wifi: rtl8xxxu: enable channel switch support

The CSA countdown in the beacon frames, which are sent out by firmware,
needs to get updated by the driver. To achieve this, convert
update_beacon_work to delayed_work and schedule it with the beacon
interval in case CSA is active and the countdown is not complete.

Signed-off-by: Martin Kaistra <martin.kaistra@linutronix.de>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240111163628.320697-3-martin.kaistra@linutronix.de

wifi: rtl8xxxu: add cancel_work_sync() for c2hcmd_work

The workqueue might still be running, when the driver is stopped. To
avoid a use-after-free, call cancel_work_sync() in rtl8xxxu_stop().

Fixes: e542e66b7c2e ("rtl8xxxu: add bluetooth co-existence support for single antenna")
Signed-off-by: Martin Kaistra <martin.kaistra@linutronix.de>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240111163628.320697-2-martin.kaistra@linutronix.de

wifi: rtlwifi: rtl_usb: Use sync register writes

Currently rtl_usb performs register writes using the async
usb_submit_urb() function. This appears to work fine for the RTL8192CU,
but the RTL8192DU (soon to be supported by rtlwifi) has a problem:
it transmits everything at the 1M rate in the 2.4 GHz band. (The 5 GHz
band is still untested.)

With this patch, rtl_usb performs the register writes using the
synchronous usb_control_msg() function, and the RTL8192DU works
normally. The RTL8192CU still works.

The vendor drivers use the async writes in only one function,
rtl8192du_trigger_gpio_0 / rtl8192cu_trigger_gpio_0, which probably
doesn't even run in real life. They use sync writes everywhere else.

Also, remove "sync" and "async" from the names of the members of
struct rtl_io to avoid confusion:

write{8,16,32}_async -> write{8,16,32}
 read{8,16,32}_sync  ->  read{8,16,32}

Signed-off-by: Bitterblue Smith <rtl8821cerfe2@gmail.com>
Acked-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/fb71bae6-8b19-4b6e-b4a6-0d260f2139e1@gmail.com

wifi: rtlwifi: cleanup few rtlxxx_tx_fill_desc() routines

Remove unreachable branches in 'rtl92ce_tx_fill_desc()',
'rtl92cu_tx_fill_desc()' and 'rtl8723e_tx_fill_desc()'.
Compile tested only.

Found by Linux Verification Center (linuxtesting.org) with SVACE.

Signed-off-by: Dmitry Antipov <dmantipov@yandex.ru>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240109104735.140550-1-dmantipov@yandex.ru

wifi: rtw89: add chip_ops::update_beacon to abstract update beacon operation

Since coming WiFi 7 and existing chips use different update_beacon()
format, add to abstract selection of H2C command.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240108091359.67636-1-pkshih@realtek.com

wifi: rtw89: add H2C command to download beacon frame for WiFi 7 chips

The firmware of coming WiFi 7 chips can support more AP functions, like
virtual AP, so extend format of download beacon frame to configure them.
Currently rtw89 only enables AP mode as old chips, so leave new fields
zeros.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240108091333.67484-1-pkshih@realtek.com

wifi: rtw89: use struct to fill H2C command to download beacon frame

Download beacon frame via H2C command for AP mode, and then firmware can
issues beacon periodically. Originally TIM offset minus fixed 24 bytes of
frame header implicitly in macro, and this patch explicitly uses
ieee80211_hdrlen() to get header length, but expected to change nothing
at all.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240108091325.67424-1-pkshih@realtek.com

wifi: rtw89: add new H2C command to pause/sleep transmitting by MAC ID

New H2C command is introduced to pause/sleep transmitting. That extends
more bits to support more MAC ID, and currently we still configure 256
MAC ID at most.

This new command is always used by coming WiFi 7 chips, and existing
chips can use old or new command because firmware still preserves the old
one. Add a firmware feature flag to determine which command is adopted
according to firmware version.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240108091315.67358-1-pkshih@realtek.com

wifi: rtw89: refine H2C command that pause transmitting by MAC ID

To reuse this function to support extended H2C command that is used by
newer chip, change to use a pointer to fill pause ID and mask.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240108091307.67296-1-pkshih@realtek.com

wifi: rtw89: fw: use struct to fill BA CAM H2C commands

The WiFi 6 chips use these commands to create BA CAM entry when RX BA
session is established, this BA CAM can record whether packets are received
or not, then reply BA for current status. This patch doesn't change logic.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240108091245.67220-1-pkshih@realtek.com

wifi: rtw89: 8922a: update BA CAM number to 24

The total BA CAM number of 8922a is 32 instead of 16, and initial 8
entries are dynamic, so update the entry number to 24.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240108091134.67007-5-pkshih@realtek.com

wifi: rtw89: add chip_ops::h2c_ba_cam() to configure BA CAM

Since chips could use different version of BA CAM H2C command, add a
chip_ops to abstract the operation.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240108091134.67007-4-pkshih@realtek.com

wifi: rtw89: mac: add feature_init to initialize BA CAM V1

Add a call of feature_init() when bringing interface up. For now, the
feature is to reset BA CAM V1 that is only used by upcoming 8922A.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240108091134.67007-3-pkshih@realtek.com

wifi: rtw89: add firmware H2C command of BA CAM V1

BA CAM is used to generate BA frame for received AMPDU packets. To support
WiFi 7, change format from V0 to have more fields and enlarge entry number
for new need.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240108091134.67007-2-pkshih@realtek.com

wifi: wilc1000: fix RCU usage in connect path

With lockdep enabled, calls to the connect function from cfg802.11 layer
lead to the following warning:

=============================
WARNING: suspicious RCU usage
6.7.0-rc1-wt+ #333 Not tainted
-----------------------------
drivers/net/wireless/microchip/wilc1000/hif.c:386
suspicious rcu_dereference_check() usage!
[...]
stack backtrace:
CPU: 0 PID: 100 Comm: wpa_supplicant Not tainted 6.7.0-rc1-wt+ #333
Hardware name: Atmel SAMA5
 unwind_backtrace from show_stack+0x18/0x1c
 show_stack from dump_stack_lvl+0x34/0x48
 dump_stack_lvl from wilc_parse_join_bss_param+0x7dc/0x7f4
 wilc_parse_join_bss_param from connect+0x2c4/0x648
 connect from cfg80211_connect+0x30c/0xb74
 cfg80211_connect from nl80211_connect+0x860/0xa94
 nl80211_connect from genl_rcv_msg+0x3fc/0x59c
 genl_rcv_msg from netlink_rcv_skb+0xd0/0x1f8
 netlink_rcv_skb from genl_rcv+0x2c/0x3c
 genl_rcv from netlink_unicast+0x3b0/0x550
 netlink_unicast from netlink_sendmsg+0x368/0x688
 netlink_sendmsg from ____sys_sendmsg+0x190/0x430
 ____sys_sendmsg from ___sys_sendmsg+0x110/0x158
 ___sys_sendmsg from sys_sendmsg+0xe8/0x150
 sys_sendmsg from ret_fast_syscall+0x0/0x1c

This warning is emitted because in the connect path, when trying to parse
target BSS parameters, we dereference a RCU pointer whithout being in RCU
critical section.
Fix RCU dereference usage by moving it to a RCU read critical section. To
avoid wrapping the whole wilc_parse_join_bss_param under the critical
section, just use the critical section to copy ies data

Fixes: c460495ee072 ("staging: wilc1000: fix incorrent type in initializer")
Signed-off-by: Alexis Lothoré <alexis.lothore@bootlin.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240105075733.36331-3-alexis.lothore@bootlin.com

wifi: wilc1000: fix declarations ordering

Reorder parameters declaration in wilc_parse_join_bss_param to enforce
reverse christmas tree

Signed-off-by: Alexis Lothoré <alexis.lothore@bootlin.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240105075733.36331-2-alexis.lothore@bootlin.com

wifi: b43: Disable QoS for bcm4331

bcm4331 seems to not function correctly with QoS support. This may be due
to issues with currently available firmware or potentially a device
specific issue.

When queues that are not of the default "best effort" priority are
selected, traffic appears to not transmit out of the hardware while no
errors are returned. This behavior is present among all the other priority
queues: video, voice, and background. While this can be worked around by
setting a kernel parameter, the default behavior is problematic for most
users and may be difficult to debug. This patch offers a working out-of-box
experience for bcm4331 users.

Log of the issue (using ssh low-priority traffic as an example):
    ssh -T -vvvv git@github.com
    OpenSSH_9.6p1, OpenSSL 3.0.12 24 Oct 2023
    debug1: Reading configuration data /etc/ssh/ssh_config
    debug2: checking match for 'host * exec "/nix/store/q1c2flcykgr4wwg5a6h450hxbk4ch589-bash-5.2-p15/bin/bash -c '/nix/store/c015armnkhr6v18za0rypm7sh1i8js8w-gnupg-2.4.1/bin/gpg-connect-agent --quiet updatestartuptty /bye >/dev/null 2>&1'"' host github.com originally github.com
    debug3: /etc/ssh/ssh_config line 5: matched 'host "github.com"'
    debug1: Executing command: '/nix/store/q1c2flcykgr4wwg5a6h450hxbk4ch589-bash-5.2-p15/bin/bash -c '/nix/store/c015armnkhr6v18za0rypm7sh1i8js8w-gnupg-2.4.1/bin/gpg-connect-agent --quiet updatestartuptty /bye >/dev/null 2>&1''
    debug3: command returned status 0
    debug3: /etc/ssh/ssh_config line 5: matched 'exec "/nix/store/q1c2flcykgr4wwg5a6h450hxbk4ch589-bash-5.2-p15/bin/bash -c '/nix/store/c015armnkhr6v18za0r"'
    debug2: match found
    debug1: /etc/ssh/ssh_config line 9: Applying options for *
    debug3: expanded UserKnownHostsFile '~/.ssh/known_hosts' -> '/home/binary-eater/.ssh/known_hosts'
    debug3: expanded UserKnownHostsFile '~/.ssh/known_hosts2' -> '/home/binary-eater/.ssh/known_hosts2'
    debug2: resolving "github.com" port 22
    debug3: resolve_host: lookup github.com:22
    debug3: channel_clear_timeouts: clearing
    debug3: ssh_connect_direct: entering
    debug1: Connecting to github.com [192.30.255.113] port 22.
    debug3: set_sock_tos: set socket 3 IP_TOS 0x48

Fixes: e6f5b934fba8 ("b43: Add QOS support")
Signed-off-by: Rahul Rameshbabu <sergeantsagara@protonmail.com>
Reviewed-by: Julian Calaby <julian.calaby@gmail.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231231050300.122806-5-sergeantsagara@protonmail.com

wifi: b43: Stop correct queue in DMA worker when QoS is disabled

When QoS is disabled, the queue priority value will not map to the correct
ieee80211 queue since there is only one queue. Stop queue 0 when QoS is
disabled to prevent trying to stop a non-existent queue and failing to stop
the actual queue instantiated.

Fixes: bad691946966 ("b43: avoid packet losses in the dma worker code.")
Signed-off-by: Rahul Rameshbabu <sergeantsagara@protonmail.com>
Reviewed-by: Julian Calaby <julian.calaby@gmail.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231231050300.122806-4-sergeantsagara@protonmail.com

wifi: b43: Stop/wake correct queue in PIO Tx path when QoS is disabled

When QoS is disabled, the queue priority value will not map to the correct
ieee80211 queue since there is only one queue. Stop/wake queue 0 when QoS
is disabled to prevent trying to stop/wake a non-existent queue and failing
to stop/wake the actual queue instantiated.

Fixes: 5100d5ac81b9 ("b43: Add PIO support for PCMCIA devices")
Signed-off-by: Rahul Rameshbabu <sergeantsagara@protonmail.com>
Reviewed-by: Julian Calaby <julian.calaby@gmail.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231231050300.122806-3-sergeantsagara@protonmail.com

wifi: b43: Stop/wake correct queue in DMA Tx path when QoS is disabled

When QoS is disabled, the queue priority value will not map to the correct
ieee80211 queue since there is only one queue. Stop/wake queue 0 when QoS
is disabled to prevent trying to stop/wake a non-existent queue and failing
to stop/wake the actual queue instantiated.

Log of issue before change (with kernel parameter qos=0):
    [  +5.112651] ------------[ cut here ]------------
    [  +0.000005] WARNING: CPU: 7 PID: 25513 at net/mac80211/util.c:449 __ieee80211_wake_queue+0xd5/0x180 [mac80211]
    [  +0.000067] Modules linked in: b43(O) snd_seq_dummy snd_hrtimer snd_seq snd_seq_device nft_chain_nat xt_MASQUERADE nf_nat xfrm_user xfrm_algo xt_addrtype overlay ccm af_packet amdgpu snd_hda_codec_cirrus snd_hda_codec_generic ledtrig_audio drm_exec amdxcp gpu_sched xt_conntrack nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 ip6t_rpfilter ipt_rpfilter xt_pkttype xt_LOG nf_log_syslog xt_tcpudp nft_compat nf_tables nfnetlink sch_fq_codel btusb uinput iTCO_wdt ctr btrtl intel_pmc_bxt i915 intel_rapl_msr mei_hdcp mei_pxp joydev at24 watchdog btintel atkbd libps2 serio radeon btbcm vivaldi_fmap btmtk intel_rapl_common snd_hda_codec_hdmi bluetooth uvcvideo nls_iso8859_1 applesmc nls_cp437 x86_pkg_temp_thermal snd_hda_intel intel_powerclamp vfat videobuf2_vmalloc coretemp fat snd_intel_dspcfg crc32_pclmul uvc polyval_clmulni snd_intel_sdw_acpi loop videobuf2_memops snd_hda_codec tun drm_suballoc_helper polyval_generic drm_ttm_helper drm_buddy tap ecdh_generic videobuf2_v4l2 gf128mul macvlan ttm ghash_clmulni_intel ecc tg3
    [  +0.000044]  videodev bridge snd_hda_core rapl crc16 drm_display_helper cec mousedev snd_hwdep evdev intel_cstate bcm5974 hid_appleir videobuf2_common stp mac_hid libphy snd_pcm drm_kms_helper acpi_als mei_me intel_uncore llc mc snd_timer intel_gtt industrialio_triggered_buffer apple_mfi_fastcharge i2c_i801 mei snd lpc_ich agpgart ptp i2c_smbus thunderbolt apple_gmux i2c_algo_bit kfifo_buf video industrialio soundcore pps_core wmi tiny_power_button sbs sbshc button ac cordic bcma mac80211 cfg80211 ssb rfkill libarc4 kvm_intel kvm drm irqbypass fuse backlight firmware_class efi_pstore configfs efivarfs dmi_sysfs ip_tables x_tables autofs4 dm_crypt cbc encrypted_keys trusted asn1_encoder tee tpm rng_core input_leds hid_apple led_class hid_generic usbhid hid sd_mod t10_pi crc64_rocksoft crc64 crc_t10dif crct10dif_generic ahci libahci libata uhci_hcd ehci_pci ehci_hcd crct10dif_pclmul crct10dif_common sha512_ssse3 sha512_generic sha256_ssse3 sha1_ssse3 aesni_intel usbcore scsi_mod libaes crypto_simd cryptd scsi_common
    [  +0.000055]  usb_common rtc_cmos btrfs blake2b_generic libcrc32c crc32c_generic crc32c_intel xor raid6_pq dm_snapshot dm_bufio dm_mod dax [last unloaded: b43(O)]
    [  +0.000009] CPU: 7 PID: 25513 Comm: irq/17-b43 Tainted: G        W  O       6.6.7 #1-NixOS
    [  +0.000003] Hardware name: Apple Inc. MacBookPro8,3/Mac-942459F5819B171B, BIOS 87.0.0.0.0 06/13/2019
    [  +0.000001] RIP: 0010:__ieee80211_wake_queue+0xd5/0x180 [mac80211]
    [  +0.000046] Code: 00 45 85 e4 0f 85 9b 00 00 00 48 8d bd 40 09 00 00 f0 48 0f ba ad 48 09 00 00 00 72 0f 5b 5d 41 5c 41 5d 41 5e e9 cb 6d 3c d0 <0f> 0b 5b 5d 41 5c 41 5d 41 5e c3 cc cc cc cc 48 8d b4 16 94 00 00
    [  +0.000002] RSP: 0018:ffffc90003c77d60 EFLAGS: 00010097
    [  +0.000001] RAX: 0000000000000001 RBX: 0000000000000002 RCX: 0000000000000000
    [  +0.000001] RDX: 0000000000000000 RSI: 0000000000000002 RDI: ffff88820b924900
    [  +0.000002] RBP: ffff88820b924900 R08: ffffc90003c77d90 R09: 000000000003bfd0
    [  +0.000001] R10: ffff88820b924900 R11: ffffc90003c77c68 R12: 0000000000000000
    [  +0.000001] R13: 0000000000000000 R14: ffffc90003c77d90 R15: ffffffffc0fa6f40
    [  +0.000001] FS:  0000000000000000(0000) GS:ffff88846fb80000(0000) knlGS:0000000000000000
    [  +0.000001] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
    [  +0.000001] CR2: 00007fafda7ae008 CR3: 000000046d220005 CR4: 00000000000606e0
    [  +0.000002] Call Trace:
    [  +0.000003]  <TASK>
    [  +0.000001]  ? __ieee80211_wake_queue+0xd5/0x180 [mac80211]
    [  +0.000044]  ? __warn+0x81/0x130
    [  +0.000005]  ? __ieee80211_wake_queue+0xd5/0x180 [mac80211]
    [  +0.000045]  ? report_bug+0x171/0x1a0
    [  +0.000004]  ? handle_bug+0x41/0x70
    [  +0.000004]  ? exc_invalid_op+0x17/0x70
    [  +0.000003]  ? asm_exc_invalid_op+0x1a/0x20
    [  +0.000005]  ? __ieee80211_wake_queue+0xd5/0x180 [mac80211]
    [  +0.000043]  ieee80211_wake_queue+0x4a/0x80 [mac80211]
    [  +0.000044]  b43_dma_handle_txstatus+0x29c/0x3a0 [b43]
    [  +0.000016]  ? __pfx_irq_thread_fn+0x10/0x10
    [  +0.000002]  b43_handle_txstatus+0x61/0x80 [b43]
    [  +0.000012]  b43_interrupt_thread_handler+0x3f9/0x6b0 [b43]
    [  +0.000011]  irq_thread_fn+0x23/0x60
    [  +0.000002]  irq_thread+0xfe/0x1c0
    [  +0.000002]  ? __pfx_irq_thread_dtor+0x10/0x10
    [  +0.000001]  ? __pfx_irq_thread+0x10/0x10
    [  +0.000001]  kthread+0xe8/0x120
    [  +0.000003]  ? __pfx_kthread+0x10/0x10
    [  +0.000003]  ret_from_fork+0x34/0x50
    [  +0.000002]  ? __pfx_kthread+0x10/0x10
    [  +0.000002]  ret_from_fork_asm+0x1b/0x30
    [  +0.000004]  </TASK>
    [  +0.000001] ---[ end trace 0000000000000000 ]---

    [  +0.000065] ------------[ cut here ]------------
    [  +0.000001] WARNING: CPU: 0 PID: 56077 at net/mac80211/util.c:514 __ieee80211_stop_queue+0xcc/0xe0 [mac80211]
    [  +0.000077] Modules linked in: b43(O) snd_seq_dummy snd_hrtimer snd_seq snd_seq_device nft_chain_nat xt_MASQUERADE nf_nat xfrm_user xfrm_algo xt_addrtype overlay ccm af_packet amdgpu snd_hda_codec_cirrus snd_hda_codec_generic ledtrig_audio drm_exec amdxcp gpu_sched xt_conntrack nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 ip6t_rpfilter ipt_rpfilter xt_pkttype xt_LOG nf_log_syslog xt_tcpudp nft_compat nf_tables nfnetlink sch_fq_codel btusb uinput iTCO_wdt ctr btrtl intel_pmc_bxt i915 intel_rapl_msr mei_hdcp mei_pxp joydev at24 watchdog btintel atkbd libps2 serio radeon btbcm vivaldi_fmap btmtk intel_rapl_common snd_hda_codec_hdmi bluetooth uvcvideo nls_iso8859_1 applesmc nls_cp437 x86_pkg_temp_thermal snd_hda_intel intel_powerclamp vfat videobuf2_vmalloc coretemp fat snd_intel_dspcfg crc32_pclmul uvc polyval_clmulni snd_intel_sdw_acpi loop videobuf2_memops snd_hda_codec tun drm_suballoc_helper polyval_generic drm_ttm_helper drm_buddy tap ecdh_generic videobuf2_v4l2 gf128mul macvlan ttm ghash_clmulni_intel ecc tg3
    [  +0.000073]  videodev bridge snd_hda_core rapl crc16 drm_display_helper cec mousedev snd_hwdep evdev intel_cstate bcm5974 hid_appleir videobuf2_common stp mac_hid libphy snd_pcm drm_kms_helper acpi_als mei_me intel_uncore llc mc snd_timer intel_gtt industrialio_triggered_buffer apple_mfi_fastcharge i2c_i801 mei snd lpc_ich agpgart ptp i2c_smbus thunderbolt apple_gmux i2c_algo_bit kfifo_buf video industrialio soundcore pps_core wmi tiny_power_button sbs sbshc button ac cordic bcma mac80211 cfg80211 ssb rfkill libarc4 kvm_intel kvm drm irqbypass fuse backlight firmware_class efi_pstore configfs efivarfs dmi_sysfs ip_tables x_tables autofs4 dm_crypt cbc encrypted_keys trusted asn1_encoder tee tpm rng_core input_leds hid_apple led_class hid_generic usbhid hid sd_mod t10_pi crc64_rocksoft crc64 crc_t10dif crct10dif_generic ahci libahci libata uhci_hcd ehci_pci ehci_hcd crct10dif_pclmul crct10dif_common sha512_ssse3 sha512_generic sha256_ssse3 sha1_ssse3 aesni_intel usbcore scsi_mod libaes crypto_simd cryptd scsi_common
    [  +0.000084]  usb_common rtc_cmos btrfs blake2b_generic libcrc32c crc32c_generic crc32c_intel xor raid6_pq dm_snapshot dm_bufio dm_mod dax [last unloaded: b43]
    [  +0.000012] CPU: 0 PID: 56077 Comm: kworker/u16:17 Tainted: G        W  O       6.6.7 #1-NixOS
    [  +0.000003] Hardware name: Apple Inc. MacBookPro8,3/Mac-942459F5819B171B, BIOS 87.0.0.0.0 06/13/2019
    [  +0.000001] Workqueue: phy7 b43_tx_work [b43]
    [  +0.000019] RIP: 0010:__ieee80211_stop_queue+0xcc/0xe0 [mac80211]
    [  +0.000076] Code: 74 11 48 8b 78 08 0f b7 d6 89 e9 4c 89 e6 e8 ab f4 00 00 65 ff 0d 9c b7 34 3f 0f 85 55 ff ff ff 0f 1f 44 00 00 e9 4b ff ff ff <0f> 0b 5b 5d 41 5c 41 5d c3 cc cc cc cc 0f 1f 80 00 00 00 00 90 90
    [  +0.000002] RSP: 0000:ffffc90004157d50 EFLAGS: 00010097
    [  +0.000002] RAX: 0000000000000001 RBX: 0000000000000002 RCX: 0000000000000000
    [  +0.000002] RDX: 0000000000000000 RSI: 0000000000000002 RDI: ffff8882d65d0900
    [  +0.000002] RBP: 0000000000000000 R08: 0000000000000001 R09: 0000000000000001
    [  +0.000001] R10: 00000000000000ff R11: ffff88814d0155a0 R12: ffff8882d65d0900
    [  +0.000002] R13: 0000000000000000 R14: ffff8881002d2800 R15: 00000000000000d0
    [  +0.000002] FS:  0000000000000000(0000) GS:ffff88846f800000(0000) knlGS:0000000000000000
    [  +0.000003] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
    [  +0.000002] CR2: 00007f2e8c10c880 CR3: 0000000385b66005 CR4: 00000000000606f0
    [  +0.000002] Call Trace:
    [  +0.000001]  <TASK>
    [  +0.000001]  ? __ieee80211_stop_queue+0xcc/0xe0 [mac80211]
    [  +0.000075]  ? __warn+0x81/0x130
    [  +0.000004]  ? __ieee80211_stop_queue+0xcc/0xe0 [mac80211]
    [  +0.000075]  ? report_bug+0x171/0x1a0
    [  +0.000005]  ? handle_bug+0x41/0x70
    [  +0.000003]  ? exc_invalid_op+0x17/0x70
    [  +0.000004]  ? asm_exc_invalid_op+0x1a/0x20
    [  +0.000004]  ? __ieee80211_stop_queue+0xcc/0xe0 [mac80211]
    [  +0.000076]  ieee80211_stop_queue+0x36/0x50 [mac80211]
    [  +0.000077]  b43_dma_tx+0x550/0x780 [b43]
    [  +0.000023]  b43_tx_work+0x90/0x130 [b43]
    [  +0.000018]  process_one_work+0x174/0x340
    [  +0.000003]  worker_thread+0x27b/0x3a0
    [  +0.000004]  ? __pfx_worker_thread+0x10/0x10
    [  +0.000002]  kthread+0xe8/0x120
    [  +0.000003]  ? __pfx_kthread+0x10/0x10
    [  +0.000004]  ret_from_fork+0x34/0x50
    [  +0.000002]  ? __pfx_kthread+0x10/0x10
    [  +0.000003]  ret_from_fork_asm+0x1b/0x30
    [  +0.000006]  </TASK>
    [  +0.000001] ---[ end trace 0000000000000000 ]---

Fixes: e6f5b934fba8 ("b43: Add QOS support")
Signed-off-by: Rahul Rameshbabu <sergeantsagara@protonmail.com>
Reviewed-by: Julian Calaby <julian.calaby@gmail.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231231050300.122806-2-sergeantsagara@protonmail.com

wifi: rtl8xxxu: Fix off by one initial RTS rate

rtl8xxxu_set_basic_rates() sets the wrong initial RTS rate. It sets the
next higher rate than the one it should set, e.g. 36M instead of 24M.

The while loop was supposed to find the index of the most significant
bit which is 1, but it was copied incorrectly from the vendor driver.
Use __fls() instead.

Signed-off-by: Bitterblue Smith <rtl8821cerfe2@gmail.com>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/761e6836-6cd6-4930-91b6-0446834655c5@gmail.com

wifi: rtl8xxxu: Fix LED control code of RTL8192FU

Some devices, like the Comfast CF-826F, use LED1, which already works.
Others, like Asus USB-N13 C1, use LED0, which doesn't work correctly.

Write the right values to the LED control registers to make LED0 work
as well.

This is unfortunately tested only with the Comfast CF-826F.

Signed-off-by: Bitterblue Smith <rtl8821cerfe2@gmail.com>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/7a2c3158-3a45-4466-b11e-fc09802b20e2@gmail.com

wifi: rtl8xxxu: declare concurrent mode support for 8188f

Everything is in place now for concurrent mode, we can tell the system
that we support it.
We will allow a maximum of 2 virtual interfaces, one of them can be in
AP mode.

Signed-off-by: Martin Kaistra <martin.kaistra@linutronix.de>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231222101442.626837-22-martin.kaistra@linutronix.de

wifi: rtl8xxxu: make supporting AP mode only on port 0 transparent

When the driver is used for concurrent mode, both virtual interfaces can
be set to station or AP mode, though only one can be in AP mode at the
same time.

In order to keep the code simple, use only hw port 0 for AP mode. When
an interface is added in AP mode which would be assigned to port 1, use
a switch_port function to transparently swap the mapping between virtual
interface and hw port.

Signed-off-by: Martin Kaistra <martin.kaistra@linutronix.de>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231222101442.626837-21-martin.kaistra@linutronix.de

wifi: rtl8xxxu: add hw crypto support for AP mode

Add a custom function for allocating entries in the sec cam. This allows
us to store multiple keys with the same keyidx.

The maximum number of sec cam entries for 8188f is 16 according to the
vendor driver. Add the number to rtl8xxxu_fileops, so that other chips
which might support more entries, can set a different number there.

Set the bssid as mac address for group keys instead of just using the
ethernet broadcast address and use BIT(6) in the sec cam ctrl entry
for differentiating them from pairwise keys like in the vendor driver.

Add the TXDESC_EN_DESC_ID bit and the hw_key_idx to tx
broadcast/multicast packets in AP mode.

Finally, allow the usage of rtl8xxxu_set_key() for AP mode.

Signed-off-by: Martin Kaistra <martin.kaistra@linutronix.de>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231222101442.626837-20-martin.kaistra@linutronix.de

wifi: rtl8xxxu: remove obsolete priv->vif

Now that all uses of priv->vif have been converted to priv->vifs[]
remove the old attribute.

Signed-off-by: Martin Kaistra <martin.kaistra@linutronix.de>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231222101442.626837-19-martin.kaistra@linutronix.de

wifi: rtl8xxxu: add macids for STA mode

Until now, the driver only assigned a dedicated macid for connections
made in AP mode, in STA mode the return value of rtl8xxxu_get_macid()
was simply 0.
To differentiate between port 0 and 1, when both are in STA mode,
allocate a second macid (with value 1) and set sta_info->macid according
to the used port_num in rtl8xxxu_sta_add().

Signed-off-by: Martin Kaistra <martin.kaistra@linutronix.de>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231222101442.626837-18-martin.kaistra@linutronix.de

wifi: rtl8xxxu: support multiple interface in start_ap()

Call set_bssid() with the correct port_num now.

Signed-off-by: Martin Kaistra <martin.kaistra@linutronix.de>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231222101442.626837-17-martin.kaistra@linutronix.de

wifi: rtl8xxxu: support multiple interfaces in bss_info_changed()

Call set_linktype and set_bssid now with correct port_num. Call
stop_tx_beacon only for port 0, as we don't support beacons on port 1.
Explicit changes to BEACON will only happen for AP type interfaces, so
we don't need an additional check there.

Signed-off-by: Martin Kaistra <martin.kaistra@linutronix.de>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231222101442.626837-16-martin.kaistra@linutronix.de

wifi: rtl8xxxu: support multiple interfaces in {add,remove}_interface()

Add a custom struct to store in vif->drv_priv with a reference to
port_num and fill it when a new interface is added. Choose a free
port_num for the newly added interface.

As we only want to support AP mode/sending beacons on port 0, only change
the beacon settings if a new interface is actually assigned to port 0.

Call set_linktype() and set_mac() with the appropriate port_num.

Signed-off-by: Martin Kaistra <martin.kaistra@linutronix.de>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231222101442.626837-15-martin.kaistra@linutronix.de

wifi: rtl8xxxu: support multiple interfaces in watchdog_callback()

Check first whether priv->vifs[0] exists and is of type STATION, then go
to priv->vifs[1]. Make sure to call refresh_rate_mask for both
interfaces.

Signed-off-by: Martin Kaistra <martin.kaistra@linutronix.de>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231222101442.626837-14-martin.kaistra@linutronix.de

wifi: rtl8xxxu: support multiple interfaces in configure_filter()

As we only want to support AP mode/sending beacons on port 0, change
from priv->vif to priv->vifs[0] in the check for AP mode.
Additionally, if we are in AP mode, don't filter RX beacon and probe
response frames to still allow working STATION mode on the other
interface.

Signed-off-by: Martin Kaistra <martin.kaistra@linutronix.de>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231222101442.626837-13-martin.kaistra@linutronix.de

wifi: rtl8xxxu: support multiple interfaces in update_beacon_work_callback()

As we only want to support AP mode/sending beacons on port 0, it is
enough to replace priv->vif with priv->vifs[0].

Signed-off-by: Martin Kaistra <martin.kaistra@linutronix.de>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231222101442.626837-12-martin.kaistra@linutronix.de

wifi: rtl8xxxu: support multiple interfaces in set_aifs()

In concurrent mode supported by this driver, both interfaces will use
the same channel and same wireless mode.
It is therefore possible to get the wireless mode by checking the first
connected interface.

Signed-off-by: Martin Kaistra <martin.kaistra@linutronix.de>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231222101442.626837-11-martin.kaistra@linutronix.de

wifi: rtl8xxxu: support setting bssid register for multiple interfaces

To prepare for concurrent mode, enhance rtl8xxxu_set_bssid() to write the
BSSID of the respective interface to REG_BSSID or REG_BSSID1.

Like done with rtl8xxxu_set_mac(), call rtl8xxxu_set_bssid() with
port_num = 0, until the callers also support multiple interfaces.

Signed-off-by: Martin Kaistra <martin.kaistra@linutronix.de>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231222101442.626837-10-martin.kaistra@linutronix.de

wifi: rtl8xxxu: don't parse CFO, if both interfaces are connected in STA mode

If both interfaces are in STATION mode and both are connected to an AP,
there might be conflicting CFO values for the two connections. Ignore
the CFO information in this case.

Signed-off-by: Martin Kaistra <martin.kaistra@linutronix.de>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231222101442.626837-9-martin.kaistra@linutronix.de

wifi: rtl8xxxu: extend check for matching bssid to both interfaces

The driver will support two interfaces soon, which both can be in
station mode, so extend the check, whether cfo information should be
parsed, to cover both interfaces.

For better code readability put the lines with priv->vifs[port_num] in a
separate function.

Signed-off-by: Martin Kaistra <martin.kaistra@linutronix.de>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231222101442.626837-8-martin.kaistra@linutronix.de

wifi: rtl8xxxu: extend wifi connected check to both interfaces

There are multiple places in the code where the current connection
status of wifi is checked. The driver will support two interfaces soon
and either one of them (or both) could be connected.

Convert all uses of (vif && vif->cfg.assoc) to a new helper
function rtl8xxxu_is_assoc() which checks both interfaces.

Signed-off-by: Martin Kaistra <martin.kaistra@linutronix.de>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231222101442.626837-7-martin.kaistra@linutronix.de

wifi: rtl8xxxu: support setting mac address register for both interfaces

To prepare for concurrent mode, enhance rtl8xxxu_set_mac() to write the
mac address of the respective interface to REG_MACID or REG_MACID1.

Remove the call to rtl8xxxu_set_mac() from the init function as we set
it in rtl8xxxu_add_interface() later anyway.

Until rtl8xxxu_add_interface() can handle both interfaces, call
rtl8xxxu_set_mac() with port_num = 0.

Signed-off-by: Martin Kaistra <martin.kaistra@linutronix.de>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231222101442.626837-6-martin.kaistra@linutronix.de

wifi: rtl8xxxu: 8188e: convert usage of priv->vif to priv->vifs[0]

The driver currently does not support AP or concurrent mode for 8188e,
so just use priv->vifs[0] instead of priv->vif for now.

Signed-off-by: Martin Kaistra <martin.kaistra@linutronix.de>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231222101442.626837-5-martin.kaistra@linutronix.de

wifi: rtl8xxxu: support setting linktype for both interfaces

To prepare for concurrent mode, enhance the set_linktype function to be
able to set the linktype in the MSR register for both hardware ports.

Until the users of set_linktype can handle multiple interfaces, use
port_num = 0.

Signed-off-by: Martin Kaistra <martin.kaistra@linutronix.de>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231222101442.626837-4-martin.kaistra@linutronix.de

wifi: rtl8xxxu: prepare supporting two virtual interfaces

To prepare for concurrent mode, add an array ("vifs") to rtl8xxxu_priv
to keep track of both interfaces.

Keep the old priv->vif as long there are still users of it and let
priv->vifs[0] point to the same location.

Signed-off-by: Martin Kaistra <martin.kaistra@linutronix.de>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231222101442.626837-3-martin.kaistra@linutronix.de

wifi: rtl8xxxu: remove assignment of priv->vif in rtl8xxxu_bss_info_changed()

priv->vif gets already set in rtl8xxxu_add_interface, there is no need
to set it also in rtl8xxxu_bss_info_changed().

Signed-off-by: Martin Kaistra <martin.kaistra@linutronix.de>
Reviewed-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231222101442.626837-2-martin.kaistra@linutronix.de

wifi: rtw88: 8822ce: refine power parameters for RFE type 5

Refine the power parameters for better step response especially at high
current ramp case that is caused by power inductor variation.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240103070155.119488-1-pkshih@realtek.com

wifi: rtw89: mac: Fix spelling mistakes "notfify" -> "notify"

There are two spelling mistakes in rtw89_err error messages. Fix these
and also add space between [ERR] and message text.

Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Acked-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20231220141831.10063-1-colin.i.king@gmail.com

wifi: rtw89: phy: set channel_info for WiFi 7 chips

The channel_info is hardware settings to reflect operational status, such
as scale factor, report unit, buffer matrix size, RU size and so on. Then,
we can get desired reports to do further tuning.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240105064440.36926-1-pkshih@realtek.com

wifi: rtw89: phy: add BB wrapper of TX power for WiFi 7 chips

TX power is controlled by BB layer basically, but it should interact with
MAC layer, so these registers are put on MAC register domain and called
BB wrapper, which contains TX power for each MAC ID, OFDMA RU power, and
consideration of power type table.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240105064433.36870-1-pkshih@realtek.com

wifi: rtw89: 8922a: add NCTL pre-settings for WiFi 7 chips

NCTL standing for nano-controller is used to assist RF calibration.
Basically, we write settings from a table, but format of the table can't
describe register mask and additional conditions, so add a function to
set this kind of settings.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240105064422.36812-1-pkshih@realtek.com

wifi: rtw89: phy: ignore special data from BB parameter file

BB parameter file is a list of tuple {addr, val} with conditional hardware
version. However, tuples within a condition can't be empty, so insert a
special dummy tuple for this case. Then, ignore this tuple when writing.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240105064407.36750-1-pkshih@realtek.com

wifi: rtw89: 8922a: update the register used in DIG and the DIG flow

DIG standing for dynamic initial gain that is used to adjust RX coverage,
and PD lower threshold is packet detection power level by received signal
strength to avoid false detection of the WiFi packet.

Because of the hardware is different between WiFi 7 and 6 ICs, we adjust
flow and add register definition for 8922A.

Signed-off-by: Cheng-Chieh Hsieh <cj.hsieh@realtek.com>
Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240105064228.36580-5-pkshih@realtek.com

wifi: rtw89: 8922a: set RX gain along with set_channel operation

Set values of LNA, TIA and RPL gain compensation read from firmware file
when channel is changed.

Signed-off-by: Chung-Hsuan Hung <hsuan8331@realtek.com>
Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240105064228.36580-4-pkshih@realtek.com

wifi: rtw89: phy: add parser to support RX gain dynamic setting flow

Add RX gain offset dynamic setting flow according to different bands
and bandwidths. RX gain offset values will be different according to
different channel bands, therefore, this dynamic mechanism is needed
while channel is changed. Add this to parse data from the element of
firmware file, and then we can use them easier at runtime.

Signed-off-by: Chung-Hsuan Hung <hsuan8331@realtek.com>
Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240105064228.36580-3-pkshih@realtek.com

wifi: rtw89: phy: move bb_gain_info used by WiFi 6 chips to union

WiFi 7 chips use different bb_gain_info struct, so move existing struct to
a union in advance. This doesn't change logic at all.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240105064228.36580-2-pkshih@realtek.com

wifi: rtw89: 8851b: update TX power tables to R37

Update TX power tables to RF version R37. Mainly update configurations for
Canada 5.9 GHz (U-NII 4) according to IC (Industry Canada) certification.

Signed-off-by: Zong-Zhe Yang <kevin_yang@realtek.com>
Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240103014114.9558-2-pkshih@realtek.com

wifi: rtw89: 8852b: update TX power tables to R36

Update TX power tables to RF version R36. Mainly update configurations for
Canada 5.9 GHz (U-NII 4) according to IC (Industry Canada) certification.

Signed-off-by: Zong-Zhe Yang <kevin_yang@realtek.com>
Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240103014114.9558-1-pkshih@realtek.com

wifi: rtw89: pci: use DBI function for 8852AE/8852BE/8851BE

Sometimes driver can't use kernel API pci_read/write_config_byte
to access the PCI config space of above address 0x100 due to
the negotiated PCI setting. 8852AE/8852BE/8851BE provide another
way called DBI function, which belongs to WiFi mac and could
access all PCI config space for this case.

Link: https://lore.kernel.org/linux-wireless/79fe81b7db7148b9a7da2353c16d70fb@realtek.com/T/#t
Signed-off-by: Chin-Yen Lee <timlee@realtek.com>
Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://msgid.link/20240103012346.6822-1-pkshih@realtek.com

Revert "mlx5 updates 2023-12-20"

Revert "net/mlx5: Implement management PF Ethernet profile"
This reverts commit 22c4640698a1d47606b5a4264a584e8046641784.
Revert "net/mlx5: Enable SD feature"
This reverts commit c88c49ac9c18fb7c3fa431126de1d8f8f555e912.
Revert "net/mlx5e: Block TLS device offload on combined SD netdev"
This reverts commit 83a59ce0057b7753d7fbece194b89622c663b2a6.
Revert "net/mlx5e: Support per-mdev queue counter"
This reverts commit d72baceb92539a178d2610b0e9ceb75706a75b55.
Revert "net/mlx5e: Support cross-vhca RSS"
This reverts commit c73a3ab8fa6e93a783bd563938d7cf00d62d5d34.
Revert "net/mlx5e: Let channels be SD-aware"
This reverts commit e4f9686bdee7b4dd89e0ed63cd03606e4bda4ced.
Revert "net/mlx5e: Create EN core HW resources for all secondary devices"
This reverts commit c4fb94aa822d6c9d05fc3c5aee35c7e339061dc1.
Revert "net/mlx5e: Create single netdev per SD group"
This reverts commit e2578b4f983cfcd47837bbe3bcdbf5920e50b2ad.
Revert "net/mlx5: SD, Add informative prints in kernel log"
This reverts commit c82d360325112ccc512fc11a3b68cdcdf04a1478.
Revert "net/mlx5: SD, Implement steering for primary and secondaries"
This reverts commit 605fcce33b2d1beb0139b6e5913fa0b2062116b2.
Revert "net/mlx5: SD, Implement devcom communication and primary election"
This reverts commit a45af9a96740873db9a4b5bb493ce2ad81ccb4d5.
Revert "net/mlx5: SD, Implement basic query and instantiation"
This reverts commit 63b9ce944c0e26c44c42cdd5095c2e9851c1a8ff.
Revert "net/mlx5: SD, Introduce SD lib"
This reverts commit 4a04a31f49320d078b8078e1da4b0e2faca5dfa3.
Revert "net/mlx5: Fix query of sd_group field"
This reverts commit e04984a37398b3f4f5a79c993b94c6b1224184cc.
Revert "net/mlx5e: Use the correct lag ports number when creating TISes"
This reverts commit a7e7b40c4bc115dbf2a2bb453d7bbb2e0ea99703.

There are some unanswered questions on the list, and we don't
have any docs. Given the lack of replies so far and the fact
that v6.8 merge window has started - let's revert this and
revisit for v6.9.

Link: https://lore.kernel.org/all/20231221005721.186607-1-saeed@kernel.org/
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Revert "net: stmmac: Enable Per DMA Channel interrupt"

Revert "net: stmmac: Use interrupt mode INTM=1 for per channel irq"
This reverts commit 36af9f25ddfd311da82628f194c794786467cb12.
Revert "net: stmmac: Add support for TX/RX channel interrupt"
This reverts commit 9072e03d32088137a435ddf3aa95fd6e038d69d8.
Revert "net: stmmac: Make MSI interrupt routine generic"
This reverts commit 477bd4beb93bf9ace9bda71f1437b191befa9cf4.
Revert "dt-bindings: net: snps,dwmac: per channel irq"
This reverts commit 67d47c8ada0f8795bfcdb85cc8f2ad3ce556674b.

Device tree bindings need to be reviewed.

Link: https://lore.kernel.org/all/2df9fe3e-7971-4aa2-89a9-0e085b3b00d7@linaro.org/
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

ipvlan: Remove usage of the deprecated ida_simple_xx() API

ida_alloc() and ida_free() should be preferred to the deprecated
ida_simple_get() and ida_simple_remove().

This is less verbose.

Note that the upper bound of ida_alloc_range() is inclusive while the one
of ida_simple_get() was exclusive. So calls have been updated accordingly.

Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipvlan: Fix a typo in a comment

s/diffentiate/differentiate/

Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/sched: Remove ipt action tests

Commit ba24ea129126 ("net/sched: Retire ipt action") removed the ipt action
but not the testcases. This patch removes the outstanding tdc tests.

Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'stmmac-per-dma-channel-interrupt'

Swee Leong Ching says:

====================
net: stmmac: Enable Per DMA Channel interrupt

Add Per DMA Channel interrupt feature for DWXGMAC IP.

Patchset (link below) contains per DMA channel interrupt, But it was
achieved.
https://lore.kernel.org/lkml/20230821203328.GA2197059-
robh@kernel.org/t/#m849b529a642e1bff89c05a07efc25d6a94c8bfb4

Some of the changes in this patchset are based on reviewer comment on
patchset mentioned beforehand.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

net: stmmac: Use interrupt mode INTM=1 for per channel irq

Enable per DMA channel interrupt that uses shared peripheral
interrupt (SPI), so only per channel TX and RX intr (TI/RI)
are handled by TX/RX ISR without calling common interrupt ISR.

Signed-off-by: Teoh Ji Sheng <ji.sheng.teoh@intel.com>
Signed-off-by: Swee Leong Ching <leong.ching.swee@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: stmmac: Add support for TX/RX channel interrupt

Enable TX/RX channel interrupt registration for MAC that interrupts CPU
through shared peripheral interrupt (SPI).

Per channel interrupts and interrupt-names are registered through,
Eg: 4 tx and 4 rx channels:
interrupts = <GIC_SPI 100 IRQ_TYPE_LEVEL_HIGH>,
             <GIC_SPI 101 IRQ_TYPE_LEVEL_HIGH>,
             <GIC_SPI 102 IRQ_TYPE_LEVEL_HIGH>,
             <GIC_SPI 103 IRQ_TYPE_LEVEL_HIGH>;
             <GIC_SPI 104 IRQ_TYPE_LEVEL_HIGH>;
             <GIC_SPI 105 IRQ_TYPE_LEVEL_HIGH>;
             <GIC_SPI 106 IRQ_TYPE_LEVEL_HIGH>;
             <GIC_SPI 107 IRQ_TYPE_LEVEL_HIGH>;
interrupt-names = "dma_tx0",
                  "dma_tx1",
                  "dma_tx2",
                  "dma_tx3",
                  "dma_rx0",
                  "dma_rx1",
                  "dma_rx2",
                  "dma_rx3";

Signed-off-by: Teoh Ji Sheng <ji.sheng.teoh@intel.com>
Signed-off-by: Swee Leong Ching <leong.ching.swee@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: stmmac: Make MSI interrupt routine generic

There is no support for per DMA channel interrupt for non-MSI platform,
where the MAC's per channel interrupt hooks up to interrupt controller(GIC)
through shared peripheral interrupt(SPI) to handle interrupt from TX/RX
transmit channel.

This patch generalize the existing MSI ISR to also support non-MSI
platform.

Signed-off-by: Teoh Ji Sheng <ji.sheng.teoh@intel.com>
Signed-off-by: Swee Leong Ching <leong.ching.swee@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

dt-bindings: net: snps,dwmac: per channel irq

Add dt-bindings for per channel irq.

Signed-off-by: Rohan G Thomas <rohan.g.thomas@intel.com>
Signed-off-by: Swee Leong Ching <leong.ching.swee@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'at803x-more-generalization'

Christian Marangi says:

====================
net: phy: at803x: even more generalization

This is part 3 of at803x required patches to split the PHY driver
in more specific PHY Family driver.

While adding support for a new PHY Family qca807x it was notice lots
of similarities with the qca808x cdt function. Hence this series
is done to make things easier in the future when qca807x PHY will be
submitted.

Changes v4:
- Fix Smatch warning
Changes v3:
- Rebase on top of net-next
Changes v2:
- Address request from Russell in a previous series on cdt get
  status improvement
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

net: phy: at803x: make read_status more generic

Make read_status more generic in preparation on moving it to shared
library as other PHY Family Driver will have the exact same
implementation.

The only specific part was a check for AR8031/33 if 1000basex was used.
The check is moved to a dedicated function specific for those PHYs.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: phy: at803x: add support for cdt cross short test for qca808x

QCA808x PHY Family supports Cable Diagnostic Test also for Cross Pair
Short.

Add all the define to make enable and support these additional tests.

Cross Short test was previously disabled by default, this is now changed
and enabled by default. In this mode, the mask changed a bit and length
is shifted based on the fault condition.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: phy: at803x: refactor qca808x cable test get status function

Refactor qca808x cable test get status function to remove code
duplication and clean things up.

The same logic is applied to each pair hence it can be generalized and
moved to a common function.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: phy: at803x: generalize cdt fault length function

Generalize cable test fault length function since they all base on the
same magic values (already reverse engineered to understand the meaning
of it) to have consistenct values on every PHY.

Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

fib: rules: remove repeated assignment in fib_nl2rule

In fib_nl2rule(), 'err' variable has been set to -EINVAL during
declaration, and no need to set the 'err' variable to -EINVAL again.
So, remove it.

Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/sched: simplify tc_action_load_ops parameters

Instead of using two bools derived from a flags passed as arguments to
the parent function of tc_action_load_ops, just pass the flags itself
to tc_action_load_ops to simplify its parameters.

Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Pedro Tammela <pctammela@mojatatu.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

nfp: flower: Remove usage of the deprecated ida_simple_xx() API

ida_alloc() and ida_free() should be preferred to the deprecated
ida_simple_get() and ida_simple_remove().

This is less verbose.

Note that the upper bound of ida_alloc_range() is inclusive while the one
of ida_simple_get() was exclusive.
So NFP_FL_LAG_GROUP_MAX has been decreased by 1. It now better watch the
comment stating that "1 to 31 are valid".

The only other user of NFP_FL_LAG_GROUP_MAX has been updated accordingly in
nfp_fl_lag_put_unprocessed().

Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: ethtool: reject unsupported RSS input xfrm values

RXFH input_xfrm currently has three supported values: 0 (clear all),
symmetric_xor and NO_CHANGE.

Reject any other value sent from user-space.

Fixes: 13e59344fb9d ("net: ethtool: add support for symmetric-xor RSS hash")
Suggested-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Ahmed Zaki <ahmed.zaki@intel.com>
Link: https://lore.kernel.org/r/20240104212653.394424-1-ahmed.zaki@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

selftests: forwarding: Avoid failures to source net/lib.sh

The expression "source ../lib.sh" added to net/forwarding/lib.sh in commit
25ae948b4478 ("selftests/net: add lib.sh") does not work for tests outside
net/forwarding which source net/forwarding/lib.sh (1). It also does not
work in some cases where only a subset of tests are exported (2).

Avoid the problems mentioned above by replacing the faulty expression with
a copy of the content from net/lib.sh which is used by files under
net/forwarding.

A more thorough solution which avoids duplicating content between
net/lib.sh and net/forwarding/lib.sh has been posted here:
https://lore.kernel.org/netdev/20231222135836.992841-1-bpoirier@nvidia.com/

The approach in the current patch is a stopgap solution to avoid submitting
large changes at the eleventh hour of this development cycle.

Example of problem 1)

tools/testing/selftests/drivers/net/bonding$ ./dev_addr_lists.sh
./net_forwarding_lib.sh: line 41: ../lib.sh: No such file or directory
TEST: bonding cleanup mode active-backup                            [ OK ]
TEST: bonding cleanup mode 802.3ad                                  [ OK ]
TEST: bonding LACPDU multicast address to slave (from bond down)    [ OK ]
TEST: bonding LACPDU multicast address to slave (from bond up)      [ OK ]

An error message is printed but since the test does not use functions from
net/lib.sh, the test results are not affected.

Example of problem 2)

tools/testing/selftests$ make install TARGETS="net/forwarding"
tools/testing/selftests$ cd kselftest_install/net/forwarding/
tools/testing/selftests/kselftest_install/net/forwarding$ ./pedit_ip.sh veth{0..3}
lib.sh: line 41: ../lib.sh: No such file or directory
TEST: ping                                                          [ OK ]
TEST: ping6                                                         [ OK ]
./pedit_ip.sh: line 135: busywait: command not found
TEST: dev veth1 ingress pedit ip src set 198.51.100.1               [FAIL]
        Expected to get 10 packets, but got .
./pedit_ip.sh: line 135: busywait: command not found
TEST: dev veth2 egress pedit ip src set 198.51.100.1                [FAIL]
        Expected to get 10 packets, but got .
./pedit_ip.sh: line 135: busywait: command not found
TEST: dev veth1 ingress pedit ip dst set 198.51.100.1               [FAIL]
        Expected to get 10 packets, but got .
./pedit_ip.sh: line 135: busywait: command not found
TEST: dev veth2 egress pedit ip dst set 198.51.100.1                [FAIL]
        Expected to get 10 packets, but got .
./pedit_ip.sh: line 135: busywait: command not found
TEST: dev veth1 ingress pedit ip6 src set 2001:db8:2::1             [FAIL]
        Expected to get 10 packets, but got .
./pedit_ip.sh: line 135: busywait: command not found
TEST: dev veth2 egress pedit ip6 src set 2001:db8:2::1              [FAIL]
        Expected to get 10 packets, but got .
./pedit_ip.sh: line 135: busywait: command not found
TEST: dev veth1 ingress pedit ip6 dst set 2001:db8:2::1             [FAIL]
        Expected to get 10 packets, but got .
./pedit_ip.sh: line 135: busywait: command not found
TEST: dev veth2 egress pedit ip6 dst set 2001:db8:2::1              [FAIL]
        Expected to get 10 packets, but got .

In this case, the test results are affected.

Fixes: 25ae948b4478 ("selftests/net: add lib.sh")
Suggested-by: Ido Schimmel <idosch@nvidia.com>
Suggested-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Tested-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Benjamin Poirier <bpoirier@nvidia.com>
Reviewed-by: Hangbin Liu <liuhangbin@gmail.com>
Link: https://lore.kernel.org/r/20240104141109.100672-1-bpoirier@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge tag 'for-netdev' of https://git./linux/kernel/git/bpf/bpf-next

Daniel Borkmann says:

====================
pull-request: bpf-next 2024-01-05

We've added 40 non-merge commits during the last 2 day(s) which contain
a total of 73 files changed, 1526 insertions(+), 951 deletions(-).

The main changes are:

1) Fix a memory leak when streaming AF_UNIX sockets were inserted
   into multiple sockmap slots/maps, from John Fastabend.

2) Fix gotol in s390 BPF JIT with large offsets, from Ilya Leoshkevich.

3) Fix reattachment branch in bpf_tracing_prog_attach() and reject
   the request if there is no valid attach_btf, from Jiri Olsa.

4) Remove deprecated bpfilter kernel leftovers given the project
   is developed in user space (https://github.com/facebook/bpfilter),
   from Quentin Deslandes.

5) Relax tracing BPF program recursive attach rules given right now
   it is not possible to create tracing program call cycles,
   from Dmitrii Dolgov.

6) Fix excessive memory consumption for the bpf_global_percpu_ma
   for systems with a large number of CPUs, from Yonghong Song.

7) Small x86 BPF JIT cleanup to reuse emit_nops instead of open-coding
   memcpy of x86_nops, from Leon Hwang.

8) Follow-up for libbpf to support __arg_ctx global function argument tag
   semantics to complement the merged kernel side, from Andrii Nakryiko.

9) Introduce "volatile compare" macros for BPF selftests in order
   to make the latter more robust against compiler optimization,
   from Alexei Starovoitov.

10) Small simplification in verifier's size checking of helper accesses
    along with additional selftests, from Andrei Matei.

* tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next: (40 commits)
  selftests/bpf: Test re-attachment fix for bpf_tracing_prog_attach
  bpf: Fix re-attachment branch in bpf_tracing_prog_attach
  selftests/bpf: Add test for recursive attachment of tracing progs
  bpf: Relax tracing prog recursive attach rules
  bpf, x86: Use emit_nops to replace memcpy x86_nops
  selftests/bpf: Test gotol with large offsets
  selftests/bpf: Double the size of test_loader log
  s390/bpf: Fix gotol with large offsets
  bpfilter: remove bpfilter
  bpf: Remove unnecessary cpu == 0 check in memalloc
  selftests/bpf: add __arg_ctx BTF rewrite test
  selftests/bpf: add arg:ctx cases to test_global_funcs tests
  libbpf: implement __arg_ctx fallback logic
  libbpf: move BTF loading step after relocation step
  libbpf: move exception callbacks assignment logic into relocation step
  libbpf: use stable map placeholder FDs
  libbpf: don't rely on map->fd as an indicator of map being created
  libbpf: use explicit map reuse flag to skip map creation steps
  libbpf: make uniform use of btf__fd() accessor inside libbpf
  selftests/bpf: Add a selftest with > 512-byte percpu allocation size
  ...
====================

Link: https://lore.kernel.org/r/20240105170105.21070-1-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

ptp_ocp: adjust MAINTAINERS and mailmap

The fb.com domain is going to be deprecated.
Use personal one for kernel contributions.

Signed-off-by: Vadim Fedorenko <vadfed@fb.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20240104172540.2379128-1-vadfed@meta.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

geneve: use DEV_STATS_INC()

geneve updates dev->stats fields locklessly.

Adopt DEV_STATS_INC() to avoid races.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20240104163633.2070538-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'net-gro-reduce-extension-header-parsing-overhead'

Richard Gobert says:

====================
net: gro: reduce extension header parsing overhead

This series attempts to reduce the parsing overhead of IPv6 extension
headers in GRO and GSO, by removing extension header specific code and
enabling the frag0 fast path.

The following changes were made:

 - Removed some unnecessary HBH conditionals by adding HBH offload
   to inet6_offloads
 - Added a utility function to support frag0 fast path in ipv6_gro_receive
 - Added selftests for IPv6 packets with extension headers in GRO

v2: https://lore.kernel.org/netdev/127b8199-1cd4-42d7-9b2b-875abaad93fe@gmail.com/
v1: https://lore.kernel.org/netdev/f4eff69d-3917-4c42-8c6b-d09597ac4437@gmail.com/
====================

Link: https://lore.kernel.org/r/ac6fb684-c00e-449c-92c3-99358a927ade@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

selftests/net: fix GRO coalesce test and add ext header coalesce tests

Currently there is no test which checks that IPv6 extension header packets
successfully coalesce. This commit adds a test, which verifies two IPv6
packets with HBH extension headers do coalesce, and another test which
checks that packets with different extension header data do not coalesce
in GRO.

I changed the receive socket filter to accept a packet with one extension
header. This change exposed a bug in the fragment test -- the old BPF did
not accept the fragment packet. I updated correct_num_packets in the
fragment test accordingly.

Signed-off-by: Richard Gobert <richardbgobert@gmail.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Link: https://lore.kernel.org/r/69282fed-2415-47e8-b3d3-34939ec3eb56@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: gro: parse ipv6 ext headers without frag0 invalidation

The existing code always pulls the IPv6 header and sets the transport
offset initially. Then optionally again pulls any extension headers in
ipv6_gso_pull_exthdrs and sets the transport offset again on return from
that call. skb->data is set at the start of the first extension header
before calling ipv6_gso_pull_exthdrs, and must disable the frag0
optimization because that function uses pskb_may_pull/pskb_pull instead of
skb_gro_ helpers. It sets the GRO offset to the TCP header with
skb_gro_pull and sets the transport header. Then returns skb->data to its
position before this block.

This commit introduces a new helper function - ipv6_gro_pull_exthdrs -
which is used in ipv6_gro_receive to pull ipv6 ext headers instead of
ipv6_gso_pull_exthdrs. Thus, there is no modification of skb->data, all
operations use skb_gro_* helpers, and the frag0 fast path can be taken for
IPv6 packets with ext headers.

Signed-off-by: Richard Gobert <richardbgobert@gmail.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/504130f6-b56c-4dcc-882c-97942c59f5b7@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: gso: add HBH extension header offload support

This commit adds net_offload to IPv6 Hop-by-Hop extension headers (as it
is done for routing and dstopts) since it is supported in GSO and GRO.
This allows to remove specific HBH conditionals in GSO and GRO when
pulling and parsing an incoming packet.

Signed-off-by: Richard Gobert <richardbgobert@gmail.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/d4f8825a-1d55-4b12-9d67-a254dbbfa6ae@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: fill in MODULE_DESCRIPTION()s for CAIF

W=1 builds now warn if module is built without a MODULE_DESCRIPTION().
Add descriptions to all the CAIF sub-modules.

Link: https://lore.kernel.org/r/20240104144855.1320993-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: fill in MODULE_DESCRIPTION() for AF_PACKET

W=1 builds now warn if module is built without a MODULE_DESCRIPTION().
Add description to net/packet/af_packet.c

Acked-by: Willem de Bruijn <willemb@google.com>
Link: https://lore.kernel.org/r/20240104144119.1319055-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: fill in MODULE_DESCRIPTION()s for DSA tags

W=1 builds now warn if module is built without a MODULE_DESCRIPTION().
Add descriptions to all the DSA tag modules.

The descriptions are copy/pasted Kconfig names, with s/^Tag/DSA tag/.

Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Acked-by: Arun Ramadoss <arun.ramadoss@microchip.com>
Acked-by: Arınç ÜNAL <arinc.unal@arinc9.com>
Acked-by: Kurt Kanzenbach <kurt@linutronix.de>
Link: https://lore.kernel.org/r/20240104143759.1318137-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: fill in MODULE_DESCRIPTION()s for ATM

W=1 builds now warn if module is built without a MODULE_DESCRIPTION().
Add descriptions to all the ATM modules and drivers.

Link: https://lore.kernel.org/r/20240104143737.1317945-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'dpll-expose-fractional-frequency-offset-value-to-user'

Jiri Pirko says:

====================
dpll: expose fractional frequency offset value to user

Allow to expose pin fractional frequency offset value over new DPLL
generic netlink attribute. Add an op to get the value from the driver.
Implement this new op in mlx5 driver.
====================

Link: https://lore.kernel.org/r/20240103132838.1501801-1-jiri@resnulli.us
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net/mlx5: DPLL, Implement fractional frequency offset get pin op

Implement ffo_get() pin op filling it up to MSEED.frequency_diff value.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Arkadiusz Kubalewski <arkadiusz.kubalewski@intel.com>
Acked-by: Vadim Fedorenko <vadim.fedorenko@linux.dev>
Acked-by: Arkadiusz Kubalewski <arkadiusz.kubalewski@intel.com>
Link: https://lore.kernel.org/r/20240103132838.1501801-4-jiri@resnulli.us
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net/mlx5: DPLL, Use struct to get values from mlx5_dpll_synce_status_get()

Instead of passing separate args, introduce
struct mlx5_dpll_synce_status to hold the values obtained by
mlx5_dpll_synce_status_get().

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Arkadiusz Kubalewski <arkadiusz.kubalewski@intel.com>
Acked-by: Vadim Fedorenko <vadim.fedorenko@linux.dev>
Acked-by: Arkadiusz Kubalewski <arkadiusz.kubalewski@intel.com>
Link: https://lore.kernel.org/r/20240103132838.1501801-3-jiri@resnulli.us
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

dpll: expose fractional frequency offset value to user

Add a new netlink attribute to expose fractional frequency offset value
for a pin. Add an op to get the value from the driver.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Acked-by: Vadim Fedorenko <vadim.fedorenko@linux.dev>
Acked-by: Arkadiusz Kubalewski <arkadiusz.kubalewski@intel.com>
Link: https://lore.kernel.org/r/20240103132838.1501801-2-jiri@resnulli.us
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'user_mii_bus-cleanup-part-one'

Vladimir Oltean says:

====================
ds->user_mii_bus cleanup (part 1)

There are some drivers which assign ds->user_mii_bus when they
don't really need its specific functionality, aka non-OF based
dsa_user_phy_connect(). There was some confusion regarding the
fact that yes, this is why ds->user_mii_bus really exists, so
I've started a cleanup series which aims to eliminate the usage
of ds->user_mii_bus from drivers when there is nothing to gain
from it.

Today's drivers are lantiq_gswip, qca8k and bcm_sf2. The work is
not done here, but a "part 2" may or may not come, depending on
other priorities.

All patches were only compile-tested.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: bcm_sf2: drop priv->master_mii_dn

There used to be a of_node_put(priv->master_mii_dn) call in
bcm_sf2_mdio_unregister(), which was accidentally deleted in commit
6ca80638b90c ("net: dsa: Use conduit and user terms").

But it's not needed - we don't need to hold a reference on the
"brcm,unimac-mdio" OF node for that long, since we don't do anything
with it. We can release it as soon as we finish bcm_sf2_mdio_register().

Also reduce "if (err && dn)" to just "if (err)". We know "dn", aka the
former priv->master_mii_dn, is non-NULL. Otherwise, of_mdio_find_bus(dn)
would not have been able to find the bus behind "brcm,unimac-mdio".

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Alvin Šipraga <alsi@bang-olufsen.dk>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Tested-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Luiz Angelo Daros de Luca <luizluca@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: bcm_sf2: stop assigning an OF node to the ds->user_mii_bus

The bcm_sf2 driver does something strange. Instead of calling
of_mdiobus_register() with an OF node argument, it manually assigns the
bus->dev->of_node and then calls the non-OF mdiobus_register(). This
circumvents some code from __of_mdiobus_register() from running, which
sets the auto-scan mask, parses some device tree properties, etc.

I'm going to go out on a limb and say that the OF node isn't, in fact,
needed at all, and can be removed. The MDIO diversion as initially
implemented in commit 461cd1b03e32 ("net: dsa: bcm_sf2: Register our
slave MDIO bus") looked quite different than it is now, after commit
771089c2a485 ("net: dsa: bcm_sf2: Ensure that MDIO diversion is used").
Initially, it made sense, as bcm_sf2 was registering another set of
driver ops for the "brcm,unimac-mdio" OF node. But now, it deletes all
phandles, which makes "phy-handle"s unable to find PHYs, which means
that it always goes through the OF-unaware dsa_user_phy_connect().

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Alvin Šipraga <alsi@bang-olufsen.dk>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Tested-by: Florian Fainelli <florian.fainelli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: qca8k: use "dev" consistently within qca8k_mdio_register()

Accessed either through priv->dev or ds->dev, it is the same device
structure. Keep a single variable which holds a reference to it, and use
it consistently.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Alvin Šipraga <alsi@bang-olufsen.dk>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Luiz Angelo Daros de Luca <luizluca@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: qca8k: consolidate calls to a single devm_of_mdiobus_register()

__of_mdiobus_register() already calls __mdiobus_register() if the
OF node provided as argument is NULL. We can take advantage of that
and simplify the 2 code path, calling devm_of_mdiobus_register() only
once for both cases.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Alvin Šipraga <alsi@bang-olufsen.dk>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Luiz Angelo Daros de Luca <luizluca@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: qca8k: assign ds->user_mii_bus only for the non-OF case

To simplify reasoning about why the DSA framework provides the
ds->user_mii_bus functionality, drivers should only use it if they
need to. The qca8k driver appears to also use it simply as storage
for a pointer, which is not a good enough reason to make the core
much more difficult to follow.

ds->user_mii_bus is useful for only 2 cases:

1. The driver probes on platform_data (no OF)
2. The driver probes on OF, but there is no OF node for the MDIO bus.

It is unclear if case (1) is supported with qca8k. It might not be:
the driver might crash when of_device_get_match_data() returns NULL
and then it dereferences priv->info without NULL checking.

Anyway, let us limit the ds->user_mii_bus usage only to the above cases,
and not assign it when an OF node is present.

The bus->phy_mask assignment follows along with the movement, because
__of_mdiobus_register() overwrites this bus field anyway. The value set
by the driver only matters for the non-OF code path.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Alvin Šipraga <alsi@bang-olufsen.dk>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Luiz Angelo Daros de Luca <luizluca@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: qca8k: skip MDIO bus creation if its OF node has status = "disabled"

Currently the driver calls the non-OF devm_mdiobus_register() rather
than devm_of_mdiobus_register() for this case, but it seems to rather
be a confusing coincidence, and not a real use case that needs to be
supported.

If the device tree says status = "disabled" for the MDIO bus, we
shouldn't need an MDIO bus at all. Instead, just exit as early as
possible and do not call any MDIO API.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Alvin Šipraga <alsi@bang-olufsen.dk>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: qca8k: put MDIO bus OF node on qca8k_mdio_register() failure

of_get_child_by_name() gives us an OF node with an elevated refcount,
which should be dropped when we're done with it. This is so that,
if (of_node_check_flag(node, OF_DYNAMIC)) is true, the node's memory can
eventually be freed.

There are 2 distinct paths to be considered in qca8k_mdio_register():

- devm_of_mdiobus_register() succeeds: since commit 3b73a7b8ec38 ("net:
  mdio_bus: add refcounting for fwnodes to mdiobus"), the MDIO core
  treats this well.

- devm_of_mdiobus_register() or anything up to that point fails: it is
  the duty of the qca8k driver to release the OF node.

This change addresses the second case by making sure that the OF node
reference is not leaked.

The "mdio" node may be NULL, but of_node_put(NULL) is safe.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Alvin Šipraga <alsi@bang-olufsen.dk>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: lantiq_gswip: ignore MDIO buses disabled in OF

If the "lantiq,xrx200-mdio" child has status = "disabled", the MDIO bus
creation should be avoided. Use of_device_is_available() to check for
that, and take advantage of 2 facts:

- of_device_is_available(NULL) returns false
- of_node_put(NULL) is a no-op

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Alvin Šipraga <alsi@bang-olufsen.dk>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>