openwrt/staging/blogic.git
6 years agortlwifi: btcoex: Add customer_id to do special deal to oem vendor
Ping-Ke Shih [Tue, 6 Mar 2018 01:25:43 +0000 (09:25 +0800)]
rtlwifi: btcoex: Add customer_id to do special deal to oem vendor

Add customer_id field in structure, and then this is a clue to implement
the behavior defined by vendor LENOVO_CHINA.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Acked-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
6 years agortlwifi: btcoex: Add enum DM_INFO for btcoex to query dm's counters
Ping-Ke Shih [Tue, 6 Mar 2018 01:25:42 +0000 (09:25 +0800)]
rtlwifi: btcoex: Add enum DM_INFO for btcoex to query dm's counters

btcoex uses dm's counters to check the environment is noisy or not. If it
is clean, btcoex set more time slots to WiFi so that it can transmit as
soon as possible. That will save time, and BT will have more time slots
after WiFi traffic is complete.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Acked-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
6 years agortlwifi: remove redundant statement found by static checker
Ping-Ke Shih [Tue, 6 Mar 2018 01:25:41 +0000 (09:25 +0800)]
rtlwifi: remove redundant statement found by static checker

smatch reports "rtl_is_special_data() warn: inconsistent indenting", but
it is an obvious redundant statement so remove it.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Acked-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
6 years agortlwifi: Add modifier static to functions reported by sparse
Ping-Ke Shih [Tue, 6 Mar 2018 01:25:40 +0000 (09:25 +0800)]
rtlwifi: Add modifier static to functions reported by sparse

sparse reports some functions were not declared, so add 'static' as
modifier. Remove an unused function btc8821a1ant_is_wifi_status_changed()
exposed due to 'static'.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Acked-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
6 years agorsi: fix kernel panic observed on 64bit machine
Amitkumar Karwar [Tue, 20 Mar 2018 13:40:41 +0000 (19:10 +0530)]
rsi: fix kernel panic observed on 64bit machine

Following kernel panic is observed on 64bit machine while loading
the driver. It is fixed if we pass dynamically allocated memory to
SDIO for DMA.

BUG: unable to handle kernel paging request at ffffeb04000172e0
IP: sg_miter_stop+0x56/0x70
PGD 0 P4D 0
Oops: 0000 [#1] SMP PTI
Modules linked in: rsi_sdio(OE+) rsi_91x(OE) btrsi(OE) rfcomm bluetooth
ecdh_generic mac80211 mmc_block fuse xt_CHECKSUM iptable_mangle
drm_kms_helper mmc_core serio_raw drm firewire_ohci tg3
CPU: 0 PID: 4003 Comm: insmod Tainted: G           OE    4.16.0-rc1+ #27
Hardware name: Dell Inc. Latitude E5500                  /0DW634, BIOS
A19 06/13/2013
RIP: 0010:sg_miter_stop+0x56/0x70
RSP: 0018:ffff88007d003e78 EFLAGS: 00010002
RAX: 0000000000000003 RBX: 0000000000000004 RCX: 0000000000000000
RDX: ffffeb04000172c0 RSI: ffff88002f58002c RDI: ffff88007d003e80
RBP: 0000000000000004 R08: ffff88007d003e80 R09: 0000000000000008
R10: 0000000000000003 R11: 0000000000000001 R12: 0000000000000004
R13: ffff88002f580028 R14: 0000000000000000 R15: 0000000000000004
FS:  00007f35c29db700(0000) GS:ffff88007d000000(0000)
knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: ffffeb04000172e0 CR3: 000000007038e000 CR4: 00000000000406f0
Call Trace:
<IRQ>
sg_copy_buffer+0xc6/0xf0
sdhci_tasklet_finish+0x170/0x260 [sdhci]
tasklet_action+0xf4/0x100
__do_softirq+0xef/0x26e
irq_exit+0xbe/0xd0
do_IRQ+0x4a/0xc0
common_interrupt+0xa2/0xa2
</IRQ>

Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
6 years agorsi: fix error path handling in SDIO probe
Amitkumar Karwar [Tue, 20 Mar 2018 13:40:40 +0000 (19:10 +0530)]
rsi: fix error path handling in SDIO probe

We miss to release IRQ in certain error path in SDIO probe which
causes following kernel panic. This patch corrects error path
handling

BUG: unable to handle kernel NULL pointer dereference at        (null)
IP:           (null)
PGD 0 P4D 0
Oops: 0010 [#1] SMP PTI
Call Trace:
 <IRQ>
 ? call_timer_fn+0x29/0x120
 ? run_timer_softirq+0x1da/0x420
 ? timer_interrupt+0x11/0x20
 ? __do_softirq+0xef/0x26e
 ? irq_exit+0xbe/0xd0
 ? do_IRQ+0x4a/0xc0
 ? common_interrupt+0xa2/0xa2
 </IRQ>
 ? cpuidle_enter_state+0x118/0x250
 ? do_idle+0x186/0x1e0
 ? cpu_startup_entry+0x6f/0x80
 ? start_kernel+0x47c/0x49c
 ? secondary_startup_64+0xa5/0xb0

Fixes: 50117605770c ("rsi: improve RX handling in SDIO interface")
Signed-off-by: Amitkumar Karwar <amit.karwar@redpinesignals.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
6 years agorsi: remove redundant duplicate assignment of buffer_size
Colin Ian King [Wed, 14 Mar 2018 14:12:39 +0000 (14:12 +0000)]
rsi: remove redundant duplicate assignment of buffer_size

Variable buffer_size is re-assigned the same value, this duplicated
assignment is redundant and can be removed.

Cleans up clang warning:
drivers/net/wireless/rsi/rsi_91x_usb.c:140:4: warning: Value stored
to 'buffer_size' is never read

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
6 years agorsi: Remove stack VLA usage
Tobin C. Harding [Thu, 15 Mar 2018 02:31:25 +0000 (13:31 +1100)]
rsi: Remove stack VLA usage

The use of stack Variable Length Arrays needs to be avoided, as they
can be a vector for stack exhaustion, which can be both a runtime bug
(kernel Oops) or a security flaw (overwriting memory beyond the
stack). Also, in general, as code evolves it is easy to lose track of
how big a VLA can get. Thus, we can end up having runtime failures
that are hard to debug. As part of the directive[1] to remove all VLAs
from the kernel, and build with -Wvla.

Currently rsi code uses a VLA based on a function argument to
`rsi_sdio_load_data_master_write()`.  The function call chain is

Both these functions

rsi_sdio_reinit_device()
rsi_probe()

start the call chain:

rsi_hal_device_init()
rsi_load_fw()
auto_fw_upgrade()
ping_pong_write()
rsi_sdio_load_data_master_write()

[Without familiarity with the code] it appears that none of the 4 locks

mutex
rx_mutex
tx_mutex
tx_bus_mutex

are held when `rsi_sdio_load_data_master_write()` is called.  It is therefore
safe to use kmalloc with GFP_KERNEL.

We can avoid using the VLA by using `kmalloc()` and free'ing the memory on all
exit paths.

Change buffer from 'u8 array' to 'u8 *'.  Call `kmalloc()` to allocate memory for
the buffer.  Using goto statement to call `kfree()` on all return paths.

It can be expected that this patch will result in a small increase in overhead
due to the use of `kmalloc()` however this code is only called on initialization
(and re-initialization) so this overhead should not degrade performance.

[1] https://lkml.org/lkml/2018/3/7/621

Signed-off-by: Tobin C. Harding <me@tobin.cc>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
6 years agomwifiex: remove warnings in mwifiex_cmd_append_11n_tlv()
Ganapathi Bhat [Thu, 15 Mar 2018 06:20:32 +0000 (11:50 +0530)]
mwifiex: remove warnings in mwifiex_cmd_append_11n_tlv()

Fix the following sparse warning in mwifiex_cmd_append_11n_tlv:

drivers/net/wireless/marvell/mwifiex/11n.c:358:65: warning: invalid assignment: &=
drivers/net/wireless/marvell/mwifiex/11n.c:358:65:    left side has type restricted __le16
drivers/net/wireless/marvell/mwifiex/11n.c:358:65:    right side has type int
drivers/net/wireless/marvell/mwifiex/11n.c:360:65: warning: invalid assignment: &=
drivers/net/wireless/marvell/mwifiex/11n.c:360:65:    left side has type restricted __le16
drivers/net/wireless/marvell/mwifiex/11n.c:360:65:    right side has type int
drivers/net/wireless/marvell/mwifiex/11n.c:366:65: warning: invalid assignment: &=
drivers/net/wireless/marvell/mwifiex/11n.c:366:65:    left side has type restricted __le16
drivers/net/wireless/marvell/mwifiex/11n.c:366:65:    right side has type int
drivers/net/wireless/marvell/mwifiex/11n.c:368:65: warning: invalid assignment: &=
drivers/net/wireless/marvell/mwifiex/11n.c:368:65:    left side has type restricted __le16
drivers/net/wireless/marvell/mwifiex/11n.c:368:65:    right side has type int

Fixes: 77423fa73927 ("mwifiex: fix incorrect ht capability problem")
Signed-off-by: Ganapathi Bhat <gbhat@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
6 years agowireless: Use octal not symbolic permissions
Joe Perches [Fri, 23 Mar 2018 22:54:37 +0000 (15:54 -0700)]
wireless: Use octal not symbolic permissions

Prefer the direct use of octal for permissions.

Done with checkpatch -f --types=SYMBOLIC_PERMS --fix-inplace
and some typing.

Miscellanea:

o Whitespace neatening around these conversions.

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
6 years agoBluetooth: btrsi: rework dependencies
Arnd Bergmann [Thu, 15 Mar 2018 21:18:24 +0000 (22:18 +0100)]
Bluetooth: btrsi: rework dependencies

The linkage between the bluetooth driver and the wireless
driver is not defined properly, leading to build problems
such as:

warning: (BT_HCIRSI) selects RSI_COEX which has unmet direct dependencies (NETDEVICES && WLAN && WLAN_VENDOR_RSI && BT_HCIRSI && RSI_91X)
drivers/net/wireless/rsi/rsi_91x_main.o: In function `rsi_read_pkt':
(.text+0x205): undefined reference to `rsi_bt_ops'

As the dependency is actually the reverse (RSI_91X uses
the BT_RSI driver, not the other way round), this changes
the dependency to match, and enables the bluetooth driver
from the RSI_COEX symbol.

Fixes: 38aa4da50483 ("Bluetooth: btrsi: add new rsi bluetooth driver")
Acked-by; Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
6 years agoMerge ath-next from git://git.kernel.org/pub/scm/linux/kernel/git/kvalo/ath.git
Kalle Valo [Tue, 27 Mar 2018 07:06:18 +0000 (10:06 +0300)]
Merge ath-next from git://git./linux/kernel/git/kvalo/ath.git

ath.git patches for 4.17. Major changes:

wil6210

* support multiple virtual interfaces

6 years agowcn36xx: Check DXE IRQ reason
Ramon Fried [Sun, 11 Mar 2018 12:01:43 +0000 (14:01 +0200)]
wcn36xx: Check DXE IRQ reason

IRQ reason was not cheked for errors.
Although error handing is not currently supported, it
will be nice to output an error value to the log if the
DMA operation failed.

Signed-off-by: Ramon Fried <rfried@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
6 years agowcn36xx: calculate DXE default channel values
Ramon Fried [Sun, 11 Mar 2018 12:01:18 +0000 (14:01 +0200)]
wcn36xx: calculate DXE default channel values

DXE channel defaults used hardcoded magic values.
Added bit definitions of the control register and
calculate this values in compilation for clarity.

Signed-off-by: Ramon Fried <rfried@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
6 years agowcn36xx: calculate DXE control registers values
Ramon Fried [Sun, 4 Mar 2018 16:31:34 +0000 (18:31 +0200)]
wcn36xx: calculate DXE control registers values

DXE descriptor control registers used hardcoded magic values.  Added bit
definitions of the control register and calculate this values in compilation
for clarity. No functional changes.

Signed-off-by: Ramon Fried <rfried@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
6 years agoath9k: spelling s/premble/preamble/
Peter Große [Tue, 6 Mar 2018 14:57:18 +0000 (15:57 +0100)]
ath9k: spelling s/premble/preamble/

Signed-off-by: Peter Große <pegro@friiks.de>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
6 years agoath9k: fix crash in spectral scan
Sebastian Gottschall [Sat, 3 Mar 2018 04:10:44 +0000 (05:10 +0100)]
ath9k: fix crash in spectral scan

Fixes crash seen on arm smp systems (gateworks ventana imx6):

Unable to handle kernel NULL pointer dereference at virtual address 00000014
pgd = 80004000
[00000014] *pgd=00000000
Internal error: Oops - BUG: 17 [#1] PREEMPT SMP ARM
Modules linked in: ip6table_filter nf_conntrack_ipv6 ip6_tables nf_log_ipv6 nf_defrag_ipv6 shortcut_fe ipcomp6 xfrm_ipcomp xfrm6_tunnel xfrm6_mode_tunnel xfrm6_mode_transport xfrm6_mode_ro xfrm6_mode_beet ip6_tunnel tunnel6 mip6 ah6 esp6 xfrm_algo sit ip_tunnel tunnel4 ipv6 ath10k_pci ath10k_core ath9k ath mac80211 cfg80211 compat ath_pci ath_hal(P) caamalg authencesn authenc caamrng caamhash caam_jr caam cdc_ncm usbnet usbcore sky2 imx2_wdt
CPU: 0 PID: 3 Comm: ksoftirqd/0 Tainted: P                4.9.85 #19
Hardware name: Freescale i.MX6 Quad/DualLite (Device Tree)
task: bf064980 task.stack: bf07c000
PC is at relay_buf_full+0xc/0x30
LR is at _674+0x740/0xf10 [ath9k]
pc : [<8018bce0>]    lr : [<7f1aa604>]    psr: 80000013
sp : bf07dbf0  ip : bf07dc00  fp : bf07dbfc
r10: 0000003f  r9 : bf130e00  r8 : 809044b0
r7 : 00000000  r6 : be67a9f0  r5 : 00000000  r4 : 809043e4
r3 : c0864c24  r2 : 00000000  r1 : 00000004  r0 : 00000000
Flags: Nzcv  IRQs on  FIQs on  Mode SVC_32  ISA ARM  Segment user
Control: 10c5387d  Table: 4e6a004a  DAC: 00000055
Process ksoftirqd/0 (pid: 3, stack limit = 0xbf07c210)
Stack: (0xbf07dbf0 to 0xbf07e000)
dbe0:                                     bf07dd04 bf07dc00 7f1aa604 8018bce0
dc00: 00004014 be59e010 bf07dc34 bf07dc18 7f1a7084 7f19c07c be59c010 be6470a0
dc20: 0000096c be648954 bf07dc6c bf07dc38 7f1c286c bf07dd90 bf07dc5c bf07dc48
dc40: 8029ea4c 0000003c 00000001 be59c010 00000094 00000000 00000000 00000000
dc60: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
dc80: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
dca0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
dcc0: 00000000 00000000 00000000 00000000 00000000 00000000 8010ef24 00000030
dce0: be94f5e8 be6485a0 bddf0200 be59c010 be6465a0 be6415a0 bf07ddf4 bf07dd08
dd00: 7f1cf800 7f1aa55c 1fc38c4c 00000000 bf07dd58 cccccccd 66666667 be640bc0
dd20: bf07dd54 be6415a0 1fc38c4c 00000000 00000000 be59c038 be67a9c0 be59e010
dd40: be67a9f0 be647170 8090c904 be59c010 00000000 00000001 1fc38e84 00000000
dd60: be640bc0 bddf0200 00000200 00000010 0000003f 00000002 20000013 be59c010
dd80: 8092d940 bf7ca2c0 bf07ddb4 bf07dd98 1fc38c4c 2602003f 0100ff1b 80ff1b00
dda0: 00808080 00000000 00000000 80808080 80808080 80808080 80808080 00008080
ddc0: 00000000 00000000 7f1b62b8 00000002 be6470ec be6470f0 00000000 bf07de98
dde0: 8092d940 be6415a0 bf07de94 bf07ddf8 7f1d1ed8 7f1cf1fc 00000000 00000000
de00: bf7cc4c0 00000400 be6470f0 bf07de18 8015165c be59c010 8090453c 8090453c
de20: bf07dec4 be6465a0 8014f614 80148884 0000619a 00000001 bf07c000 00000100
de40: bf07de78 00000001 7f327850 00000002 afb50401 bf064980 bf07de9c bf07de68
de60: bf064a00 803cc668 bf064a00 be6470b4 be6470b8 80844180 00000000 bf07de98
de80: 8092d940 bf07c000 bf07dec4 bf07de98 80124d18 7f1d1c44 80124c94 00000000
dea0: 00000006 80902098 80902080 40000006 00000100 bf07c000 bf07df24 bf07dec8
dec0: 8012501c 80124ca0 bf7cc4c0 bf064980 be95e1c0 04208040 80902d00 000061c7
dee0: 0000000a 80600b54 8092d940 808441f8 80902080 bf07dec8 bf03b200 bf07c000
df00: bf03b200 8090fe54 00000000 00000000 00000000 00000000 bf07df34 bf07df28
df20: 80125148 80124f28 bf07df5c bf07df38 8013deb4 8012511c 00000000 bf03b240
df40: bf03b200 8013dc90 00000000 00000000 bf07dfac bf07df60 8013ad40 8013dc9c
df60: 70448040 00000001 00000000 bf03b200 00000000 00030003 bf07df78 bf07df78
df80: 00000000 00000000 bf07df88 bf07df88 bf03b240 8013ac48 00000000 00000000
dfa0: 00000000 bf07dfb0 80107760 8013ac54 00000000 00000000 00000000 00000000
dfc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
dfe0: 00000000 00000000 00000000 00000000 00000013 00000000 8c120004 1190ad04
Backtrace:
[<8018bcd4>] (relay_buf_full) from [<7f1aa604>] (_674+0x740/0xf10 [ath9k])
[<7f1aa550>] (_674 [ath9k]) from [<7f1cf800>] (_582+0x14b4/0x3708 [ath9k])
 r10:be6415a0 r9:be6465a0 r8:be59c010 r7:bddf0200 r6:be6485a0 r5:be94f5e8
 r4:00000030
[<7f1cf1f0>] (_582 [ath9k]) from [<7f1d1ed8>] (_735+0x2a0/0xec4 [ath9k])
 r10:be6415a0 r9:8092d940 r8:bf07de98 r7:00000000 r6:be6470f0 r5:be6470ec
 r4:00000002
[<7f1d1c38>] (_735 [ath9k]) from [<80124d18>] (tasklet_action+0x84/0xf8)
 r10:bf07c000 r9:8092d940 r8:bf07de98 r7:00000000 r6:80844180 r5:be6470b8
 r4:be6470b4
[<80124c94>] (tasklet_action) from [<8012501c>] (__do_softirq+0x100/0x1f4)
 r10:bf07c000 r9:00000100 r8:40000006 r7:80902080 r6:80902098 r5:00000006
 r4:00000000 r3:80124c94
[<80124f1c>] (__do_softirq) from [<80125148>] (run_ksoftirqd+0x38/0x4c)
 r10:00000000 r9:00000000 r8:00000000 r7:00000000 r6:8090fe54 r5:bf03b200
 r4:bf07c000
[<80125110>] (run_ksoftirqd) from [<8013deb4>] (smpboot_thread_fn+0x224/0x260)
[<8013dc90>] (smpboot_thread_fn) from [<8013ad40>] (kthread+0xf8/0x100)
 r9:00000000 r8:00000000 r7:8013dc90 r6:bf03b200 r5:bf03b240 r4:00000000
[<8013ac48>] (kthread) from [<80107760>] (ret_from_fork+0x14/0x34)
 r7:00000000 r6:00000000 r5:8013ac48 r4:bf03b240
Code: e89da800 e1a0c00d e92dd800 e24cb004 (e5901014)
---[ end trace dddf11ac9111b272 ]---
Kernel panic - not syncing: Fatal exception in interrupt
CPU1: stopping
CPU: 1 PID: 0 Comm: swapper/1 Tainted: P      D         4.9.85 #19
Hardware name: Freescale i.MX6 Quad/DualLite (Device Tree)
Backtrace:
[<8010a708>] (dump_backtrace) from [<8010a99c>] (show_stack+0x18/0x1c)
 r7:bf093f58 r6:20000193 r5:809168e8 r4:00000000
[<8010a984>] (show_stack) from [<802a09c4>] (dump_stack+0x94/0xa8)
[<802a0930>] (dump_stack) from [<8010d184>] (handle_IPI+0xe8/0x180)
 r7:bf093f58 r6:00000000 r5:00000001 r4:808478c4
[<8010d09c>] (handle_IPI) from [<801013e8>] (gic_handle_irq+0x78/0x7c)
 r7:f4000100 r6:bf093f58 r5:f400010c r4:8090467c
[<80101370>] (gic_handle_irq) from [<8010b378>] (__irq_svc+0x58/0x8c)
Exception stack(0xbf093f58 to 0xbf093fa0)
3f40:                                                       bf7d62a0 00000000
3f60: 0010a5f4 80113460 bf092000 809043e4 00000002 80904434 bf092008 412fc09a
3f80: 00000000 bf093fb4 bf093fb8 bf093fa8 8010804c 80108050 60000013 ffffffff
 r9:bf092000 r8:bf092008 r7:bf093f8c r6:ffffffff r5:60000013 r4:80108050
[<80108014>] (arch_cpu_idle) from [<80553c2c>] (default_idle_call+0x30/0x34)
[<80553bfc>] (default_idle_call) from [<80158394>] (cpu_startup_entry+0xc4/0xfc)
[<801582d0>] (cpu_startup_entry) from [<8010ce40>] (secondary_start_kernel+0x168/0x174)
 r7:8092d2f8 r4:80913568
[<8010ccd8>] (secondary_start_kernel) from [<10101488>] (0x10101488)
 r5:00000055 r4:4f07806a
Rebooting in 10 seconds..
Reboot failed -- System halted

Signed-off-by: Sebastian Gottschall <s.gottschall@dd-wrt.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
6 years agoath10k: Fix kernel panic while using worker (ath10k_sta_rc_update_wk)
Karthikeyan Periyasamy [Mon, 12 Mar 2018 11:39:40 +0000 (17:09 +0530)]
ath10k: Fix kernel panic while using worker (ath10k_sta_rc_update_wk)

When attempt to run worker (ath10k_sta_rc_update_wk) after the station object
(ieee80211_sta) delete will trigger the kernel panic.

This problem arise in AP + Mesh configuration, Where the current node AP VAP
and neighbor node mesh VAP MAC address are same. When the current mesh node
try to establish the mesh link with neighbor node, driver peer creation for
the neighbor mesh node fails due to duplication MAC address. Already the AP
VAP created with same MAC address.

It is caused by the following scenario steps.

Steps:
1. In above condition, ath10k driver sta_state callback (ath10k_sta_state)
   fails to do the state change for a station from IEEE80211_STA_NOTEXIST
   to IEEE80211_STA_NONE due to peer creation fails. Sta_state callback is
   called from ieee80211_add_station() to handle the new station
   (neighbor mesh node) request from the wpa_supplicant.
2. Concurrently ath10k receive the sta_rc_update callback notification from
   the mesh_neighbour_update() to handle the beacon frames of the above
   neighbor mesh node. since its atomic callback, ath10k driver queue the
   work (ath10k_sta_rc_update_wk) to handle rc update.
3. Due to driver sta_state callback fails (step 1), mac80211 free the station
   object.
4. When the worker (ath10k_sta_rc_update_wk) scheduled to run, it will access
   the station object which is already deleted. so it will trigger kernel
   panic.

Added the peer exist check in sta_rc_update callback before queue the work.

Kernel Panic log:

Unable to handle kernel NULL pointer dereference at virtual address 00000000
pgd = c0204000
[00000000] *pgd=00000000
Internal error: Oops: 17 [#1] PREEMPT SMP ARM
CPU: 1 PID: 1833 Comm: kworker/u4:2 Not tainted 3.14.77 #1
task: dcef0000 ti: d72b6000 task.ti: d72b6000
PC is at pwq_activate_delayed_work+0x10/0x40
LR is at pwq_activate_delayed_work+0xc/0x40
pc : [<c023f988>]    lr : [<c023f984>]    psr: 40000193
sp : d72b7f18  ip : 0000007a  fp : d72b6000
r10: 00000000  r9 : dd404414  r8 : d8c31998
r7 : d72b6038  r6 : 00000004  r5 : d4907ec8  r4 : dcee1300
r3 : ffffffe0  r2 : 00000000  r1 : 00000001  r0 : 00000000
Flags: nZcv  IRQs off  FIQs on  Mode SVC_32  ISA ARM  Segment kernel
Control: 10c5787d  Table: 595bc06a  DAC: 00000015
...
Process kworker/u4:2 (pid: 1833, stack limit = 0xd72b6238)
Stack: (0xd72b7f18 to 0xd72b8000)
7f00:                                                       00000001 dcee1300
7f20: 00000001 c02410dc d8c31980 dd404400 dd404400 c0242790 d8c31980 00000089
7f40: 00000000 d93e1340 00000000 d8c31980 c0242568 00000000 00000000 00000000
7f60: 00000000 c02474dc 00000000 00000000 000000f8 d8c31980 00000000 00000000
7f80: d72b7f80 d72b7f80 00000000 00000000 d72b7f90 d72b7f90 d72b7fac d93e1340
7fa0: c0247404 00000000 00000000 c0208d20 00000000 00000000 00000000 00000000
7fc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
7fe0: 00000000 00000000 00000000 00000000 00000013 00000000 00000000 00000000
[<c023f988>] (pwq_activate_delayed_work) from [<c02410dc>] (pwq_dec_nr_in_flight+0x58/0xc4)
[<c02410dc>] (pwq_dec_nr_in_flight) from [<c0242790>] (worker_thread+0x228/0x360)
[<c0242790>] (worker_thread) from [<c02474dc>] (kthread+0xd8/0xec)
[<c02474dc>] (kthread) from [<c0208d20>] (ret_from_fork+0x14/0x34)
Code: e92d4038 e1a05000 ebffffbc[69210.619376] SMP: failed to stop secondary CPUs
Rebooting in 3 seconds..

Signed-off-by: Karthikeyan Periyasamy <periyasa@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
6 years agoath10k: fix use-after-free in ath10k_wmi_cmd_send_nowait
Carl Huang [Mon, 5 Mar 2018 06:44:02 +0000 (14:44 +0800)]
ath10k: fix use-after-free in ath10k_wmi_cmd_send_nowait

The skb may be freed in tx completion context before
trace_ath10k_wmi_cmd is called. This can be easily captured when
KASAN(Kernel Address Sanitizer) is enabled. The fix is to move
trace_ath10k_wmi_cmd before the send operation. As the ret has no
meaning in trace_ath10k_wmi_cmd then, so remove this parameter too.

Signed-off-by: Carl Huang <cjhuang@codeaurora.org>
Tested-by: Brian Norris <briannorris@chromium.org>
Reviewed-by: Brian Norris <briannorris@chromium.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
6 years agowcn36xx: reduce verbosity of drivers messages
Ramon Fried [Wed, 14 Mar 2018 10:14:11 +0000 (12:14 +0200)]
wcn36xx: reduce verbosity of drivers messages

Whenever the WLAN interface is started the FW
version and caps are printed.
The caps now will be displayed only in debug mode.
Firmware version will be displayed only once on first
startup of the interface.

Change-Id: I4db6ea7f384fe15eebe4c3ddb1d1ccab00094332
Signed-off-by: Ramon Fried <rfried@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
6 years agoath: fix false radar detection in JP region
Sriram R [Wed, 14 Mar 2018 10:14:10 +0000 (12:14 +0200)]
ath: fix false radar detection in JP region

This fixes false radar detection (of radar type 7)
in Japan region by correcting the radar pulse type
to Chirp as per specification.

Signed-off-by: Sriram R <srirrama@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
6 years agoath10k: debugfs support to get final TPC stats for 10.4 variants
Maharaja Kennadyrajan [Wed, 14 Mar 2018 10:14:08 +0000 (12:14 +0200)]
ath10k: debugfs support to get final TPC stats for 10.4 variants

Export the final Transmit Power Control (TPC) value, which is the
minimum of control power and existing TPC value to user space via
a new debugfs file "tpc_stats_final" to help with debugging.
It works with the new wmi cmd and event introduced in 10.4 firmware
branch.

WMI command ID: WMI_PDEV_GET_TPC_TABLE_CMDID
WMI event ID: WMI_PDEV_TPC_TABLE_EVENTID

cat /sys/kernel/debug/ieee80211/phyX/ath10k/tpc_stats_final

$ cat /sys/kernel/debug/ieee80211/phyX/ath10k/tpc_stats_final

TPC config for channel 5180 mode 10

CTL             =  0x 0 Reg. Domain             = 58
Antenna Gain    =  0 Reg. Max Antenna Gain      =   0
Power Limit     = 60 Reg. Max Power             = 60
Num tx chains   =  2 Num supported rates        = 109

******************* CDD POWER TABLE ****************

No.  Preamble Rate_code tpc_value1 tpc_value2 tpc_value3
0    CCK      0x40        0          0
1    CCK      0x41        0          0
[...]
107  HTCUP    0x 0       46          46
108  HTCUP    0x 0       46          46

******************* STBC POWER TABLE ****************

No.  Preamble Rate_code tpc_value1 tpc_value2 tpc_value3
0    CCK      0x40        0          0
1    CCK      0x41        0          0
[...]
107  HTCUP    0x 0        46         46
108  HTCUP    0x 0        46         46

***********************************
TXBF not supported
**********************************

The existing tpc_stats debugfs file provides the dump
which is minimum of target power and regulatory domain.

cat /sys/kernel/debug/ieee80211/phyX/ath10k/tpc_stats

Hardware_used: QCA4019
Firmware version: firmware-5.bin_10.4-3.0-00209

Signed-off-by: Maharaja Kennadyrajan <mkenna@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
6 years agoath10k: add sta rx packet stats per tid
Vasanthakumar Thiagarajan [Wed, 14 Mar 2018 10:14:06 +0000 (12:14 +0200)]
ath10k: add sta rx packet stats per tid

Added per tid sta counters for the following

- Total number MSDUs received from firmware
- Number of MSDUs received with errors like decryption, crc, mic ,etc.
- Number of MSDUs dropped in the driver
- A-MPDU/A-MSDU subframe stats
- Number of MSDUS passed to mac80211

All stats other than A-MPDU stats are only for received data frames.
A-MPDU stats might have stats for management frames when monitor
interface is active where management frames are notified both in wmi
and HTT interfaces.

These per tid stats can be enabled with tid bitmask through a debugfs
like below

 echo <tid_bitmask> > /sys/kernel/debug/ieee80211/phyX/ath10k/sta_tid_stats_mask

 tid 16 (tid_bitmask 0x10000) is used for non-qos data/management frames

The stats are read from
/sys/kernel/debug/ieee80211/phyX/netdev\:wlanX/stations/<sta_mac>/dump_tid_stats

Sample output:

 To enable rx stats for tid 0, 5 and 6,

 echo 0x00000061 > /sys/kernel/debug/ieee80211/phy0/ath10k/sta_tid_stats_mask

cat /sys/kernel/debug/ieee80211/phy0/netdev\:wlan15/stations/8c\:fd\:f0\:0a\:8e\:df/dump_tid_stats

   Driver Rx pkt stats per tid, ([tid] count)
                ------------------------------------------
MSDUs from FW                   [00] 2567        [05] 3178        [06] 1089
MSDUs unchained                 [00] 0           [05] 0           [06] 0
MSDUs locally dropped:chained   [00] 0           [05] 0           [06] 0
MSDUs locally dropped:filtered  [00] 0           [05] 0           [06] 0
MSDUs queued for mac80211       [00] 2567        [05] 3178        [06] 1089
MSDUs with error:fcs_err        [00] 0           [05] 0           [06] 2
MSDUs with error:tkip_err       [00] 0           [05] 0           [06] 0
MSDUs with error:crypt_err      [00] 0           [05] 0           [06] 0
MSDUs with error:peer_idx_inval [00] 0           [05] 0           [06] 0

A-MPDU num subframes upto 10    [00] 2567        [05] 3178        [06] 1087
A-MPDU num subframes 11-20      [00] 0           [05] 0           [06] 0
A-MPDU num subframes 21-30      [00] 0           [05] 0           [06] 0
A-MPDU num subframes 31-40      [00] 0           [05] 0           [06] 0
A-MPDU num subframes 41-50      [00] 0           [05] 0           [06] 0
A-MPDU num subframes 51-60      [00] 0           [05] 0           [06] 0
A-MPDU num subframes >60        [00] 0           [05] 0           [06] 0

A-MSDU num subframes 1          [00] 2567        [05] 3178        [06] 1089
A-MSDU num subframes 2          [00] 0           [05] 0           [06] 0
A-MSDU num subframes 3          [00] 0           [05] 0           [06] 0
A-MSDU num subframes 4          [00] 0           [05] 0           [06] 0
A-MSDU num subframes >4         [00] 0           [05] 0           [06] 0

Signed-off-by: Vasanthakumar Thiagarajan <vthiagar@codeaurora.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
6 years agoMerge tag 'wireless-drivers-next-for-davem-2018-03-24' of git://git.kernel.org/pub...
David S. Miller [Mon, 26 Mar 2018 01:27:38 +0000 (21:27 -0400)]
Merge tag 'wireless-drivers-next-for-davem-2018-03-24' of git://git./linux/kernel/git/kvalo/wireless-drivers-next

Kalle Valo says:

====================
wireless-drivers-next patches for 4.17

The biggest changes are the bluetooth related patches to the rsi
driver. It adds a new bluetooth driver which communicates directly
with the wireless driver and the interface is defined in
include/net/rsi_91x.h.

Major changes:

wl1251

* read the MAC address from the NVS file

rtlwifi

* enable mac80211 fast-tx support

mt76

* add capability to select tx/rx antennas

mt7601

* let mac80211 validate rx CCMP Packet Number (PN)

rsi

* bluetooth: add new btrsi driver

* btcoex support with the new btrsi driver
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotipc: tipc_disc_addr_trial_msg() can be static
kbuild test robot [Fri, 23 Mar 2018 19:47:42 +0000 (03:47 +0800)]
tipc: tipc_disc_addr_trial_msg() can be static

Fixes: 25b0b9c4e835 ("tipc: handle collisions of 32-bit node address hash values")
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Acked-by: Jon Maloy jon.maloy@ericsson.com
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoibmvnic: Potential NULL dereference in clean_one_tx_pool()
Dan Carpenter [Fri, 23 Mar 2018 11:36:15 +0000 (14:36 +0300)]
ibmvnic: Potential NULL dereference in clean_one_tx_pool()

There is an && vs || typo here, which potentially leads to a NULL
dereference.

Fixes: e9e1e97884b7 ("ibmvnic: Update TX pool cleaning routine")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agocxgb4: support new ISSI flash parts
Ganesh Goudar [Fri, 23 Mar 2018 11:35:49 +0000 (17:05 +0530)]
cxgb4: support new ISSI flash parts

Add support for new 32MB and 64MB ISSI (Integrated Silicon
Solution, Inc.) FLASH parts.

Signed-off-by: Casey Leedom <leedom@chelsio.com>
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agocxgb4: depend on firmware event for link status
Ganesh Goudar [Fri, 23 Mar 2018 11:33:10 +0000 (17:03 +0530)]
cxgb4: depend on firmware event for link status

Depend on the firmware sending us link status changes,
rather than assuming that the link goes down upon L1
configuration.

Signed-off-by: Casey Leedom <leedom@chelsio.com>
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agocxgb4: copy vlan_id in ndo_get_vf_config
Arjun Vynipadath [Fri, 23 Mar 2018 10:18:46 +0000 (15:48 +0530)]
cxgb4: copy vlan_id in ndo_get_vf_config

Copy vlan_id to get it displayed in vf info.

Signed-off-by: Arjun Vynipadath <arjun@chelsio.com>
Signed-off-by: Casey Leedom <leedom@chelsio.com>
Signed-off-by: Ganesh Goudhar <ganeshgr@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agocxgb4: Setup FW queues before registering netdev
Arjun Vynipadath [Fri, 23 Mar 2018 09:55:10 +0000 (15:25 +0530)]
cxgb4: Setup FW queues before registering netdev

When NetworkManager is enabled, there are chances that interface up
is called even before probe completes. This means we have not yet
allocated the FW sge queues, hence rest of ingress queue allocation
wont be proper. Fix this by calling setup_fw_sge_queues() before
register_netdev().

Fixes: 0fbc81b3ad51 ('chcr/cxgb4i/cxgbit/RDMA/cxgb4: Allocate resources dynamically for all cxgb4 ULD's')
Signed-off-by: Arjun Vynipadath <arjun@chelsio.com>
Signed-off-by: Casey Leedom <leedom@chelsio.com>
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'broadcom-Adaptive-interrupt-coalescing'
David S. Miller [Mon, 26 Mar 2018 00:48:26 +0000 (20:48 -0400)]
Merge branch 'broadcom-Adaptive-interrupt-coalescing'

Florian Fainelli says:

====================
net: broadcom: Adaptive interrupt coalescing

This patch series adds adaptive interrupt coalescing for the Gigabit Ethernet
drivers SYSTEMPORT and GENET.

This really helps lower the interrupt count and system load, as measured by
vmstat for a Gigabit TCP RX session:

SYSTEMPORT:

without:

 1  0      0 192188      0  25472    0    0     0     0 122100 38870  1 42 57  0  0
[ ID] Interval       Transfer     Bandwidth
[  4]  0.0-10.0 sec  1.03 GBytes   884 Mbits/sec

with:

 1  0      0 192288      0  25468    0    0     0     0 58806 44401  0 100  0  0  0
[  5]  0.0-10.0 sec  1.04 GBytes   888 Mbits/sec

GENET:

without:

 1  0      0 1170404      0  25420    0    0     0     0 130785 63402  2 85 12  0  0
[ ID] Interval       Transfer     Bandwidth
[  4]  0.0-10.0 sec  1.04 GBytes   888 Mbits/sec

with:

 1  0      0 1170560      0  25420    0    0     0     0 50610 48477  0 100  0  0  0
[  5]  0.0-10.0 sec  1.05 GBytes   899 Mbits/sec

Please look at the implementation and let me know if you see any problems, this
was largely inspired by bnxt_en.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: bcmgenet: Add support for adaptive RX coalescing
Florian Fainelli [Fri, 23 Mar 2018 01:19:33 +0000 (18:19 -0700)]
net: bcmgenet: Add support for adaptive RX coalescing

Unlike the moder modern SYSTEMPORT hardware, we do not have a
configurable TDMA timeout, which limits us to implement adaptive RX
interrupt coalescing only. We have each of our RX rings implement a
bcmgenet_net_dim structure which holds an interrupt counter, number of
packets, bytes, and a container for a net_dim instance.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: systemport: Implement adaptive interrupt coalescing
Florian Fainelli [Fri, 23 Mar 2018 01:19:32 +0000 (18:19 -0700)]
net: systemport: Implement adaptive interrupt coalescing

Implement support for adaptive RX and TX interrupt coalescing using
net_dim. We have each of our TX ring and our single RX ring implement a
bcm_sysport_net_dim structure which holds an interrupt counter, number
of packets, bytes, and a container for a net_dim instance.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'mv88e6xxx-module-reloading'
David S. Miller [Mon, 26 Mar 2018 00:43:42 +0000 (20:43 -0400)]
Merge branch 'mv88e6xxx-module-reloading'

Andrew Lunn says:

====================
Fixes to allow mv88e6xxx module to be reloaded

As reported by Uwe Kleine-König, the interrupt trigger is first
configured by DT and then reconfigured to edge. This results in a
failure on EPROBE_DEFER, or if the module is unloaded and reloaded.

A second crash happens on module reload due to a missing call to the
common IRQ free code when using polled interrupts.

With these fixes in place, it becomes possible to load and unload the
kernel modules a few times without it crashing.

v2: Fix the ü in Künig a couple of times
v3: But the ü should be an ö!
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: mv88e6xxx: Call the common IRQ free code
Andrew Lunn [Sun, 25 Mar 2018 21:43:15 +0000 (23:43 +0200)]
net: dsa: mv88e6xxx: Call the common IRQ free code

When free'ing the polled IRQs, call the common irq free code.
Otherwise the interrupts are left registered, and when we come to load
the driver a second time, we get an Opps.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: dsa: mv88e6xxx: Use the DT IRQ trigger mode
Andrew Lunn [Sun, 25 Mar 2018 21:43:14 +0000 (23:43 +0200)]
net: dsa: mv88e6xxx: Use the DT IRQ trigger mode

By calling request_threaded_irq() with the flag IRQF_TRIGGER_FALLING
we override the trigger mode provided in device tree. And the
interrupt is actually active low, which is what all the current device
tree descriptions use.

Suggested-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Acked-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Tested-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotc-testing: updated police, mirred, skbedit and skbmod with more tests
Roman Mashak [Sun, 25 Mar 2018 21:20:06 +0000 (17:20 -0400)]
tc-testing: updated police, mirred, skbedit and skbmod with more tests

Added extra test cases for control actions (reclassify, pipe etc.),
cookies, max index value and police args sanity check.

Signed-off-by: Roman Mashak <mrv@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'hv_netvsc-Fix-improve-RX-path-error-handling'
David S. Miller [Sun, 25 Mar 2018 21:07:41 +0000 (17:07 -0400)]
Merge branch 'hv_netvsc-Fix-improve-RX-path-error-handling'

Haiyang Zhang says:

====================
hv_netvsc: Fix/improve RX path error handling

Fix the status code returned to the host. Also add range
check for rx packet offset and length.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agohv_netvsc: Add range checking for rx packet offset and length
Haiyang Zhang [Thu, 22 Mar 2018 19:01:14 +0000 (12:01 -0700)]
hv_netvsc: Add range checking for rx packet offset and length

This patch adds range checking for rx packet offset and length.
It may only happen if there is a host side bug.

Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agohv_netvsc: Fix the return status in RX path
Haiyang Zhang [Thu, 22 Mar 2018 19:01:13 +0000 (12:01 -0700)]
hv_netvsc: Fix the return status in RX path

As defined in hyperv_net.h, the NVSP_STAT_SUCCESS is one not zero.
Some functions returns 0 when it actually means NVSP_STAT_SUCCESS.
This patch fixes them.

In netvsc_receive(), it puts the last RNDIS packet's receive status
for all packets in a vmxferpage which may contain multiple RNDIS
packets.
This patch puts NVSP_STAT_FAIL in the receive completion if one of
the packets in a vmxferpage fails.

Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'net-permit-skb_segment-on-head_frag-frag_list-skb'
David S. Miller [Sun, 25 Mar 2018 20:46:05 +0000 (16:46 -0400)]
Merge branch 'net-permit-skb_segment-on-head_frag-frag_list-skb'

Yonghong Song says:

====================
net: permit skb_segment on head_frag frag_list skb

One of our in-house projects, bpf-based NAT, hits a kernel BUG_ON at
function skb_segment(), line 3667. The bpf program attaches to
clsact ingress, calls bpf_skb_change_proto to change protocol
from ipv4 to ipv6 or from ipv6 to ipv4, and then calls bpf_redirect
to send the changed packet out.
 ...
    3665                 while (pos < offset + len) {
    3666                         if (i >= nfrags) {
    3667                                 BUG_ON(skb_headlen(list_skb));
 ...

The triggering input skb has the following properties:
    list_skb = skb->frag_list;
    skb->nfrags != NULL && skb_headlen(list_skb) != 0
and skb_segment() is not able to handle a frag_list skb
if its headlen (list_skb->len - list_skb->data_len) is not 0.

Patch #1 provides a simple solution to avoid BUG_ON. If
list_skb->head_frag is true, its page-backed frag will
be processed before the list_skb->frags.
Patch #2 provides a test case in test_bpf module which
constructs a skb and calls skb_segment() directly. The test
case is able to trigger the BUG_ON without Patch #1.

The patch has been tested in the following setup:
  ipv6_host <-> nat_server <-> ipv4_host
where nat_server has a bpf program doing ipv4<->ipv6
translation and forwarding through clsact hook
bpf_skb_change_proto.

Changelog:
v5 -> v6:
  . Added back missed BUG_ON(!nfrags) for zero
    skb_headlen(skb) case, plus a couple of
    cosmetic changes, from Alexander.
v4 -> v5:
  . Replace local variable head_frag with
    a static inline function skb_head_frag_to_page_desc
    which gets the head_frag on-demand. This makes
    code more readable and also does not increase
    the stack size, from Alexander.
  . Remove the "if(nfrags)" guard for skb_orphan_frags
    and skb_zerocopy_clone as I found that they can
    handle zero-frag skb (with non-zero skb_headlen(skb))
    properly.
  . Properly release segment list from skb_segment()
    in the test, from Eric.
v3 -> v4:
  . Remove dynamic memory allocation and use rewinding
    for both index and frag to remove one branch in fast path,
    from Alexander.
  . Fix a bunch of issues in test_bpf skb_segment() test,
    including proper way to allocate skb, proper function
    argument for skb_add_rx_frag and not freeint skb, etc.,
    from Eric.
v2 -> v3:
  . Use starting frag index -1 (instead of 0) to
    special process head_frag before other frags in the skb,
    from Alexander Duyck.
v1 -> v2:
  . Removed never-hit BUG_ON, spotted by Linyu Yuan.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: bpf: add a test for skb_segment in test_bpf module
Yonghong Song [Wed, 21 Mar 2018 23:31:04 +0000 (16:31 -0700)]
net: bpf: add a test for skb_segment in test_bpf module

Without the previous commit,
"modprobe test_bpf" will have the following errors:
...
[   98.149165] ------------[ cut here ]------------
[   98.159362] kernel BUG at net/core/skbuff.c:3667!
[   98.169756] invalid opcode: 0000 [#1] SMP PTI
[   98.179370] Modules linked in:
[   98.179371]  test_bpf(+)
...
which triggers the bug the previous commit intends to fix.

The skbs are constructed to mimic what mlx5 may generate.
The packet size/header may not mimic real cases in production. But
the processing flow is similar.

Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: permit skb_segment on head_frag frag_list skb
Yonghong Song [Wed, 21 Mar 2018 23:31:03 +0000 (16:31 -0700)]
net: permit skb_segment on head_frag frag_list skb

One of our in-house projects, bpf-based NAT, hits a kernel BUG_ON at
function skb_segment(), line 3667. The bpf program attaches to
clsact ingress, calls bpf_skb_change_proto to change protocol
from ipv4 to ipv6 or from ipv6 to ipv4, and then calls bpf_redirect
to send the changed packet out.

3472 struct sk_buff *skb_segment(struct sk_buff *head_skb,
3473                             netdev_features_t features)
3474 {
3475         struct sk_buff *segs = NULL;
3476         struct sk_buff *tail = NULL;
...
3665                 while (pos < offset + len) {
3666                         if (i >= nfrags) {
3667                                 BUG_ON(skb_headlen(list_skb));
3668
3669                                 i = 0;
3670                                 nfrags = skb_shinfo(list_skb)->nr_frags;
3671                                 frag = skb_shinfo(list_skb)->frags;
3672                                 frag_skb = list_skb;
...

call stack:
...
 #1 [ffff883ffef03558] __crash_kexec at ffffffff8110c525
 #2 [ffff883ffef03620] crash_kexec at ffffffff8110d5cc
 #3 [ffff883ffef03640] oops_end at ffffffff8101d7e7
 #4 [ffff883ffef03668] die at ffffffff8101deb2
 #5 [ffff883ffef03698] do_trap at ffffffff8101a700
 #6 [ffff883ffef036e8] do_error_trap at ffffffff8101abfe
 #7 [ffff883ffef037a0] do_invalid_op at ffffffff8101acd0
 #8 [ffff883ffef037b0] invalid_op at ffffffff81a00bab
    [exception RIP: skb_segment+3044]
    RIP: ffffffff817e4dd4  RSP: ffff883ffef03860  RFLAGS: 00010216
    RAX: 0000000000002bf6  RBX: ffff883feb7aaa00  RCX: 0000000000000011
    RDX: ffff883fb87910c0  RSI: 0000000000000011  RDI: ffff883feb7ab500
    RBP: ffff883ffef03928   R8: 0000000000002ce2   R9: 00000000000027da
    R10: 000001ea00000000  R11: 0000000000002d82  R12: ffff883f90a1ee80
    R13: ffff883fb8791120  R14: ffff883feb7abc00  R15: 0000000000002ce2
    ORIG_RAX: ffffffffffffffff  CS: 0010  SS: 0018
 #9 [ffff883ffef03930] tcp_gso_segment at ffffffff818713e7

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch '10GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next...
David S. Miller [Sun, 25 Mar 2018 20:24:34 +0000 (16:24 -0400)]
Merge branch '10GbE' of git://git./linux/kernel/git/jkirsher/next-queue

Jeff Kirsher says:

====================
10GbE Intel Wired LAN Driver Updates 2018-03-23

This series contains updates to ixgbe and ixgbevf only.

Paul adds status register reads to reduce a potential race condition
where registers can read 0xFFFFFFFF during a PCI reset, which in turn
causes the driver to remove the adapter.  Then fixes an assignment
operation with an "OR" operation.

Shannon Nelson provides several IPsec offload cleanups to ixgbe, as well as a
patch to enable TSO with IPsec offload.

Tony provides the much anticipated XDP support for ixgbevf.  Currently,
pass, drop and XDP_TX actions are supported, as well as meta data and
stats reporting.

Björn Töpel tweaks the page counting for XDP_REDIRECT, since a page can
have its reference count decreased via the xdp_do_redirect() call.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'liquidio-Tx-queue-cleanup'
David S. Miller [Sun, 25 Mar 2018 20:18:55 +0000 (16:18 -0400)]
Merge branch 'liquidio-Tx-queue-cleanup'

Intiyaz Basha says:

====================
liquidio: Tx queue cleanup

Moved some common function to octeon_network.h
Removed some unwanted functions and checks.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoliquidio: Renamed txqs_start to start_txqs
Intiyaz Basha [Sat, 24 Mar 2018 00:37:44 +0000 (17:37 -0700)]
liquidio: Renamed txqs_start to start_txqs

For consistency renaming txqs_start to start_txqs

Signed-off-by: Intiyaz Basha <intiyaz.basha@cavium.com>
Acked-by: Derek Chickles <derek.chickles@cavium.com>
Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoliquidio: Renamed txqs_stop to stop_txqs
Intiyaz Basha [Sat, 24 Mar 2018 00:37:41 +0000 (17:37 -0700)]
liquidio: Renamed txqs_stop to stop_txqs

For consistency renaming txqs_stop to stop_txqs

Signed-off-by: Intiyaz Basha <intiyaz.basha@cavium.com>
Acked-by: Derek Chickles <derek.chickles@cavium.com>
Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoliquidio: Renamed txqs_wake to wake_txqs
Intiyaz Basha [Sat, 24 Mar 2018 00:37:39 +0000 (17:37 -0700)]
liquidio: Renamed txqs_wake to wake_txqs

For consistency renaming txqs_wake to wake_txqs

Signed-off-by: Intiyaz Basha <intiyaz.basha@cavium.com>
Acked-by: Derek Chickles <derek.chickles@cavium.com>
Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoliquidio: Function call skb_iq for deriving queue from skb
Intiyaz Basha [Sat, 24 Mar 2018 00:37:36 +0000 (17:37 -0700)]
liquidio: Function call skb_iq for deriving queue from skb

Using skb_iq function for deriving queue from skb

Signed-off-by: Intiyaz Basha <intiyaz.basha@cavium.com>
Acked-by: Derek Chickles <derek.chickles@cavium.com>
Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoliquidio: Removed one line function wake_q
Intiyaz Basha [Sat, 24 Mar 2018 00:37:33 +0000 (17:37 -0700)]
liquidio: Removed one line function wake_q

Removing one line function wake_q

Signed-off-by: Intiyaz Basha <intiyaz.basha@cavium.com>
Acked-by: Derek Chickles <derek.chickles@cavium.com>
Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoliquidio: Removed one line function stop_q
Intiyaz Basha [Sat, 24 Mar 2018 00:37:30 +0000 (17:37 -0700)]
liquidio: Removed one line function stop_q

Removing one line function stop_q

Signed-off-by: Intiyaz Basha <intiyaz.basha@cavium.com>
Acked-by: Derek Chickles <derek.chickles@cavium.com>
Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoliquidio: Removed netif_is_multiqueue check
Intiyaz Basha [Sat, 24 Mar 2018 00:37:28 +0000 (17:37 -0700)]
liquidio: Removed netif_is_multiqueue check

Removing checks for netif_is_multiqueue.
Configuring single queue will be a multiqueue netdev with one queues.

Signed-off-by: Intiyaz Basha <intiyaz.basha@cavium.com>
Acked-by: Derek Chickles <derek.chickles@cavium.com>
Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoliquidio: Removed start_txq function
Intiyaz Basha [Sat, 24 Mar 2018 00:37:25 +0000 (17:37 -0700)]
liquidio: Removed start_txq function

Removing start_txq function from VF and PF files

Signed-off-by: Intiyaz Basha <intiyaz.basha@cavium.com>
Acked-by: Derek Chickles <derek.chickles@cavium.com>
Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoliquidio: Removed one line function stop_txq
Intiyaz Basha [Sat, 24 Mar 2018 00:37:20 +0000 (17:37 -0700)]
liquidio: Removed one line function stop_txq

Removing one line function stop_txq

Signed-off-by: Intiyaz Basha <intiyaz.basha@cavium.com>
Acked-by: Derek Chickles <derek.chickles@cavium.com>
Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoliquidio: Moved common function skb_iq to to octeon_network.h
Intiyaz Basha [Sat, 24 Mar 2018 00:37:17 +0000 (17:37 -0700)]
liquidio: Moved common function skb_iq to to octeon_network.h

Moving common function skb_iq to to octeon_network.h

Signed-off-by: Intiyaz Basha <intiyaz.basha@cavium.com>
Acked-by: Derek Chickles <derek.chickles@cavium.com>
Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoliquidio: Moved common function txqs_start to octeon_network.h
Intiyaz Basha [Sat, 24 Mar 2018 00:37:07 +0000 (17:37 -0700)]
liquidio: Moved common function txqs_start to octeon_network.h

Moving common function txqs_start to octeon_network.h

Signed-off-by: Intiyaz Basha <intiyaz.basha@cavium.com>
Acked-by: Derek Chickles <derek.chickles@cavium.com>
Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoliquidio: Moved common function txqs_wake to octeon_network.h
Intiyaz Basha [Sat, 24 Mar 2018 00:36:58 +0000 (17:36 -0700)]
liquidio: Moved common function txqs_wake to octeon_network.h

Moving common function txqs_wake to octeon_network.h

Signed-off-by: Intiyaz Basha <intiyaz.basha@cavium.com>
Acked-by: Derek Chickles <derek.chickles@cavium.com>
Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoliquidio: Moved common function txqs_stop to octeon_network.h
Intiyaz Basha [Sat, 24 Mar 2018 00:36:56 +0000 (17:36 -0700)]
liquidio: Moved common function txqs_stop to octeon_network.h

Moving common function txqs_stop to octeon_network.h

Signed-off-by: Intiyaz Basha <intiyaz.basha@cavium.com>
Acked-by: Derek Chickles <derek.chickles@cavium.com>
Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/sched: act_vlan: declare push_vid with host byte order
Davide Caratti [Fri, 23 Mar 2018 18:31:30 +0000 (19:31 +0100)]
net/sched: act_vlan: declare push_vid with host byte order

use u16 in place of __be16 to suppress the following sparse warnings:

 net/sched/act_vlan.c:150:26: warning: incorrect type in assignment (different base types)
 net/sched/act_vlan.c:150:26: expected restricted __be16 [usertype] push_vid
 net/sched/act_vlan.c:150:26: got unsigned short
 net/sched/act_vlan.c:151:21: warning: restricted __be16 degrades to integer
 net/sched/act_vlan.c:208:26: warning: incorrect type in assignment (different base types)
 net/sched/act_vlan.c:208:26: expected unsigned short [unsigned] [usertype] tcfv_push_vid
 net/sched/act_vlan.c:208:26: got restricted __be16 [usertype] push_vid

Signed-off-by: Davide Caratti <dcaratti@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/sched: remove tcf_idr_cleanup()
Davide Caratti [Fri, 23 Mar 2018 18:09:39 +0000 (19:09 +0100)]
net/sched: remove tcf_idr_cleanup()

tcf_idr_cleanup() is no more used, so remove it.

Suggested-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: Davide Caratti <dcaratti@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agomlxsw: spectrum_span: Prevent duplicate mirrors
Ido Schimmel [Fri, 23 Mar 2018 18:03:58 +0000 (21:03 +0300)]
mlxsw: spectrum_span: Prevent duplicate mirrors

In net commit 8175f7c4736f ("mlxsw: spectrum: Prevent duplicate
mirrors") we prevented the user from mirroring more than once from a
single binding point (port-direction pair).

The fix was essentially reverted in a merge conflict resolution when net
was merged into net-next. Restore it.

Fixes: 03fe2debbb27 ("Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net")
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoixgbe: tweak page counting for XDP_REDIRECT
Björn Töpel [Thu, 22 Mar 2018 09:02:36 +0000 (10:02 +0100)]
ixgbe: tweak page counting for XDP_REDIRECT

The current page counting scheme assumes that the reference count
cannot decrease until the received frame is sent to the upper layers
of the networking stack. This assumption does not hold for the
XDP_REDIRECT action, since a page (pointed out by xdp_buff) can have
its reference count decreased via the xdp_do_redirect call.

To work around that, we now start off by a large page count and then
don't allow a refcount less than two.

Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoixgbevf: Add XDP queue stats reporting
Tony Nguyen [Fri, 16 Mar 2018 22:34:06 +0000 (15:34 -0700)]
ixgbevf: Add XDP queue stats reporting

XDP stats are included in TX stats, however, they are not
reported in TX queue stats since they are setup on different
queues.  Add reporting for XDP queue stats to provide
consistency between the total stats and per queue stats.

Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoixgbevf: Add support for meta data
Tony Nguyen [Fri, 16 Mar 2018 22:34:05 +0000 (15:34 -0700)]
ixgbevf: Add support for meta data

Add support for XDP meta data when using build skb.

Based on commit 366a88fe2f40 ("bpf, ixgbe: add meta data support")

Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoixgbevf: Delay tail write for XDP packets
Tony Nguyen [Fri, 16 Mar 2018 22:34:04 +0000 (15:34 -0700)]
ixgbevf: Delay tail write for XDP packets

Current XDP implementation hits the tail on every XDP_TX; change the
driver to only hit the tail after packet processing is complete.

Based on
commit 7379f97a4fce ("ixgbe: delay tail write to every 'n' packets")

Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoixgbevf: Add support for XDP_TX action
Tony Nguyen [Fri, 16 Mar 2018 22:34:03 +0000 (15:34 -0700)]
ixgbevf: Add support for XDP_TX action

This implements the XDP_TX action which is modeled on the ixgbe
implementation. However instead of using CPU id to determine which XDP
queue to use, this uses the received RX queue index, which is similar
to i40e. Doing this eliminates the restriction that number of CPUs not
exceed number of XDP queues that ixgbe has.

Also, based on the number of queues available, the number of TX queues
may be reduced when an XDP program is loaded in order to accommodate the
XDP queues.

Based largely on
commit 33fdc82f0883 ("ixgbe: add support for XDP_TX action")

Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoixgbevf: Add XDP support for pass and drop actions
Tony Nguyen [Fri, 16 Mar 2018 22:34:02 +0000 (15:34 -0700)]
ixgbevf: Add XDP support for pass and drop actions

Implement XDP_PASS and XDP_DROP based on the ixgbe implementation.

Based largely on commit 924708081629 ("ixgbe: add XDP support for pass and
drop actions").

Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoixgbe: enable TSO with IPsec offload
Shannon Nelson [Fri, 16 Mar 2018 18:09:07 +0000 (11:09 -0700)]
ixgbe: enable TSO with IPsec offload

Fix things up to support TSO offload in conjunction
with IPsec hw offload.  This raises throughput with
IPsec offload on to nearly line rate.

Signed-off-by: Shannon Nelson <shannon.nelson@oracle.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoixgbe: no need for esp trailer if GSO
Shannon Nelson [Fri, 16 Mar 2018 18:09:06 +0000 (11:09 -0700)]
ixgbe: no need for esp trailer if GSO

There is no need to calculate the trailer length if we're doing
a GSO/TSO, as there is no trailer added to the packet data.
Also, don't bother clearing the flags field as it was already
cleared earlier.

Signed-off-by: Shannon Nelson <shannon.nelson@oracle.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoixgbe: remove unneeded ipsec test in TX path
Shannon Nelson [Fri, 16 Mar 2018 18:09:05 +0000 (11:09 -0700)]
ixgbe: remove unneeded ipsec test in TX path

Since the ipsec data fields will be zero anyway in the non-ipsec
case, we can remove the conditional jump.

Suggested-by: Alexander Duyck <alexander.duyck@gmail.com>
Signed-off-by: Shannon Nelson <shannon.nelson@oracle.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoixgbe: no need for ipsec csum feature check
Shannon Nelson [Fri, 16 Mar 2018 18:09:04 +0000 (11:09 -0700)]
ixgbe: no need for ipsec csum feature check

With the patch
commit f8aa2696b4af ("esp: check the NETIF_F_HW_ESP_TX_CSUM bit before segmenting")
we no longer need to protect ourself from checksum
offload requests on IPsec packets, so we can remove
the check in our .ndo_features_check callback.

Signed-off-by: Shannon Nelson <shannon.nelson@oracle.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoixgbe: fix read-modify-write in x550 phy setup
Paul Greenwalt [Thu, 15 Mar 2018 12:22:07 +0000 (08:22 -0400)]
ixgbe: fix read-modify-write in x550 phy setup

Replaced an assignment operation with an OR operation.

The variable assignment was overwriting the value read from the PHY
register. The OR operation sets only the intended register bits.

The bits that were being overwritten are reserved, so the assignment had no
functional impact.

Reported by: Shannon Nelson <shannon.nelson@oracle.com>
Signed-off-by: Paul Greenwalt <paul.greenwalt@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoixgbe: add status reg reads to ixgbe_check_remove
Paul Greenwalt [Mon, 12 Mar 2018 13:22:55 +0000 (09:22 -0400)]
ixgbe: add status reg reads to ixgbe_check_remove

Add status register reads and delay between reads to ixgbe_check_remove.
Registers can read 0xFFFFFFFF during PCI reset, which causes the driver
to remove the adapter. The additional status register reads can reduce the
chance of this race condition.

If the status register is not 0xFFFFFFFF, then ixgbe_check_remove returns
the value of the register being read.

Signed-off-by: Paul Greenwalt <paul.greenwalt@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agonet: phy: intel-xway: add VR9 v1.1 phy ids
Mathias Kresin [Thu, 22 Mar 2018 22:31:39 +0000 (23:31 +0100)]
net: phy: intel-xway: add VR9 v1.1 phy ids

The phys embedded into the v1.1 of the VR9 SoC are using different phy
ids. Add the phy ids to use the driver for this VR9 version as well.

Signed-off-by: Mathias Kresin <dev@kresin.me>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: phy: intel-xway: add VR9 version number
Mathias Kresin [Thu, 22 Mar 2018 22:31:38 +0000 (23:31 +0100)]
net: phy: intel-xway: add VR9 version number

The VR9 phy ids are matching only for the SoC version 1.2. Rename the
macros and change the names to take this into account.

Signed-off-by: Mathias Kresin <dev@kresin.me>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: hns3: hclge_inform_reset_assert_to_vf() can be static
kbuild test robot [Thu, 22 Mar 2018 21:31:07 +0000 (05:31 +0800)]
net: hns3: hclge_inform_reset_assert_to_vf() can be static

Fixes: 2bfbd35d8ecd ("net: hns3: Changes required in PF mailbox to support VF reset")
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoqed: Use true and false for boolean values
Gustavo A. R. Silva [Thu, 22 Mar 2018 20:08:49 +0000 (15:08 -0500)]
qed: Use true and false for boolean values

Assign true or false to boolean variables instead of an integer value.

This issue was detected with the help of Coccinelle.

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Acked-by: Sudarsana Kalluru <Sudarsana.Kalluru@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agodpaa_eth: use true and false for boolean values
Gustavo A. R. Silva [Thu, 22 Mar 2018 19:59:27 +0000 (14:59 -0500)]
dpaa_eth: use true and false for boolean values

Assign true or false to boolean variables instead of an integer value.

This issue was detected with the help of Coccinelle.

Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'tipc-introduce-128-bit-auto-configurable-node-id'
David S. Miller [Fri, 23 Mar 2018 17:12:19 +0000 (13:12 -0400)]
Merge branch 'tipc-introduce-128-bit-auto-configurable-node-id'

Jon Maloy says:

====================
tipc: introduce 128-bit auto-configurable node id

We introduce a 128-bit free-format node identity as an alternative to
the legacy <Zone.Cluster.Node> structured 32-bit node address.

We also make configuration of this identity optional; if a bearer is
enabled without a pre-configured node id it will be set automatically
based on the used interface's MAC or IP address.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotipc: obtain node identity from interface by default
Jon Maloy [Thu, 22 Mar 2018 19:42:52 +0000 (20:42 +0100)]
tipc: obtain node identity from interface by default

Selecting and explicitly configuring a TIPC node identity may be
unwanted in some cases.

In this commit we introduce a default setting if the identity has not
been set at the moment the first bearer is enabled. We do this by
using a raw copy of a unique identifier from the used interface: MAC
address in the case of an L2 bearer, IPv4/IPv6 address in the case
of a UDP bearer.

Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotipc: handle collisions of 32-bit node address hash values
Jon Maloy [Thu, 22 Mar 2018 19:42:51 +0000 (20:42 +0100)]
tipc: handle collisions of 32-bit node address hash values

When a 32-bit node address is generated from a 128-bit identifier,
there is a risk of collisions which must be discovered and handled.

We do this as follows:
- We don't apply the generated address immediately to the node, but do
  instead initiate a 1 sec trial period to allow other cluster members
  to discover and handle such collisions.

- During the trial period the node periodically sends out a new type
  of message, DSC_TRIAL_MSG, using broadcast or emulated broadcast,
  to all the other nodes in the cluster.

- When a node is receiving such a message, it must check that the
  presented 32-bit identifier either is unused, or was used by the very
  same peer in a previous session. In both cases it accepts the request
  by not responding to it.

- If it finds that the same node has been up before using a different
  address, it responds with a DSC_TRIAL_FAIL_MSG containing that
  address.

- If it finds that the address has already been taken by some other
  node, it generates a new, unused address and returns it to the
  requester.

- During the trial period the requesting node must always be prepared
  to accept a failure message, i.e., a message where a peer suggests a
  different (or equal)  address to the one tried. In those cases it
  must apply the suggested value as trial address and restart the trial
  period.

This algorithm ensures that in the vast majority of cases a node will
have the same address before and after a reboot. If a legacy user
configures the address explicitly, there will be no trial period and
messages, so this protocol addition is completely backwards compatible.

Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotipc: add 128-bit node identifier
Jon Maloy [Thu, 22 Mar 2018 19:42:50 +0000 (20:42 +0100)]
tipc: add 128-bit node identifier

We add a 128-bit node identity, as an alternative to the currently used
32-bit node address.

For the sake of compatibility and to minimize message header changes
we retain the existing 32-bit address field. When not set explicitly by
the user, this field will be filled with a hash value generated from the
much longer node identity, and be used as a shorthand value for the
latter.

We permit either the address or the identity to be set by configuration,
but not both, so when the address value is set by a legacy user the
corresponding 128-bit node identity is generated based on the that value.

Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotipc: remove direct accesses to own_addr field in struct tipc_net
Jon Maloy [Thu, 22 Mar 2018 19:42:49 +0000 (20:42 +0100)]
tipc: remove direct accesses to own_addr field in struct tipc_net

As a preparation to changing the addressing structure of TIPC we replace
all direct accesses to the tipc_net::own_addr field with the function
dedicated for this, tipc_own_addr().

There are no changes to program logics in this commit.

Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotipc: allow closest-first lookup algorithm when legacy address is configured
Jon Maloy [Thu, 22 Mar 2018 19:42:48 +0000 (20:42 +0100)]
tipc: allow closest-first lookup algorithm when legacy address is configured

The removal of an internal structure of the node address has an unwanted
side effect.
- Currently, if a user is sending an anycast message with destination
  domain 0, the tipc_namebl_translate() function will use the 'closest-
  first' algorithm to first look for a node local destination, and only
  when no such is found, will it resort to the cluster global 'round-
  robin' lookup algorithm.
- Current users can get around this, and enforce unconditional use of
  global round-robin by indicating a destination as Z.0.0 or Z.C.0.
- This option disappears when we make the node address flat, since the
  lookup algorithm has no way of recognizing this case. So, as long as
  there are node local destinations, the algorithm will always select
  one of those, and there is nothing the sender can do to change this.

We solve this by eliminating the 'closest-first' option, which was never
a good idea anyway, for non-legacy users, but only for those. To
distinguish between legacy users and non-legacy users we introduce a new
flag 'legacy_addr_format' in struct tipc_core, to be set when the user
configures a legacy-style Z.C.N node address. Hence, when a legacy user
indicates a zero lookup domain 'closest-first' is selected, and in all
other cases we use 'round-robin'.

Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotipc: remove restrictions on node address values
Jon Maloy [Thu, 22 Mar 2018 19:42:47 +0000 (20:42 +0100)]
tipc: remove restrictions on node address values

Nominally, TIPC organizes network nodes into a three-level network
hierarchy consisting of the levels 'zone', 'cluster' and 'node'. This
hierarchy is reflected in the node address format, - it is sub-divided
into an 8-bit zone id, and 12 bit cluster id, and a 12-bit node id.

However, the 'zone' and 'cluster' levels have in reality never been
fully implemented,and never will be. The result of this has been
that the first 20 bits the node identity structure have been wasted,
and the usable node identity range within a cluster has been limited
to 12 bits. This is starting to become a problem.

In the following commits, we will need to be able to connect between
nodes which are using the whole 32-bit value space of the node address.
We therefore remove the restrictions on which values can be assigned
to node identity, -it is from now on only a 32-bit integer with no
assumed internal structure.

Isolation between clusters is now achieved only by setting different
values for the 'network id' field used during neighbor discovery, in
practice leading to the latter becoming the new cluster identity.

The rules for accepting discovery requests/responses from neighboring
nodes now become:

- If the user is using legacy address format on both peers, reception
  of discovery messages is subject to the legacy lookup domain check
  in addition to the cluster id check.

- Otherwise, the discovery request/response is always accepted, provided
  both peers have the same network id.

This secures backwards compatibility for users who have been using zone
or cluster identities as cluster separators, instead of the intended
'network id'.

Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotipc: some cleanups in the file discover.c
Jon Maloy [Thu, 22 Mar 2018 19:42:46 +0000 (20:42 +0100)]
tipc: some cleanups in the file discover.c

To facilitate the coming changes in the neighbor discovery functionality
we make some renaming and refactoring of that code. The functional changes
in this commit are trivial, e.g., that we move the message sending call in
tipc_disc_timeout() outside the spinlock protected region.

Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotipc: refactor function tipc_enable_bearer()
Jon Maloy [Thu, 22 Mar 2018 19:42:45 +0000 (20:42 +0100)]
tipc: refactor function tipc_enable_bearer()

As a preparation for the next commits we try to reduce the footprint of
the function tipc_enable_bearer(), while hopefully making is simpler to
follow.

Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet/mlx5: Fix use-after-free
Gustavo A. R. Silva [Thu, 22 Mar 2018 18:44:56 +0000 (13:44 -0500)]
net/mlx5: Fix use-after-free

_rule_ is being freed and then dereferenced by accessing rule->ctx

Fix this by copying the value returned by PTR_ERR(rule->ctx) into a local
variable for its safe use after freeing _rule_

Addresses-Coverity-ID: 1466041 ("Read from pointer after free")
Fixes: 05564d0ae075 ("net/mlx5: Add flow-steering commands for FPGA IPSec implementation")
Reviewed-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'pernet-convert-part11'
David S. Miller [Fri, 23 Mar 2018 17:00:47 +0000 (13:00 -0400)]
Merge branch 'pernet-convert-part11'

Kirill Tkhai says:

====================
Converting pernet_operations (part #11)

this series continues to review and to convert pernet_operations
to make them possible to be executed in parallel for several
net namespaces at the same time.

I thought last series was last, but there is one
new pernet_operations came to kernel. This is
udp_sysctl_ops, and here we convert it.

Also, David Howells acked rxrpc_net_ops, so I resend
the patch in case of it should be queued by patchwork:

https://www.spinics.net/lists/netdev/msg490678.html
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: Convert rxrpc_net_ops
Kirill Tkhai [Thu, 22 Mar 2018 18:34:55 +0000 (21:34 +0300)]
net: Convert rxrpc_net_ops

These pernet_operations modifies rxrpc_net_id-pointed
per-net entities. There is external link to AF_RXRPC
in fs/afs/Kconfig, but it seems there is no other
pernet_operations interested in that per-net entities.

Signed-off-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Acked-by: David Howells <dhowells@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: Convert udp_sysctl_ops
Kirill Tkhai [Thu, 22 Mar 2018 18:34:46 +0000 (21:34 +0300)]
net: Convert udp_sysctl_ops

These pernet_operations just initialize udp4 defaults.

Signed-off-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agomlxsw: spectrum_span: Fix initialization of struct mlxsw_sp_span_parms
Petr Machata [Thu, 22 Mar 2018 18:14:47 +0000 (20:14 +0200)]
mlxsw: spectrum_span: Fix initialization of struct mlxsw_sp_span_parms

Since the first element of struct mlxsw_sp_span_parms is a pointer,
to zero-initialize this structure the correct notation is not = {0}, but
rather = {NULL}, as reported by sparse.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotc-testing: add selftests for 'bpf' action
Davide Caratti [Thu, 22 Mar 2018 18:12:19 +0000 (19:12 +0100)]
tc-testing: add selftests for 'bpf' action

Test d959: Add cBPF action with valid bytecode
Test f84a: Add cBPF action with invalid bytecode
Test e939: Add eBPF action with valid object-file
Test 282d: Add eBPF action with invalid object-file
Test d819: Replace cBPF bytecode and action control
Test 6ae3: Delete cBPF action
Test 3e0d: List cBPF actions
Test 55ce: Flush BPF actions
Test ccc3: Add cBPF action with duplicate index
Test 89c7: Add cBPF action with invalid index
Test 7ab9: Add cBPF action with cookie

Changes since v1:
 - use index=2^32-1 in test ccc3, add tests 7a89, 89c7 (thanks Roman Mashak)
 - added test 282d

Signed-off-by: Davide Caratti <dcaratti@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: bridge: fix direct access to bridge vlan_enabled and use helper
Nikolay Aleksandrov [Fri, 23 Mar 2018 16:27:06 +0000 (18:27 +0200)]
net: bridge: fix direct access to bridge vlan_enabled and use helper

We need to use br_vlan_enabled() helper otherwise we'll break builds
without bridge vlans:
net/bridge//br_if.c: In function ‘br_mtu’:
net/bridge//br_if.c:458:8: error: ‘const struct net_bridge’ has no
member named ‘vlan_enabled’
  if (br->vlan_enabled)
        ^
net/bridge//br_if.c:462:1: warning: control reaches end of non-void
function [-Wreturn-type]
 }
 ^
scripts/Makefile.build:324: recipe for target 'net/bridge//br_if.o'
failed

Fixes: 419d14af9e07 ("bridge: Allow max MTU when multiple VLANs present")
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'tls-RX'
David S. Miller [Fri, 23 Mar 2018 16:25:55 +0000 (12:25 -0400)]
Merge branch 'tls-RX'

Dave Watson says:

====================
TLS Rx

TLS tcp socket RX implementation, to match existing TX code.

This patchset completes the software TLS socket, allowing full
bi-directional communication over TLS using normal socket syscalls,
after the handshake has been done in userspace.  Only the symmetric
encryption is done in the kernel.

This allows usage of TLS sockets from within the kernel (for example
with network block device, or from bpf).  Performance can be better
than userspace, with appropriate crypto routines [1].

sk->sk_socket->ops must be overridden to implement splice_read and
poll, but otherwise the interface & implementation match TX closely.
strparser is used to parse TLS framing on receive.

There are Openssl RX patches that work with this interface [2], as
well as a testing tool using the socket interface directly (without
cmsg support) [3].  An example tcp socket setup is:

  // Normal tcp socket connect/accept, and TLS handshake
  // using any TLS library.
  setsockopt(sock, SOL_TCP, TCP_ULP, "tls", sizeof("tls"));

  struct tls12_crypto_info_aes_gcm_128 crypto_info_rx;
  // Fill in crypto_info based on negotiated keys.

  setsockopt(sock, SOL_TLS, TLS_RX, &crypto_info, sizeof(crypto_info_rx));
  // You can optionally TLX_TX as well.

  char buffer[16384];
  int ret = recv(sock, buffer, 16384);

  // cmsg can be received using recvmsg and a msg_control
  // of type TLS_GET_RECORD_TYPE will be set.

V1 -> V2

* For too-small framing errors, return EBADMSG, to match openssl error
  code semantics.  Docs and commit logs about this also updated.

RFC -> V1

* Refactor 'tx' variable names to drop tx
* Error return codes changed per discussion
* Only call skb_cow_data based on in-place decryption,
  drop unnecessary frag list check.

[1] Recent crypto patchset to remove copies, resulting in optimally
    zero copies vs. userspace's one, vs. previous kernel's two.

https://marc.info/?l=linux-crypto-vger&m=151931242406416&w=2

[2] https://github.com/Mellanox/openssl/commits/tls_rx2

[3] https://github.com/ktls/af_ktls-tool/tree/RX
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotls: Add receive path documentation
Dave Watson [Thu, 22 Mar 2018 17:10:44 +0000 (10:10 -0700)]
tls: Add receive path documentation

Add documentation on rx path setup and cmsg interface.

Signed-off-by: Dave Watson <davejwatson@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotls: RX path for ktls
Dave Watson [Thu, 22 Mar 2018 17:10:35 +0000 (10:10 -0700)]
tls: RX path for ktls

Add rx path for tls software implementation.

recvmsg, splice_read, and poll implemented.

An additional sockopt TLS_RX is added, with the same interface as
TLS_TX.  Either TLX_RX or TLX_TX may be provided separately, or
together (with two different setsockopt calls with appropriate keys).

Control messages are passed via CMSG in a similar way to transmit.
If no cmsg buffer is passed, then only application data records
will be passed to userspace, and EIO is returned for other types of
alerts.

EBADMSG is passed for decryption errors, and EMSGSIZE is passed for
framing too big, and EBADMSG for framing too small (matching openssl
semantics). EINVAL is returned for TLS versions that do not match the
original setsockopt call.  All are unrecoverable.

strparser is used to parse TLS framing.   Decryption is done directly
in to userspace buffers if they are large enough to support it, otherwise
sk_cow_data is called (similar to ipsec), and buffers are decrypted in
place and copied.  splice_read always decrypts in place, since no
buffers are provided to decrypt in to.

sk_poll is overridden, and only returns POLLIN if a full TLS message is
received.  Otherwise we wait for strparser to finish reading a full frame.
Actual decryption is only done during recvmsg or splice_read calls.

Signed-off-by: Dave Watson <davejwatson@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotls: Refactor variable names
Dave Watson [Thu, 22 Mar 2018 17:10:26 +0000 (10:10 -0700)]
tls: Refactor variable names

Several config variables are prefixed with tx, drop the prefix
since these will be used for both tx and rx.

Signed-off-by: Dave Watson <davejwatson@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotls: Pass error code explicitly to tls_err_abort
Dave Watson [Thu, 22 Mar 2018 17:10:15 +0000 (10:10 -0700)]
tls: Pass error code explicitly to tls_err_abort

Pass EBADMSG explicitly to tls_err_abort.  Receive path will
pass additional codes - EMSGSIZE if framing is larger than max
TLS record size, EINVAL if TLS version mismatch.

Signed-off-by: Dave Watson <davejwatson@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotls: Move cipher info to a separate struct
Dave Watson [Thu, 22 Mar 2018 17:10:06 +0000 (10:10 -0700)]
tls: Move cipher info to a separate struct

Separate tx crypto parameters to a separate cipher_context struct.
The same parameters will be used for rx using the same struct.

tls_advance_record_sn is modified to only take the cipher info.

Signed-off-by: Dave Watson <davejwatson@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>