[PATCH net-next v13 00/11] enic: SR-IOV V2 admin channel and MBOX protocol

[PATCH net-next v13 00/11] enic: SR-IOV V2 admin channel and MBOX protocol

[Satish Kharat]

This series adds the admin channel infrastructure and mailbox (MBOX)
protocol needed for V2 SR-IOV support in the enic driver.

The V2 SR-IOV design uses a direct PF-VF communication channel built on
dedicated WQ/RQ/CQ hardware resources and an MSI-X interrupt.

This series touches enic_remove(), which is also modified by a fix
already applied to net:

  enic: fix tx_hang_reset use-after-free on device removal
  https://git.kernel.org/netdev/net/c/ec680ea4ba1b

That fix was previously carried at the head of this series (patch 1 in
v10-v12); it has been split out and applied to net separately.  This
net-next series no longer includes it and applies cleanly on current
net-next, which already contains that fix (via the net -> net-next
merge); the overlapping enic_remove() teardown-ordering change is
incorporated in this series.

Firmware capability and admin channel infrastructure (patches 1-4):
  - Probe-time firmware feature check for V2 SR-IOV support
  - Admin channel open/close, RQ buffer management, CQ service
    with MSI-X interrupt and workqueue-based polling

MBOX protocol and VF enable (patches 5-10):
  - MBOX message types, core send/receive, PF and VF handlers
  - V2 SR-IOV enable wiring with admin channel setup
  - V2 VF probe with admin channel and PF registration

Patch 11 completes reset recovery for V2 VFs: the reset paths added
earlier in the series re-establish the admin channel only for the PF,
which left a VF unregistered and unable to exchange MBOX traffic after
a reset taken on the VF.

Known follow-ups (not blocking this series; the V2 PF data path is not
yet activated and V2 VFs are brought up only in a later series that wires
.sriov_configure):
  - Full request/reply correlation by msg_num requires the MBOX peer to
    echo the request number; this series retains the existing
    single-outstanding-request model with reply-type gating.
  - PF-side registration synchronization, initial link-state retry, and
    admin control-message throttling will be completed alongside the
    .sriov_configure enablement.

Signed-off-by: Satish Kharat <satishkh@cisco.com>
---
Changes in v13:
- Drop the tx_hang_reset use-after-free fix that was patch 1 in v10-v12;
  it is a fix (not net-next material) and has been applied to net
  separately, using disable_work_sync() to also close the watchdog/notify
  requeue race:
  https://git.kernel.org/netdev/net/c/ec680ea4ba1b
  [Jakub Kicinski]
- patch 3 (admin RQ buffers): allocate with kzalloc() so a short DMA
  cannot expose uninitialised heap in a received admin message.
  [Sashiko]
- patch 4 (admin CQ service): keep the admin MSI-X vector masked from
  request_irq() until the rings are initialised and filled during
  channel open, so an early or spurious interrupt cannot run the poll
  handler against uninitialised rings; reword the credit-return
  description to match the code (the unmask re-arms the vector
  independently of the returned credit count). [Sashiko; Jakub Kicinski]
- patch 4 (admin CQ service): on an admin RQ refill failure, reschedule
  the admin CQ poll work (a delayed_work) after a short delay to repost
  buffers and re-arm the RQ, so sustained memory pressure cannot drain
  the RQ to zero and permanently stall the admin channel.  Routing the
  retry through the poll work keeps the admin RQ ring owned by a single
  context, so refills never run concurrently. [Jakub Kicinski; Sashiko]
- patch 6 (MBOX core send): compute total_len as size_t and reject a
  payload larger than the admin buffer before taking mbox_lock,
  avoiding a u16 length overflow in the exported enic_mbox_send_msg().
  [Sashiko]
- patch 8 (MBOX VF handlers): annotate the single-outstanding mbox
  reply state (mbox_expected_reply) with READ_ONCE()/WRITE_ONCE(); it
  is written by the request thread and read by the admin CQ poll/
  dispatch context that runs the receive handlers.
- patch 9 (V2 SR-IOV enable): enic_link_check() returns early on a V2
  VF, since the PF admin-channel link-state notification is the sole
  carrier authority for such VFs; adaptive Rx coalescing is
  intentionally a PF-only responsibility.
- patch 9 (V2 SR-IOV enable): 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 are brought up via sysfs (.sriov_configure) in a follow-up, not
  here; neither must be auto-enabled through the legacy
  pci_enable_sriov() path at probe.
- patches 9-11: install the admin_rq/MBOX receive handler (via
  enic_mbox_init()) before the admin channel is opened and its
  interrupt unmasked, at the enable, VF-probe and reset-reopen sites,
  and close the admin channel on the reset-reopen error paths (patch
  11) so it is not left half-open.
- patch 11 (reset recovery): invalidate the VF's local registration
  flag before reopening the admin channel, so a failed reopen or
  re-handshake cannot leave a stale registered state that a later
  teardown would act on over a dead channel.
- Link to v12: https://patch.msgid.link/20260719-enic-sriov-v2-admin-channel-v2-v12-0-6ff2c617001d@cisco.com

Changes in v12:
- Rebased on net-next.
- patch 2: reworded the commit message for accuracy (no code change) and
  added Reviewed-by: Breno Leitao.
- patch 5: return the actual admin CQ interrupt credit count instead of a
  forced minimum of one, and correct the accompanying comment; the MSI-X
  vector is re-armed by the unmask bit independently of the credit value,
  so there is no functional change.
- Link to v11: https://patch.msgid.link/20260703-enic-sriov-v2-admin-channel-v2-v11-0-5b739f1fe9e5@cisco.com

Changes in v11:
- Set mbox_send_disabled at the start of enic_admin_channel_open() so an
  early error return cannot leave MBOX sends enabled against a half-open
  or freed admin_wq (patch 7)
- Document that enic_sriov_configure() is not yet wired (__maybe_unused)
  and that rtnl serialization vs the reset paths is added when it is
  wired in a follow-up (patch 10)
- Document the admin CQ poll-handler credit return: the pre-drain credit
  snapshot and the minimum one-credit unmask that re-arms the MSI-X
  vector; no functional change (patch 5)
- Improve the VF capability-check log so a PF rejection is not reported
  as "version too old" (patch 9)
- Clarify the enic_remove() teardown comment and note in the commit
  message that a V2 VF without admin WQ/RQ/CQ resources fails probe with
  -ENODEV (patch 11)
- Link to v10: https://patch.msgid.link/20260629-enic-sriov-v2-admin-channel-v2-v10-0-62569af83417@cisco.com

Changes in v10:
- Cancel tx_hang_reset work in enic_remove() to fix a pre-existing
  use-after-free when a TX timeout fires during device removal; carried
  as an independent fix at the head of the series (new patch 1)
  [Sashiko]
- Clarify in the patch 2 commit message that V2 VFs are only enabled via
  .sriov_configure, which rejects firmware without V2 support, so such
  firmware never exposes VFs (patch 2) [Sashiko]
- Track admin-channel up/down state and gate admin/MBOX operations on
  it, fixing a NULL pointer dereference when close() runs after a failed
  open() and when a reset fails to reopen the channel (patch 3)
  [Sashiko]
- Bound the admin message list with ENIC_ADMIN_MSG_MAX (256) to prevent
  a malicious VF from exhausting PF memory (patch 5) [Sashiko]
- Name the admin MSI-X interrupt with pci_name() instead of the
  not-yet-registered netdev name so it no longer appears as
  "eth%d-admin" in /proc/interrupts (patch 5) [Sashiko]
- Document the in-order admin CQ/RQ completion guarantee in a comment
  (patch 5) [Sashiko]
- On a VF, validate that admin MBOX messages are sourced from the PF
  before acting on them, rejecting spoofed link-state messages (patch 9)
  [Sashiko]
- Move the link_notify_work initialisation ahead of the VF setup block
  so a VF probe error path cannot cancel_work_sync() an uninitialised
  work item (patch 11) [Sashiko]
- Cancel link_notify_work in enic_remove() after unregister_netdev() to
  close the narrow window where enic_link_check() could schedule it just
  as SR-IOV was disabled, leaving the work to outlive vf_state (patch 11)
  [Sashiko]
- Re-establish the V2 VF admin channel and re-run PF registration after
  a driver-initiated device reset (the soft reset from a WQ/RQ error and
  the tx-hang reset from a TX timeout); previously only the PF recovered,
  so a reset taken on a VF left it unable to exchange MBOX traffic
  (new patch 12) [Sashiko]

Testing:
- Exercised on a Cisco VIC with multiple V2 VFs under a KASAN + lockdep
  + DMA-API-debug kernel.  VF resets (soft and tx-hang) and PF reset,
  including a 10x reset stress loop, re-established the admin channel and
  re-registered the VFs with no use-after-free, lockdep, or DMA-API
  warnings.  The admin channel, MBOX VF registration and PF link-state
  notification were verified to survive resets.  (The MAC/VLAN/spoofchk/
  trust/MTU control-plane operations are added by a follow-up series and
  are not part of this submission.)

- Link to v9: https://patch.msgid.link/20260617-enic-sriov-v2-admin-channel-v2-v9-0-37f5f5af4c93@cisco.com

Changes in v9:
- Use dma_rmb() instead of rmb() when reading admin RQ completion
  descriptors written by DMA (patch 4) [Sashiko]
- Use GFP_KERNEL instead of GFP_ATOMIC for admin RQ refill and for
  received-message allocation; both run in workqueue (process)
  context after the v8 NAPI-to-workqueue switch (patch 4) [Sashiko]
- Correct the enic_admin_msg comment to describe the workqueue
  enqueue path rather than NAPI (patch 4) [Sashiko]
- Set mbox_send_disabled in enic_admin_channel_close() so a MBOX
  send cannot race with channel teardown (patch 6) [Sashiko]
- Send the actual PF carrier state to a VF on registration instead
  of unconditionally reporting link up (patch 7) [Sashiko]
- Call reinit_completion() before setting mbox_expected_reply so a
  reply arriving between the two is not missed (patch 8) [Sashiko]
- Defer PF->VF link state notification to a workqueue and gate it on
  carrier transitions; enic_link_check() runs in the notify (atomic)
  context while the MBOX send sleeps on a mutex/completion (patch 9)
  [Sashiko]
- Clear ENIC_SRIOV_ENABLED and cancel the link-notify work before
  freeing per-VF state in the SR-IOV disable path, closing a
  use-after-free window against a concurrent link notification
  (patch 9) [Sashiko]
- Link to v8: https://patch.msgid.link/20260609-enic-sriov-v2-admin-channel-v2-v8-0-8ad8babbb826@cisco.com

Changes in v8:
- Replace NAPI polling with workqueue for admin CQ service — admin
  channel is low-frequency control traffic, not data path (patch 4)
  [Jakub Kicinski]
- Use explicit enum value (= 4) for VIC_FEATURE_SRIOV instead of
  placeholder VIC_FEATURE_PTP entry (patch 1) [Breno Leitao]
- Remove unnecessary rmb() in WQ CQ service (patch 4) [Jakub Kicinski]
- Remove admin_msg_drop_cnt counter (patch 4) [Simon Horman]
- Drop NAPI reschedule on RQ refill failure — the NAPI-to-workqueue
  switch removes the livelock and budget issues (patch 4) [Simon Horman]
- Remove unnecessary READ_ONCE/WRITE_ONCE on admin_rq_handler — all
  access is serialized by probe/remove (patch 6) [Jakub Kicinski]
- Fix checkpatch line-length warnings (patches 3, 5, 6)
- Rate-limit link state send failure and ACK error warnings (patch 7)
  [Jakub Kicinski]
- Correct enic_link_check comment to describe actual PF link state
  notification flow (patch 7) [Simon Horman]
- Correct mbox_expected_reply comment — serialization is by
  RTNL/probe, not mbox_lock (patch 8) [Jakub Kicinski]
- Wire enic_mbox_send_link_state() from enic_link_check() so PF
  notifies VFs on carrier change (patch 9) [Simon Horman]
- Fix commit message wording about MSI-X reservation (patch 10)
  [Simon Horman]
- Link to v7: https://patch.msgid.link/20260513-enic-sriov-v2-admin-channel-v2-v7-0-68b9f4141f4c@cisco.com

Changes in v7:
- Replace magic numbers in admin channel init with named macros
  and inline comments for MBOX descriptor encoding
  (patches 2, 6) [Paolo Abeni]
- Add defense-in-depth bounds check on admin RQ bytes_written (patch 4)
- Force NAPI reschedule on admin RQ refill failure (patch 4)
- Always unmask admin interrupt even with zero credits (patch 4)
- Reorder NAPI init before request_irq in admin channel open (patch 4)
- Remove redundant netdev_warn on admin msg enqueue kmalloc failure
  (patch 4) [Paolo Abeni]
- Add netdev_warn on admin WQ/RQ disable failure in close path
  (patch 2)
- Remove incorrect RES_TYPE_SRIOV_INTR interrupt allocation from
  admin channel open (patch 2); interrupt setup handled entirely
  in patch 4 using RES_TYPE_INTR_CTRL
- Rate-limit VF register/unregister log messages (patch 7) [Paolo Abeni]
- Add __aligned(8) to admin message data[] for strict-alignment
  safety (patch 4)
- Rate-limit MBOX handler error warnings (patch 7)
- Pre-allocate port profile array before pci_disable_sriov in V1
  disable path to avoid half-torn-down state on alloc failure (patch 9)
- Account for admin channel interrupt reservation in
  enic_set_intr_mode() and enic_adjust_resources() (patch 9) [Paolo Abeni]
- Clear admin_rq_handler in enic_admin_channel_close (patch 9)
- Quiesce admin channel (mask interrupt, disable NAPI, block MBOX
  sends) around soft reset (patch 9)
- Use WRITE_ONCE/READ_ONCE for mbox_send_disabled and
  admin_rq_handler across data-path/reset boundaries
  (patches 4, 6, 9)
- Fix commit message: reference enic_adjust_resources() alongside
  enic_set_intr_mode() (patch 10)
Investigated findings from automated review (Simon Horman / Sashiko):
- Race between probe-time feature check and VF proxy: false positive;
  detection runs at probe, enable runs from sriov_configure
- Struct alignment of __le32 after 2-byte mbox_hdr_embed: compiler
  inserts correct padding, no manual alignment needed
- Stale MBOX reply matching / reinit_completion race: single-flight
  design with mutex serialization prevents this
- cancel_work_sync vs MBOX unregister race: work cannot be
  re-triggered during the close window
- Link to v6: https://patch.msgid.link/20260503-enic-sriov-v2-admin-channel-v2-v6-0-0af4fbc2d86d@cisco.com

Changes in v6:
- Add explanatory comments documenting admin_cq[0] (WQ CQE size) and
  admin_cq[1] (RQ CQE size matching firmware enic_ext_cq() programming)
  allocations (patch 2)
- Enforce bytes_written from CQ descriptor when enqueuing admin RQ
  message; previously buf->len (allocation size) was passed, exposing
  uninitialized buffer memory beyond the real payload (patch 4)
- Drop admin RQ messages with TRUNCATED set or FCS_OK clear, gated by
  netdev_warn_once() (patch 4)
- Disable interrupt_enable on admin_cq[0]: WQ completions are polled
  synchronously inside enic_mbox_send_msg() and never raise an
  interrupt; matches admin_cq[1] (RQ) which does NAPI polling (patch 4)
- Add mbox_expected_reply gating in VF reply handlers (capability,
  register, unregister): drop replies whose type does not match the
  current waiter's expected type, avoiding spurious wakeup of an
  unrelated waiter from a stale reply that arrives after timeout
  (patch 8)
- Distinguish error returns in enic_mbox_vf_unregister(): -ETIMEDOUT
  (no reply received), -EACCES (PF rejected the unregister), 0 on
  success.  Previously all paths collapsed to a single -ETIMEDOUT
  (patch 8)
- Reserve one extra MSI-X slot in enic_set_intr_mode() when
  has_admin_channel is set so enic_admin_setup_intr() always has room
  to allocate at intr_count without exceeding intr_avail bounds when
  data queue count is maxed out (patch 10)
- Clarify in commit messages that .sriov_configure is intentionally
  not yet wired in this series and will be added in a follow-up after
  the necessary devcmd hardening lands (patch 9)
- Link to v5: https://patch.msgid.link/20260423-enic-sriov-v2-admin-channel-v2-v5-0-caa9f504a3dc@cisco.com

Changes in v5:
- Fix DMA-into-freed-memory race: call enic_admin_qp_type_set() before
  disabling RQ/WQ in both error and close paths (patch 3)
- Fix DMA mapping leak: enic_admin_wq_buf_clean() now unmaps and frees
  WQ buffers still held at close time after a send timeout (patch 3)
- Log rate-limited warning on admin RQ refill failure (patch 4)
- Add missing linux/types.h and linux/bits.h includes to enic_mbox.h
  (patch 5)
- Guard mbox_lock/mbox_comp init with mbox_initialized flag to prevent
  re-initialization on sriov_configure re-entry (patch 7)
- Clear VF registered state before sending unregister reply so PF does
  not treat a dead VF as still registered (patch 8)
- Gate VF-facing log messages with net_ratelimit() to prevent malicious
  VF from flooding PF dmesg (patch 8)
- Reject VF port profile requests when V2 SR-IOV is active since
  enic->pp is not reallocated for V2 VFs (patch 9)
- Move enic_sriov_detect_vf_type() before auto-enable check; skip
  probe-time auto-enable for V2 VFs (patch 9)
- Move admin channel close and VF unregister before unregister_netdev()
  in enic_remove() to prevent use-after-free on netdev (patch 10)
- Add comment in enic_reset() documenting that admin channel is not
  recovered after soft reset (patch 10)
- Bypass RES_TYPE_SRIOV_INTR check for V2 VFs in admin channel
  capability detection (patch 10)
- Link to v4: https://patch.msgid.link/20260411-enic-sriov-v2-admin-channel-v2-v4-0-f052326c2a57@cisco.com

Changes in v4:
- Fix reverse xmas tree variable ordering (patches 1, 6)
- Use kzalloc_obj instead of kzalloc with sizeof (patch 9)
- Add NULL check for pp allocation in V1 SR-IOV disable path (patch 9)
- Link to v3: https://lore.kernel.org/r/20260408-enic-sriov-v2-admin-channel-v2-v3-0-1d4999a03cec@cisco.com

Changes in v3:
- Use early-return pattern in enic_sriov_detect_vf_type to reduce
  nesting (patch 1) [Breno Leitao]
- Link to v2: https://lore.kernel.org/r/20260408-enic-sriov-v2-admin-channel-v2-v2-0-d05dd3623fd3@cisco.com

Changes in v2:
- Fix lines exceeding 80 columns (patches 4, 6, 7, 8)
- Add __maybe_unused to enic_sriov_configure and enic_sriov_v2_enable;
  .sriov_configure wiring deferred to a later series after devcmd
  hardening is in place (patch 9)
- Guard probe-time auto-enable to skip V2 VFs (patch 9)
- Link to v1: https://lore.kernel.org/r/20260406-enic-sriov-v2-admin-channel-v2-v1-0-82cc47636a78@cisco.com

---
Satish Kharat (11):
      enic: verify firmware supports V2 SR-IOV at probe time
      enic: add admin channel open and close for SR-IOV
      enic: add admin RQ buffer management
      enic: add admin CQ service with MSI-X interrupt and workqueue polling
      enic: define MBOX message types and header structures
      enic: add MBOX core send and receive for admin channel
      enic: add MBOX PF handlers for VF register and capability
      enic: add MBOX VF handlers for capability, register and link state
      enic: wire V2 SR-IOV enable with admin channel and MBOX
      enic: add V2 VF probe with admin channel and PF registration
      enic: re-establish V2 VF admin channel and PF registration after reset

 drivers/net/ethernet/cisco/enic/Makefile      |   3 +-
 drivers/net/ethernet/cisco/enic/enic.h        |  41 +-
 drivers/net/ethernet/cisco/enic/enic_admin.c  | 667 ++++++++++++++++++++++++++
 drivers/net/ethernet/cisco/enic/enic_admin.h  |  27 ++
 drivers/net/ethernet/cisco/enic/enic_main.c   | 431 ++++++++++++++++-
 drivers/net/ethernet/cisco/enic/enic_mbox.c   | 647 +++++++++++++++++++++++++
 drivers/net/ethernet/cisco/enic/enic_mbox.h   |  95 ++++
 drivers/net/ethernet/cisco/enic/enic_pp.c     |   5 +
 drivers/net/ethernet/cisco/enic/enic_res.c    |   4 +-
 drivers/net/ethernet/cisco/enic/vnic_cq.h     |   9 +
 drivers/net/ethernet/cisco/enic/vnic_devcmd.h |  13 +
 drivers/net/ethernet/cisco/enic/vnic_enet.h   |   4 +-
 12 files changed, 1924 insertions(+), 22 deletions(-)
---
base-commit: ac155a26750a595703e7dadff84735456d75a479
change-id: 20260404-enic-sriov-v2-admin-channel-v2-c0aa3e988833

Best regards,
--  
Satish Kharat <satishkh@cisco.com>

[PATCH net-next v13 01/11] enic: verify firmware supports V2 SR-IOV at probe time

[Satish Kharat]

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>
---
 drivers/net/ethernet/cisco/enic/enic_main.c   | 21 ++++++++++++++++++++-
 drivers/net/ethernet/cisco/enic/vnic_devcmd.h |  2 ++
 2 files changed, 22 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/cisco/enic/enic_main.c b/drivers/net/ethernet/cisco/enic/enic_main.c
index d98f7e7ccab9..b56e5c75ade3 100644
--- a/drivers/net/ethernet/cisco/enic/enic_main.c
+++ b/drivers/net/ethernet/cisco/enic/enic_main.c
@@ -2641,8 +2641,10 @@ static void enic_iounmap(struct enic *enic)
 static void enic_sriov_detect_vf_type(struct enic *enic)
 {
 	struct pci_dev *pdev = enic->pdev;
-	int pos;
+	u64 supported_versions, a1 = 0;
 	u16 vf_dev_id;
+	int pos;
+	int err;
 
 	if (enic_is_sriov_vf(enic) || enic_is_dynamic(enic))
 		return;
@@ -2669,6 +2671,23 @@ static void enic_sriov_detect_vf_type(struct enic *enic)
 		enic->vf_type = ENIC_VF_TYPE_NONE;
 		break;
 	}
+
+	if (enic->vf_type != ENIC_VF_TYPE_V2)
+		return;
+
+	/* A successful command means firmware recognizes
+	 * VIC_FEATURE_SRIOV; supported_versions is available
+	 * for sub-feature versioning in the future.
+	 */
+	err = vnic_dev_get_supported_feature_ver(enic->vdev,
+						 VIC_FEATURE_SRIOV,
+						 &supported_versions,
+						 &a1);
+	if (err) {
+		dev_warn(&pdev->dev,
+			 "SR-IOV V2 not supported by current firmware. Upgrade to VIC FW 5.3(4.72) or higher.\n");
+		enic->vf_type = ENIC_VF_TYPE_NONE;
+	}
 }
 #endif
 
diff --git a/drivers/net/ethernet/cisco/enic/vnic_devcmd.h b/drivers/net/ethernet/cisco/enic/vnic_devcmd.h
index 605ef17f967e..3b6efa743dba 100644
--- a/drivers/net/ethernet/cisco/enic/vnic_devcmd.h
+++ b/drivers/net/ethernet/cisco/enic/vnic_devcmd.h
@@ -734,6 +734,8 @@ enum vic_feature_t {
 	VIC_FEATURE_VXLAN,
 	VIC_FEATURE_RDMA,
 	VIC_FEATURE_VXLAN_PATCH,
+	/* slot 3 reserved for firmware VIC_FEATURE_PTP */
+	VIC_FEATURE_SRIOV	= 4,
 	VIC_FEATURE_MAX,
 };
 

-- 
2.43.0

[PATCH net-next v13 02/11] enic: add admin channel open and close for SR-IOV

[Satish Kharat]

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>
---
 drivers/net/ethernet/cisco/enic/Makefile      |   3 +-
 drivers/net/ethernet/cisco/enic/enic.h        |   5 +
 drivers/net/ethernet/cisco/enic/enic_admin.c  | 227 ++++++++++++++++++++++++++
 drivers/net/ethernet/cisco/enic/enic_admin.h  |  15 ++
 drivers/net/ethernet/cisco/enic/vnic_cq.h     |   9 +
 drivers/net/ethernet/cisco/enic/vnic_devcmd.h |  11 ++
 6 files changed, 269 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/cisco/enic/Makefile b/drivers/net/ethernet/cisco/enic/Makefile
index a96b8332e6e2..7ae72fefc99a 100644
--- a/drivers/net/ethernet/cisco/enic/Makefile
+++ b/drivers/net/ethernet/cisco/enic/Makefile
@@ -3,5 +3,6 @@ obj-$(CONFIG_ENIC) := enic.o
 
 enic-y := enic_main.o vnic_cq.o vnic_intr.o vnic_wq.o \
 	enic_res.o enic_dev.o enic_pp.o vnic_dev.o vnic_rq.o vnic_vic.o \
-	enic_ethtool.o enic_api.o enic_clsf.o enic_rq.o enic_wq.o
+	enic_ethtool.o enic_api.o enic_clsf.o enic_rq.o enic_wq.o \
+	enic_admin.o
 
diff --git a/drivers/net/ethernet/cisco/enic/enic.h b/drivers/net/ethernet/cisco/enic/enic.h
index 08472420f3a1..398227448b37 100644
--- a/drivers/net/ethernet/cisco/enic/enic.h
+++ b/drivers/net/ethernet/cisco/enic/enic.h
@@ -292,6 +292,11 @@ struct enic {
 
 	/* Admin channel resources for SR-IOV MBOX */
 	bool has_admin_channel;
+	/* true only while the admin WQ/RQ/CQ are allocated and enabled; gates
+	 * enic_admin_channel_close() so it is a no-op after a failed (re)open
+	 * left the resources freed.
+	 */
+	bool admin_chan_up;
 	struct vnic_wq admin_wq;
 	struct vnic_rq admin_rq;
 	struct vnic_cq admin_cq[2];
diff --git a/drivers/net/ethernet/cisco/enic/enic_admin.c b/drivers/net/ethernet/cisco/enic/enic_admin.c
new file mode 100644
index 000000000000..50b46b92c88f
--- /dev/null
+++ b/drivers/net/ethernet/cisco/enic/enic_admin.c
@@ -0,0 +1,227 @@
+// SPDX-License-Identifier: GPL-2.0-only
+// Copyright 2025 Cisco Systems, Inc.  All rights reserved.
+
+#include <linux/kernel.h>
+#include <linux/netdevice.h>
+
+#include "vnic_dev.h"
+#include "vnic_wq.h"
+#include "vnic_rq.h"
+#include "vnic_cq.h"
+#include "vnic_intr.h"
+#include "vnic_resource.h"
+#include "vnic_devcmd.h"
+#include "enic.h"
+#include "enic_admin.h"
+#include "cq_desc.h"
+#include "wq_enet_desc.h"
+#include "rq_enet_desc.h"
+
+/* Clean up any admin WQ buffers still held by hardware at close time.
+ * Normally buffers are freed inline after send completion, but a timed-out
+ * send intentionally leaves the buffer live until the queue is stopped.
+ */
+static void enic_admin_wq_buf_clean(struct vnic_wq *wq,
+				    struct vnic_wq_buf *buf)
+{
+	struct enic *enic = vnic_dev_priv(wq->vdev);
+
+	if (buf->os_buf) {
+		dma_unmap_single(&enic->pdev->dev, buf->dma_addr,
+				 buf->len, DMA_TO_DEVICE);
+		kfree(buf->os_buf);
+		buf->os_buf = NULL;
+	}
+}
+
+/* No-op: admin RQ buffer teardown is handled in enic_admin_channel_close */
+static void enic_admin_rq_buf_clean(struct vnic_rq *rq,
+				    struct vnic_rq_buf *buf)
+{
+}
+
+static int enic_admin_qp_type_set(struct enic *enic, u32 enable)
+{
+	u64 a0 = QP_TYPE_ADMIN, a1 = enable;
+	int wait = 1000;
+	int err;
+
+	spin_lock_bh(&enic->devcmd_lock);
+	err = vnic_dev_cmd(enic->vdev, CMD_QP_TYPE_SET, &a0, &a1, wait);
+	spin_unlock_bh(&enic->devcmd_lock);
+
+	return err;
+}
+
+static int enic_admin_alloc_resources(struct enic *enic)
+{
+	int err;
+
+	err = vnic_wq_alloc_with_type(enic->vdev, &enic->admin_wq, 0,
+				      ENIC_ADMIN_DESC_COUNT,
+				      sizeof(struct wq_enet_desc),
+				      RES_TYPE_ADMIN_WQ);
+	if (err)
+		return err;
+
+	err = vnic_rq_alloc_with_type(enic->vdev, &enic->admin_rq, 0,
+				      ENIC_ADMIN_DESC_COUNT,
+				      sizeof(struct rq_enet_desc),
+				      RES_TYPE_ADMIN_RQ);
+	if (err)
+		goto free_wq;
+
+	/* admin_cq[0] is the WQ completion queue.  WQ CQEs are always
+	 * 16 bytes wide; firmware always writes 16-byte CQEs for WQ
+	 * completions on every WQ, including the admin channel WQ.
+	 * Use sizeof(struct cq_desc) accordingly.
+	 */
+	err = vnic_cq_alloc_with_type(enic->vdev, &enic->admin_cq[0], 0,
+				      ENIC_ADMIN_DESC_COUNT,
+				      sizeof(struct cq_desc),
+				      RES_TYPE_ADMIN_CQ);
+	if (err)
+		goto free_rq;
+
+	/* admin_cq[1] is the RQ completion queue.  Its descriptor size
+	 * must match what firmware writes.  enic_ext_cq() called earlier
+	 * in probe issues CMD_CQ_ENTRY_SIZE_SET for VNIC_RQ_ALL,
+	 * programming firmware to write CQ entries of (16 << enic->ext_cq)
+	 * bytes for every RQ CQ on the vNIC, including the admin RQ CQ.
+	 * Allocating with the same size keeps the host poller and
+	 * firmware in lockstep:
+	 *
+	 *   - The color/valid bit lives at byte (desc_size - 1) of every
+	 *     cq_enet_rq_desc[_32|_64] variant, so enic_admin_cq_color()
+	 *     reads it from the correct offset.
+	 *   - Only the first 15 bytes of the descriptor (vlan,
+	 *     bytes_written_flags, ...) are accessed by the admin path;
+	 *     these fields are identical across all three variants (see
+	 *     comment in enic_rq.c above cq_enet_rq_desc_dec()).
+	 */
+	err = vnic_cq_alloc_with_type(enic->vdev, &enic->admin_cq[1], 1,
+				      ENIC_ADMIN_DESC_COUNT,
+				      16 << enic->ext_cq,
+				      RES_TYPE_ADMIN_CQ);
+	if (err)
+		goto free_cq0;
+
+	return 0;
+
+free_cq0:
+	vnic_cq_free(&enic->admin_cq[0]);
+free_rq:
+	vnic_rq_free(&enic->admin_rq);
+free_wq:
+	vnic_wq_free(&enic->admin_wq);
+	return err;
+}
+
+static void enic_admin_free_resources(struct enic *enic)
+{
+	vnic_cq_free(&enic->admin_cq[1]);
+	vnic_cq_free(&enic->admin_cq[0]);
+	vnic_rq_free(&enic->admin_rq);
+	vnic_wq_free(&enic->admin_wq);
+}
+
+static void enic_admin_init_resources(struct enic *enic)
+{
+	vnic_wq_init(&enic->admin_wq,
+		     0, 0, 0); /* cq_index, err_intr_enable, err_intr_offset */
+	vnic_rq_init(&enic->admin_rq,
+		     1, 0, 0); /* cq_index, err_intr_enable, err_intr_offset */
+	vnic_cq_init(&enic->admin_cq[0],
+		     VNIC_CQ_FC_DISABLE,
+		     VNIC_CQ_COLOR_ENABLE,
+		     0, 0, 1, /* cq_head, cq_tail, cq_tail_color */
+		     VNIC_CQ_INTR_DISABLE,
+		     VNIC_CQ_ENTRY_ENABLE,
+		     VNIC_CQ_MSG_DISABLE,
+		     0, /* interrupt_offset */
+		     0 /* cq_message_addr */);
+	vnic_cq_init(&enic->admin_cq[1],
+		     VNIC_CQ_FC_DISABLE,
+		     VNIC_CQ_COLOR_ENABLE,
+		     0, 0, 1, /* cq_head, cq_tail, cq_tail_color */
+		     VNIC_CQ_INTR_DISABLE,
+		     VNIC_CQ_ENTRY_ENABLE,
+		     VNIC_CQ_MSG_DISABLE,
+		     0, /* interrupt_offset */
+		     0 /* cq_message_addr */);
+}
+
+int enic_admin_channel_open(struct enic *enic)
+{
+	int err;
+
+	if (!enic->has_admin_channel)
+		return -ENODEV;
+
+	err = enic_admin_alloc_resources(enic);
+	if (err) {
+		netdev_err(enic->netdev,
+			   "Failed to alloc admin channel resources: %d\n",
+			   err);
+		return err;
+	}
+
+	enic_admin_init_resources(enic);
+
+	vnic_wq_enable(&enic->admin_wq);
+	vnic_rq_enable(&enic->admin_rq);
+
+	err = enic_admin_qp_type_set(enic, QP_ENABLE);
+	if (err) {
+		netdev_err(enic->netdev,
+			   "Failed to set admin QP type: %d\n", err);
+		goto disable_queues;
+	}
+
+	enic->admin_chan_up = true;
+
+	return 0;
+
+disable_queues:
+	enic_admin_qp_type_set(enic, QP_DISABLE);
+	if (vnic_wq_disable(&enic->admin_wq))
+		netdev_warn(enic->netdev, "Failed to disable admin WQ\n");
+	if (vnic_rq_disable(&enic->admin_rq))
+		netdev_warn(enic->netdev, "Failed to disable admin RQ\n");
+	enic_admin_free_resources(enic);
+	return err;
+}
+
+void enic_admin_channel_close(struct enic *enic)
+{
+	int err;
+
+	if (!enic->has_admin_channel)
+		return;
+
+	/* Nothing to tear down if the channel was never (re)opened, e.g. a
+	 * failed enic_admin_channel_open() in probe or in the reset path;
+	 * otherwise the disable/clean calls below dereference freed resources.
+	 */
+	if (!enic->admin_chan_up)
+		return;
+
+	enic_admin_qp_type_set(enic, QP_DISABLE);
+
+	err = vnic_wq_disable(&enic->admin_wq);
+	if (err)
+		netdev_warn(enic->netdev,
+			    "Failed to disable admin WQ: %d\n", err);
+	err = vnic_rq_disable(&enic->admin_rq);
+	if (err)
+		netdev_warn(enic->netdev,
+			    "Failed to disable admin RQ: %d\n", err);
+
+	vnic_wq_clean(&enic->admin_wq, enic_admin_wq_buf_clean);
+	vnic_rq_clean(&enic->admin_rq, enic_admin_rq_buf_clean);
+	vnic_cq_clean(&enic->admin_cq[0]);
+	vnic_cq_clean(&enic->admin_cq[1]);
+	enic_admin_free_resources(enic);
+
+	enic->admin_chan_up = false;
+}
diff --git a/drivers/net/ethernet/cisco/enic/enic_admin.h b/drivers/net/ethernet/cisco/enic/enic_admin.h
new file mode 100644
index 000000000000..569aadeb9312
--- /dev/null
+++ b/drivers/net/ethernet/cisco/enic/enic_admin.h
@@ -0,0 +1,15 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/* Copyright 2025 Cisco Systems, Inc.  All rights reserved. */
+
+#ifndef _ENIC_ADMIN_H_
+#define _ENIC_ADMIN_H_
+
+#define ENIC_ADMIN_DESC_COUNT	64
+#define ENIC_ADMIN_BUF_SIZE	2048
+
+struct enic;
+
+int enic_admin_channel_open(struct enic *enic);
+void enic_admin_channel_close(struct enic *enic);
+
+#endif /* _ENIC_ADMIN_H_ */
diff --git a/drivers/net/ethernet/cisco/enic/vnic_cq.h b/drivers/net/ethernet/cisco/enic/vnic_cq.h
index d46d4d2ef6bb..35ffa3230713 100644
--- a/drivers/net/ethernet/cisco/enic/vnic_cq.h
+++ b/drivers/net/ethernet/cisco/enic/vnic_cq.h
@@ -76,6 +76,15 @@ int vnic_cq_alloc(struct vnic_dev *vdev, struct vnic_cq *cq, unsigned int index,
 int vnic_cq_alloc_with_type(struct vnic_dev *vdev, struct vnic_cq *cq,
 			    unsigned int index, unsigned int desc_count,
 			    unsigned int desc_size, unsigned int res_type);
+#define VNIC_CQ_FC_ENABLE	1
+#define VNIC_CQ_FC_DISABLE	0
+#define VNIC_CQ_COLOR_ENABLE	1
+#define VNIC_CQ_INTR_ENABLE	1
+#define VNIC_CQ_INTR_DISABLE	0
+#define VNIC_CQ_ENTRY_ENABLE	1
+#define VNIC_CQ_MSG_ENABLE	1
+#define VNIC_CQ_MSG_DISABLE	0
+
 void vnic_cq_init(struct vnic_cq *cq, unsigned int flow_control_enable,
 	unsigned int color_enable, unsigned int cq_head, unsigned int cq_tail,
 	unsigned int cq_tail_color, unsigned int interrupt_enable,
diff --git a/drivers/net/ethernet/cisco/enic/vnic_devcmd.h b/drivers/net/ethernet/cisco/enic/vnic_devcmd.h
index 3b6efa743dba..90ca06691ebd 100644
--- a/drivers/net/ethernet/cisco/enic/vnic_devcmd.h
+++ b/drivers/net/ethernet/cisco/enic/vnic_devcmd.h
@@ -455,8 +455,19 @@ enum vnic_devcmd_cmd {
 	 */
 	CMD_CQ_ENTRY_SIZE_SET = _CMDC(_CMD_DIR_WRITE, _CMD_VTYPE_ENET, 90),
 
+	/*
+	 * Set queue pair type (admin or data)
+	 * in: (u32) a0 = queue pair type (0 = admin, 1 = data)
+	 * in: (u32) a1 = enable (1) / disable (0)
+	 */
+	CMD_QP_TYPE_SET = _CMDC(_CMD_DIR_WRITE, _CMD_VTYPE_ENET, 97),
 };
 
+#define QP_TYPE_ADMIN	0
+#define QP_TYPE_DATA	1
+#define QP_ENABLE	1
+#define QP_DISABLE	0
+
 /* CMD_ENABLE2 flags */
 #define CMD_ENABLE2_STANDBY 0x0
 #define CMD_ENABLE2_ACTIVE  0x1

-- 
2.43.0

[PATCH net-next v13 03/11] enic: add admin RQ buffer management

[Satish Kharat]

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>
---
 drivers/net/ethernet/cisco/enic/enic_admin.c | 66 +++++++++++++++++++++++++++-
 1 file changed, 64 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/cisco/enic/enic_admin.c b/drivers/net/ethernet/cisco/enic/enic_admin.c
index 50b46b92c88f..edd0a2d3994b 100644
--- a/drivers/net/ethernet/cisco/enic/enic_admin.c
+++ b/drivers/net/ethernet/cisco/enic/enic_admin.c
@@ -3,6 +3,7 @@
 
 #include <linux/kernel.h>
 #include <linux/netdevice.h>
+#include <linux/dma-mapping.h>
 
 #include "vnic_dev.h"
 #include "vnic_wq.h"
@@ -34,10 +35,63 @@ static void enic_admin_wq_buf_clean(struct vnic_wq *wq,
 	}
 }
 
-/* No-op: admin RQ buffer teardown is handled in enic_admin_channel_close */
 static void enic_admin_rq_buf_clean(struct vnic_rq *rq,
 				    struct vnic_rq_buf *buf)
 {
+	struct enic *enic = vnic_dev_priv(rq->vdev);
+
+	if (!buf->os_buf)
+		return;
+
+	dma_unmap_single(&enic->pdev->dev, buf->dma_addr, buf->len,
+			 DMA_FROM_DEVICE);
+	kfree(buf->os_buf);
+	buf->os_buf = NULL;
+}
+
+static int enic_admin_rq_post_one(struct enic *enic, gfp_t gfp)
+{
+	struct vnic_rq *rq = &enic->admin_rq;
+	struct rq_enet_desc *desc;
+	dma_addr_t dma_addr;
+	void *buf;
+
+	buf = kzalloc(ENIC_ADMIN_BUF_SIZE, gfp);
+	if (!buf)
+		return -ENOMEM;
+
+	dma_addr = dma_map_single(&enic->pdev->dev, buf, ENIC_ADMIN_BUF_SIZE,
+				  DMA_FROM_DEVICE);
+	if (dma_mapping_error(&enic->pdev->dev, dma_addr)) {
+		kfree(buf);
+		return -ENOMEM;
+	}
+
+	desc = vnic_rq_next_desc(rq);
+	rq_enet_desc_enc(desc, (u64)dma_addr | VNIC_PADDR_TARGET,
+			 RQ_ENET_TYPE_ONLY_SOP, ENIC_ADMIN_BUF_SIZE);
+	vnic_rq_post(rq, buf, 0, dma_addr, ENIC_ADMIN_BUF_SIZE, 0);
+
+	return 0;
+}
+
+static int enic_admin_rq_fill(struct enic *enic, gfp_t gfp)
+{
+	struct vnic_rq *rq = &enic->admin_rq;
+	int err;
+
+	while (vnic_rq_desc_avail(rq) > 0) {
+		err = enic_admin_rq_post_one(enic, gfp);
+		if (err)
+			return err;
+	}
+
+	return 0;
+}
+
+static void enic_admin_rq_drain(struct enic *enic)
+{
+	vnic_rq_clean(&enic->admin_rq, enic_admin_rq_buf_clean);
 }
 
 static int enic_admin_qp_type_set(struct enic *enic, u32 enable)
@@ -171,6 +225,13 @@ int enic_admin_channel_open(struct enic *enic)
 	vnic_wq_enable(&enic->admin_wq);
 	vnic_rq_enable(&enic->admin_rq);
 
+	err = enic_admin_rq_fill(enic, GFP_KERNEL);
+	if (err) {
+		netdev_err(enic->netdev,
+			   "Failed to fill admin RQ buffers: %d\n", err);
+		goto disable_queues;
+	}
+
 	err = enic_admin_qp_type_set(enic, QP_ENABLE);
 	if (err) {
 		netdev_err(enic->netdev,
@@ -188,6 +249,7 @@ int enic_admin_channel_open(struct enic *enic)
 		netdev_warn(enic->netdev, "Failed to disable admin WQ\n");
 	if (vnic_rq_disable(&enic->admin_rq))
 		netdev_warn(enic->netdev, "Failed to disable admin RQ\n");
+	enic_admin_rq_drain(enic);
 	enic_admin_free_resources(enic);
 	return err;
 }
@@ -218,7 +280,7 @@ void enic_admin_channel_close(struct enic *enic)
 			    "Failed to disable admin RQ: %d\n", err);
 
 	vnic_wq_clean(&enic->admin_wq, enic_admin_wq_buf_clean);
-	vnic_rq_clean(&enic->admin_rq, enic_admin_rq_buf_clean);
+	enic_admin_rq_drain(enic);
 	vnic_cq_clean(&enic->admin_cq[0]);
 	vnic_cq_clean(&enic->admin_cq[1]);
 	enic_admin_free_resources(enic);

-- 
2.43.0

[PATCH net-next v13 04/11] enic: add admin CQ service with MSI-X interrupt and workqueue polling

[Satish Kharat]

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>
---
 drivers/net/ethernet/cisco/enic/enic.h       |   8 +
 drivers/net/ethernet/cisco/enic/enic_admin.c | 345 ++++++++++++++++++++++++++-
 drivers/net/ethernet/cisco/enic/enic_admin.h |  12 +
 3 files changed, 361 insertions(+), 4 deletions(-)

diff --git a/drivers/net/ethernet/cisco/enic/enic.h b/drivers/net/ethernet/cisco/enic/enic.h
index 398227448b37..0f08de228b36 100644
--- a/drivers/net/ethernet/cisco/enic/enic.h
+++ b/drivers/net/ethernet/cisco/enic/enic.h
@@ -301,6 +301,14 @@ struct enic {
 	struct vnic_rq admin_rq;
 	struct vnic_cq admin_cq[2];
 	struct vnic_intr admin_intr;
+	struct delayed_work admin_poll_work;
+	unsigned int admin_intr_index;
+	struct work_struct admin_msg_work;
+	spinlock_t admin_msg_lock;	/* protects admin_msg_list */
+	struct list_head admin_msg_list;
+	unsigned int admin_msg_count;	/* current depth of admin_msg_list */
+	void (*admin_rq_handler)(struct enic *enic, void *buf,
+				 unsigned int len);
 };
 
 static inline struct net_device *vnic_get_netdev(struct vnic_dev *vdev)
diff --git a/drivers/net/ethernet/cisco/enic/enic_admin.c b/drivers/net/ethernet/cisco/enic/enic_admin.c
index edd0a2d3994b..65aec9651a6f 100644
--- a/drivers/net/ethernet/cisco/enic/enic_admin.c
+++ b/drivers/net/ethernet/cisco/enic/enic_admin.c
@@ -4,6 +4,7 @@
 #include <linux/kernel.h>
 #include <linux/netdevice.h>
 #include <linux/dma-mapping.h>
+#include <linux/interrupt.h>
 
 #include "vnic_dev.h"
 #include "vnic_wq.h"
@@ -15,9 +16,15 @@
 #include "enic.h"
 #include "enic_admin.h"
 #include "cq_desc.h"
+#include "cq_enet_desc.h"
 #include "wq_enet_desc.h"
 #include "rq_enet_desc.h"
 
+/* Retry interval for a failed admin RQ refill.  Short so the control
+ * channel recovers quickly once memory is available again.
+ */
+#define ENIC_ADMIN_RQ_REFILL_RETRY_MS	100
+
 /* Clean up any admin WQ buffers still held by hardware at close time.
  * Normally buffers are freed inline after send completion, but a timed-out
  * send intentionally leaves the buffer live until the queue is stopped.
@@ -94,6 +101,283 @@ static void enic_admin_rq_drain(struct enic *enic)
 	vnic_rq_clean(&enic->admin_rq, enic_admin_rq_buf_clean);
 }
 
+static unsigned int enic_admin_cq_color(void *cq_desc, unsigned int desc_size)
+{
+	u8 type_color = *((u8 *)cq_desc + desc_size - 1);
+
+	return (type_color >> CQ_DESC_COLOR_SHIFT) & CQ_DESC_COLOR_MASK;
+}
+
+unsigned int enic_admin_wq_cq_service(struct enic *enic)
+{
+	struct vnic_cq *cq = &enic->admin_cq[0];
+	unsigned int work = 0;
+	void *desc;
+
+	desc = vnic_cq_to_clean(cq);
+	while (enic_admin_cq_color(desc, cq->ring.desc_size) !=
+	       cq->last_color) {
+		vnic_cq_inc_to_clean(cq);
+		work++;
+		desc = vnic_cq_to_clean(cq);
+	}
+
+	return work;
+}
+
+/* Upper bound on pending admin messages.  A buggy or hostile VF could flood
+ * the PF admin channel faster than admin_msg_work drains it; cap the backlog
+ * so a guest cannot drive the host out of memory.
+ */
+#define ENIC_ADMIN_MSG_MAX	256
+
+static void enic_admin_msg_enqueue(struct enic *enic, void *buf,
+				   unsigned int len)
+{
+	struct enic_admin_msg *msg;
+
+	msg = kmalloc(struct_size(msg, data, len), GFP_KERNEL);
+	if (!msg)
+		return;
+
+	msg->len = len;
+	memcpy(msg->data, buf, len);
+
+	spin_lock(&enic->admin_msg_lock);
+	if (enic->admin_msg_count >= ENIC_ADMIN_MSG_MAX) {
+		spin_unlock(&enic->admin_msg_lock);
+		kfree(msg);
+		if (net_ratelimit())
+			netdev_warn(enic->netdev,
+				    "admin msg backlog full (%u); dropping\n",
+				    ENIC_ADMIN_MSG_MAX);
+		return;
+	}
+	list_add_tail(&msg->list, &enic->admin_msg_list);
+	enic->admin_msg_count++;
+	spin_unlock(&enic->admin_msg_lock);
+}
+
+unsigned int enic_admin_rq_cq_service(struct enic *enic)
+{
+	struct vnic_cq *cq = &enic->admin_cq[1];
+	struct vnic_rq *rq = &enic->admin_rq;
+	struct cq_enet_rq_desc *rq_desc;
+	struct vnic_rq_buf *buf;
+	u16 bwf, bytes_written;
+	unsigned int work = 0;
+	void *desc;
+
+	/* The admin RQ and its CQ form a single in-order channel: firmware
+	 * posts exactly one CQE per consumed RQ descriptor, in submission
+	 * order.  Each CQE therefore pairs with rq->to_clean below without a
+	 * completed_index cross-check, mirroring the in-order assumption of
+	 * the main enic RX path.
+	 */
+	desc = vnic_cq_to_clean(cq);
+	while (enic_admin_cq_color(desc, cq->ring.desc_size) !=
+	       cq->last_color) {
+		/* Ensure DMA descriptor fields are read after
+		 * the color/valid check.  dma_rmb() is the
+		 * correct barrier for DMA-written descriptors.
+		 */
+		dma_rmb();
+		buf = rq->to_clean;
+
+		/* Decode the actual number of bytes hardware wrote into
+		 * the RX buffer.  buf->len is the static allocation size
+		 * (ENIC_ADMIN_BUF_SIZE); copying that many bytes would read
+		 * beyond the actual DMA payload.  bytes_written_flags is at
+		 * the same offset in every cq_enet_rq_desc[_32|_64] variant.
+		 */
+		rq_desc = desc;
+		bwf = le16_to_cpu(rq_desc->bytes_written_flags);
+		bytes_written = bwf & CQ_ENET_RQ_DESC_BYTES_WRITTEN_MASK;
+		if (bytes_written > buf->len)
+			goto next_desc;
+
+		dma_sync_single_for_cpu(&enic->pdev->dev,
+					buf->dma_addr, buf->len,
+					DMA_FROM_DEVICE);
+
+		/* Drop on hardware error indications.  Admin messages
+		 * are internal to the VIC, not received over the wire.
+		 * Firmware sets TRUNCATED when the message does not fit
+		 * in the posted buffer, and FCS_OK is always set on
+		 * healthy admin completions.
+		 */
+		if (bwf & CQ_ENET_RQ_DESC_FLAGS_TRUNCATED) {
+			netdev_warn_once(enic->netdev,
+					 "admin RQ: truncated message dropped\n");
+			goto next_desc;
+		}
+		if (!(rq_desc->flags & CQ_ENET_RQ_DESC_FLAGS_FCS_OK)) {
+			netdev_warn_once(enic->netdev,
+					 "admin RQ: bad FCS, dropping message\n");
+			goto next_desc;
+		}
+
+		enic_admin_msg_enqueue(enic, buf->os_buf, bytes_written);
+
+next_desc:
+		enic_admin_rq_buf_clean(rq, rq->to_clean);
+		rq->to_clean = rq->to_clean->next;
+		rq->ring.desc_avail++;
+
+		vnic_cq_inc_to_clean(cq);
+		work++;
+		desc = vnic_cq_to_clean(cq);
+	}
+
+	if (enic_admin_rq_fill(enic, GFP_KERNEL)) {
+		/* Some RX buffers could not be reposted (transient memory
+		 * pressure).  If the ring is left empty the channel would stall
+		 * with no completion to drive the next refill, so arm a delayed
+		 * re-run of this poll work (the sole owner of the admin RQ ring)
+		 * to repost buffers and re-arm the RQ from a single context.
+		 */
+		if (net_ratelimit())
+			netdev_warn(enic->netdev,
+				    "admin RQ refill failed; scheduling retry\n");
+		schedule_delayed_work(&enic->admin_poll_work,
+				      msecs_to_jiffies(ENIC_ADMIN_RQ_REFILL_RETRY_MS));
+	}
+
+	return work;
+}
+
+static irqreturn_t enic_admin_isr_msix(int irq, void *data)
+{
+	struct enic *enic = data;
+
+	schedule_delayed_work(&enic->admin_poll_work, 0);
+
+	return IRQ_HANDLED;
+}
+
+static void enic_admin_msg_work_handler(struct work_struct *work)
+{
+	struct enic *enic = container_of(work, struct enic, admin_msg_work);
+	struct enic_admin_msg *msg, *tmp;
+	LIST_HEAD(local_list);
+
+	spin_lock_bh(&enic->admin_msg_lock);
+	list_splice_init(&enic->admin_msg_list, &local_list);
+	enic->admin_msg_count = 0;
+	spin_unlock_bh(&enic->admin_msg_lock);
+
+	list_for_each_entry_safe(msg, tmp, &local_list, list) {
+		if (enic->admin_rq_handler)
+			enic->admin_rq_handler(enic, msg->data, msg->len);
+		list_del(&msg->list);
+		kfree(msg);
+	}
+}
+
+static void enic_admin_poll_work_handler(struct work_struct *work)
+{
+	struct enic *enic = container_of(to_delayed_work(work), struct enic,
+					 admin_poll_work);
+	unsigned int credits;
+	unsigned int rq_work;
+
+	/* Snapshot the pending credit count before draining so we acknowledge
+	 * exactly what the hardware reported for this interrupt.  Credits that
+	 * accrue while enic_admin_rq_cq_service() runs are left for the next
+	 * interrupt, which is harmless on this low-rate control path.
+	 */
+	credits = vnic_intr_credits(&enic->admin_intr);
+
+	rq_work = enic_admin_rq_cq_service(enic);
+
+	if (rq_work > 0)
+		schedule_work(&enic->admin_msg_work);
+
+	/* Acknowledge the snapshotted credits and unmask the vector.  Unlike
+	 * the NAPI data path, the admin channel is not re-polled, so the vector
+	 * must be re-armed here to receive the next completion.  The unmask is
+	 * applied through the interrupt mask register independently of the
+	 * credit count, so returning zero credits on a spurious wakeup still
+	 * re-arms the vector.
+	 */
+	vnic_intr_return_credits(&enic->admin_intr,
+				 credits,
+				 1 /* unmask */, 0);
+}
+
+static int enic_admin_setup_intr(struct enic *enic)
+{
+	unsigned int intr_index = enic->intr_count;
+	int err;
+
+	if (vnic_dev_get_intr_mode(enic->vdev) != VNIC_DEV_INTR_MODE_MSIX ||
+	    intr_index >= enic->intr_avail)
+		return -ENODEV;
+
+	/* The admin INTR uses a slot in the same RES_TYPE_INTR_CTRL
+	 * strided array of per-vector control blocks (mask, coalescing
+	 * timer, credit return) that the data-path IRQs occupy in BAR0.
+	 * vnic_intr_alloc() defaults to RES_TYPE_INTR_CTRL, which is what
+	 * we want here.
+	 */
+	err = vnic_intr_alloc(enic->vdev, &enic->admin_intr, intr_index);
+	if (err) {
+		netdev_warn(enic->netdev,
+			    "Failed to alloc admin intr at index %u: %d\n",
+			    intr_index, err);
+		return err;
+	}
+
+	enic->admin_intr_index = intr_index;
+
+	/* Mask the admin vector before requesting the IRQ so an early or
+	 * spurious completion cannot run the poll handler against the
+	 * not-yet-initialised admin rings.  enic_admin_channel_open() unmasks
+	 * it only after the rings are initialised and filled.
+	 */
+	vnic_intr_mask(&enic->admin_intr);
+
+	/* A V2 VF opens the admin channel during probe, before
+	 * register_netdev() resolves the "eth%d" name template, so using
+	 * netdev->name here would register the literal "eth%d-admin" in
+	 * /proc/interrupts.  Use the already-stable PCI device name instead.
+	 */
+	snprintf(enic->msix[intr_index].devname,
+		 sizeof(enic->msix[intr_index].devname),
+		 "%s-admin", pci_name(enic->pdev));
+	enic->msix[intr_index].isr = enic_admin_isr_msix;
+	enic->msix[intr_index].devid = enic;
+
+	err = request_irq(enic->msix_entry[intr_index].vector,
+			  enic->msix[intr_index].isr, 0,
+			  enic->msix[intr_index].devname,
+			  enic->msix[intr_index].devid);
+	if (err) {
+		netdev_warn(enic->netdev,
+			    "Failed to request admin MSI-X irq: %d\n", err);
+		vnic_intr_free(&enic->admin_intr);
+		return err;
+	}
+
+	enic->msix[intr_index].requested = 1;
+
+	netdev_dbg(enic->netdev,
+		   "admin channel using MSI-X interrupt (index %u)\n",
+		   intr_index);
+
+	return 0;
+}
+
+static void enic_admin_teardown_intr(struct enic *enic)
+{
+	unsigned int intr_index = enic->admin_intr_index;
+
+	free_irq(enic->msix_entry[intr_index].vector,
+		 enic->msix[intr_index].devid);
+	cancel_delayed_work_sync(&enic->admin_poll_work);
+	enic->msix[intr_index].requested = 0;
+}
+
 static int enic_admin_qp_type_set(struct enic *enic, u32 enable)
 {
 	u64 a0 = QP_TYPE_ADMIN, a1 = enable;
@@ -173,6 +457,7 @@ static int enic_admin_alloc_resources(struct enic *enic)
 
 static void enic_admin_free_resources(struct enic *enic)
 {
+	vnic_intr_free(&enic->admin_intr);
 	vnic_cq_free(&enic->admin_cq[1]);
 	vnic_cq_free(&enic->admin_cq[0]);
 	vnic_rq_free(&enic->admin_rq);
@@ -181,6 +466,8 @@ static void enic_admin_free_resources(struct enic *enic)
 
 static void enic_admin_init_resources(struct enic *enic)
 {
+	unsigned int intr_offset = enic->admin_intr_index;
+
 	vnic_wq_init(&enic->admin_wq,
 		     0, 0, 0); /* cq_index, err_intr_enable, err_intr_offset */
 	vnic_rq_init(&enic->admin_rq,
@@ -189,20 +476,35 @@ static void enic_admin_init_resources(struct enic *enic)
 		     VNIC_CQ_FC_DISABLE,
 		     VNIC_CQ_COLOR_ENABLE,
 		     0, 0, 1, /* cq_head, cq_tail, cq_tail_color */
-		     VNIC_CQ_INTR_DISABLE,
+		     VNIC_CQ_INTR_DISABLE, /* polled synchronously by mbox send */
 		     VNIC_CQ_ENTRY_ENABLE,
 		     VNIC_CQ_MSG_DISABLE,
-		     0, /* interrupt_offset */
+		     intr_offset,
 		     0 /* cq_message_addr */);
 	vnic_cq_init(&enic->admin_cq[1],
 		     VNIC_CQ_FC_DISABLE,
 		     VNIC_CQ_COLOR_ENABLE,
 		     0, 0, 1, /* cq_head, cq_tail, cq_tail_color */
-		     VNIC_CQ_INTR_DISABLE,
+		     VNIC_CQ_INTR_ENABLE,
 		     VNIC_CQ_ENTRY_ENABLE,
 		     VNIC_CQ_MSG_DISABLE,
-		     0, /* interrupt_offset */
+		     intr_offset,
 		     0 /* cq_message_addr */);
+	vnic_intr_init(&enic->admin_intr,
+		       0, 0, 1); /* coalescing_timer, coalescing_type, mask_on_assertion */
+}
+
+static void enic_admin_msg_drain(struct enic *enic)
+{
+	struct enic_admin_msg *msg, *tmp;
+
+	spin_lock_bh(&enic->admin_msg_lock);
+	list_for_each_entry_safe(msg, tmp, &enic->admin_msg_list, list) {
+		list_del(&msg->list);
+		kfree(msg);
+	}
+	enic->admin_msg_count = 0;
+	spin_unlock_bh(&enic->admin_msg_lock);
 }
 
 int enic_admin_channel_open(struct enic *enic)
@@ -220,6 +522,19 @@ int enic_admin_channel_open(struct enic *enic)
 		return err;
 	}
 
+	spin_lock_init(&enic->admin_msg_lock);
+	INIT_LIST_HEAD(&enic->admin_msg_list);
+	INIT_WORK(&enic->admin_msg_work, enic_admin_msg_work_handler);
+	INIT_DELAYED_WORK(&enic->admin_poll_work, enic_admin_poll_work_handler);
+
+	err = enic_admin_setup_intr(enic);
+	if (err) {
+		netdev_err(enic->netdev,
+			   "Admin channel requires MSI-X, SR-IOV unavailable: %d\n",
+			   err);
+		goto free_resources;
+	}
+
 	enic_admin_init_resources(enic);
 
 	vnic_wq_enable(&enic->admin_wq);
@@ -239,17 +554,31 @@ int enic_admin_channel_open(struct enic *enic)
 		goto disable_queues;
 	}
 
+	vnic_intr_unmask(&enic->admin_intr);
+
+	netdev_dbg(enic->netdev,
+		   "admin channel open: intr=%u wq_avail=%u rq_avail=%u cq0_color=%u cq1_color=%u\n",
+		   enic->admin_intr_index,
+		   vnic_wq_desc_avail(&enic->admin_wq),
+		   vnic_rq_desc_avail(&enic->admin_rq),
+		   enic->admin_cq[0].last_color,
+		   enic->admin_cq[1].last_color);
+
 	enic->admin_chan_up = true;
 
 	return 0;
 
 disable_queues:
+	enic_admin_teardown_intr(enic);
 	enic_admin_qp_type_set(enic, QP_DISABLE);
 	if (vnic_wq_disable(&enic->admin_wq))
 		netdev_warn(enic->netdev, "Failed to disable admin WQ\n");
 	if (vnic_rq_disable(&enic->admin_rq))
 		netdev_warn(enic->netdev, "Failed to disable admin RQ\n");
+	cancel_work_sync(&enic->admin_msg_work);
+	enic_admin_msg_drain(enic);
 	enic_admin_rq_drain(enic);
+free_resources:
 	enic_admin_free_resources(enic);
 	return err;
 }
@@ -268,6 +597,13 @@ void enic_admin_channel_close(struct enic *enic)
 	if (!enic->admin_chan_up)
 		return;
 
+	netdev_dbg(enic->netdev, "admin channel close\n");
+
+	vnic_intr_mask(&enic->admin_intr);
+	enic_admin_teardown_intr(enic);
+	cancel_work_sync(&enic->admin_msg_work);
+	enic_admin_msg_drain(enic);
+
 	enic_admin_qp_type_set(enic, QP_DISABLE);
 
 	err = vnic_wq_disable(&enic->admin_wq);
@@ -283,6 +619,7 @@ void enic_admin_channel_close(struct enic *enic)
 	enic_admin_rq_drain(enic);
 	vnic_cq_clean(&enic->admin_cq[0]);
 	vnic_cq_clean(&enic->admin_cq[1]);
+	vnic_intr_clean(&enic->admin_intr);
 	enic_admin_free_resources(enic);
 
 	enic->admin_chan_up = false;
diff --git a/drivers/net/ethernet/cisco/enic/enic_admin.h b/drivers/net/ethernet/cisco/enic/enic_admin.h
index 569aadeb9312..62c80220b0ca 100644
--- a/drivers/net/ethernet/cisco/enic/enic_admin.h
+++ b/drivers/net/ethernet/cisco/enic/enic_admin.h
@@ -9,7 +9,19 @@
 
 struct enic;
 
+/* Wrapper for received admin messages queued for deferred processing.
+ * The admin CQ poll work handler enqueues these; a separate work handler
+ * processes them where sleeping (mutex, GFP_KERNEL) is safe.
+ */
+struct enic_admin_msg {
+	struct list_head list;
+	unsigned int len;
+	u8 data[] __aligned(8);
+};
+
 int enic_admin_channel_open(struct enic *enic);
 void enic_admin_channel_close(struct enic *enic);
+unsigned int enic_admin_wq_cq_service(struct enic *enic);
+unsigned int enic_admin_rq_cq_service(struct enic *enic);
 
 #endif /* _ENIC_ADMIN_H_ */

-- 
2.43.0

[PATCH net-next v13 05/11] enic: define MBOX message types and header structures

[Satish Kharat]

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>
---
 drivers/net/ethernet/cisco/enic/enic_mbox.h | 83 +++++++++++++++++++++++++++++
 1 file changed, 83 insertions(+)

diff --git a/drivers/net/ethernet/cisco/enic/enic_mbox.h b/drivers/net/ethernet/cisco/enic/enic_mbox.h
new file mode 100644
index 000000000000..a52f1d25cb21
--- /dev/null
+++ b/drivers/net/ethernet/cisco/enic/enic_mbox.h
@@ -0,0 +1,83 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/* Copyright 2025 Cisco Systems, Inc.  All rights reserved. */
+
+#ifndef _ENIC_MBOX_H_
+#define _ENIC_MBOX_H_
+
+#include <linux/bits.h>
+#include <linux/types.h>
+
+/*
+ * Mailbox protocol for PF-VF communication over the admin channel.
+ *
+ * Even numbers are requests, odd numbers are replies/acks.
+ * The prefix indicates the initiator: VF_ = VF-initiated, PF_ = PF-initiated.
+ */
+enum enic_mbox_msg_type {
+	ENIC_MBOX_VF_CAPABILITY_REQUEST		= 0,
+	ENIC_MBOX_VF_CAPABILITY_REPLY		= 1,
+	ENIC_MBOX_VF_REGISTER_REQUEST		= 2,
+	ENIC_MBOX_VF_REGISTER_REPLY		= 3,
+	ENIC_MBOX_VF_UNREGISTER_REQUEST		= 4,
+	ENIC_MBOX_VF_UNREGISTER_REPLY		= 5,
+	ENIC_MBOX_PF_LINK_STATE_NOTIF		= 6,
+	ENIC_MBOX_PF_LINK_STATE_ACK		= 7,
+	ENIC_MBOX_MAX
+};
+
+struct enic_mbox_hdr {
+	__le16 src_vnic_id;
+	__le16 dst_vnic_id;
+	u8 msg_type;
+	u8 flags;
+	__le16 msg_len;
+	__le64 msg_num;
+};
+
+struct enic_mbox_generic_reply {
+	__le16 ret_major;
+	__le16 ret_minor;
+};
+
+#define ENIC_MBOX_ERR_GENERIC		BIT(0)
+#define ENIC_MBOX_ERR_VF_NOT_REGISTERED	BIT(1)
+#define ENIC_MBOX_ERR_MSG_NOT_SUPPORTED	BIT(2)
+
+/* ENIC_MBOX_VF_CAPABILITY_REQUEST / _REPLY */
+#define ENIC_MBOX_CAP_VERSION_0		0
+#define ENIC_MBOX_CAP_VERSION_1		1
+
+struct enic_mbox_vf_capability_msg {
+	__le32 version;
+	__le32 reserved[32];
+};
+
+/* The embedded enic_mbox_generic_reply has 2-byte alignment, but the
+ * __le32 members give this struct 4-byte natural alignment.  Receive
+ * buffers come from kmalloc (>= 8-byte aligned), so there is no
+ * misaligned access risk when casting from the receive buffer.
+ */
+struct enic_mbox_vf_capability_reply_msg {
+	struct enic_mbox_generic_reply reply;
+	__le32 version;
+	__le32 reserved[32];
+};
+
+/* ENIC_MBOX_VF_REGISTER / _UNREGISTER */
+struct enic_mbox_vf_register_reply_msg {
+	struct enic_mbox_generic_reply reply;
+};
+
+/* ENIC_MBOX_PF_LINK_STATE_NOTIF / _ACK */
+#define ENIC_MBOX_LINK_STATE_DISABLE	0
+#define ENIC_MBOX_LINK_STATE_ENABLE	1
+
+struct enic_mbox_pf_link_state_notif_msg {
+	__le32 link_state;
+};
+
+struct enic_mbox_pf_link_state_ack_msg {
+	struct enic_mbox_generic_reply ack;
+};
+
+#endif /* _ENIC_MBOX_H_ */

-- 
2.43.0

[PATCH net-next v13 06/11] enic: add MBOX core send and receive for admin channel

[Satish Kharat]

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>
---
 drivers/net/ethernet/cisco/enic/Makefile     |   2 +-
 drivers/net/ethernet/cisco/enic/enic.h       |   6 +
 drivers/net/ethernet/cisco/enic/enic_admin.c |  42 ++++++-
 drivers/net/ethernet/cisco/enic/enic_mbox.c  | 177 +++++++++++++++++++++++++++
 drivers/net/ethernet/cisco/enic/enic_mbox.h  |   8 ++
 5 files changed, 232 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/cisco/enic/Makefile b/drivers/net/ethernet/cisco/enic/Makefile
index 7ae72fefc99a..e38aaf34c148 100644
--- a/drivers/net/ethernet/cisco/enic/Makefile
+++ b/drivers/net/ethernet/cisco/enic/Makefile
@@ -4,5 +4,5 @@ obj-$(CONFIG_ENIC) := enic.o
 enic-y := enic_main.o vnic_cq.o vnic_intr.o vnic_wq.o \
 	enic_res.o enic_dev.o enic_pp.o vnic_dev.o vnic_rq.o vnic_vic.o \
 	enic_ethtool.o enic_api.o enic_clsf.o enic_rq.o enic_wq.o \
-	enic_admin.o
+	enic_admin.o enic_mbox.o
 
diff --git a/drivers/net/ethernet/cisco/enic/enic.h b/drivers/net/ethernet/cisco/enic/enic.h
index 0f08de228b36..525a67414e63 100644
--- a/drivers/net/ethernet/cisco/enic/enic.h
+++ b/drivers/net/ethernet/cisco/enic/enic.h
@@ -297,6 +297,8 @@ struct enic {
 	 * left the resources freed.
 	 */
 	bool admin_chan_up;
+	/* set on send timeout; cleared on channel re-open */
+	bool mbox_send_disabled;
 	struct vnic_wq admin_wq;
 	struct vnic_rq admin_rq;
 	struct vnic_cq admin_cq[2];
@@ -309,6 +311,10 @@ struct enic {
 	unsigned int admin_msg_count;	/* current depth of admin_msg_list */
 	void (*admin_rq_handler)(struct enic *enic, void *buf,
 				 unsigned int len);
+
+	/* MBOX protocol state — mbox_lock serializes admin WQ sends */
+	struct mutex mbox_lock;
+	u64 mbox_msg_num;
 };
 
 static inline struct net_device *vnic_get_netdev(struct vnic_dev *vdev)
diff --git a/drivers/net/ethernet/cisco/enic/enic_admin.c b/drivers/net/ethernet/cisco/enic/enic_admin.c
index 65aec9651a6f..f4a409fec983 100644
--- a/drivers/net/ethernet/cisco/enic/enic_admin.c
+++ b/drivers/net/ethernet/cisco/enic/enic_admin.c
@@ -19,6 +19,7 @@
 #include "cq_enet_desc.h"
 #include "wq_enet_desc.h"
 #include "rq_enet_desc.h"
+#include "enic_mbox.h"
 
 /* Retry interval for a failed admin RQ refill.  Short so the control
  * channel recovers quickly once memory is available again.
@@ -217,7 +218,26 @@ unsigned int enic_admin_rq_cq_service(struct enic *enic)
 			goto next_desc;
 		}
 
-		enic_admin_msg_enqueue(enic, buf->os_buf, bytes_written);
+		if (enic->admin_rq_handler) {
+			u16 sender_vlan;
+
+			/* Firmware sets the CQ VLAN field to identify the
+			 * sender: 0 = PF, 1-based = VF index.  Overwrite
+			 * the untrusted src_vnic_id in the MBOX header with
+			 * the hardware-verified value.
+			 */
+			sender_vlan = le16_to_cpu(rq_desc->vlan);
+			if (bytes_written >= sizeof(struct enic_mbox_hdr)) {
+				struct enic_mbox_hdr *hdr = buf->os_buf;
+
+				hdr->src_vnic_id = (sender_vlan == 0) ?
+					cpu_to_le16(ENIC_MBOX_DST_PF) :
+					cpu_to_le16(sender_vlan - 1);
+			}
+
+			enic_admin_msg_enqueue(enic, buf->os_buf,
+					       bytes_written);
+		}
 
 next_desc:
 		enic_admin_rq_buf_clean(rq, rq->to_clean);
@@ -490,8 +510,9 @@ static void enic_admin_init_resources(struct enic *enic)
 		     VNIC_CQ_MSG_DISABLE,
 		     intr_offset,
 		     0 /* cq_message_addr */);
+	/* coalescing_timer, coalescing_type, mask_on_assertion */
 	vnic_intr_init(&enic->admin_intr,
-		       0, 0, 1); /* coalescing_timer, coalescing_type, mask_on_assertion */
+		       0, 0, 1);
 }
 
 static void enic_admin_msg_drain(struct enic *enic)
@@ -514,6 +535,13 @@ int enic_admin_channel_open(struct enic *enic)
 	if (!enic->has_admin_channel)
 		return -ENODEV;
 
+	/* Keep MBOX sends disabled for the entire open sequence.  It is
+	 * cleared only after every resource is allocated and enabled below,
+	 * so any early error return here leaves sends disabled and a
+	 * concurrent sender cannot touch a half-open or freed admin_wq.
+	 */
+	WRITE_ONCE(enic->mbox_send_disabled, true);
+
 	err = enic_admin_alloc_resources(enic);
 	if (err) {
 		netdev_err(enic->netdev,
@@ -556,6 +584,14 @@ int enic_admin_channel_open(struct enic *enic)
 
 	vnic_intr_unmask(&enic->admin_intr);
 
+	/* Only now that the admin WQ/RQ/CQ and interrupt are fully allocated,
+	 * programmed and enabled is it safe to allow MBOX sends.  Clearing this
+	 * earlier opened a window where a concurrent sender (e.g. link-notify
+	 * work scheduled by a post-reset link-up) could call enic_mbox_send_msg()
+	 * against a not-yet-allocated admin_wq and crash.
+	 */
+	WRITE_ONCE(enic->mbox_send_disabled, false);
+
 	netdev_dbg(enic->netdev,
 		   "admin channel open: intr=%u wq_avail=%u rq_avail=%u cq0_color=%u cq1_color=%u\n",
 		   enic->admin_intr_index,
@@ -597,6 +633,8 @@ void enic_admin_channel_close(struct enic *enic)
 	if (!enic->admin_chan_up)
 		return;
 
+	WRITE_ONCE(enic->mbox_send_disabled, true);
+
 	netdev_dbg(enic->netdev, "admin channel close\n");
 
 	vnic_intr_mask(&enic->admin_intr);
diff --git a/drivers/net/ethernet/cisco/enic/enic_mbox.c b/drivers/net/ethernet/cisco/enic/enic_mbox.c
new file mode 100644
index 000000000000..c1680f77abac
--- /dev/null
+++ b/drivers/net/ethernet/cisco/enic/enic_mbox.c
@@ -0,0 +1,177 @@
+// SPDX-License-Identifier: GPL-2.0-only
+// Copyright 2025 Cisco Systems, Inc.  All rights reserved.
+
+#include <linux/kernel.h>
+#include <linux/netdevice.h>
+#include <linux/dma-mapping.h>
+#include <linux/delay.h>
+
+#include "vnic_dev.h"
+#include "vnic_wq.h"
+#include "vnic_cq.h"
+#include "enic.h"
+#include "enic_admin.h"
+#include "enic_mbox.h"
+#include "wq_enet_desc.h"
+
+#define ENIC_MBOX_POLL_TIMEOUT_US	5000000
+#define ENIC_MBOX_POLL_INTERVAL_US	100
+
+static void enic_mbox_fill_hdr(struct enic *enic, struct enic_mbox_hdr *hdr,
+			       u8 msg_type, u16 dst_vnic_id, u16 msg_len)
+{
+	memset(hdr, 0, sizeof(*hdr));
+	hdr->dst_vnic_id = cpu_to_le16(dst_vnic_id);
+	hdr->msg_type = msg_type;
+	hdr->msg_len = cpu_to_le16(msg_len);
+	hdr->msg_num = cpu_to_le64(++enic->mbox_msg_num);
+}
+
+int enic_mbox_send_msg(struct enic *enic, u8 msg_type, u16 dst_vnic_id,
+		       void *payload, u16 payload_len)
+{
+	size_t total_len = sizeof(struct enic_mbox_hdr) + payload_len;
+	struct vnic_wq *wq = &enic->admin_wq;
+	struct wq_enet_desc *desc;
+	unsigned long timeout;
+	dma_addr_t dma_addr;
+	u16 vlan_tag;
+	void *buf;
+	int err;
+
+	/* Reject payloads that cannot fit in a single admin buffer.  Checked
+	 * before taking mbox_lock; total_len is computed as size_t so the
+	 * sizeof() + payload_len sum cannot wrap.
+	 */
+	if (payload_len > ENIC_ADMIN_BUF_SIZE - sizeof(struct enic_mbox_hdr))
+		return -EINVAL;
+
+	/* Serialize MBOX sends. The admin channel is a low-frequency
+	 * control path; holding the mutex across the poll is acceptable.
+	 */
+	mutex_lock(&enic->mbox_lock);
+
+	if (!enic->has_admin_channel || READ_ONCE(enic->mbox_send_disabled)) {
+		err = -ENODEV;
+		goto unlock;
+	}
+
+	if (vnic_wq_desc_avail(wq) == 0) {
+		err = -ENOSPC;
+		goto unlock;
+	}
+
+	buf = kmalloc(total_len, GFP_KERNEL);
+	if (!buf) {
+		err = -ENOMEM;
+		goto unlock;
+	}
+
+	enic_mbox_fill_hdr(enic, buf, msg_type, dst_vnic_id, total_len);
+	if (payload_len) {
+		void *dst = buf + sizeof(struct enic_mbox_hdr);
+
+		memcpy(dst, payload, payload_len);
+	}
+
+	dma_addr = dma_map_single(&enic->pdev->dev, buf, total_len,
+				  DMA_TO_DEVICE);
+	if (dma_mapping_error(&enic->pdev->dev, dma_addr)) {
+		kfree(buf);
+		err = -ENOMEM;
+		goto unlock;
+	}
+
+	/* Firmware uses vlan field for routing: 0 = PF, 1-based = VF index */
+	if (dst_vnic_id == ENIC_MBOX_DST_PF)
+		vlan_tag = 0;
+	else
+		vlan_tag = dst_vnic_id + 1;
+
+	desc = vnic_wq_next_desc(wq);
+	wq_enet_desc_enc(desc, (u64)dma_addr | VNIC_PADDR_TARGET,
+			 total_len,
+			 0, 0, 0,       /* mss, hdr_len, offload_mode */
+			 1, 1,          /* eop, cq_entry */
+			 0,             /* fcoe_encap */
+			 1, vlan_tag,   /* vlan_tag_insert, vlan_tag */
+			 0);            /* loopback */
+	vnic_wq_post(wq, buf, dma_addr, total_len,
+		     1, 1,              /* sop, eop */
+		     1, 1,              /* desc_skip_cnt, cq_entry */
+		     0, 0);             /* compressed_send, wrid */
+	vnic_wq_doorbell(wq);
+
+	timeout = jiffies + usecs_to_jiffies(ENIC_MBOX_POLL_TIMEOUT_US);
+	err = -ETIMEDOUT;
+	while (time_before(jiffies, timeout)) {
+		if (enic_admin_wq_cq_service(enic)) {
+			err = 0;
+			break;
+		}
+		usleep_range(ENIC_MBOX_POLL_INTERVAL_US,
+			     ENIC_MBOX_POLL_INTERVAL_US + 50);
+	}
+	/* Final check in case completion arrived during the last sleep */
+	if (err && enic_admin_wq_cq_service(enic))
+		err = 0;
+
+	if (!err) {
+		wq->to_clean = wq->to_clean->next;
+		wq->ring.desc_avail++;
+		dma_unmap_single(&enic->pdev->dev, dma_addr, total_len,
+				 DMA_TO_DEVICE);
+		kfree(buf);
+	} else {
+		netdev_err(enic->netdev,
+			   "MBOX send timed out (type %u dst %u), disabling channel\n",
+			   msg_type, dst_vnic_id);
+		/*
+		 * The WQ descriptor is still live in hardware. Do not unmap
+		 * or free the buffer: the device may still DMA from dma_addr.
+		 * Mark the channel unusable so no further sends are attempted.
+		 */
+		WRITE_ONCE(enic->mbox_send_disabled, true);
+	}
+
+	netdev_dbg(enic->netdev,
+		   "MBOX send msg_type %u dst %u vlan %u err %d\n",
+		   msg_type, dst_vnic_id, vlan_tag, err);
+unlock:
+	mutex_unlock(&enic->mbox_lock);
+	return err;
+}
+
+static void enic_mbox_recv_handler(struct enic *enic, void *buf,
+				   unsigned int len)
+{
+	struct enic_mbox_hdr *hdr = buf;
+
+	if (len < sizeof(*hdr)) {
+		if (net_ratelimit())
+			netdev_warn(enic->netdev,
+				    "MBOX: truncated message (len %u < %zu)\n",
+				    len, sizeof(*hdr));
+		return;
+	}
+
+	if (hdr->msg_type >= ENIC_MBOX_MAX) {
+		if (net_ratelimit())
+			netdev_warn(enic->netdev,
+				    "MBOX: unknown msg type %u\n",
+				    hdr->msg_type);
+		return;
+	}
+
+	netdev_dbg(enic->netdev,
+		   "MBOX recv: type %u from vnic %u len %u\n",
+		   hdr->msg_type, le16_to_cpu(hdr->src_vnic_id),
+		   le16_to_cpu(hdr->msg_len));
+}
+
+void enic_mbox_init(struct enic *enic)
+{
+	enic->mbox_msg_num = 0;
+	mutex_init(&enic->mbox_lock);
+	enic->admin_rq_handler = enic_mbox_recv_handler;
+}
diff --git a/drivers/net/ethernet/cisco/enic/enic_mbox.h b/drivers/net/ethernet/cisco/enic/enic_mbox.h
index a52f1d25cb21..73fd7f783ee2 100644
--- a/drivers/net/ethernet/cisco/enic/enic_mbox.h
+++ b/drivers/net/ethernet/cisco/enic/enic_mbox.h
@@ -80,4 +80,12 @@ struct enic_mbox_pf_link_state_ack_msg {
 	struct enic_mbox_generic_reply ack;
 };
 
+#define ENIC_MBOX_DST_PF	0xFFFF
+
+struct enic;
+
+void enic_mbox_init(struct enic *enic);
+int enic_mbox_send_msg(struct enic *enic, u8 msg_type, u16 dst_vnic_id,
+		       void *payload, u16 payload_len);
+
 #endif /* _ENIC_MBOX_H_ */

-- 
2.43.0

[PATCH net-next v13 07/11] enic: add MBOX PF handlers for VF register and capability

[Satish Kharat]

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>
---
 drivers/net/ethernet/cisco/enic/enic.h      |   7 +-
 drivers/net/ethernet/cisco/enic/enic_mbox.c | 190 +++++++++++++++++++++++++++-
 drivers/net/ethernet/cisco/enic/enic_mbox.h |   1 +
 3 files changed, 194 insertions(+), 4 deletions(-)

diff --git a/drivers/net/ethernet/cisco/enic/enic.h b/drivers/net/ethernet/cisco/enic/enic.h
index 525a67414e63..6261f3e8e4f8 100644
--- a/drivers/net/ethernet/cisco/enic/enic.h
+++ b/drivers/net/ethernet/cisco/enic/enic.h
@@ -256,9 +256,7 @@ struct enic {
 	struct enic_rx_coal rx_coalesce_setting;
 	u32 rx_coalesce_usecs;
 	u32 tx_coalesce_usecs;
-#ifdef CONFIG_PCI_IOV
 	u16 num_vfs;
-#endif
 	enum enic_vf_type vf_type;
 	unsigned int enable_count;
 	spinlock_t enic_api_lock;
@@ -315,6 +313,11 @@ struct enic {
 	/* MBOX protocol state — mbox_lock serializes admin WQ sends */
 	struct mutex mbox_lock;
 	u64 mbox_msg_num;
+
+	/* PF: per-VF MBOX state, allocated when SRIOV V2 is enabled */
+	struct enic_vf_state {
+		bool registered;
+	} *vf_state;
 };
 
 static inline struct net_device *vnic_get_netdev(struct vnic_dev *vdev)
diff --git a/drivers/net/ethernet/cisco/enic/enic_mbox.c b/drivers/net/ethernet/cisco/enic/enic_mbox.c
index c1680f77abac..d1b13737e77d 100644
--- a/drivers/net/ethernet/cisco/enic/enic_mbox.c
+++ b/drivers/net/ethernet/cisco/enic/enic_mbox.c
@@ -142,10 +142,183 @@ int enic_mbox_send_msg(struct enic *enic, u8 msg_type, u16 dst_vnic_id,
 	return err;
 }
 
+int enic_mbox_send_link_state(struct enic *enic, u16 vf_id, u32 link_state)
+{
+	struct enic_mbox_pf_link_state_notif_msg notif = {};
+
+	if (!enic->vf_state || vf_id >= enic->num_vfs ||
+	    !enic->vf_state[vf_id].registered) {
+		netdev_dbg(enic->netdev,
+			   "MBOX: skip link state to unregistered VF %u\n",
+			   vf_id);
+		return 0;
+	}
+
+	notif.link_state = cpu_to_le32(link_state);
+	return enic_mbox_send_msg(enic, ENIC_MBOX_PF_LINK_STATE_NOTIF, vf_id,
+				  &notif, sizeof(notif));
+}
+
+static int enic_mbox_pf_handle_capability(struct enic *enic, void *msg,
+					  u16 vf_id, u64 msg_num)
+{
+	struct enic_mbox_vf_capability_reply_msg reply = {};
+
+	reply.reply.ret_major = cpu_to_le16(0);
+	reply.version = cpu_to_le32(ENIC_MBOX_CAP_VERSION_1);
+
+	return enic_mbox_send_msg(enic, ENIC_MBOX_VF_CAPABILITY_REPLY, vf_id,
+				  &reply, sizeof(reply));
+}
+
+static int enic_mbox_pf_handle_register(struct enic *enic, void *msg,
+					u16 vf_id, u64 msg_num)
+{
+	struct enic_mbox_vf_register_reply_msg reply = {};
+	u32 link_state;
+	int err;
+
+	if (!enic->vf_state || vf_id >= enic->num_vfs) {
+		if (net_ratelimit())
+			netdev_warn(enic->netdev,
+				    "MBOX: register from invalid VF %u\n",
+				    vf_id);
+		return -EINVAL;
+	}
+
+	/* VF re-registering (e.g. guest reboot without clean unregister):
+	 * mark the previous registration inactive before accepting the new one.
+	 */
+	if (enic->vf_state[vf_id].registered) {
+		netdev_dbg(enic->netdev,
+			   "MBOX: VF %u re-register, cleaning previous state\n",
+			   vf_id);
+		enic->vf_state[vf_id].registered = false;
+	}
+
+	reply.reply.ret_major = cpu_to_le16(0);
+	err = enic_mbox_send_msg(enic, ENIC_MBOX_VF_REGISTER_REPLY, vf_id,
+				 &reply, sizeof(reply));
+	if (err)
+		return err;
+
+	enic->vf_state[vf_id].registered = true;
+	if (net_ratelimit())
+		netdev_info(enic->netdev, "VF %u registered via MBOX\n", vf_id);
+
+	link_state = netif_carrier_ok(enic->netdev) ?
+		ENIC_MBOX_LINK_STATE_ENABLE :
+		ENIC_MBOX_LINK_STATE_DISABLE;
+	err = enic_mbox_send_link_state(enic, vf_id, link_state);
+	if (err && net_ratelimit())
+		netdev_warn(enic->netdev,
+			    "VF %u: failed to send initial link state: %d\n",
+			    vf_id, err);
+	/* Registration succeeded; initial link state notification attempted
+	 * above.  Subsequent link state changes are sent from the PF
+	 * when enic_link_check() detects carrier changes.
+	 */
+	return 0;
+}
+
+static int enic_mbox_pf_handle_unregister(struct enic *enic, void *msg,
+					  u16 vf_id, u64 msg_num)
+{
+	struct enic_mbox_vf_register_reply_msg reply = {};
+	int err;
+
+	if (!enic->vf_state || vf_id >= enic->num_vfs) {
+		if (net_ratelimit())
+			netdev_warn(enic->netdev,
+				    "MBOX: unregister from invalid VF %u\n",
+				    vf_id);
+		return -EINVAL;
+	}
+
+	/* VF is unloading; clear local state regardless of whether
+	 * the reply is successfully delivered to avoid the PF treating
+	 * a dead VF as still registered.
+	 */
+	enic->vf_state[vf_id].registered = false;
+
+	reply.reply.ret_major = cpu_to_le16(0);
+	err = enic_mbox_send_msg(enic, ENIC_MBOX_VF_UNREGISTER_REPLY, vf_id,
+				 &reply, sizeof(reply));
+
+	if (net_ratelimit())
+		netdev_info(enic->netdev,
+			    "VF %u unregistered via MBOX\n", vf_id);
+
+	return err;
+}
+
+static void enic_mbox_pf_process_msg(struct enic *enic,
+				     struct enic_mbox_hdr *hdr, void *payload)
+{
+	u16 vf_id = le16_to_cpu(hdr->src_vnic_id);
+	u16 msg_len = le16_to_cpu(hdr->msg_len);
+	int err = 0;
+
+	if (!enic->vf_state) {
+		netdev_dbg(enic->netdev,
+			   "MBOX: PF received msg but SRIOV not active\n");
+		return;
+	}
+
+	if (vf_id >= enic->num_vfs) {
+		if (net_ratelimit())
+			netdev_warn(enic->netdev,
+				    "MBOX: PF received msg from invalid VF %u\n",
+				    vf_id);
+		return;
+	}
+
+	switch (hdr->msg_type) {
+	case ENIC_MBOX_VF_CAPABILITY_REQUEST:
+		err = enic_mbox_pf_handle_capability(enic, payload, vf_id,
+						     le64_to_cpu(hdr->msg_num));
+		break;
+	case ENIC_MBOX_VF_REGISTER_REQUEST:
+		err = enic_mbox_pf_handle_register(enic, payload, vf_id,
+						   le64_to_cpu(hdr->msg_num));
+		break;
+	case ENIC_MBOX_VF_UNREGISTER_REQUEST:
+		err = enic_mbox_pf_handle_unregister(enic, payload, vf_id,
+						     le64_to_cpu(hdr->msg_num));
+		break;
+	case ENIC_MBOX_PF_LINK_STATE_ACK: {
+		struct enic_mbox_pf_link_state_ack_msg *ack = payload;
+
+		if (msg_len < sizeof(*hdr) + sizeof(*ack))
+			break;
+		if (le16_to_cpu(ack->ack.ret_major) && net_ratelimit())
+			netdev_warn(enic->netdev,
+				    "MBOX: VF %u link state ACK error %u/%u\n",
+				    vf_id,
+				    le16_to_cpu(ack->ack.ret_major),
+				    le16_to_cpu(ack->ack.ret_minor));
+		break;
+	}
+	default:
+		netdev_dbg(enic->netdev,
+			   "MBOX: PF unhandled msg type %u from VF %u\n",
+			   hdr->msg_type, vf_id);
+		err = -EOPNOTSUPP;
+		break;
+	}
+
+	if (err && net_ratelimit())
+		netdev_warn(enic->netdev,
+			    "MBOX: PF handler for msg type %u from VF %u failed: %d\n",
+			    hdr->msg_type, vf_id, err);
+}
+
 static void enic_mbox_recv_handler(struct enic *enic, void *buf,
 				   unsigned int len)
 {
 	struct enic_mbox_hdr *hdr = buf;
+	void *payload;
+	u16 msg_len;
 
 	if (len < sizeof(*hdr)) {
 		if (net_ratelimit())
@@ -163,10 +336,23 @@ static void enic_mbox_recv_handler(struct enic *enic, void *buf,
 		return;
 	}
 
+	msg_len = le16_to_cpu(hdr->msg_len);
+	if (msg_len < sizeof(*hdr) || msg_len > len) {
+		if (net_ratelimit())
+			netdev_warn(enic->netdev,
+				    "MBOX: invalid msg_len %u (buf len %u)\n",
+				    msg_len, len);
+		return;
+	}
+
 	netdev_dbg(enic->netdev,
 		   "MBOX recv: type %u from vnic %u len %u\n",
-		   hdr->msg_type, le16_to_cpu(hdr->src_vnic_id),
-		   le16_to_cpu(hdr->msg_len));
+		   hdr->msg_type, le16_to_cpu(hdr->src_vnic_id), msg_len);
+
+	payload = buf + sizeof(*hdr);
+
+	if (enic->vf_state)
+		enic_mbox_pf_process_msg(enic, hdr, payload);
 }
 
 void enic_mbox_init(struct enic *enic)
diff --git a/drivers/net/ethernet/cisco/enic/enic_mbox.h b/drivers/net/ethernet/cisco/enic/enic_mbox.h
index 73fd7f783ee2..f1de67db1273 100644
--- a/drivers/net/ethernet/cisco/enic/enic_mbox.h
+++ b/drivers/net/ethernet/cisco/enic/enic_mbox.h
@@ -87,5 +87,6 @@ struct enic;
 void enic_mbox_init(struct enic *enic);
 int enic_mbox_send_msg(struct enic *enic, u8 msg_type, u16 dst_vnic_id,
 		       void *payload, u16 payload_len);
+int enic_mbox_send_link_state(struct enic *enic, u16 vf_id, u32 link_state);
 
 #endif /* _ENIC_MBOX_H_ */

-- 
2.43.0

[PATCH net-next v13 08/11] enic: add MBOX VF handlers for capability, register and link state

[Satish Kharat]

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>
---
 drivers/net/ethernet/cisco/enic/enic.h      |  12 ++
 drivers/net/ethernet/cisco/enic/enic_mbox.c | 277 +++++++++++++++++++++++++++-
 drivers/net/ethernet/cisco/enic/enic_mbox.h |   3 +
 3 files changed, 291 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/cisco/enic/enic.h b/drivers/net/ethernet/cisco/enic/enic.h
index 6261f3e8e4f8..db4d52a4158f 100644
--- a/drivers/net/ethernet/cisco/enic/enic.h
+++ b/drivers/net/ethernet/cisco/enic/enic.h
@@ -258,6 +258,8 @@ struct enic {
 	u32 tx_coalesce_usecs;
 	u16 num_vfs;
 	enum enic_vf_type vf_type;
+	bool vf_registered;
+	u32 pf_cap_version;
 	unsigned int enable_count;
 	spinlock_t enic_api_lock;
 	bool enic_api_busy;
@@ -313,6 +315,16 @@ struct enic {
 	/* MBOX protocol state — mbox_lock serializes admin WQ sends */
 	struct mutex mbox_lock;
 	u64 mbox_msg_num;
+	/* MBOX request-reply state.  mbox_expected_reply is written and
+	 * cleared by the process-context request helpers (capability/register/
+	 * unregister) and only read by the admin_msg_work receive handlers, so
+	 * it is annotated with READ_ONCE()/WRITE_ONCE() rather than locked:
+	 * only one request is in flight at a time (requesters run under RTNL or
+	 * single-threaded probe/remove), so each request is serialized and its
+	 * reply completes mbox_comp before the next request is issued.
+	 */
+	struct completion mbox_comp;
+	u8 mbox_expected_reply;
 
 	/* PF: per-VF MBOX state, allocated when SRIOV V2 is enabled */
 	struct enic_vf_state {
diff --git a/drivers/net/ethernet/cisco/enic/enic_mbox.c b/drivers/net/ethernet/cisco/enic/enic_mbox.c
index d1b13737e77d..6b8b44c4d96e 100644
--- a/drivers/net/ethernet/cisco/enic/enic_mbox.c
+++ b/drivers/net/ethernet/cisco/enic/enic_mbox.c
@@ -5,6 +5,7 @@
 #include <linux/netdevice.h>
 #include <linux/dma-mapping.h>
 #include <linux/delay.h>
+#include <linux/completion.h>
 
 #include "vnic_dev.h"
 #include "vnic_wq.h"
@@ -142,6 +143,16 @@ int enic_mbox_send_msg(struct enic *enic, u8 msg_type, u16 dst_vnic_id,
 	return err;
 }
 
+static int enic_mbox_wait_reply(struct enic *enic, unsigned long timeout_ms)
+{
+	unsigned long left;
+
+	left = wait_for_completion_timeout(&enic->mbox_comp,
+					   msecs_to_jiffies(timeout_ms));
+
+	return left ? 0 : -ETIMEDOUT;
+}
+
 int enic_mbox_send_link_state(struct enic *enic, u16 vf_id, u32 link_state)
 {
 	struct enic_mbox_pf_link_state_notif_msg notif = {};
@@ -313,6 +324,166 @@ static void enic_mbox_pf_process_msg(struct enic *enic,
 			    hdr->msg_type, vf_id, err);
 }
 
+static void enic_mbox_vf_handle_capability_reply(struct enic *enic,
+						 void *payload)
+{
+	struct enic_mbox_vf_capability_reply_msg *reply = payload;
+
+	if (READ_ONCE(enic->mbox_expected_reply) != ENIC_MBOX_VF_CAPABILITY_REPLY) {
+		netdev_warn(enic->netdev,
+			    "MBOX: stale capability reply (expected %u), drop\n",
+			    READ_ONCE(enic->mbox_expected_reply));
+		return;
+	}
+
+	if (le16_to_cpu(reply->reply.ret_major) == 0)
+		enic->pf_cap_version = le32_to_cpu(reply->version);
+	else
+		netdev_warn(enic->netdev,
+			    "MBOX: PF rejected capability request: %u/%u\n",
+			    le16_to_cpu(reply->reply.ret_major),
+			    le16_to_cpu(reply->reply.ret_minor));
+	complete(&enic->mbox_comp);
+}
+
+static void enic_mbox_vf_handle_register_reply(struct enic *enic,
+					       void *payload)
+{
+	struct enic_mbox_vf_register_reply_msg *reply = payload;
+
+	if (READ_ONCE(enic->mbox_expected_reply) != ENIC_MBOX_VF_REGISTER_REPLY) {
+		netdev_warn(enic->netdev,
+			    "MBOX: stale register reply (expected %u), drop\n",
+			    READ_ONCE(enic->mbox_expected_reply));
+		return;
+	}
+
+	if (le16_to_cpu(reply->reply.ret_major)) {
+		netdev_warn(enic->netdev,
+			    "MBOX: VF register rejected by PF: %u/%u\n",
+			    le16_to_cpu(reply->reply.ret_major),
+			    le16_to_cpu(reply->reply.ret_minor));
+	} else {
+		enic->vf_registered = true;
+	}
+	complete(&enic->mbox_comp);
+}
+
+static void enic_mbox_vf_handle_unregister_reply(struct enic *enic,
+						 void *payload)
+{
+	struct enic_mbox_vf_register_reply_msg *reply = payload;
+
+	if (READ_ONCE(enic->mbox_expected_reply) != ENIC_MBOX_VF_UNREGISTER_REPLY) {
+		netdev_warn(enic->netdev,
+			    "MBOX: stale unregister reply (expected %u), drop\n",
+			    READ_ONCE(enic->mbox_expected_reply));
+		return;
+	}
+
+	if (le16_to_cpu(reply->reply.ret_major)) {
+		netdev_warn(enic->netdev,
+			    "MBOX: VF unregister rejected by PF: %u/%u\n",
+			    le16_to_cpu(reply->reply.ret_major),
+			    le16_to_cpu(reply->reply.ret_minor));
+	} else {
+		enic->vf_registered = false;
+	}
+	complete(&enic->mbox_comp);
+}
+
+static void enic_mbox_vf_handle_link_state(struct enic *enic, void *payload)
+{
+	struct enic_mbox_pf_link_state_notif_msg *notif = payload;
+	struct enic_mbox_pf_link_state_ack_msg ack = {};
+	int err;
+
+	switch (le32_to_cpu(notif->link_state)) {
+	case ENIC_MBOX_LINK_STATE_ENABLE:
+		if (!netif_carrier_ok(enic->netdev))
+			netif_carrier_on(enic->netdev);
+		netdev_dbg(enic->netdev, "MBOX: link state -> UP\n");
+		break;
+	case ENIC_MBOX_LINK_STATE_DISABLE:
+		if (netif_carrier_ok(enic->netdev))
+			netif_carrier_off(enic->netdev);
+		netdev_dbg(enic->netdev, "MBOX: link state -> DOWN\n");
+		break;
+	default:
+		netdev_warn(enic->netdev, "MBOX: unknown link state %u\n",
+			    le32_to_cpu(notif->link_state));
+		ack.ack.ret_major = cpu_to_le16(ENIC_MBOX_ERR_GENERIC);
+		break;
+	}
+
+	err = enic_mbox_send_msg(enic, ENIC_MBOX_PF_LINK_STATE_ACK,
+				 ENIC_MBOX_DST_PF, &ack, sizeof(ack));
+	if (err && net_ratelimit())
+		netdev_warn(enic->netdev,
+			    "MBOX: failed to send link state ACK: %d\n", err);
+}
+
+static bool enic_mbox_vf_payload_ok(struct enic *enic, u8 msg_type,
+				    u16 payload_len, size_t min_len)
+{
+	if (payload_len < min_len) {
+		netdev_warn(enic->netdev,
+			    "MBOX: short payload for type %u (%u < %zu)\n",
+			    msg_type, payload_len, min_len);
+		return false;
+	}
+	return true;
+}
+
+static void enic_mbox_vf_process_msg(struct enic *enic,
+				     struct enic_mbox_hdr *hdr, void *payload,
+				     u16 payload_len)
+{
+	switch (hdr->msg_type) {
+	case ENIC_MBOX_VF_CAPABILITY_REPLY: {
+		size_t exp = sizeof(struct enic_mbox_vf_capability_reply_msg);
+
+		if (!enic_mbox_vf_payload_ok(enic, hdr->msg_type,
+					     payload_len, exp))
+			return;
+		enic_mbox_vf_handle_capability_reply(enic, payload);
+		break;
+	}
+	case ENIC_MBOX_VF_REGISTER_REPLY: {
+		size_t exp = sizeof(struct enic_mbox_vf_register_reply_msg);
+
+		if (!enic_mbox_vf_payload_ok(enic, hdr->msg_type,
+					     payload_len, exp))
+			return;
+		enic_mbox_vf_handle_register_reply(enic, payload);
+		break;
+	}
+	case ENIC_MBOX_VF_UNREGISTER_REPLY: {
+		size_t exp = sizeof(struct enic_mbox_vf_register_reply_msg);
+
+		if (!enic_mbox_vf_payload_ok(enic, hdr->msg_type,
+					     payload_len, exp))
+			return;
+		enic_mbox_vf_handle_unregister_reply(enic, payload);
+		break;
+	}
+	case ENIC_MBOX_PF_LINK_STATE_NOTIF: {
+		size_t exp = sizeof(struct enic_mbox_pf_link_state_notif_msg);
+
+		if (!enic_mbox_vf_payload_ok(enic, hdr->msg_type,
+					     payload_len, exp))
+			return;
+		enic_mbox_vf_handle_link_state(enic, payload);
+		break;
+	}
+	default:
+		netdev_dbg(enic->netdev,
+			   "MBOX: VF unhandled msg type %u\n",
+			   hdr->msg_type);
+		break;
+	}
+}
+
 static void enic_mbox_recv_handler(struct enic *enic, void *buf,
 				   unsigned int len)
 {
@@ -351,13 +522,117 @@ static void enic_mbox_recv_handler(struct enic *enic, void *buf,
 
 	payload = buf + sizeof(*hdr);
 
-	if (enic->vf_state)
+	if (enic->vf_state) {
 		enic_mbox_pf_process_msg(enic, hdr, payload);
+	} else if (le16_to_cpu(hdr->src_vnic_id) == ENIC_MBOX_DST_PF) {
+		/* src_vnic_id was overwritten from the hardware-verified CQ
+		 * VLAN sender field, so a VF only accepts messages that the
+		 * adapter attributes to the PF.  Its sole admin-channel peer is
+		 * the PF; drop anything else as a spoofed notification.
+		 */
+		enic_mbox_vf_process_msg(enic, hdr, payload,
+					 msg_len - (u16)sizeof(*hdr));
+	} else if (net_ratelimit()) {
+		netdev_warn(enic->netdev,
+			    "MBOX: VF dropping non-PF message from vnic %u\n",
+			    le16_to_cpu(hdr->src_vnic_id));
+	}
+}
+
+int enic_mbox_vf_capability_check(struct enic *enic)
+{
+	struct enic_mbox_vf_capability_msg req = {};
+	int err;
+
+	enic->pf_cap_version = 0;
+	reinit_completion(&enic->mbox_comp);
+	WRITE_ONCE(enic->mbox_expected_reply, ENIC_MBOX_VF_CAPABILITY_REPLY);
+	req.version = cpu_to_le32(ENIC_MBOX_CAP_VERSION_1);
+
+	err = enic_mbox_send_msg(enic, ENIC_MBOX_VF_CAPABILITY_REQUEST,
+				 ENIC_MBOX_DST_PF, &req, sizeof(req));
+	if (err) {
+		WRITE_ONCE(enic->mbox_expected_reply, 0);
+		return err;
+	}
+
+	err = enic_mbox_wait_reply(enic, 3000);
+	WRITE_ONCE(enic->mbox_expected_reply, 0);
+	if (err) {
+		netdev_warn(enic->netdev,
+			    "MBOX: no capability reply from PF\n");
+		return err;
+	}
+
+	if (enic->pf_cap_version < ENIC_MBOX_CAP_VERSION_1) {
+		netdev_warn(enic->netdev,
+			    "MBOX: PF rejected capability request or reported unsupported version %u\n",
+			    enic->pf_cap_version);
+		return -EOPNOTSUPP;
+	}
+
+	return 0;
+}
+
+int enic_mbox_vf_register(struct enic *enic)
+{
+	int err;
+
+	enic->vf_registered = false;
+	reinit_completion(&enic->mbox_comp);
+	WRITE_ONCE(enic->mbox_expected_reply, ENIC_MBOX_VF_REGISTER_REPLY);
+
+	err = enic_mbox_send_msg(enic, ENIC_MBOX_VF_REGISTER_REQUEST,
+				 ENIC_MBOX_DST_PF, NULL, 0);
+	if (err) {
+		WRITE_ONCE(enic->mbox_expected_reply, 0);
+		return err;
+	}
+
+	err = enic_mbox_wait_reply(enic, 3000);
+	WRITE_ONCE(enic->mbox_expected_reply, 0);
+	if (err) {
+		netdev_warn(enic->netdev,
+			    "MBOX: VF registration with PF timed out\n");
+		return err;
+	}
+
+	if (!enic->vf_registered)
+		return -ENODEV;
+
+	return 0;
+}
+
+int enic_mbox_vf_unregister(struct enic *enic)
+{
+	int err;
+
+	if (!enic->vf_registered)
+		return 0;
+
+	reinit_completion(&enic->mbox_comp);
+	WRITE_ONCE(enic->mbox_expected_reply, ENIC_MBOX_VF_UNREGISTER_REPLY);
+
+	err = enic_mbox_send_msg(enic, ENIC_MBOX_VF_UNREGISTER_REQUEST,
+				 ENIC_MBOX_DST_PF, NULL, 0);
+	if (err) {
+		WRITE_ONCE(enic->mbox_expected_reply, 0);
+		return err;
+	}
+
+	err = enic_mbox_wait_reply(enic, 3000);
+	WRITE_ONCE(enic->mbox_expected_reply, 0);
+	if (err)
+		return err;
+	if (enic->vf_registered)
+		return -EACCES;
+	return 0;
 }
 
 void enic_mbox_init(struct enic *enic)
 {
 	enic->mbox_msg_num = 0;
 	mutex_init(&enic->mbox_lock);
+	init_completion(&enic->mbox_comp);
 	enic->admin_rq_handler = enic_mbox_recv_handler;
 }
diff --git a/drivers/net/ethernet/cisco/enic/enic_mbox.h b/drivers/net/ethernet/cisco/enic/enic_mbox.h
index f1de67db1273..15e30ee2b0ed 100644
--- a/drivers/net/ethernet/cisco/enic/enic_mbox.h
+++ b/drivers/net/ethernet/cisco/enic/enic_mbox.h
@@ -88,5 +88,8 @@ void enic_mbox_init(struct enic *enic);
 int enic_mbox_send_msg(struct enic *enic, u8 msg_type, u16 dst_vnic_id,
 		       void *payload, u16 payload_len);
 int enic_mbox_send_link_state(struct enic *enic, u16 vf_id, u32 link_state);
+int enic_mbox_vf_capability_check(struct enic *enic);
+int enic_mbox_vf_register(struct enic *enic);
+int enic_mbox_vf_unregister(struct enic *enic);
 
 #endif /* _ENIC_MBOX_H_ */

-- 
2.43.0

[PATCH net-next v13 09/11] enic: wire V2 SR-IOV enable with admin channel and MBOX

[Satish Kharat]

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>
---
 drivers/net/ethernet/cisco/enic/enic.h       |   3 +
 drivers/net/ethernet/cisco/enic/enic_admin.c |   3 +
 drivers/net/ethernet/cisco/enic/enic_main.c  | 288 +++++++++++++++++++++++++--
 drivers/net/ethernet/cisco/enic/enic_mbox.c  |  13 +-
 drivers/net/ethernet/cisco/enic/enic_pp.c    |   5 +
 drivers/net/ethernet/cisco/enic/enic_res.c   |   1 +
 drivers/net/ethernet/cisco/enic/vnic_enet.h  |   4 +-
 7 files changed, 301 insertions(+), 16 deletions(-)

diff --git a/drivers/net/ethernet/cisco/enic/enic.h b/drivers/net/ethernet/cisco/enic/enic.h
index db4d52a4158f..4a67947cfb9f 100644
--- a/drivers/net/ethernet/cisco/enic/enic.h
+++ b/drivers/net/ethernet/cisco/enic/enic.h
@@ -305,6 +305,7 @@ struct enic {
 	struct vnic_intr admin_intr;
 	struct delayed_work admin_poll_work;
 	unsigned int admin_intr_index;
+	struct work_struct link_notify_work;
 	struct work_struct admin_msg_work;
 	spinlock_t admin_msg_lock;	/* protects admin_msg_list */
 	struct list_head admin_msg_list;
@@ -325,6 +326,7 @@ struct enic {
 	 */
 	struct completion mbox_comp;
 	u8 mbox_expected_reply;
+	bool mbox_initialized;
 
 	/* PF: per-VF MBOX state, allocated when SRIOV V2 is enabled */
 	struct enic_vf_state {
@@ -451,6 +453,7 @@ void enic_reset_addr_lists(struct enic *enic);
 int enic_sriov_enabled(struct enic *enic);
 int enic_is_valid_vf(struct enic *enic, int vf);
 int enic_is_dynamic(struct enic *enic);
+int enic_is_sriov_vf_v2(struct enic *enic);
 void enic_set_ethtool_ops(struct net_device *netdev);
 int __enic_set_rsskey(struct enic *enic);
 void enic_ext_cq(struct enic *enic);
diff --git a/drivers/net/ethernet/cisco/enic/enic_admin.c b/drivers/net/ethernet/cisco/enic/enic_admin.c
index f4a409fec983..7188f1b81c04 100644
--- a/drivers/net/ethernet/cisco/enic/enic_admin.c
+++ b/drivers/net/ethernet/cisco/enic/enic_admin.c
@@ -639,6 +639,7 @@ void enic_admin_channel_close(struct enic *enic)
 
 	vnic_intr_mask(&enic->admin_intr);
 	enic_admin_teardown_intr(enic);
+	cancel_work_sync(&enic->link_notify_work);
 	cancel_work_sync(&enic->admin_msg_work);
 	enic_admin_msg_drain(enic);
 
@@ -658,6 +659,8 @@ void enic_admin_channel_close(struct enic *enic)
 	vnic_cq_clean(&enic->admin_cq[0]);
 	vnic_cq_clean(&enic->admin_cq[1]);
 	vnic_intr_clean(&enic->admin_intr);
+
+	enic->admin_rq_handler = NULL;
 	enic_admin_free_resources(enic);
 
 	enic->admin_chan_up = false;
diff --git a/drivers/net/ethernet/cisco/enic/enic_main.c b/drivers/net/ethernet/cisco/enic/enic_main.c
index b56e5c75ade3..27b31d5a95dd 100644
--- a/drivers/net/ethernet/cisco/enic/enic_main.c
+++ b/drivers/net/ethernet/cisco/enic/enic_main.c
@@ -60,6 +60,8 @@
 #include "enic_clsf.h"
 #include "enic_rq.h"
 #include "enic_wq.h"
+#include "enic_admin.h"
+#include "enic_mbox.h"
 
 #define ENIC_NOTIFY_TIMER_PERIOD	(2 * HZ)
 
@@ -314,6 +316,11 @@ static int enic_is_sriov_vf(struct enic *enic)
 	       enic->pdev->device == PCI_DEVICE_ID_CISCO_VIC_ENET_VF_V2;
 }
 
+int enic_is_sriov_vf_v2(struct enic *enic)
+{
+	return enic->pdev->device == PCI_DEVICE_ID_CISCO_VIC_ENET_VF_V2;
+}
+
 int enic_is_valid_vf(struct enic *enic, int vf)
 {
 #ifdef CONFIG_PCI_IOV
@@ -411,18 +418,50 @@ static void enic_set_rx_coal_setting(struct enic *enic)
 	rx_coal->use_adaptive_rx_coalesce = 1;
 }
 
+static void enic_link_notify_work_handler(struct work_struct *work)
+{
+	struct enic *enic = container_of(work, struct enic,
+					 link_notify_work);
+	u32 state;
+	u16 i;
+
+	if (!enic_sriov_enabled(enic) || !enic->vf_state)
+		return;
+
+	state = netif_carrier_ok(enic->netdev) ?
+		ENIC_MBOX_LINK_STATE_ENABLE :
+		ENIC_MBOX_LINK_STATE_DISABLE;
+
+	for (i = 0; i < enic->num_vfs; i++)
+		enic_mbox_send_link_state(enic, i, state);
+}
+
 static void enic_link_check(struct enic *enic)
 {
-	int link_status = vnic_dev_link_status(enic->vdev);
-	int carrier_ok = netif_carrier_ok(enic->netdev);
+	int link_status;
+	int carrier_ok;
+
+	/* A V2 SR-IOV VF's carrier is driven by PF link-state MBOX
+	 * notifications, not by its own vnic link status; skip the
+	 * autonomous check so it cannot flap the VF carrier.
+	 */
+	if (enic_is_sriov_vf_v2(enic))
+		return;
+
+	link_status = vnic_dev_link_status(enic->vdev);
+	carrier_ok = netif_carrier_ok(enic->netdev);
 
 	if (link_status && !carrier_ok) {
 		netdev_info(enic->netdev, "Link UP\n");
 		netif_carrier_on(enic->netdev);
 		enic_set_rx_coal_setting(enic);
+		if (enic_sriov_enabled(enic) && enic->vf_state)
+			schedule_work(&enic->link_notify_work);
 	} else if (!link_status && carrier_ok) {
 		netdev_info(enic->netdev, "Link DOWN\n");
 		netif_carrier_off(enic->netdev);
+		if (enic_sriov_enabled(enic) && enic->vf_state)
+			schedule_work(&enic->link_notify_work);
 	}
 }
 
@@ -2154,15 +2193,47 @@ static void enic_reset(struct work_struct *work)
 	/* Stop any activity from infiniband */
 	enic_set_api_busy(enic, true);
 
+	/* Fully tear down the V2 admin/MBOX channel before the soft reset.
+	 * The reset wipes all hardware queues including the admin WQ/RQ;
+	 * closing first tells firmware to stop the admin QP (so it no longer
+	 * DMAs from the about-to-be-reset rings) and frees the admin resources
+	 * so they are cleanly re-allocated afterwards.
+	 */
+	if (enic_sriov_enabled(enic) &&
+	    enic->vf_type == ENIC_VF_TYPE_V2)
+		enic_admin_channel_close(enic);
+
 	enic_stop(enic->netdev);
+
 	enic_dev_soft_reset(enic);
 	enic_reset_addr_lists(enic);
 	enic_init_vnic_resources(enic);
 	enic_set_rss_nic_cfg(enic);
 	enic_dev_set_ig_vlan_rewrite_mode(enic);
 	enic_ext_cq(enic);
+
 	enic_open(enic->netdev);
 
+	/* Re-establish the admin/MBOX channel after the data path is back up,
+	 * mirroring the SR-IOV enable path (channel open + mbox init).  The
+	 * channel was fully torn down by enic_admin_channel_close() above.
+	 */
+	if (enic_sriov_enabled(enic) &&
+	    enic->vf_type == ENIC_VF_TYPE_V2) {
+		if (enic_admin_channel_open(enic)) {
+			netdev_err(enic->netdev,
+				   "admin channel reopen after reset failed\n");
+		} else {
+			enic_mbox_init(enic);
+			/* The link came back up during enic_open() above
+			 * while MBOX sends were still disabled (channel not
+			 * yet reopened), so that link-notify was dropped.
+			 * Re-push current link state to registered VFs now.
+			 */
+			schedule_work(&enic->link_notify_work);
+		}
+	}
+
 	/* Allow infiniband to fiddle with the device again */
 	enic_set_api_busy(enic, false);
 
@@ -2180,16 +2251,46 @@ static void enic_tx_hang_reset(struct work_struct *work)
 	/* Stop any activity from infiniband */
 	enic_set_api_busy(enic, true);
 
+	/* Fully tear down the V2 admin/MBOX channel before the hang reset, for
+	 * the same reason as the soft reset path: stop the admin QP and free
+	 * the admin resources before the hardware queues are wiped.
+	 */
+	if (enic_sriov_enabled(enic) &&
+	    enic->vf_type == ENIC_VF_TYPE_V2)
+		enic_admin_channel_close(enic);
+
 	enic_dev_hang_notify(enic);
 	enic_stop(enic->netdev);
+
 	enic_dev_hang_reset(enic);
 	enic_reset_addr_lists(enic);
 	enic_init_vnic_resources(enic);
 	enic_set_rss_nic_cfg(enic);
 	enic_dev_set_ig_vlan_rewrite_mode(enic);
 	enic_ext_cq(enic);
+
 	enic_open(enic->netdev);
 
+	/* Re-establish the admin/MBOX channel after the data path is back up,
+	 * mirroring the SR-IOV enable path (channel open + mbox init).  The
+	 * channel was fully torn down by enic_admin_channel_close() above.
+	 */
+	if (enic_sriov_enabled(enic) &&
+	    enic->vf_type == ENIC_VF_TYPE_V2) {
+		if (enic_admin_channel_open(enic)) {
+			netdev_err(enic->netdev,
+				   "admin channel reopen after reset failed\n");
+		} else {
+			enic_mbox_init(enic);
+			/* The link came back up during enic_open() above
+			 * while MBOX sends were still disabled (channel not
+			 * yet reopened), so that link-notify was dropped.
+			 * Re-push current link state to registered VFs now.
+			 */
+			schedule_work(&enic->link_notify_work);
+		}
+	}
+
 	/* Allow infiniband to fiddle with the device again */
 	enic_set_api_busy(enic, false);
 
@@ -2200,6 +2301,8 @@ static void enic_tx_hang_reset(struct work_struct *work)
 
 static int enic_set_intr_mode(struct enic *enic)
 {
+	unsigned int admin_reserve = enic->has_admin_channel ? 1 : 0;
+	unsigned int min_intr = ENIC_MSIX_MIN_INTR + admin_reserve;
 	unsigned int i;
 	int num_intr;
 
@@ -2210,12 +2313,12 @@ static int enic_set_intr_mode(struct enic *enic)
 	 */
 
 	if (enic->config.intr_mode < 1 &&
-	    enic->intr_avail >= ENIC_MSIX_MIN_INTR) {
+	    enic->intr_avail >= min_intr) {
 		for (i = 0; i < enic->intr_avail; i++)
 			enic->msix_entry[i].entry = i;
 
 		num_intr = pci_enable_msix_range(enic->pdev, enic->msix_entry,
-						 ENIC_MSIX_MIN_INTR,
+						 min_intr,
 						 enic->intr_avail);
 		if (num_intr > 0) {
 			vnic_dev_set_intr_mode(enic->vdev,
@@ -2310,7 +2413,13 @@ static int enic_adjust_resources(struct enic *enic)
 		enic->cq_count = 2;
 		enic->intr_count = enic->intr_avail;
 		break;
-	case VNIC_DEV_INTR_MODE_MSIX:
+	case VNIC_DEV_INTR_MODE_MSIX: {
+		/* Reserve one MSI-X slot for the admin channel interrupt
+		 * when V2 SR-IOV admin channel resources are present.
+		 */
+		unsigned int admin_reserve =
+			enic->has_admin_channel ? 1 : 0;
+
 		/* Adjust the number of wqs/rqs/cqs/interrupts that will be
 		 * used based on which resource is the most constrained
 		 */
@@ -2319,7 +2428,8 @@ static int enic_adjust_resources(struct enic *enic)
 				 ENIC_RQ_MIN_DEFAULT);
 		rq_avail = min3(enic->rq_avail, ENIC_RQ_MAX, rq_default);
 		max_queues = min(enic->cq_avail,
-				 enic->intr_avail - ENIC_MSIX_RESERVED_INTR);
+				 enic->intr_avail - ENIC_MSIX_RESERVED_INTR -
+				 admin_reserve);
 		if (wq_avail + rq_avail <= max_queues) {
 			enic->rq_count = rq_avail;
 			enic->wq_count = wq_avail;
@@ -2337,6 +2447,7 @@ static int enic_adjust_resources(struct enic *enic)
 		enic->intr_count = enic->cq_count + ENIC_MSIX_RESERVED_INTR;
 
 		break;
+	}
 	default:
 		dev_err(enic_get_dev(enic), "Unknown interrupt mode\n");
 		return -EINVAL;
@@ -2689,6 +2800,144 @@ static void enic_sriov_detect_vf_type(struct enic *enic)
 		enic->vf_type = ENIC_VF_TYPE_NONE;
 	}
 }
+
+static int __maybe_unused
+enic_sriov_v2_enable(struct enic *enic, int num_vfs)
+{
+	int err;
+
+	if (!enic->has_admin_channel) {
+		netdev_err(enic->netdev,
+			   "V2 SR-IOV requires admin channel resources\n");
+		return -EOPNOTSUPP;
+	}
+
+	enic->vf_state = kcalloc(num_vfs, sizeof(*enic->vf_state), GFP_KERNEL);
+	if (!enic->vf_state)
+		return -ENOMEM;
+
+	/* Install the MBOX receive handler before the admin interrupt is
+	 * unmasked in enic_admin_channel_open(), so no early completion is
+	 * dropped.
+	 */
+	enic_mbox_init(enic);
+
+	err = enic_admin_channel_open(enic);
+	if (err) {
+		netdev_err(enic->netdev,
+			   "Failed to open admin channel: %d\n", err);
+		goto free_vf_state;
+	}
+
+	enic->num_vfs = num_vfs;
+
+	err = pci_enable_sriov(enic->pdev, num_vfs);
+	if (err) {
+		netdev_err(enic->netdev,
+			   "pci_enable_sriov failed: %d\n", err);
+		goto close_admin;
+	}
+
+	enic->priv_flags |= ENIC_SRIOV_ENABLED;
+	return num_vfs;
+
+close_admin:
+	enic->num_vfs = 0;
+	enic_admin_channel_close(enic);
+free_vf_state:
+	kfree(enic->vf_state);
+	enic->vf_state = NULL;
+	return err;
+}
+
+static void enic_sriov_v2_disable(struct enic *enic)
+{
+	/* Stop new VF link-state broadcasts before tearing down vf_state.
+	 * Clearing ENIC_SRIOV_ENABLED makes enic_link_check() (called from
+	 * the notify timer/ISR) skip the VF notify path, and cancelling
+	 * link_notify_work ensures any already-queued broadcast has finished
+	 * before vf_state is freed, closing a use-after-free window.
+	 */
+	enic->priv_flags &= ~ENIC_SRIOV_ENABLED;
+	cancel_work_sync(&enic->link_notify_work);
+
+	pci_disable_sriov(enic->pdev);
+	enic_admin_channel_close(enic);
+	kfree(enic->vf_state);
+	enic->vf_state = NULL;
+	enic->num_vfs = 0;
+}
+
+/*
+ * enic_sriov_configure() and its V2 helpers are defined but not yet wired
+ * into enic_driver via .sriov_configure (see the __maybe_unused annotations);
+ * V2 enable/disable is activated in a follow-up series.  Because the callback
+ * is not registered, it cannot run concurrently with the rtnl-protected reset
+ * paths (enic_reset(), enic_tx_hang_reset()) yet.  Serialization against those
+ * paths is added together with the .sriov_configure wiring in that series.
+ */
+static int __maybe_unused
+enic_sriov_configure(struct pci_dev *pdev, int num_vfs)
+{
+	struct net_device *netdev = pci_get_drvdata(pdev);
+	struct enic *enic = netdev_priv(netdev);
+	struct enic_port_profile *pp;
+	int err;
+
+	if (num_vfs > 0) {
+		if (enic->config.mq_subvnic_count) {
+			netdev_err(netdev,
+				   "SR-IOV not supported with multi-queue sub-vnics\n");
+			return -EOPNOTSUPP;
+		}
+
+		if (enic->vf_type == ENIC_VF_TYPE_NONE) {
+			netdev_err(netdev,
+				   "SR-IOV not supported on this firmware version\n");
+			return -EOPNOTSUPP;
+		}
+
+		if (enic->vf_type == ENIC_VF_TYPE_V2)
+			return enic_sriov_v2_enable(enic, num_vfs);
+
+		pp = kcalloc(num_vfs, sizeof(*pp), GFP_KERNEL);
+		if (!pp)
+			return -ENOMEM;
+
+		err = pci_enable_sriov(pdev, num_vfs);
+		if (err) {
+			kfree(pp);
+			return err;
+		}
+
+		kfree(enic->pp);
+		enic->pp = pp;
+		enic->num_vfs = num_vfs;
+		enic->priv_flags |= ENIC_SRIOV_ENABLED;
+		return num_vfs;
+	}
+
+	if (!enic_sriov_enabled(enic))
+		return 0;
+
+	if (enic->vf_type == ENIC_VF_TYPE_V2) {
+		enic_sriov_v2_disable(enic);
+		return 0;
+	}
+
+	pp = kzalloc_obj(*enic->pp, GFP_KERNEL);
+	if (!pp)
+		return -ENOMEM;
+
+	pci_disable_sriov(pdev);
+	enic->num_vfs = 0;
+	enic->priv_flags &= ~ENIC_SRIOV_ENABLED;
+
+	kfree(enic->pp);
+	enic->pp = pp;
+
+	return 0;
+}
 #endif
 
 static int enic_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
@@ -2787,12 +3036,23 @@ static int enic_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
 		goto err_out_vnic_unregister;
 
 #ifdef CONFIG_PCI_IOV
-	/* Get number of subvnics */
+	enic_sriov_detect_vf_type(enic);
+
+	/* Auto-enable SR-IOV only for the legacy VF types.  V2 VFs require
+	 * the admin channel, which is not yet set up at probe time (V2 SR-IOV
+	 * will be enabled through the sysfs .sriov_configure callback once a
+	 * follow-up series wires it up); and a V2-capable device whose
+	 * firmware lacks V2 support is downgraded to ENIC_VF_TYPE_NONE by
+	 * enic_sriov_detect_vf_type() and must not be brought up through the
+	 * legacy pci_enable_sriov() path either.
+	 */
 	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_SRIOV);
 	if (pos) {
 		pci_read_config_word(pdev, pos + PCI_SRIOV_TOTAL_VF,
 			&enic->num_vfs);
-		if (enic->num_vfs) {
+		if (enic->num_vfs &&
+		    (enic->vf_type == ENIC_VF_TYPE_V1 ||
+		     enic->vf_type == ENIC_VF_TYPE_USNIC)) {
 			err = pci_enable_sriov(pdev, enic->num_vfs);
 			if (err) {
 				dev_err(dev, "SRIOV enable failed, aborting."
@@ -2804,7 +3064,6 @@ static int enic_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
 			num_pps = enic->num_vfs;
 		}
 	}
-	enic_sriov_detect_vf_type(enic);
 #endif
 
 	/* Allocate structure for port profiles */
@@ -2881,6 +3140,7 @@ static int enic_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
 	INIT_WORK(&enic->reset, enic_reset);
 	INIT_WORK(&enic->tx_hang_reset, enic_tx_hang_reset);
 	INIT_WORK(&enic->change_mtu_work, enic_change_mtu_work);
+	INIT_WORK(&enic->link_notify_work, enic_link_notify_work_handler);
 
 	for (i = 0; i < enic->wq_count; i++)
 		spin_lock_init(&enic->wq[i].lock);
@@ -3034,14 +3294,16 @@ static void enic_remove(struct pci_dev *pdev)
 		disable_work_sync(&enic->tx_hang_reset);
 		disable_work_sync(&enic->change_mtu_work);
 		unregister_netdev(netdev);
-		enic_dev_deinit(enic);
-		vnic_dev_close(enic->vdev);
 #ifdef CONFIG_PCI_IOV
 		if (enic_sriov_enabled(enic)) {
-			pci_disable_sriov(pdev);
-			enic->priv_flags &= ~ENIC_SRIOV_ENABLED;
+			if (enic->vf_type == ENIC_VF_TYPE_V2)
+				enic_sriov_v2_disable(enic);
+			else
+				pci_disable_sriov(pdev);
 		}
 #endif
+		enic_dev_deinit(enic);
+		vnic_dev_close(enic->vdev);
 		kfree(enic->pp);
 		vnic_dev_unregister(enic->vdev);
 		enic_iounmap(enic);
diff --git a/drivers/net/ethernet/cisco/enic/enic_mbox.c b/drivers/net/ethernet/cisco/enic/enic_mbox.c
index 6b8b44c4d96e..2fb0f1e2ff50 100644
--- a/drivers/net/ethernet/cisco/enic/enic_mbox.c
+++ b/drivers/net/ethernet/cisco/enic/enic_mbox.c
@@ -631,8 +631,17 @@ int enic_mbox_vf_unregister(struct enic *enic)
 
 void enic_mbox_init(struct enic *enic)
 {
+	/* mbox_lock and mbox_comp must be initialized exactly once per
+	 * device lifetime; the PF sriov_configure path can re-enter this
+	 * on each enable cycle where these primitives are already set up.
+	 */
+	if (!enic->mbox_initialized) {
+		mutex_init(&enic->mbox_lock);
+		init_completion(&enic->mbox_comp);
+		enic->mbox_initialized = true;
+	} else {
+		reinit_completion(&enic->mbox_comp);
+	}
 	enic->mbox_msg_num = 0;
-	mutex_init(&enic->mbox_lock);
-	init_completion(&enic->mbox_comp);
 	enic->admin_rq_handler = enic_mbox_recv_handler;
 }
diff --git a/drivers/net/ethernet/cisco/enic/enic_pp.c b/drivers/net/ethernet/cisco/enic/enic_pp.c
index 4720a952725d..3f611e240c25 100644
--- a/drivers/net/ethernet/cisco/enic/enic_pp.c
+++ b/drivers/net/ethernet/cisco/enic/enic_pp.c
@@ -25,6 +25,11 @@ int enic_is_valid_pp_vf(struct enic *enic, int vf, int *err)
 	if (vf != PORT_SELF_VF) {
 #ifdef CONFIG_PCI_IOV
 		if (enic_sriov_enabled(enic)) {
+			/* V2 SR-IOV uses MBOX, not port profiles */
+			if (enic->vf_type == ENIC_VF_TYPE_V2) {
+				*err = -EOPNOTSUPP;
+				goto err_out;
+			}
 			if (vf < 0 || vf >= enic->num_vfs) {
 				*err = -EINVAL;
 				goto err_out;
diff --git a/drivers/net/ethernet/cisco/enic/enic_res.c b/drivers/net/ethernet/cisco/enic/enic_res.c
index 2b7545d6a67f..436326ace049 100644
--- a/drivers/net/ethernet/cisco/enic/enic_res.c
+++ b/drivers/net/ethernet/cisco/enic/enic_res.c
@@ -59,6 +59,7 @@ int enic_get_vnic_config(struct enic *enic)
 	GET_CONFIG(intr_timer_usec);
 	GET_CONFIG(loop_tag);
 	GET_CONFIG(num_arfs);
+	GET_CONFIG(mq_subvnic_count);
 	GET_CONFIG(max_rq_ring);
 	GET_CONFIG(max_wq_ring);
 	GET_CONFIG(max_cq_ring);
diff --git a/drivers/net/ethernet/cisco/enic/vnic_enet.h b/drivers/net/ethernet/cisco/enic/vnic_enet.h
index 9e8e86262a3f..519d2969990b 100644
--- a/drivers/net/ethernet/cisco/enic/vnic_enet.h
+++ b/drivers/net/ethernet/cisco/enic/vnic_enet.h
@@ -21,7 +21,9 @@ struct vnic_enet_config {
 	u16 loop_tag;
 	u16 vf_rq_count;
 	u16 num_arfs;
-	u8 reserved[66];
+	u8 reserved1[32];
+	u16 mq_subvnic_count;
+	u8 reserved2[32];
 	u32 max_rq_ring;	// MAX RQ ring size
 	u32 max_wq_ring;	// MAX WQ ring size
 	u32 max_cq_ring;	// MAX CQ ring size

-- 
2.43.0

[PATCH net-next v13 10/11] enic: add V2 VF probe with admin channel and PF registration

[Satish Kharat]

When a V2 SR-IOV VF probes, initialize the MBOX protocol, open the
admin channel, perform the capability check with the PF, and register
with the PF. This establishes the PF-VF communication path that the PF
uses to send link state notifications.

The admin channel and MBOX registration happen after enic_dev_init()
(which discovers admin channel resources) and before register_netdev()
so the VF is fully initialized before the interface is visible to
userspace.

A V2 VF whose firmware did not provision admin WQ/RQ/CQ resources
fails probe with -ENODEV from enic_admin_channel_open(); the admin
channel is a hard requirement for V2 VFs.

enic_mbox_init() installs the receive handler and resets the message
sequence number before enic_admin_channel_open() unmasks the admin
interrupt, so a completion can never arrive before the handler is in
place.

On remove, the VF unregisters from the PF and closes its admin channel
before tearing down data path resources.

V2 VFs are not provisioned with an RES_TYPE_SRIOV_INTR resource by
firmware, so bypass that check in the admin channel capability
detection for V2 VFs. The PF still requires this resource.

The admin MSI-X vector reserved by enic_set_intr_mode()
is used for the admin channel interrupt.
enic_adjust_resources() ensures the reserved slot is within
intr_avail bounds even at maximum queue configurations.  The
admin INTR uses a RES_TYPE_INTR_CTRL slot shared with the
data path.

Signed-off-by: Satish Kharat <satishkh@cisco.com>
---
 drivers/net/ethernet/cisco/enic/enic_main.c | 94 ++++++++++++++++++++++++++---
 drivers/net/ethernet/cisco/enic/enic_res.c  |  3 +-
 2 files changed, 86 insertions(+), 11 deletions(-)

diff --git a/drivers/net/ethernet/cisco/enic/enic_main.c b/drivers/net/ethernet/cisco/enic/enic_main.c
index 27b31d5a95dd..537ed5ad7800 100644
--- a/drivers/net/ethernet/cisco/enic/enic_main.c
+++ b/drivers/net/ethernet/cisco/enic/enic_main.c
@@ -2414,15 +2414,19 @@ static int enic_adjust_resources(struct enic *enic)
 		enic->intr_count = enic->intr_avail;
 		break;
 	case VNIC_DEV_INTR_MODE_MSIX: {
-		/* Reserve one MSI-X slot for the admin channel interrupt
-		 * when V2 SR-IOV admin channel resources are present.
-		 */
-		unsigned int admin_reserve =
-			enic->has_admin_channel ? 1 : 0;
-
 		/* Adjust the number of wqs/rqs/cqs/interrupts that will be
-		 * used based on which resource is the most constrained
+		 * used based on which resource is the most constrained.
+		 * Reserve one extra MSI-X slot for the admin channel INTR
+		 * when has_admin_channel is set so that
+		 * enic_admin_setup_intr() can allocate at intr_count
+		 * within the intr_avail bounds even when the data queue
+		 * count is maxed out.  intr_count counts only the data-path
+		 * IRQs (registered by enic_request_intr()); the admin INTR
+		 * lives at msix index intr_count and is set up later by
+		 * enic_admin_setup_intr().
 		 */
+		unsigned int admin_reserve = enic->has_admin_channel ? 1 : 0;
+
 		wq_avail = min(enic->wq_avail, ENIC_WQ_MAX);
 		rq_default = max(netif_get_num_default_rss_queues(),
 				 ENIC_RQ_MIN_DEFAULT);
@@ -3128,6 +3132,42 @@ static int enic_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
 		goto err_out_dev_close;
 	}
 
+	/* Initialise link_notify_work before the V2-VF admin-open block below:
+	 * its error path (err_out_admin_close -> enic_admin_channel_close() ->
+	 * cancel_work_sync()) would otherwise act on an uninitialised work.
+	 */
+	INIT_WORK(&enic->link_notify_work, enic_link_notify_work_handler);
+
+	/* V2 VF: open admin channel and register with PF.
+	 * Must happen before register_netdev so the VF is fully
+	 * initialized before the interface is visible to userspace.
+	 *
+	 * enic_mbox_init() installs the receive handler and resets the
+	 * sequence number; it must run before enic_admin_channel_open()
+	 * unmasks the admin interrupt so an early completion is not dropped.
+	 */
+	if (enic_is_sriov_vf_v2(enic)) {
+		enic_mbox_init(enic);
+		err = enic_admin_channel_open(enic);
+		if (err) {
+			dev_err(dev,
+				"Failed to open admin channel: %d\n", err);
+			goto err_out_dev_deinit;
+		}
+		err = enic_mbox_vf_capability_check(enic);
+		if (err) {
+			dev_err(dev,
+				"MBOX capability check failed: %d\n", err);
+			goto err_out_admin_close;
+		}
+		err = enic_mbox_vf_register(enic);
+		if (err) {
+			dev_err(dev,
+				"MBOX VF registration failed: %d\n", err);
+			goto err_out_admin_close;
+		}
+	}
+
 	netif_set_real_num_tx_queues(netdev, enic->wq_count);
 	netif_set_real_num_rx_queues(netdev, enic->rq_count);
 
@@ -3140,7 +3180,6 @@ static int enic_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
 	INIT_WORK(&enic->reset, enic_reset);
 	INIT_WORK(&enic->tx_hang_reset, enic_tx_hang_reset);
 	INIT_WORK(&enic->change_mtu_work, enic_change_mtu_work);
-	INIT_WORK(&enic->link_notify_work, enic_link_notify_work_handler);
 
 	for (i = 0; i < enic->wq_count; i++)
 		spin_lock_init(&enic->wq[i].lock);
@@ -3153,7 +3192,7 @@ static int enic_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
 	err = enic_set_mac_addr(netdev, enic->mac_addr);
 	if (err) {
 		dev_err(dev, "Invalid MAC address, aborting\n");
-		goto err_out_dev_deinit;
+		goto err_out_admin_close;
 	}
 
 	enic->tx_coalesce_usecs = enic->config.intr_timer_usec;
@@ -3251,11 +3290,23 @@ static int enic_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
 	err = register_netdev(netdev);
 	if (err) {
 		dev_err(dev, "Cannot register net device, aborting\n");
-		goto err_out_dev_deinit;
+		goto err_out_admin_close;
 	}
 
 	return 0;
 
+err_out_admin_close:
+	if (enic_is_sriov_vf_v2(enic)) {
+		if (enic->vf_registered) {
+			int unreg_err = enic_mbox_vf_unregister(enic);
+
+			if (unreg_err)
+				netdev_warn(netdev,
+					    "Failed to unregister from PF: %d\n",
+					    unreg_err);
+		}
+		enic_admin_channel_close(enic);
+	}
 err_out_dev_deinit:
 	enic_dev_deinit(enic);
 err_out_dev_close:
@@ -3293,7 +3344,30 @@ static void enic_remove(struct pci_dev *pdev)
 		disable_work_sync(&enic->reset);
 		disable_work_sync(&enic->tx_hang_reset);
 		disable_work_sync(&enic->change_mtu_work);
+
+		/* Close the admin channel and unregister from the PF before
+		 * unregister_netdev() to prevent a late PF notification from
+		 * touching a netdev that is being torn down.
+		 */
+		if (enic_is_sriov_vf_v2(enic)) {
+			if (enic->vf_registered) {
+				int unreg_err = enic_mbox_vf_unregister(enic);
+
+				if (unreg_err)
+					netdev_warn(netdev,
+						    "Failed to unregister from PF: %d\n",
+						    unreg_err);
+			}
+			enic_admin_channel_close(enic);
+		}
+
 		unregister_netdev(netdev);
+		/* unregister_netdev() -> enic_stop() stops the notify timer, so
+		 * no new link_notify_work can be queued past this point.  Cancel
+		 * unconditionally to cover the narrow window where
+		 * enic_link_check() scheduled it just as SR-IOV was disabled.
+		 */
+		cancel_work_sync(&enic->link_notify_work);
 #ifdef CONFIG_PCI_IOV
 		if (enic_sriov_enabled(enic)) {
 			if (enic->vf_type == ENIC_VF_TYPE_V2)
diff --git a/drivers/net/ethernet/cisco/enic/enic_res.c b/drivers/net/ethernet/cisco/enic/enic_res.c
index 436326ace049..74cd2ee3af5c 100644
--- a/drivers/net/ethernet/cisco/enic/enic_res.c
+++ b/drivers/net/ethernet/cisco/enic/enic_res.c
@@ -211,7 +211,8 @@ void enic_get_res_counts(struct enic *enic)
 		vnic_dev_get_res_count(enic->vdev, RES_TYPE_ADMIN_RQ) >= 1 &&
 		vnic_dev_get_res_count(enic->vdev, RES_TYPE_ADMIN_CQ) >=
 			ARRAY_SIZE(enic->admin_cq) &&
-		vnic_dev_get_res_count(enic->vdev, RES_TYPE_SRIOV_INTR) >= 1;
+		(enic_is_sriov_vf_v2(enic) ||
+		 vnic_dev_get_res_count(enic->vdev, RES_TYPE_SRIOV_INTR) >= 1);
 
 	dev_info(enic_get_dev(enic),
 		"vNIC resources avail: wq %d rq %d cq %d intr %d admin %s\n",

-- 
2.43.0

[PATCH net-next v13 11/11] enic: re-establish V2 VF admin channel and PF registration after reset

[Satish Kharat]

The reset paths (enic_reset/enic_tx_hang_reset) tore down and re-opened
the V2 admin/MBOX channel only for the PF: the close/reopen was gated on
enic_sriov_enabled() && vf_type == ENIC_VF_TYPE_V2, which is never true on
a VF (vf_type is set only on the PF; VFs are identified by
enic_is_sriov_vf_v2()).  A VF-initiated reset therefore left the VF admin
QP wiped by the reset but never re-opened, and the VF never re-registered
with the PF, so VF<->PF MBOX traffic (currently link state)
stopped working until the VF was re-probed.

Factor the decision into enic_has_admin_chan() (true for a V2 PF while
SR-IOV is enabled and for every V2 VF) and the reopen sequence into
enic_admin_chan_reopen().  For a VF the helper additionally re-runs the
probe-time handshake (enic_mbox_vf_capability_check() +
enic_mbox_vf_register()) so the PF learns about the VF again; for a PF it
re-pushes the current link state as before.

Before reopening, invalidate the VF's local registration flag.  The reset
only wipes the VF's admin QP, not the PF's software vf_state (that changes
only via the register/unregister MBOX handlers), so the PF may still hold
a stale "registered" until the VF re-registers.  Locally, a failed reopen
or re-handshake must not leave a stale registered state that a later
teardown would try to unregister over a dead channel.

Signed-off-by: Satish Kharat <satishkh@cisco.com>
---
 drivers/net/ethernet/cisco/enic/enic_main.c | 122 +++++++++++++++++++---------
 1 file changed, 82 insertions(+), 40 deletions(-)

diff --git a/drivers/net/ethernet/cisco/enic/enic_main.c b/drivers/net/ethernet/cisco/enic/enic_main.c
index 537ed5ad7800..0baef7a120ec 100644
--- a/drivers/net/ethernet/cisco/enic/enic_main.c
+++ b/drivers/net/ethernet/cisco/enic/enic_main.c
@@ -2181,6 +2181,74 @@ static void enic_set_api_busy(struct enic *enic, bool busy)
 	spin_unlock(&enic->enic_api_lock);
 }
 
+/* The admin/MBOX channel exists on a V2 PF while SR-IOV is enabled and on
+ * every V2 VF.  A reset wipes the admin WQ/RQ/CQ, so such devices must tear
+ * the channel down before the reset and re-establish it afterwards.
+ */
+static bool enic_has_admin_chan(struct enic *enic)
+{
+	return enic_is_sriov_vf_v2(enic) ||
+	       (enic_sriov_enabled(enic) && enic->vf_type == ENIC_VF_TYPE_V2);
+}
+
+/* Re-establish the admin/MBOX channel after a reset has re-created the data
+ * path.  Mirrors the relevant part of the probe / SR-IOV-enable sequence:
+ * reinitialise MBOX and reopen the channel, then for a VF re-run the PF
+ * handshake (the reset wiped the VF's admin QP, so the VF must register
+ * again), or for a PF re-push the current link state to registered VFs.
+ */
+static void enic_admin_chan_reopen(struct enic *enic)
+{
+	int err;
+
+	/* Install the MBOX receive handler and reset the sequence number
+	 * before opening the channel, so the handler is in place before the
+	 * admin interrupt is unmasked and no early completion is dropped.
+	 */
+	enic_mbox_init(enic);
+
+	/* A reset destroys the VF's local admin QP, so the VF can no longer
+	 * rely on its previous registration.  The PF may retain stale software
+	 * registration state until the VF successfully registers again.
+	 * Clear the local flag before reopening so a failed reopen or
+	 * re-handshake cannot leave the VF believing it has a usable PF
+	 * registration over a dead channel.
+	 */
+	if (enic_is_sriov_vf_v2(enic))
+		enic->vf_registered = false;
+
+	err = enic_admin_channel_open(enic);
+	if (err) {
+		netdev_err(enic->netdev,
+			   "admin channel reopen after reset failed: %d\n", err);
+		return;
+	}
+
+	if (enic_is_sriov_vf_v2(enic)) {
+		err = enic_mbox_vf_capability_check(enic);
+		if (err) {
+			netdev_err(enic->netdev,
+				   "MBOX capability check after reset failed: %d\n",
+				   err);
+			enic_admin_channel_close(enic);
+			return;
+		}
+		err = enic_mbox_vf_register(enic);
+		if (err) {
+			netdev_err(enic->netdev,
+				   "MBOX VF re-registration after reset failed: %d\n",
+				   err);
+			enic_admin_channel_close(enic);
+		}
+	} else {
+		/* The link came back up during enic_open() above while MBOX
+		 * sends were still disabled (channel not yet reopened), so that
+		 * link-notify was dropped.  Re-push current link state now.
+		 */
+		schedule_work(&enic->link_notify_work);
+	}
+}
+
 static void enic_reset(struct work_struct *work)
 {
 	struct enic *enic = container_of(work, struct enic, reset);
@@ -2199,8 +2267,7 @@ static void enic_reset(struct work_struct *work)
 	 * DMAs from the about-to-be-reset rings) and frees the admin resources
 	 * so they are cleanly re-allocated afterwards.
 	 */
-	if (enic_sriov_enabled(enic) &&
-	    enic->vf_type == ENIC_VF_TYPE_V2)
+	if (enic_has_admin_chan(enic))
 		enic_admin_channel_close(enic);
 
 	enic_stop(enic->netdev);
@@ -2214,25 +2281,13 @@ static void enic_reset(struct work_struct *work)
 
 	enic_open(enic->netdev);
 
-	/* Re-establish the admin/MBOX channel after the data path is back up,
-	 * mirroring the SR-IOV enable path (channel open + mbox init).  The
-	 * channel was fully torn down by enic_admin_channel_close() above.
+	/* Re-establish the admin/MBOX channel after the data path is back up.
+	 * It was fully torn down by enic_admin_channel_close() above;
+	 * enic_admin_chan_reopen() reopens it and, for a PF re-pushes link
+	 * state, or for a VF re-runs the probe-time PF handshake.
 	 */
-	if (enic_sriov_enabled(enic) &&
-	    enic->vf_type == ENIC_VF_TYPE_V2) {
-		if (enic_admin_channel_open(enic)) {
-			netdev_err(enic->netdev,
-				   "admin channel reopen after reset failed\n");
-		} else {
-			enic_mbox_init(enic);
-			/* The link came back up during enic_open() above
-			 * while MBOX sends were still disabled (channel not
-			 * yet reopened), so that link-notify was dropped.
-			 * Re-push current link state to registered VFs now.
-			 */
-			schedule_work(&enic->link_notify_work);
-		}
-	}
+	if (enic_has_admin_chan(enic))
+		enic_admin_chan_reopen(enic);
 
 	/* Allow infiniband to fiddle with the device again */
 	enic_set_api_busy(enic, false);
@@ -2255,8 +2310,7 @@ static void enic_tx_hang_reset(struct work_struct *work)
 	 * the same reason as the soft reset path: stop the admin QP and free
 	 * the admin resources before the hardware queues are wiped.
 	 */
-	if (enic_sriov_enabled(enic) &&
-	    enic->vf_type == ENIC_VF_TYPE_V2)
+	if (enic_has_admin_chan(enic))
 		enic_admin_channel_close(enic);
 
 	enic_dev_hang_notify(enic);
@@ -2271,25 +2325,13 @@ static void enic_tx_hang_reset(struct work_struct *work)
 
 	enic_open(enic->netdev);
 
-	/* Re-establish the admin/MBOX channel after the data path is back up,
-	 * mirroring the SR-IOV enable path (channel open + mbox init).  The
-	 * channel was fully torn down by enic_admin_channel_close() above.
+	/* Re-establish the admin/MBOX channel after the data path is back up.
+	 * It was fully torn down by enic_admin_channel_close() above;
+	 * enic_admin_chan_reopen() reopens it and, for a PF re-pushes link
+	 * state, or for a VF re-runs the probe-time PF handshake.
 	 */
-	if (enic_sriov_enabled(enic) &&
-	    enic->vf_type == ENIC_VF_TYPE_V2) {
-		if (enic_admin_channel_open(enic)) {
-			netdev_err(enic->netdev,
-				   "admin channel reopen after reset failed\n");
-		} else {
-			enic_mbox_init(enic);
-			/* The link came back up during enic_open() above
-			 * while MBOX sends were still disabled (channel not
-			 * yet reopened), so that link-notify was dropped.
-			 * Re-push current link state to registered VFs now.
-			 */
-			schedule_work(&enic->link_notify_work);
-		}
-	}
+	if (enic_has_admin_chan(enic))
+		enic_admin_chan_reopen(enic);
 
 	/* Allow infiniband to fiddle with the device again */
 	enic_set_api_busy(enic, false);

-- 
2.43.0

Re: [PATCH net-next v13 00/11] enic: SR-IOV V2 admin channel and MBOX protocol

[patchwork-bot+netdevbpf]

Hello:

This series was applied to netdev/net-next.git (main)
by Jakub Kicinski <kuba@kernel.org>:

On Wed, 12 Aug 2026 05:48:00 -0700 you wrote:
> This series adds the admin channel infrastructure and mailbox (MBOX)
> protocol needed for V2 SR-IOV support in the enic driver.
> 
> The V2 SR-IOV design uses a direct PF-VF communication channel built on
> dedicated WQ/RQ/CQ hardware resources and an MSI-X interrupt.
> 
> This series touches enic_remove(), which is also modified by a fix
> already applied to net:
> 
> [...]

Here is the summary with links:
  - [net-next,v13,01/11] enic: verify firmware supports V2 SR-IOV at probe time
    https://git.kernel.org/netdev/net-next/c/a16975e2231b
  - [net-next,v13,02/11] enic: add admin channel open and close for SR-IOV
    https://git.kernel.org/netdev/net-next/c/3258931d4052
  - [net-next,v13,03/11] enic: add admin RQ buffer management
    https://git.kernel.org/netdev/net-next/c/a0da4ea750a2
  - [net-next,v13,04/11] enic: add admin CQ service with MSI-X interrupt and workqueue polling
    https://git.kernel.org/netdev/net-next/c/cb1dba54c723
  - [net-next,v13,05/11] enic: define MBOX message types and header structures
    https://git.kernel.org/netdev/net-next/c/1584793311ce
  - [net-next,v13,06/11] enic: add MBOX core send and receive for admin channel
    https://git.kernel.org/netdev/net-next/c/1f0c856b5963
  - [net-next,v13,07/11] enic: add MBOX PF handlers for VF register and capability
    https://git.kernel.org/netdev/net-next/c/06bdfb16216b
  - [net-next,v13,08/11] enic: add MBOX VF handlers for capability, register and link state
    https://git.kernel.org/netdev/net-next/c/72b65c94058e
  - [net-next,v13,09/11] enic: wire V2 SR-IOV enable with admin channel and MBOX
    https://git.kernel.org/netdev/net-next/c/4ff204cef01e
  - [net-next,v13,10/11] enic: add V2 VF probe with admin channel and PF registration
    https://git.kernel.org/netdev/net-next/c/b2dfdc966b65
  - [net-next,v13,11/11] enic: re-establish V2 VF admin channel and PF registration after reset
    https://git.kernel.org/netdev/net-next/c/885f462fe90e

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html