[PATCH v3 0/8] wifi: brcm80211: performance and stability improvements

[Shivesh <chanelshivesh@gmail.com>]

From: Shivesh <chanelshivesh@gmail.com>

This series addresses several performance bottlenecks and correctness
issues in both brcmfmac (PCIe/SDIO) and brcmsmac drivers.

Changes since v2:
- core.c: Add skb_cow_head() check before skb_push() to prevent a
  kernel panic when the skb arrives with zero headroom in monitor mode
  receive path.
- cfg80211.c: Fix a silent PMKID list corruption bug in brcmf_pmksa_v2_op.
  The shadow entry counter (npmk) was stored in list->length, which was
  then overwritten with the wire-format byte-length at the send: label.
  On subsequent calls, le16_to_cpu(list->length) would decode the byte-
  length as an entry count, silently corrupting the PMKID cache.

Changes since v1:
- Sent with correct author/Signed-off-by address (chanelshivesh@gmail.com).

Patch overview:
  1. flowring: Replace O(N) blocked-ring scan with an atomic per-interface
     counter, fixing a teardown counter leak that permanently stopped netif.
  2. sdio: Coalesce N+1 separate sdio_claim_host/release pairs in the
     rxglom path into a single pair covering all hdparse calls.
  3. core: Populate the radiotap header with RSSI from wlc_d11rxhdr in
     monitor mode, fixing missing signal strength in packet captures.
     Add skb_cow_head() to prevent panic on zero-headroom skbs.
  4. cfg80211: Implement the PMKID_V2 set/del/flush paths for FILS
     fast-roaming support on firmware revision 12 devices. Fix the V2
     shadow counter overwrite bug. Replace mdelay() busy-wait in
     brcmf_delay() with usleep_range().
  5. msgbuf: Increase NR_TX_PKTIDS (2048->4096) and TX flush thresholds
     (32/96->64/128). Fix silent TX stall under high load by using
     test_and_set_bit to guarantee worker scheduling.
  6. pcie: Replace fixed msleep(10) mailbox polling with usleep_range
     exponential backoff. Replace msleep(50) IRQ teardown polling with
     usleep_range for faster teardown.
  7. fwsignal: Document that a duplicate MAC handle ADD is a safe no-op.
  8. brcmsmac/ampdu: Clarify IEEE 802.11n TID compliance comment.

Shivesh (8):
  wifi: brcmfmac: flowring: replace O(N) loop with atomic counter
  wifi: brcmfmac: sdio: coalesce host locks in rx path
  wifi: brcmfmac: core: populate radiotap header with RSSI
  wifi: brcmfmac: cfg80211: implement PMKID_V2 and fix delay busy-wait
  wifi: brcmfmac: msgbuf: tune thresholds and optimize sleep latency
  wifi: brcmfmac: pcie: optimize latency and irq teardown
  wifi: brcmfmac: fwsignal: safe no-op on duplicate MAC add
  wifi: brcmsmac: ampdu: clarify standard compliance on QoS change

 .../broadcom/brcm80211/brcmfmac/cfg80211.c    | 123 ++++++++++++++++--
 .../broadcom/brcm80211/brcmfmac/cfg80211.h    |   4 +-
 .../broadcom/brcm80211/brcmfmac/core.c        |  44 ++++---
 .../broadcom/brcm80211/brcmfmac/flowring.c    |  65 ++++++---
 .../broadcom/brcm80211/brcmfmac/flowring.h    |  18 +++
 .../broadcom/brcm80211/brcmfmac/fwsignal.c    |   4 +-
 .../broadcom/brcm80211/brcmfmac/msgbuf.c      |  69 ++++++++--
 .../broadcom/brcm80211/brcmfmac/pcie.c        |  66 ++++++++--
 .../broadcom/brcm80211/brcmfmac/sdio.c        |  62 +++++----
 .../broadcom/brcm80211/brcmsmac/ampdu.c       |  11 +-
 10 files changed, 375 insertions(+), 91 deletions(-)


base-commit: 5e0bcb4920bcb9e0ef6c03a91730fe45f643f3a4
-- 
2.53.0