[PATCH v1 0/3] wifi: brcm80211: avoid memcpy warning when CONFIG_FORTIFY_SOURCE=y

[Petr Tesarik <petrtesarik@huaweicloud.com>]

From: Petr Tesarik <petr.tesarik.ext@huawei.com>

When built with CONFIG_FORTIFY_SOURCE=y, the brcmfmac module triggers a
memcpy() warning like this:

 ------------[ cut here ]------------
 memcpy: detected field-spanning write (size 76) of single field "&params_le->channel_list[0]" at drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c:1072 (size 2)
 WARNING: CPU: 2 PID: 991 at drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c:1072 brcmf_scan_params_v2_to_v1+0xd4/0x118 [brcmfmac]
 Modules linked in: qrtr(E) algif_hash(E) aes_neon_bs(E) aes_neon_blk(E) algif_skcipher(E) af_alg(E) bnep(E) brcmfmac_wcc(E) bcm2835_v4l2(CE) bcm2835_mmal_vchiq(CE) videobuf2_vmalloc(E) videobuf2_memops(E) videobuf2_v4l2(E) videodev(E) hci_uart(E) btsdio(E) videobuf2_common(E) btbcm(E) mc(E) snd_bcm2835(CE) bluetooth(E) snd_pcm(E) brcmfmac(E) snd_timer(E) cpufreq_dt(E) snd(E) soundcore(E) cfg80211(E) ecdh_generic(E) brcmutil(E) raspberrypi_cpufreq(E) rfkill(E) vchiq(CE) bcm2711_thermal(E) leds_gpio(E) fuse(E) efi_pstore(E) dmi_sysfs(E) ip_tables(E) x_tables(E) rpcsec_gss_krb5(E) auth_rpcgss(E) nfsv4(E) dns_resolver(E) nfs(E) lockd(E) grace(E) fscache(E) netfs(E) af_packet(E) mmc_block(E) xhci_pci(E) xhci_pci_renesas(E) xhci_hcd(E) usbcore(E) usb_common(E) clk_raspberrypi(E) gpio_raspberrypi_exp(E) bcm2835_dma(E) crct10dif_ce(E) virt_dma(E) pcie_brcmstb(E) sdhci_iproc(E) gpio_regulator(E) sdhci_pltfm(E) sdhci(E) mmc_core(E) fixed(E) nvmem_rmem(E) sunrpc(E) sg(E) dm_multipath(E) dm_mod(E) efivarfs(E)
 Unloaded tainted modules: aes_ce_cipher(E):1
 CPU: 2 PID: 991 Comm: wpa_supplicant Tainted: G         C  E      6.5.0-rc4-dynswiotlb+ #27 2ec0961165cc91fdbec101d9d43b3331ba4f0927
 Hardware name: Unknown Unknown Product/Unknown Product, BIOS 2023.04 04/01/2023
 pstate: 60400005 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
 pc : brcmf_scan_params_v2_to_v1+0xd4/0x118 [brcmfmac]
 lr : brcmf_scan_params_v2_to_v1+0xd4/0x118 [brcmfmac]
 sp : ffff8000829ab590
 x29: ffff8000829ab590 x28: 0000000000000000 x27: 0000000000000001
 x26: ffff000105e7e0a4 x25: ffff00010a0bcb48 x24: ffff000101e03800
 x23: ffff000105ec8920 x22: ffff000106332980 x21: ffff00010a0bc0c0
 x20: ffff00010a0bcb90 x19: ffff00010a0bc108 x18: ffffffffffffffff
 x17: 0000000000000000 x16: 0000000000000000 x15: 616f72622f737365
 x14: 6c657269772f7465 x13: 616d666d6372622f x12: 31313230386d6372
 x11: 00000000ffffdfff x10: ffff800081ad3328 x9 : ffff800080130694
 x8 : 000000000002ffe8 x7 : c0000000ffffdfff x6 : 00000000000affa8
 x5 : ffff0001fef75e00 x4 : 0000000000000000 x3 : 0000000000000027
 x2 : 0000000000000000 x1 : 0000000000000000 x0 : ffff00010aa0e000
 Call trace:
  brcmf_scan_params_v2_to_v1+0xd4/0x118 [brcmfmac 38c4a81a3b85b4aff1650c67f95f20bc542d60c1]
  brcmf_run_escan+0x148/0x1a0 [brcmfmac 38c4a81a3b85b4aff1650c67f95f20bc542d60c1]
  brcmf_do_escan+0x74/0xe0 [brcmfmac 38c4a81a3b85b4aff1650c67f95f20bc542d60c1]
  brcmf_cfg80211_scan+0xcc/0x298 [brcmfmac 38c4a81a3b85b4aff1650c67f95f20bc542d60c1]
  rdev_scan+0x38/0x158 [cfg80211 8907673111c49ec56be88af3d38994cc1cf54cb8]
  cfg80211_scan+0x134/0x178 [cfg80211 8907673111c49ec56be88af3d38994cc1cf54cb8]
  nl80211_trigger_scan+0x3e8/0x768 [cfg80211 8907673111c49ec56be88af3d38994cc1cf54cb8]
  genl_family_rcv_msg_doit.isra.0+0xc0/0x130
  genl_rcv_msg+0x1e4/0x278
  netlink_rcv_skb+0x64/0x138
  genl_rcv+0x40/0x60
  netlink_unicast+0x1cc/0x2d8
  netlink_sendmsg+0x1d4/0x448
  sock_sendmsg+0x64/0xc0
  ____sys_sendmsg+0x260/0x2e0
  ___sys_sendmsg+0x88/0xf0
  __sys_sendmsg+0x70/0xd8
  __arm64_sys_sendmsg+0x2c/0x40
  invoke_syscall+0x78/0x100
  el0_svc_common.constprop.0+0x100/0x130
  do_el0_svc+0x40/0xa8
  el0_svc+0x34/0x138
  el0t_64_sync_handler+0x120/0x130
  el0t_64_sync+0x1a8/0x1b0
 ---[ end trace 0000000000000000 ]---

This is in fact a false positive, but the data structures can be improved
to make the checker happy.

Tested on a Raspberry Pi 4 by running scans in a loop while activating and
deactivating the wireless connection in parallel to cause scan aborts.

Petr Tesarik (3):
  wifi: brcm80211: drop struct brcmf_p2p_scan_le
  wifi: brcm80211: separate abort scan param prepare from actual scan
  wifi: brcm80211: change channel_list to a flexible array

 .../broadcom/brcm80211/brcmfmac/cfg80211.c    | 67 ++++++++++---------
 .../broadcom/brcm80211/brcmfmac/fwil_types.h  | 14 +++-
 .../broadcom/brcm80211/brcmfmac/p2p.c         | 38 +++--------
 3 files changed, 57 insertions(+), 62 deletions(-)

-- 
2.25.1