path, MPJoin required)
MPJoinAckNoCtx server-side ACK with no subflow context
MPJoinNoIdFound MP_JOIN with a valid token but no PM local ID
DssReset data mapping invalid (also fires on
MAPPING_NODSS / EMIDDLEBOX path)
MPJoinNotEstablished JOIN attempted on a not-fully-established msk
MPJoinNoIdFound covers the second half of the no-msk MP_JOIN reset:
the existing MPJoinNoTokenFound (MPTCP_MIB_JOINNOTOKEN) only counts the
missing-token case in subflow_token_join_request(), while a JOIN that
carries a valid token but for which the path manager returns no local
id reaches the same MPTCP_RST_EMPTCP in subflow_check_req() uncounted.
The aggregate MPRstTx/MPRstRx counters are unchanged.
Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/511
Signed-off-by: Shardul Bankar <shardul.b@mpiricsoftware.com>
Reviewed-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20260812-net-next-mptcp-misc-feat-7-3-v1-5-1905a818f6cb@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The MPTCP-level retransmit timers (DATA_FIN retransmissions and the
fallback timeout) used the hard-coded TCP_RTO_MIN / TCP_RTO_MAX
constants, ignoring the tcp_rto_min_us and tcp_rto_max_ms sysctls.
Make them follow the sysctls instead: seed icsk_rto_min / icsk_rto_max
on the MPTCP socket from the per-netns sysctls in __mptcp_init_sock()
-- the msk does not go through tcp_init_sock(), so these fields would
otherwise stay zero -- and read them directly where the constants were
used:
- mptcp_set_datafin_timeout(): both the backoff cap computation and
the resulting timer_ival. The two sysctls are validated
independently, so rto_min > rto_max is a valid configuration; keep
a max_t() guard so ilog2() is never called with 0.
- __mptcp_set_timeout(): the fallback when no subflow timeout is
available.
The icsk fields are read directly instead of using the
tcp_rto_min()/tcp_rto_max() helpers: the MPTCP socket does not perform
routing lookups in these paths, so the rto_min route metric checked by
tcp_rto_min() can never apply here. The TCP_RTO_MIN_US /
TCP_RTO_MAX_MS socket options are not supported by MPTCP setsockopt()
either; this can be revisited if they get supported on MPTCP sockets.
The remaining uses of TCP_RTO_MAX in net/mptcp/ctrl.c (default
add_addr_timeout) and net/mptcp/subflow.c (MP_FAIL timeout) are
intentionally left unchanged: they use the constant as a default
duration, not as an RTO bound on a retransmit timer.
Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/618
Signed-off-by: Kalpan Jani <kalpan.jani@mpiricsoftware.com>
Reviewed-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20260812-net-next-mptcp-misc-feat-7-3-v1-4-1905a818f6cb@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Only used in pm_userspace.c.
While at it, use the mptcp_userspace_pm_ prefix, like most functions in
this file: that makes it clear it is specific to this userspace PM.
Reviewed-by: Geliang Tang <geliang@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20260812-net-next-mptcp-misc-feat-7-3-v1-3-1905a818f6cb@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The data_ack and data_ack32 fields in struct mptcp_ext are no longer used
anywhere. Remove them from the structure and update mptcp_dump_mpext()
trace helper accordingly. Drop the data_ack field from the trace entry
and the corresponding output in TP_printk().
Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
Reviewed-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20260812-net-next-mptcp-misc-feat-7-3-v1-2-1905a818f6cb@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
extra_subflows is a u8 counter that can underflow if a decrement races
with or precedes an increment. While the recently fixed userspace PM
subflow creation path eliminated the primary cause, add defensive
WARN_ON_ONCE guards at both decrement sites to catch any remaining edge
cases rather than silently wrapping to 255.
Signed-off-by: Tao Cui <cuitao@kylinos.cn>
Reviewed-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20260812-net-next-mptcp-misc-feat-7-3-v1-1-1905a818f6cb@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Théo Lebrun says:
====================
net: macb: implement context swapping [part]
====================
Trivial cleanups from the larger resource management rework.
Link: https://patch.msgid.link/20260812-macb-context-v9-0-7ddbf5f715e0@bootlin.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
EMAC has never supported changing ring sizes: RX is hardcoded to 9 and
TX is the tiniest ring buffer you can imagine.
Make sure the operation fails early rather than silently succeed and
storing values in bp->configured_{rx,tx}_ring_size that are never read
in the EMAC case.
Signed-off-by: Théo Lebrun <theo.lebrun@bootlin.com>
Link: https://patch.msgid.link/20260812-macb-context-v9-7-7ddbf5f715e0@bootlin.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The tieoff descriptor is a RX DMA descriptor ring of size one. It gets
configured onto queues for Wake-on-LAN during system-wide suspend when
hardware does not support disabling individual queues
(MACB_CAPS_QUEUE_DISABLE).
MACB/GEM driver allocates it alongside the main RX ring
inside macb_alloc() at open. Free is done by macb_free() at close.
Change to allocate once at probe and free on probe failure or device
removal. This makes the tieoff descriptor lifetime much longer,
avoiding repeating coherent buffer allocation on each open/close cycle.
Main benefit: we dissociate its lifetime from the main ring's lifetime.
That way there is less work to be doing on resources (re)alloc. This
currently happens on close/open, but will soon also happen on context
swap operations (set_ringparam, change_mtu, set_channels, etc).
Reviewed-by: Nicolai Buchwitz <nb@tipi-net.de>
Signed-off-by: Théo Lebrun <theo.lebrun@bootlin.com>
Link: https://patch.msgid.link/20260812-macb-context-v9-6-7ddbf5f715e0@bootlin.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Enforce the reverse christmas tree convention in those functions:
macb_tx_error_task()
gem_rx_refill()
gem_rx()
macb_rx_frame()
macb_init_rx_ring()
macb_rx()
macb_rx_pending()
macb_start_xmit()
The goal is to minimise unrelated diff in future patches.
In macb_tx_error_task(), we fold the assignment into the declaration
statement.
Acked-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Nicolai Buchwitz <nb@tipi-net.de>
Signed-off-by: Théo Lebrun <theo.lebrun@bootlin.com>
Link: https://patch.msgid.link/20260812-macb-context-v9-5-7ddbf5f715e0@bootlin.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Variables are named q or queue_index. Types are int, unsigned int, u32
and u16. Use `unsigned int q` everywhere.
Skip over taprio functions. They use `u8 queue_id` which fits with the
`struct macb_queue_enst_config` field. Using `queue_id` everywhere
would be too verbose.
Reviewed-by: Nicolai Buchwitz <nb@tipi-net.de>
Signed-off-by: Théo Lebrun <theo.lebrun@bootlin.com>
Link: https://patch.msgid.link/20260812-macb-context-v9-4-7ddbf5f715e0@bootlin.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Follow MACB naming convention throughout on two aspects:
- Always name `struct macb *bp` rather than `lp`.
- Always name `struct macb_queue *queue` rather than `q`.
The latter is to reserve `q` for queue indexes.
Acked-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Nicolai Buchwitz <nb@tipi-net.de>
Signed-off-by: Théo Lebrun <theo.lebrun@bootlin.com>
Link: https://patch.msgid.link/20260812-macb-context-v9-3-7ddbf5f715e0@bootlin.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Here are all device pointer variable permutations inside MACB:
struct device *dev;
struct net_device *dev;
struct net_device *ndev;
struct net_device *netdev;
struct pci_dev *pdev; // inside macb_pci.c
struct phy_device *phy;
struct phy_device *phydev;
struct platform_device *pdev;
struct platform_device *plat_dev; // inside macb_pci.c
Unify to this convention:
struct device *dev;
struct net_device *netdev;
struct pci_dev *pci;
struct phy_device *phydev;
struct platform_device *pdev;
Ensure nothing slipped through using ctags tooling:
⟩ ctags -o - --kinds-c='{local}{member}{parameter}' \
--fields='{typeref}' drivers/net/ethernet/cadence/* | \
awk -F"\t" '
$NF~/struct:.*(device|dev) / {print $NF, $1}' | \
sort -u
typeref:struct:device * dev
typeref:struct:in_device * idev // ignored
typeref:struct:net_device * netdev
typeref:struct:pci_dev * pci
typeref:struct:phy_device * phydev
typeref:struct:platform_device * pdev
Also fix some printk() calls to use __func__ instead of hardcoding.
This silences some checkpatch.pl warnings and doesn't deserve a
separate commit.
Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Nicolai Buchwitz <nb@tipi-net.de>
Signed-off-by: Théo Lebrun <theo.lebrun@bootlin.com>
Link: https://patch.msgid.link/20260812-macb-context-v9-2-7ddbf5f715e0@bootlin.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Since commit 4df95131ea80 ("net/macb: change RX path for GEM") those
functions have not been only allocating or freeing consistent memory
mappings.
Rename from macb_alloc_consistent() to macb_alloc() and
from macb_free_consistent() to macb_free().
Acked-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Nicolai Buchwitz <nb@tipi-net.de>
Signed-off-by: Théo Lebrun <theo.lebrun@bootlin.com>
Link: https://patch.msgid.link/20260812-macb-context-v9-1-7ddbf5f715e0@bootlin.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2
Pull gfs2 updates from Andreas Gruenbacher:
- Don't cache unreferenced glocks: when a glock is no longer referenced
(for example, because the inode it protects is evicted), it is now
released as soon as possible instead of leaving it around until
memory pressure or an unmount forces it out.
For some workloads, this saves a lot of memory and speeds up unmounts
significantly.
- Harden gfs2_glock_hold() by making sure the caller holds a reference
and fix a related race in checking for the liveliness of glocks
between gdlm_bast() and gfs2_glock_cb().
* tag 'gfs2-for-7.3' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2:
gfs2: harden gfs2_glock_hold
gfs2: Remove the glock lru list and shrinker
gfs2: Skip dlm unlocks earlier
gfs2: Don't cache unreferenced glocks
gfs2: Enable automatic glock hash table shrinking
|
|
mv88e6352_pcs_link_check() ignores errors returned by
port_get_cmode(). If the port status register read fails,
mv88e6352_port_get_cmode() returns without setting cmode. The link check
then compares an uninitialized value and may incorrectly treat the PCS
as active.
Save the return value and fail the link check after releasing the
register lock. marvell_c22_pcs_get_state() initializes the reported link
state to down before calling the check, so a read failure is handled
safely until a later poll succeeds.
This issue was found by a static analysis checker and confirmed by manual
source review.
Fixes: 85764555442f ("net: dsa: mv88e6xxx: convert 88e6352 to phylink_pcs")
Signed-off-by: Ruoyu Wang <ruoyuw560@gmail.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Link: https://patch.msgid.link/20260813153131.3952970-1-ruoyuw560@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Pull xfs updates from Carlos Maiolino:
"There are no big standing out features on this window, so this
mostly consists on bug fixes and code refactoring.
The only user visible change that stands out is the support for
FALLOC_FL_WRITE_ZEROES added to this"
* tag 'xfs-merge-7.3' of git://git.kernel.org:/pub/scm/fs/xfs/xfs-linux: (23 commits)
xfs: validate attr entry pointer before field access
xfs: check split_sectors validity before bio_split call
xfs: use file target for post-log fsync fallback flush
xfs: restore nofs context unconditionally in xfs_trans_roll
xfs: add lockless xfs_buf_readahead_map fast path
xfs: move buffer locking out of xfs_find_get_buf
xfs: merge xfs_buf_reverify into xfs_buf_read_map
xfs: use goto based error unwinding in xfs_buf_read_map
xfs: don't reverify buffers in xfs_buf_readahead_map
xfs: use WRITE_ONCE to update b_flags
xfs: hide b_flags manipulation from code outside of xfs_buf.c
xfs: remove _XBF_LOGRECOVERY
xfs: remove spurious XBF_DONE clearing on readahead validation failure
xfs: split out a lower-level xfs_buf_get_map helper from xfs_find_get_buf
xfs: consolidate buffer locking in xfs_buf_get_map
xfs: don't get a pag reference in xfs_buf_get_map
xfs: use kmalloc_objs() instead of kmalloc() in xfs_da_grow_inode_int
xfs: mark internal metadir file creation helpers static
xfs: create rtgroup metadir inodes using xfs_metadir_create_file
xfs: create quota metadir inodes using xfs_metadir_create_file
...
|
|
A common mistake when trying to record system-wide profiles for a given
duration is running commands like 'perf record sleep 1' or 'perf stat
sleep 1' without passing '-a' / '--all-cpus'. When '-a' is omitted, perf
defaults to per-process monitoring of the sleep process itself, which
does not collect system-wide activity and records very few events.
Add a warning in evlist__prepare_workload() when the workload executable
is 'sleep' and system-wide mode is not enabled.
Assisted-by: Antigravity:gemini-3.6-flash
Signed-off-by: Ian Rogers <irogers@google.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
|
|
Describe the function view hierarchy (read-side function -> contending
writer function -> shared cachelines), the per-level indentation, and the
keys, with a worked example.
Document that reliable function attribution requires `iaddr` in
`--coalesce`, that the reader and writer may be the same function, and why
the coalesced function view cannot distinguish same-thread from
different-thread accesses in that case. Also document that verbose mode
includes code addresses in function rows.
Signed-off-by: Jiebin Sun <jiebin.sun@intel.com>
Reviewed-by: Tianyou Li <tianyou.li@intel.com>
Reviewed-by: Wangyang Guo <wangyang.guo@intel.com>
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Dapeng Mi <dapeng1.mi@linux.intel.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Thomas Falcon <thomas.falcon@intel.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
|
|
Add the browser front end: create/run/delete the hist_browser and add the
title. The d shortcut opens the existing per-cacheline detail view for the
selected level-3 cacheline. Level-3 entries retain the source cacheline
index, so the shortcut can locate the original entry without relying on a
potentially ambiguous virtual address.
Report a warning when the common model rejects a cacheline coalescing field
list without `iaddr`. Without it, the detail histograms may already have
merged samples from different functions and cannot support reliable
function attribution.
Keep visible-row accounting local to the function view by wrapping the
generic browser refresh callback and recounting the currently reachable
hierarchy before each redraw. This keeps navigation correct when a level-1
row is collapsed while level-3 descendants remain expanded, without adding
C2C-specific hooks to the shared hist_browser. Also handle Ctrl-C like the
other function-view exit keys.
Keep callchains hidden while the function browser runs, restoring the
user's setting while opening the cacheline detail view.
Wire the builder into perf_c2c__browse_function_view().
Signed-off-by: Jiebin Sun <jiebin.sun@intel.com>
Reviewed-by: Tianyou Li <tianyou.li@intel.com>
Reviewed-by: Wangyang Guo <wangyang.guo@intel.com>
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Dapeng Mi <dapeng1.mi@linux.intel.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Thomas Falcon <thomas.falcon@intel.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
|
|
Add the builder that walks the top-level cacheline entries and, for each
read-side function, correlates the functions that write the same lines
(level 2) and the specific cachelines they contend over (level 3) within
each retained detail histogram. Aggregate the write traffic per contending
function, resort by store count, and prune writers/functions with no
contention. The finalize pass then computes the Cycles % denominator from
the surviving level-1 entries after pruning, so the column shows each
function's share of the functions retained in the table rather than of the
whole recording -- the semantics documented for Cycles % in perf-c2c.txt.
Expose c2c_function__build() and c2c_function__reset() for the TUI front
end added by the next patch. The builder requires iaddr in the cacheline
coalescing fields and returns the completed hists through an output
argument. Validate the inputs before replacing an existing model.
Function-view entries do not carry callchains. Suppress callchain handling
while building and tearing down the model so the common API does not depend
on the caller's current callchain setting.
Signed-off-by: Jiebin Sun <jiebin.sun@intel.com>
Reviewed-by: Tianyou Li <tianyou.li@intel.com>
Reviewed-by: Wangyang Guo <wangyang.guo@intel.com>
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Dapeng Mi <dapeng1.mi@linux.intel.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Thomas Falcon <thomas.falcon@intel.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
|
|
Add the entry-creation layer: owned-reference child allocation and
insertion, and the level-1/2/3 lookup-or-create functions keyed by
function symbol (level 1 read-side, level 2 writer) and by the source
cacheline's existing index (level 3).
Give synthetic children normal entry operations and acquire their thread
and map-symbol references. This lets the hierarchy teardown use
hist_entry__delete() for the common fields while the function-view free
callback handles the private child tree and containing allocation.
Reuse cacheline_idx to preserve the source entry identity without adding
function-view-only state. Add c2c_function__find_cacheline() to locate the
original cacheline entry by the same index.
These are driven by the hierarchy builder in the next patch and are
__maybe_unused until then.
Signed-off-by: Jiebin Sun <jiebin.sun@intel.com>
Reviewed-by: Tianyou Li <tianyou.li@intel.com>
Reviewed-by: Wangyang Guo <wangyang.guo@intel.com>
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Dapeng Mi <dapeng1.mi@linux.intel.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Thomas Falcon <thomas.falcon@intel.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
|
|
Add the per-entry stats/cstats aggregation helpers and hierarchy teardown.
Child common fields are released through hist_entry__delete(), while the
function-view free callback handles the private child tree and containing
allocation. Also add a helper for pruning writer entries with no stores or
cacheline children.
These are used by the entry-creation and builder patches that follow and
are __maybe_unused until then.
Signed-off-by: Jiebin Sun <jiebin.sun@intel.com>
Reviewed-by: Tianyou Li <tianyou.li@intel.com>
Reviewed-by: Wangyang Guo <wangyang.guo@intel.com>
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Dapeng Mi <dapeng1.mi@linux.intel.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Thomas Falcon <thomas.falcon@intel.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
|
|
Add the parser that builds the function view's local HPP output and
sort lists from field strings. This includes dimension lookup, comparator
wrappers, c2c_fmt allocation, and the initialization entry points used by
the hierarchy builder.
The generic perf_hpp__setup_output_field() registers formats on the global
perf_hpp_list. Using it here would leave the function view's local list
without output columns and modify the cacheline view's list instead. Add
c2c_function_hists__setup_output_field() to append sort keys to the local
output list.
Signed-off-by: Jiebin Sun <jiebin.sun@intel.com>
Reviewed-by: Tianyou Li <tianyou.li@intel.com>
Reviewed-by: Wangyang Guo <wangyang.guo@intel.com>
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Dapeng Mi <dapeng1.mi@linux.intel.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Thomas Falcon <thomas.falcon@intel.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
|
|
Add renderers for the function view's Cycles %, Store count, and
hierarchy identity columns. The identity column renders the read-side
function, contending writer, or cacheline, with indentation for the
hierarchy level. Also add width and header helpers, estimated-cycle
calculation, comparators, and the dimension table that ties them together.
Clamp the identity renderer's returned length to its local buffer before
using it for pointer and padding calculations. This handles snprintf-style
would-have-been lengths without changing normal output.
The next patch connects these dimensions to the view's HPP lists, so the
symbols used only there are temporarily marked __maybe_unused.
Signed-off-by: Jiebin Sun <jiebin.sun@intel.com>
Reviewed-by: Tianyou Li <tianyou.li@intel.com>
Reviewed-by: Wangyang Guo <wangyang.guo@intel.com>
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Dapeng Mi <dapeng1.mi@linux.intel.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Thomas Falcon <thomas.falcon@intel.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
|
|
Add the initial common model for the c2c function view: model state and
small helpers shared by the hierarchy construction and formatting added
in later patches.
Build the model from util/ so it remains independent of the TUI and
command-private symbols.
Signed-off-by: Jiebin Sun <jiebin.sun@intel.com>
Reviewed-by: Tianyou Li <tianyou.li@intel.com>
Reviewed-by: Wangyang Guo <wangyang.guo@intel.com>
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Dapeng Mi <dapeng1.mi@linux.intel.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Thomas Falcon <thomas.falcon@intel.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
|
|
The VXLAN VNI filter entry policy declares the GROUP/GROUP6 address
attributes as NLA_BINARY with only a maximum length, so validate_nla()
accepts a payload shorter than the address. The GROUP consumer reads it
with nla_get_in_addr(), an unconditional 4-byte load, so a short
attribute over-reads up to 3 bytes of uninitialised slab data, which are
stored into remote_ip and echoed back via RTM_GETTUNNEL, disclosing
kernel memory.
Switch both entries to NLA_POLICY_EXACT_LEN() so the validator rejects
any GROUP/GROUP6 that is not exactly 4 / 16 bytes; a valid address is
always sent at full width.
Fixes: f9c4bb0b245c ("vxlan: vni filtering support on collect metadata device")
Reported-by: Weiming Shi <bestswngs@gmail.com>
Signed-off-by: Xiang Mei <xmei5@asu.edu>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://patch.msgid.link/20260812215341.763123-1-xmei5@asu.edu
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
ip_vs_ftp_get_addrport() accumulates decimal digits into a __u16
(hport) and into unsigned char (p[]) without checking for overflow.
A crafted FTP PASV/EPSV response with an over-long port or address
octet wraps the value, so the helper configures the data connection
with a truncated port/address.
The netfilter conntrack FTP helper had the same defect, fixed in
commit 2b413fc689ba ("netfilter: nf_conntrack_ftp: avoid u16
overflows"). Apply the equivalent fix here: widen the port accumulator
to u32 and reject values above 65535, and reject address octets above
255.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Joas Antonio dos Santos <joasantonio108@gmail.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
|
|
The rbtree set only builds the b-search array after the new ruleset has
been published through set ops .commit.
This exposes an empty set for a short time span which results in a bogus
mismatch for the following batch:
destroy table ip x
table ip x {
...
}
The same problem also affects the pipapo set backend which also provides
a set ops .commit interface too.
This patch moves the set ops .commit call right before building and
publishing the chain blob. The commit path now performs an early
handling of the DELSETELEM command to remove stale elements from the
clone before it is published via rcu. Note that DELSETELEM notifications
are still delivered in order. NEWSETELEM commands are handled after the
set is published, since this clears the previous genbit to 1 to prepare
the element for the next control plane transaction. This comes at the
cost of one extra iteration over the transaction list.
Suggested-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
|
|
This list is used to invoke the set .commit and .abort ops for the
rbtree and pipapo to run GC on expired elements and replace the current
datastructure view by the clone. For the rbtree, this also rebuild the
datapath b-search array.
From abort path, remove the set from the update_list if it is already
bound to rule, then the rule itself takes care of releasing the set and
its elements, otherwise, memleak is possible because set ops .abort
only deals with removing the set data structure, not the elements.
This is a preparation patch to call set .commit before processing the
transaction list for the rbtree, no functional changes are intended.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
|
|
Expose expectation flags included in the NF_CT_EXPECT_MASK bitmask
only. The DEAD flag is internal, do not expose it.
Fixes: b8b09dc2bf35 ("netfilter: nf_conntrack_expect: use conntrack GC to reap expectations")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
|
|
expectation
Consolidate the check for buggy expectations with DEAD flag on
insertion, which is called both by nf_ct_expect_related() and
nf_ct_expect_related_pair().
Fixes: e765c95faa10 ("netfilter: nf_conntrack_expect: bail out on insert dead expectations")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
|
|
All file:line references below are against v7.2-rc4 (ac5b0e5651b1). The
trace was captured on 7.2.0-rc6-kasan72rc6 (075b74841bd0), where the same
lines apply.
nft_obj_notify() is exported and reached from the packet path. Its only
in-tree caller is nft_quota_obj_eval() (net/netfilter/nft_quota.c:68),
which notifies with GFP_ATOMIC while evaluating a rule for a transiting
packet, holding no mutex.
Since commit 67cc570edaa0 ("netfilter: nf_tables: coalesce multiple
notifications into one skbuff") that notification is no longer sent
immediately. __nft_obj_notify() queues it onto nft_net->notify_list via
nft_notify_enqueue() (net/netfilter/nf_tables_api.c:1211), which is a bare
list_add_tail(). notify_list has no lock of its own
(include/net/netfilter/nf_tables.h:1951), it is serialised by commit_mutex:
the six other enqueue sites all run inside a netlink transaction, and the
drain in nft_commit_notify() (net/netfilter/nf_tables_api.c:10746) does
list_del() + kfree_skb() from nf_tables_commit() with commit_mutex held.
Sending packets through a chain that references a depleted quota object
therefore races an unlocked list_add_tail() against list_del() +
kfree_skb() on another CPU. The WRITE_ONCE(prev->next, new) in __list_add()
then stores through an sk_buff that has already been freed:
BUG: KASAN: slab-use-after-free in __nft_obj_notify+0x2c5/0x2d0
Write of size 8 at addr ff110001047183c0 by task poc/76
CPU: 0 UID: 1000 PID: 76 Comm: poc Tainted: G W 7.2.0-rc6-kasan72rc6 #4
Call Trace:
<IRQ>
__nft_obj_notify (include/linux/list.h:164 include/linux/list.h:191
net/netfilter/nf_tables_api.c:1211
net/netfilter/nf_tables_api.c:8743)
nft_quota_obj_eval (net/netfilter/nft_quota.c:68)
nft_do_chain_inet
nf_hook_slow
__ip_local_out
ip_push_pending_frames
udp_send_skb
udp_sendmsg
__x64_sys_sendto
Allocated by task 77:
__alloc_skb (net/core/skbuff.c:704)
__nft_obj_notify (include/net/netlink.h:1055
net/netfilter/nf_tables_api.c:8731)
nft_quota_obj_eval (net/netfilter/nft_quota.c:68)
nft_do_chain
Freed by task 79:
nf_tables_commit (include/linux/skbuff.h:1332
net/netfilter/nf_tables_api.c:10759
net/netfilter/nf_tables_api.c:11185)
nfnetlink_rcv_batch (net/netfilter/nfnetlink.c:574)
netlink_unicast
netlink_sendmsg
The buggy address belongs to the cache skbuff_head_cache of size 232
Queueing from the packet path is wrong even leaving the race aside:
notify_list is only drained by nft_commit_notify() from nf_tables_commit()
(:11185), so a notification enqueued outside a transaction is not sent
until some later netlink batch commits, if one ever does.
The gfp argument that nft_obj_notify() still takes is a leftover of the
pre-67cc570edaa0 behaviour, where this path called nfnetlink_send()
directly. Restore that: split the message construction out into
nft_obj_notify_alloc() and let each caller decide what to do with the skb.
nft_obj_notify(), the exported one reached from the packet path, sends it
straight away; nf_tables_obj_notify(), which runs under commit_mutex, keeps
queueing it, so transaction notifications are still coalesced.
Fixes: 67cc570edaa0 ("netfilter: nf_tables: coalesce multiple notifications into one skbuff")
Cc: stable@kernel.org
Reported-by: TencentOS Corvus AI <corvus@tencent.com>
Assisted-by: tencentos-corvus-ai:kimi-k3
Signed-off-by: Fourie Zhang <fouriezhang@tencent.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
|
|
Allocating mem via GFP_ATOMIC on delete is problematic, delete operations
should always succeed.
Do in-place substitution: When /cidr reaches 0 count (no more elements in
the range), move ranges stored later in the array forward and keep the
count 0 ones at the end.
INIT_CIDR() can then check count == 0 without a need to search next element
in the array.
To avoid problems on weakly ordered architectures, pack the structure so it
is only 32bit wide, then use READ/WRITE_ONCE to store both cidr and count.
atomically.
Also update comments to mention the possible presence of ignored
0-count-0-cidr structures at the end and need for seqcount.
seqcount is used to restart. This avoids bogus range misses.
Given: [0]: /29 [1]: /24
cpu1 reads slot 0. then, right after, cpu2 removes /29. count drops to 0,
so it updates array to: [0], /24, [1], /0 (count 0).
cpu1 then skips /28: slot 0 was already visited, but slot 1 already replaced.
Note that mtype_add() doesn't check mtype_add_cidr() return value.
Doing this here is useless noise as this code is extensively rewritten
in the rhashtable replacement patch.
Assisted-by: Claude:claude-sonnet-5
Fixes: 8e5fd2a55e24 ("netfilter: ipset: rework cidr bookkeeping")
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
|
|
NFQUEUE and nft_payload can hand packet data modified by userspace back
to the stack. Recent restrictions keep link and network headers stable,
but transport header fields can still be changed.
A packet can therefore keep the same network header and conntrack entry
while changing the transport header layout. For TCP, increasing doff can
make later helper or NAT code use a different transport-header base than
the parser used, and can make offsets point past skb->tail.
Extend NFQUEUE payload validation to check the final L4 protocol and
known base headers after IPv4 options or IPv6 extension headers. Reject
packets whose L4 protocol no longer matches an attached non-template
conntrack entry, and reject IP fragments that already have such a
conntrack entry before trying to validate transport headers. Unknown L4
protocols are left to their normal protocol handlers.
For nft payload writes, reject transport-header stores that overlap TCP
doff. nft_nh_write_ok() already rejects network-header protocol changes,
so keeping doff stable prevents nft payload writes from changing the TCP
header length underneath conntrack and helper users.
This patch is a follow up to commit df07998dfd40 ("netfilter: nftables:
restrict linklayer and network header writes") and commit 54f34607d184
("netfilter: nfnetlink_queue: restrict writes to network header").
Reported-by: Vega <vega@nebusec.ai>
Signed-off-by: Zhiling Zou <zhilinz@nebusec.ai>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
|
