kernel/linux.git/drivers/net/thunderbolt, branch v6.18.48

net: thunderbolt: Tear down DMA paths before stopping the rings

2026-08-19T16:17:41+00:00

[ Upstream commit 68bf02b6b4ad3f748c6db71fd77b6c0402d252f4 ] tbnet_tear_down() stops both rings and frees their frame buffers before calling tb_xdomain_disable_paths(). tb_ring_stop() zeroes the ring's descriptor base and tbnet_free_buffers() unmaps and frees the pages the frames sit in, so by the time __tb_path_deactivate_hop() polls the hop's 'pending' bit, anything still in flight has nowhere to drain to. The teardown sequence has been in this order since the driver was added. The setup path has not: commit ff7cd07f3064 ("net: thunderbolt: Enable DMA paths only after rings are enabled") moved the path enable to the end of tbnet_connected_work() and documented why: /* Both logins successful so enable the rings, high-speed DMA * paths and start the network device queue. * * Note we enable the DMA paths last to make sure we have primed * the Rx ring before any incoming packets are allowed to * arrive. */ Teardown was never updated to match, so the rings and the paths now come down in the same order they go up instead of in reverse. On an ASMedia ASM4242 host router the 'pending' bit then never clears: every teardown burns the full 500 ms timeout and __tb_path_deactivate_hop() returns -ETIMEDOUT. Raising the timeout to 5 s does not help, so the hop is not slow to drain, it never drains at all. The failure is invisible above the thunderbolt core. __tb_path_deactivate_hops() is void and only calls tb_port_warn(); tb_path_deactivate(), tb_tunnel_deactivate() and __tb_disconnect_xdomain_paths() are void as well, and tb_disconnect_xdomain_paths() ends in an unconditional "return 0". So tb_xdomain_disable_paths() reports success and the netdev_warn() below it never fires. Repeated teardowns eventually take the XDomain control channel down, after which the peer node is gone and only a power cycle brings the controller back. Deactivating the paths first fixes it. Measured with kretprobes on a stock v6.17 tree with no other patches applied, on a link that was up and had just carried traffic: before: __tb_path_deactivate_hop() returns 0 for the first hop, then -ETIMEDOUT for the second 500335 us later after: 0 for both, 525 us apart Alternating the two orderings ABBA over three load levels, four teardowns per arm: every teardown failed before the change (21 of 21 that ran), none failed after (0 of 24). The before arms ran short because the link died partway through. The same split shows up when the interface is enslaved to a bond instead of just brought down, which is how I ran into this in the first place. Throughput and latency after the change are unchanged. Hosts whose routers drain the hop despite the stale descriptor base see no functional difference, since the paths end up deactivated either way. Fixes: e69b6c02b4c3 ("net: Add support for networking over Thunderbolt cable") Signed-off-by: Fan XinRan Acked-by: Mika Westerberg Link: https://patch.msgid.link/20260803-b4-tbnet-teardown-v2-1-27de6a13ca2d@gmail.com Signed-off-by: Jakub Kicinski Signed-off-by: Sasha Levin

Revert "net: thunderbolt: Enable end-to-end flow control also in transmit"

2026-08-19T16:17:18+00:00

[ Upstream commit 1881f2efbf7f78dc0a79a387b29fde6ff56d3731 ] This reverts commit a8065af3346ebd7c76ebc113451fb3ba94cf7769. Per the USB4 spec, a Transmit Descriptor Ring with E2E flow control disabled does not require any credits to be available before the Host Interface Adapter Layer transmits a tunneled packet from it. Once E2E is enabled on that ring the controller must first obtain end-to-end credits. The ASMedia ASM4242 USB4 host router (PCI 1b21:2425) never delivers those credits. The controller does accept the configuration: reading the ring OPTIONS register back right after tb_ring_start() returns exactly what was written, including RING_FLAG_E2E_FLOW_CONTROL (bit 28) and the E2E HopID field. No credit ever arrives though, so the Tx ring's hardware consumer index never advances and the link carries no traffic at all. Measured on two hosts connected point to point, onboard ASM4242 on MSI X870E and X870, v6.17, stock drivers/net/thunderbolt/main.c with only this revert applied on top: before: 100% packet loss to the peer; thunderbolt0 is up and the XDomain handshake completes ("new host found"), but iperf3 fails with "No route to host" once the neighbour entry expires after: 0% packet loss, 0.28 ms RTT; iperf3 4.21 Gb/s one way and 5.17 Gb/s the other (5 runs each, stddev <= 0.02), 1 retransmit in 10 s An instrumented build additionally showed a frozen-Tx-consumer watchdog firing ~30k times in a 10 s window before this change. Rx-side E2E is not touched by this revert, so peers that do return credits keep receive-side flow control. ASMedia does not look like an isolated case. The out-of-tree thunderbolt-ibverbs project disables native E2E on AMD NHI by default, noting that "Strix Halo has reproduced TX completion wedges with multiple native E2E rings active" -- the same failure mode, on a different vendor. Since the driver has no way to tell in advance which host router returns the credits, going back to the previous behaviour looks safer than adding a quirk per affected part; Tx-side E2E can be reintroduced as an opt-in for controllers that are known to implement the credit return. Note that the reverted commit was not fixing a reported problem, it was derived from the spec wording alone, so this revert is not expected to regress a known workload. Cc'ing the original author in case there was one. Fixes: a8065af3346e ("net: thunderbolt: Enable end-to-end flow control also in transmit") Cc: zhangjianrong Signed-off-by: Fan Ye Acked-by: Mika Westerberg Link: https://patch.msgid.link/20260727123002.25225-1-fy15309206903@gmail.com Signed-off-by: Jakub Kicinski Signed-off-by: Sasha Levin

net: thunderbolt: Fix frags[] overflow by bounding frame_count

2026-07-24T14:17:03+00:00

commit 55d9895f89970501fe126d1026b586b04a224c27 upstream. tbnet_poll() assembles a multi-frame ThunderboltIP packet into one skb. The first frame goes into the skb linear area and every further frame is added as a page fragment. skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, page, hdr_size, frame_size, TBNET_RX_PAGE_SIZE - hdr_size); A packet of frame_count frames therefore ends up with frame_count - 1 fragments. tbnet_check_frame() only bounds the peer supplied frame_count to TBNET_RING_SIZE / 4 (64), which is far above MAX_SKB_FRAGS (17 by default). A peer that sends a packet of 19 or more small frames pushes nr_frags past MAX_SKB_FRAGS, so skb_add_rx_frag() writes past skb_shinfo()->frags[] and corrupts memory after the shared info. Tighten the start of packet bound to MAX_SKB_FRAGS + 1 so a packet can never produce more fragments than frags[] can hold. This matches the recent skb frags overflow fixes in other receive paths, for example f0813bcd2d9d ("net: wwan: t7xx: fix potential skb->frags overflow in RX path") and 600dc40554dc ("net: usb: cdc-phonet: fix skb frags[] overflow in rx_complete()"). Fixes: e69b6c02b4c3 ("net: Add support for networking over Thunderbolt cable") Cc: stable@vger.kernel.org Signed-off-by: Maoyi Xie Acked-by: Mika Westerberg Link: https://patch.msgid.link/178163152194.2486768.14724194232649760778@maoyixie.com Signed-off-by: Jakub Kicinski Signed-off-by: Greg Kroah-Hartman

net: thunderbolt: Enable end-to-end flow control also in transmit

2025-07-02T21:17:43+00:00

According to USB4 specification, if E2E flow control is disabled for the Transmit Descriptor Ring, the Host Interface Adapter Layer shall not require any credits to be available before transmitting a Tunneled Packet from this Transmit Descriptor Ring, so e2e flow control should be enabled in both directions. Acked-by: Mika Westerberg Link: https://lore.kernel.org/20250624153805.GC2824380@black.fi.intel.com Signed-off-by: zhangjianrong Link: https://patch.msgid.link/20250628093813.647005-1-zhangjianrong5@huawei.com Signed-off-by: Jakub Kicinski

net: thunderbolt: Fix the parameter passing of tb_xdomain_enable_paths()/tb_xdomain_disable_paths()

2025-07-02T21:15:09+00:00

According to the description of tb_xdomain_enable_paths(), the third parameter represents the transmit ring and the fifth parameter represents the receive ring. tb_xdomain_disable_paths() is the same case. [Jakub] Mika says: it works now because both rings ->hop is the same Acked-by: Mika Westerberg Link: https://lore.kernel.org/20250625051149.GD2824380@black.fi.intel.com Signed-off-by: zhangjianrong Link: https://patch.msgid.link/20250628094920.656658-1-zhangjianrong5@huawei.com Signed-off-by: Jakub Kicinski

net: thunderbolt: Fix TCPv6 GSO checksum calculation

2023-09-15T12:45:05+00:00

Alex reported that running ssh over IPv6 does not work with Thunderbolt/USB4 networking driver. The reason for that is that driver should call skb_is_gso() before calling skb_is_gso_v6(), and it should not return false after calculates the checksum successfully. This probably was a copy paste error from the original driver where it was done properly. Reported-by: Alex Balcanquall Fixes: e69b6c02b4c3 ("net: Add support for networking over Thunderbolt cable") Cc: stable@vger.kernel.org Signed-off-by: Mika Westerberg Reviewed-by: Eric Dumazet Reviewed-by: Jiri Pirko Reviewed-by: Jiri Pirko Signed-off-by: David S. Miller

net: thunderbolt: Fix typos in comments

2023-04-13T04:09:12+00:00

Fix two typos in comments: blongs -> belongs UPD -> UDP No functional changes. Signed-off-by: Mika Westerberg Reviewed-by: Simon Horman Reviewed-by: Andy Shevchenko Signed-off-by: Jakub Kicinski

net: thunderbolt: Fix sparse warnings in tbnet_xmit_csum_and_map()

2023-04-13T04:09:12+00:00

Fixes the following warning when the driver is built with sparse checks enabled: main.c:993:23: warning: incorrect type in initializer (different base types) main.c:993:23: expected restricted __wsum [usertype] wsum main.c:993:23: got restricted __be32 [usertype] No functional changes intended. Signed-off-by: Mika Westerberg Reviewed-by: Simon Horman Reviewed-by: Andy Shevchenko Signed-off-by: Jakub Kicinski

net: thunderbolt: Fix sparse warnings in tbnet_check_frame() and tbnet_poll()

2023-04-13T04:09:12+00:00

Fixes the following warnings when the driver is built with sparse checks enabled: main.c:767:47: warning: restricted __le32 degrades to integer main.c:775:47: warning: restricted __le16 degrades to integer main.c:776:44: warning: restricted __le16 degrades to integer main.c:876:40: warning: incorrect type in assignment (different base types) main.c:876:40: expected restricted __le32 [usertype] frame_size main.c:876:40: got unsigned int [assigned] [usertype] frame_size main.c:877:41: warning: incorrect type in assignment (different base types) main.c:877:41: expected restricted __le32 [usertype] frame_count main.c:877:41: got unsigned int [usertype] main.c:878:41: warning: incorrect type in assignment (different base types) main.c:878:41: expected restricted __le16 [usertype] frame_index main.c:878:41: got unsigned short [usertype] main.c:879:38: warning: incorrect type in assignment (different base types) main.c:879:38: expected restricted __le16 [usertype] frame_id main.c:879:38: got unsigned short [usertype] main.c:880:62: warning: restricted __le32 degrades to integer main.c:880:35: warning: restricted __le16 degrades to integer No functional changes intended. Signed-off-by: Mika Westerberg Reviewed-by: Simon Horman Reviewed-by: Andy Shevchenko Signed-off-by: Jakub Kicinski

net: thunderbolt: Add tracepoints

2023-01-13T05:19:30+00:00

These are useful when debugging various performance issues. Signed-off-by: Mika Westerberg Acked-by: Yehezkel Bernat Signed-off-by: Jakub Kicinski