platform/kernel/linux-rpi.git
5 years agosk_buff: add skb extension infrastructure
Florian Westphal [Tue, 18 Dec 2018 16:15:16 +0000 (17:15 +0100)]
sk_buff: add skb extension infrastructure

This adds an optional extension infrastructure, with ispec (xfrm) and
bridge netfilter as first users.
objdiff shows no changes if kernel is built without xfrm and br_netfilter
support.

The third (planned future) user is Multipath TCP which is still
out-of-tree.
MPTCP needs to map logical mptcp sequence numbers to the tcp sequence
numbers used by individual subflows.

This DSS mapping is read/written from tcp option space on receive and
written to tcp option space on transmitted tcp packets that are part of
and MPTCP connection.

Extending skb_shared_info or adding a private data field to skb fclones
doesn't work for incoming skb, so a different DSS propagation method would
be required for the receive side.

mptcp has same requirements as secpath/bridge netfilter:

1. extension memory is released when the sk_buff is free'd.
2. data is shared after cloning an skb (clone inherits extension)
3. adding extension to an skb will COW the extension buffer if needed.

The "MPTCP upstreaming" effort adds SKB_EXT_MPTCP extension to store the
mapping for tx and rx processing.

Two new members are added to sk_buff:
1. 'active_extensions' byte (filling a hole), telling which extensions
   are available for this skb.
   This has two purposes.
   a) avoids the need to initialize the pointer.
   b) allows to "delete" an extension by clearing its bit
   value in ->active_extensions.

   While it would be possible to store the active_extensions byte
   in the extension struct instead of sk_buff, there is one problem
   with this:
    When an extension has to be disabled, we can always clear the
    bit in skb->active_extensions.  But in case it would be stored in the
    extension buffer itself, we might have to COW it first, if
    we are dealing with a cloned skb.  On kmalloc failure we would
    be unable to turn an extension off.

2. extension pointer, located at the end of the sk_buff.
   If the active_extensions byte is 0, the pointer is undefined,
   it is not initialized on skb allocation.

This adds extra code to skb clone and free paths (to deal with
refcount/free of extension area) but this replaces similar code that
manages skb->nf_bridge and skb->sp structs in the followup patches of
the series.

It is possible to add support for extensions that are not preseved on
clones/copies.

To do this, it would be needed to define a bitmask of all extensions that
need copy/cow semantics, and change __skb_ext_copy() to check
->active_extensions & SKB_EXT_PRESERVE_ON_CLONE, then just set
->active_extensions to 0 on the new clone.

This isn't done here because all extensions that get added here
need the copy/cow semantics.

v2:
Allocate entire extension space using kmem_cache.
Upside is that this allows better tracking of used memory,
downside is that we will allocate more space than strictly needed in
most cases (its unlikely that all extensions are active/needed at same
time for same skb).
The allocated memory (except the small extension header) is not cleared,
so no additonal overhead aside from memory usage.

Avoid atomic_dec_and_test operation on skb_ext_put()
by using similar trick as kfree_skbmem() does with fclone_ref:
If recount is 1, there is no concurrent user and we can free right away.

Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonetfilter: avoid using skb->nf_bridge directly
Florian Westphal [Tue, 18 Dec 2018 16:15:15 +0000 (17:15 +0100)]
netfilter: avoid using skb->nf_bridge directly

This pointer is going to be removed soon, so use the existing helpers in
more places to avoid noise when the removal happens.

Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'dpaa2-eth-add-QBMAN-statistics'
David S. Miller [Wed, 19 Dec 2018 18:37:23 +0000 (10:37 -0800)]
Merge branch 'dpaa2-eth-add-QBMAN-statistics'

Ioana Ciornei says:

====================
dpaa2-eth: add QBMAN statistics

This patch set adds ethtool statistics for pending frames/bytes
in Rx/Tx conf FQs and number of buffers in pool.

The first patch adds support for the query APIs in the DPIO driver
while the latter actually exposes the statistics through ethtool.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodpaa2-eth: Add QBMAN related stats
Ioana Radulescu [Tue, 18 Dec 2018 15:23:01 +0000 (15:23 +0000)]
dpaa2-eth: Add QBMAN related stats

Add statistics for pending frames in Rx/Tx conf FQs and
number of buffers in pool. Available through ethtool -S.

Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: Ioana ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agosoc: fsl: dpio: Add BP and FQ query APIs
Roy Pledge [Tue, 18 Dec 2018 15:23:01 +0000 (15:23 +0000)]
soc: fsl: dpio: Add BP and FQ query APIs

Add FQ (Frame Queue) and BP (Buffer Pool) query APIs that
users of QBMan can invoke to see the status of the queues
and pools that they are using.

Signed-off-by: Roy Pledge <roy.pledge@nxp.com>
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: mscc: Fix the VSC 8531/41 Chip Init sequence
Raju Lakkaraju [Tue, 18 Dec 2018 09:27:56 +0000 (14:57 +0530)]
net: phy: mscc: Fix the VSC 8531/41 Chip Init sequence

- Turn on Broadcast writes
- UNH 1.8.1 clear bias for UNH 1000BT distortion
- UNH 1.8.7 optimize pre-emphasis for 100BasTx UNH 100W fix
- Enable Token-ring during 'Coma Mode'

Signed-off-by: Raju Lakkaraju <Raju.Lakkaraju@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'for-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/bluetoot...
David S. Miller [Wed, 19 Dec 2018 16:41:45 +0000 (08:41 -0800)]
Merge branch 'for-upstream' of git://git./linux/kernel/git/bluetooth/bluetooth-next

Johan Hedberg says:

====================
pull request: bluetooth-next 2018-12-19

Here's the main bluetooth-next pull request for 4.21:

 - Multiple fixes & improvements for Broadcom-based controllers
 - New USB ID for an Intel controller
 - Support for new Broadcom controller variants
 - Use DEFINE_SHOW_ATTRIBUTE to simplify debugfs code
 - Eliminate confusing "last event is not cmd complete" warning message
 - Added vendor suspend/resume support for H:5 (3-Wire UART) controllers
 - Various other smaller improvements & fixes

Please let me know if there are any issues pulling. Thanks.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge tag 'mac80211-next-for-davem-2018-12-19' of git://git.kernel.org/pub/scm/linux...
David S. Miller [Wed, 19 Dec 2018 16:36:18 +0000 (08:36 -0800)]
Merge tag 'mac80211-next-for-davem-2018-12-19' of git://git./linux/kernel/git/jberg/mac80211-next

Johannes Berg says:

====================
This time we have too many changes to list, highlights:
 * virt_wifi - wireless control simulation on top of
   another network interface
 * hwsim configurability to test capabilities similar
   to real hardware
 * various mesh improvements
 * various radiotap vendor data fixes in mac80211
 * finally the nl_set_extack_cookie_u64() we talked
   about previously, used for
 * peer measurement APIs, right now only with FTM
   (flight time measurement) for location
 * made nl80211 radio/interface announcements more complete
 * various new HE (802.11ax) things:
   updates, TWT support, ...
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoBluetooth: Fix unnecessary error message for HCI request completion
Johan Hedberg [Tue, 27 Nov 2018 09:37:46 +0000 (11:37 +0200)]
Bluetooth: Fix unnecessary error message for HCI request completion

In case a command which completes in Command Status was sent using the
hci_cmd_send-family of APIs there would be a misleading error in the
hci_get_cmd_complete function, since the code would be trying to fetch
the Command Complete parameters when there are none.

Avoid the misleading error and silently bail out from the function in
case the received event is a command status.

Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Acked-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: hci_bcm: Handle specific unknown packets after firmware loading
Jonathan Bakker [Wed, 19 Dec 2018 11:22:28 +0000 (12:22 +0100)]
Bluetooth: hci_bcm: Handle specific unknown packets after firmware loading

The Broadcom controller on aries S5PV210 boards sends out a couple of
unknown packets after the firmware is loaded.  This will cause
logging of errors such as:
Bluetooth: hci0: Frame reassembly failed (-84)

This is probably also the case with other boards, as there are related
Android userspace patches for custom ROMs such as
https://review.lineageos.org/#/c/LineageOS/android_system_bt/+/142721/
Since this appears to be intended behaviour, treated them as diagnostic
packets.

Note that this is another variant of commit 01d5e44ace8a
("Bluetooth: hci_bcm: Handle empty packet after firmware loading")

Signed-off-by: Jonathan Bakker <xc-racer2@live.ca>
Signed-off-by: Paweł Chmiel <pawel.mikolaj.chmiel@gmail.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: btbcm: Add entry for BCM4329B1 UART bluetooth
Paweł Chmiel [Wed, 19 Dec 2018 11:22:46 +0000 (12:22 +0100)]
Bluetooth: btbcm: Add entry for BCM4329B1 UART bluetooth

This patch adds the device ID for the BCM 4329 combo module used
in the Samsung Aries based phones (Galaxy S and it's variants).

```
[   11.508980] Bluetooth: hci0: BCM: chip id 41
[   11.518975] Bluetooth: hci0: BCM: features 0x04
[   11.550132] Bluetooth: hci0: BCM4329B1
[   11.557046] Bluetooth: hci0: BCM4329B1 (002.002.023) build 0000
[   13.737071] Bluetooth: hci0: BCM4329B1 (002.002.023) build 0744
```

Output from hciconfig

```
hci0:   Type: Primary  Bus: UART
        BD Address: 43:29:B1:55:00:00  ACL MTU: 1021:6  SCO MTU: 64:1
        UP RUNNING
        RX bytes:1675 acl:0 sco:0 events:145 errors:0
        TX bytes:20426 acl:0 sco:0 commands:146 errors:0
        Features: 0xbf 0xfe 0x8f 0xfe 0x9b 0xff 0x79 0x83
        Packet type: DM1 DM3 DM5 DH1 DH3 DH5 HV1 HV2 HV3
        Link policy: RSWITCH SNIFF
        Link mode: SLAVE ACCEPT
        Name: 'aries'
        Class: 0x000000
        Service Classes: Unspecified
        Device Class: Miscellaneous,
        HCI Version: 2.1 (0x4)  Revision: 0x2e8
        LMP Version: 2.1 (0x4)  Subversion: 0x4217
        Manufacturer: Broadcom Corporation (15)
```

Signed-off-by: Paweł Chmiel <pawel.mikolaj.chmiel@gmail.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agomac80211: Properly access radiotap vendor data
Ilan Peer [Sat, 15 Dec 2018 09:03:25 +0000 (11:03 +0200)]
mac80211: Properly access radiotap vendor data

The radiotap vendor data might be placed after some other
radiotap elements, and thus when accessing it, need to access
the correct offset in the skb data. Fix the code accordingly.

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agocfg80211: fix ieee80211_get_vht_max_nss()
Johannes Berg [Sat, 15 Dec 2018 09:03:16 +0000 (11:03 +0200)]
cfg80211: fix ieee80211_get_vht_max_nss()

Fix two bugs in ieee80211_get_vht_max_nss():
 * the spec says we should round down
   (reported by Nissim)
 * there's a double condition, the first one is wrong,
   supp_width == 0 / ext_nss_bw == 2 is valid in 80+80
   (found by smatch)

Fixes: b0aa75f0b1b2 ("ieee80211: add new VHT capability fields/parsing")
Reported-by: Nissim Bendanan <nissimx.bendanan@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: fix radiotap vendor presence bitmap handling
Johannes Berg [Sat, 15 Dec 2018 09:03:12 +0000 (11:03 +0200)]
mac80211: fix radiotap vendor presence bitmap handling

Due to the alignment handling, it actually matters where in the code
we add the 4 bytes for the presence bitmap to the length; the first
field is the timestamp with 8 byte alignment so we need to add the
space for the extra vendor namespace presence bitmap *before* we do
any alignment for the fields.

Move the presence bitmap length accounting to the right place to fix
the alignment for the data properly.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agoBluetooth: btusb: Add support for Intel bluetooth device 8087:0029
Raghuram Hegde [Wed, 19 Dec 2018 06:12:18 +0000 (11:42 +0530)]
Bluetooth: btusb: Add support for Intel bluetooth device 8087:0029

Include the new USB product ID for Intel Bluetooth device 22260
family(CcPeak)

The /sys/kernel/debug/usb/devices portion for this device is:

T:  Bus=01 Lev=01 Prnt=01 Port=02 Cnt=02 Dev#=  2 Spd=12   MxCh= 0
D:  Ver= 2.00 Cls=e0(wlcon) Sub=01 Prot=01 MxPS=64 #Cfgs=  1
P:  Vendor=8087 ProdID=0029 Rev= 0.01
C:* #Ifs= 2 Cfg#= 1 Atr=e0 MxPwr=100mA
I:* If#= 0 Alt= 0 #EPs= 3 Cls=e0(wlcon) Sub=01 Prot=01 Driver=btusb
E:  Ad=81(I) Atr=03(Int.) MxPS=  64 Ivl=1ms
E:  Ad=02(O) Atr=02(Bulk) MxPS=  64 Ivl=0ms
E:  Ad=82(I) Atr=02(Bulk) MxPS=  64 Ivl=0ms
I:* If#= 1 Alt= 0 #EPs= 2 Cls=e0(wlcon) Sub=01 Prot=01 Driver=btusb
E:  Ad=03(O) Atr=01(Isoc) MxPS=   0 Ivl=1ms
E:  Ad=83(I) Atr=01(Isoc) MxPS=   0 Ivl=1ms
I:  If#= 1 Alt= 1 #EPs= 2 Cls=e0(wlcon) Sub=01 Prot=01 Driver=btusb
E:  Ad=03(O) Atr=01(Isoc) MxPS=   9 Ivl=1ms
E:  Ad=83(I) Atr=01(Isoc) MxPS=   9 Ivl=1ms
I:  If#= 1 Alt= 2 #EPs= 2 Cls=e0(wlcon) Sub=01 Prot=01 Driver=btusb
E:  Ad=03(O) Atr=01(Isoc) MxPS=  17 Ivl=1ms
E:  Ad=83(I) Atr=01(Isoc) MxPS=  17 Ivl=1ms
I:  If#= 1 Alt= 3 #EPs= 2 Cls=e0(wlcon) Sub=01 Prot=01 Driver=btusb
E:  Ad=03(O) Atr=01(Isoc) MxPS=  25 Ivl=1ms
E:  Ad=83(I) Atr=01(Isoc) MxPS=  25 Ivl=1ms
I:  If#= 1 Alt= 4 #EPs= 2 Cls=e0(wlcon) Sub=01 Prot=01 Driver=btusb
E:  Ad=03(O) Atr=01(Isoc) MxPS=  33 Ivl=1ms
E:  Ad=83(I) Atr=01(Isoc) MxPS=  33 Ivl=1ms
I:  If#= 1 Alt= 5 #EPs= 2 Cls=e0(wlcon) Sub=01 Prot=01 Driver=btusb
E:  Ad=03(O) Atr=01(Isoc) MxPS=  49 Ivl=1ms
E:  Ad=83(I) Atr=01(Isoc) MxPS=  49 Ivl=1ms
I:  If#= 1 Alt= 6 #EPs= 2 Cls=e0(wlcon) Sub=01 Prot=01 Driver=btusb
E:  Ad=03(O) Atr=01(Isoc) MxPS=  63 Ivl=1ms
E:  Ad=83(I) Atr=01(Isoc) MxPS=  63 Ivl=1ms

Signed-off-by: Raghuram Hegde <raghuram.hegde@intel.com>
Signed-off-by: Chethan T N <chethan.tumkur.narayan@intel.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agonet: ethernet: don't set phylib state CHANGELINK in drivers
Heiner Kallweit [Tue, 18 Dec 2018 23:43:58 +0000 (00:43 +0100)]
net: ethernet: don't set phylib state CHANGELINK in drivers

After phy_start() phylib takes care of all needed actions, including
aneg settings and checking link state. There's no need to set state
PHY_CHANGELINK in drivers.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMAINTAINERS: Add a maintainer for Microsemi switches
Alexandre Belloni [Tue, 18 Dec 2018 14:25:39 +0000 (15:25 +0100)]
MAINTAINERS: Add a maintainer for Microsemi switches

Microsemi has been bought by Microchip and Microchip is supporting those
switches.

Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Acked-by: Woojung Huh <Woojung.Huh@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agotipc: handle broadcast NAME_DISTRIBUTOR packet when receiving it
Zhenbo Gao [Tue, 18 Dec 2018 09:43:52 +0000 (17:43 +0800)]
tipc: handle broadcast NAME_DISTRIBUTOR packet when receiving it

NAME_DISTRIBUTOR messages are transmitted through unicast link on TIPC
2.0, by contrast, the messages are delivered through broadcast link on
TIPC 1.7. But at present, NAME_DISTRIBUTOR messages received by
broadcast link cannot be handled in tipc_rcv() until an unicast message
arrives, which may lead to a significant delay to update name table.

To avoid this delay, we will also deal with broadcast NAME_DISTRIBUTOR
message on broadcast receive path.

Signed-off-by: Zhenbo Gao <zhenbo.gao@windriver.com>
Reviewed-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: remove unused code in phy_probe
Heiner Kallweit [Mon, 17 Dec 2018 22:43:25 +0000 (23:43 +0100)]
net: phy: remove unused code in phy_probe

3c1bcc8614db ("net: ethernet: Convert phydev advertize and supported
from u32 to link mode") left some unused code in phy_probe(), remove it.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge tag 'mlx5-uplink-rep-2018-12-15' of git://git.kernel.org/pub/scm/linux/kernel...
David S. Miller [Wed, 19 Dec 2018 00:44:45 +0000 (16:44 -0800)]
Merge tag 'mlx5-uplink-rep-2018-12-15' of git://git./linux/kernel/git/saeed/linux

Saeed Mahameed:

====================
mlx5-uplink-rep-2018-12-15

Or Gerlitz says:

This series is essentially a cleanup to align with the rest of the NIC
switchdev drivers and make us
more robust and clear/n: currently the PF netdev serves as the mlx5
e-switch uplink netdev
representor when going into switchdev mode and back as plain NIC
netdev when going out.
This causes some irregularities and misc troubles.

Move to use dedicated uplink rep, as we have for the VF vports.

The uplink rep netdev does has sysfs link and supports the sriov vf
mac ndo, these two are in
use by libvirt and other orchestrators, It also has richer ethtool
support to allow controlling the
port link & mtu along with supporting dcb and plugging into the mlx5
lag logic.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: bridge: remove unneeded variable 'err'
YueHaibing [Mon, 17 Dec 2018 09:46:23 +0000 (17:46 +0800)]
net: bridge: remove unneeded variable 'err'

function br_multicast_toggle now always return 0,
so the variable 'err' is unneeded.
Also cleanup dead branch in br_changelink.

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Acked-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoqlcnic: remove set but not used variables 'op, cmd_op'
YueHaibing [Mon, 17 Dec 2018 09:27:16 +0000 (17:27 +0800)]
qlcnic: remove set but not used variables 'op, cmd_op'

Fixes gcc '-Wunused-but-set-variable' warning:

drivers/net/ethernet/qlogic/qlcnic/qlcnic_sriov_pf.c:1070:5: warning:
 variable 'op' set but not used [-Wunused-but-set-variable]
drivers/net/ethernet/qlogic/qlcnic/qlcnic_sriov_pf.c:1342:5: warning:
 variable 'cmd_op' set but not used [-Wunused-but-set-variable]

'op' never used since introduction in commit 7cb03b2347d5 ("qlcnic:
Support VF-PF communication channel commands.")
'cmd_op' not used since commit 6226204bcf20 ("qlcnic: Fix operation
type and command type.")

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoBluetooth: hci_h5: Turn off RTL8723BS on suspend, reprobe on resume
Hans de Goede [Tue, 30 Oct 2018 13:17:23 +0000 (14:17 +0100)]
Bluetooth: hci_h5: Turn off RTL8723BS on suspend, reprobe on resume

On many devices the RTL8723BS device gets reset during suspend/resume,
causing it to lose its firmware and all state.

Testing has shown it drops back to communicating at 115200 bps and sends
sync-request packages, indicating it has been fully reset.

This commit fixes this by queueing a reprobe on resume.

This mirrors how USB RTL BT devices, which have the same problem, are
handled in the btusb driver, there we set the USB_QUIRK_RESET_RESUME for
all RTL devices, which also causes a reprobe on resume. The only difference
is that here we need to do the reprobe ourselves.

Since we are doing a full reprobe on resume now, we can also turn off the
device on suspend to save power while suspended.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: hci_h5: Add suspend / resume ops
Hans de Goede [Tue, 30 Oct 2018 13:17:22 +0000 (14:17 +0100)]
Bluetooth: hci_h5: Add suspend / resume ops

Add support for vendor specific suspend / resume callbacks.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: hci_intel: clean an indentation issue, remove extraneous spaces
Colin Ian King [Tue, 30 Oct 2018 11:34:19 +0000 (11:34 +0000)]
Bluetooth: hci_intel: clean an indentation issue, remove extraneous spaces

Trivial fix to clean up an indentation issue, remove spaces

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: clean an indentation issue, remove extraneous space
Colin Ian King [Fri, 9 Nov 2018 13:27:36 +0000 (13:27 +0000)]
Bluetooth: clean an indentation issue, remove extraneous space

Trivial fix to clean up an indentation issue

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: hci_serdev: Remove setting of HCI_QUIRK_RESET_ON_CLOSE.
Balakrishna Godavarthi [Thu, 22 Nov 2018 12:50:47 +0000 (18:20 +0530)]
Bluetooth: hci_serdev: Remove setting of HCI_QUIRK_RESET_ON_CLOSE.

HCI_QUIRK_RESET_ON_CLOSE quirk is required for BT v1.0 based devices,
to send a reset command to the chip during hci device close. Serdev
architecture is used for the latest BT chips, which doesn't require to
send the reset command during close. If still chips required reset
command during close, it would be better enabling it in the vendor
probes or in proto setup.

Signed-off-by: Balakrishna Godavarthi <bgodavar@codeaurora.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: Remove unnecessary smp_mb__after_atomic() barriers
Andrea Parri [Tue, 27 Nov 2018 11:22:25 +0000 (12:22 +0100)]
Bluetooth: Remove unnecessary smp_mb__after_atomic() barriers

The barriers are redundant because atomic_test_and_clear_bit() already
provides the required full ordering for the cases in question (that is,
when the bit is cleared).

Signed-off-by: Andrea Parri <andrea.parri@amarulasolutions.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: btbcm: Add default address for BCM43430A0
Chen-Yu Tsai [Mon, 17 Dec 2018 04:04:47 +0000 (12:04 +0800)]
Bluetooth: btbcm: Add default address for BCM43430A0

The BCM43430A0 has the default MAC address 43:43:A0:12:1F:AC if none
is given. This address was found when enabling Bluetooth on a bunch of
boards with the AMPAK AP6210 module, all sharing the same address. It
also contains the sequence 4343A0, which is suspicious as that is also
the name the chip identifies itself as.

Add this to the list of default MAC addresses and leave it to the user
to configure a valid one.

Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: hci_bcm: Add compatible string for BCM4330
Chen-Yu Tsai [Mon, 17 Dec 2018 04:04:46 +0000 (12:04 +0800)]
Bluetooth: hci_bcm: Add compatible string for BCM4330

The BCM4330 chip is a 802.11 a/b/g/n + Bluetooth 4.0 + HS controller.
This patch adds a compatible string match to the serdev driver for the
Bluetooth part of the chip.

Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: hci_bcm: Add BCM20702A1 variant
Maxime Ripard [Mon, 17 Dec 2018 04:04:45 +0000 (12:04 +0800)]
Bluetooth: hci_bcm: Add BCM20702A1 variant

The BCM20702A1 chip is a single-chip Bluetooth 4.0 controller and
transceiver. It is found in the AMPAK AP6210 WiFi+BT package.

Signed-off-by: Maxime Ripard <maxime.ripard@bootlin.com>
Tested-by: Ondrej Jirman <megous@megous.com>
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: hci_bcm: Wait for device to come out of reset after power on
Chen-Yu Tsai [Mon, 17 Dec 2018 04:04:44 +0000 (12:04 +0800)]
Bluetooth: hci_bcm: Wait for device to come out of reset after power on

The datasheets for BCM20702 and BCM43438 both have power up time
sequence graphs, however they are slightly different. Both chips
also have an internal power-on-reset, which holds the chip in reset
for a short time after the regulators are enabled.

For the BCM20702, the time period from when the regulators are enabled,
until the chip settles and comes out of sleep state, is 6564 ~ 8171 us.

For the BCM43438, the graph only shows the time period from when the
regulators are enabled until the chip responds by driving the host's
CTS line low, assuming the host has already driven its RTS line low.
This is shown to be 6.5 sleep cycles, with the sleep clock at 32.768
kHz. This is around 2 ms.

Wait a full 10 ms after the regulators are enabled to account for signal
rising times.

Tested-by: Ondrej Jirman <megous@megous.com>
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: hci_bcm: Add support for regulator supplies
Chen-Yu Tsai [Mon, 17 Dec 2018 04:04:43 +0000 (12:04 +0800)]
Bluetooth: hci_bcm: Add support for regulator supplies

The Broadcom Bluetooth chips have two power inputs, VBAT and VDDIO.
The former provides overall power for the chip, while the latter powers
the I/O pins and buffers.

Model these two as regulator supplies, and let the driver manage them
in the same way as it does the clock supply.

Tested-by: Ondrej Jirman <megous@megous.com>
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: hci_bcm: Add support for LPO clock
Chen-Yu Tsai [Mon, 17 Dec 2018 04:04:42 +0000 (12:04 +0800)]
Bluetooth: hci_bcm: Add support for LPO clock

The Broadcom Bluetooth controllers support a secondary LPO clock at
32.768 kHz. This external clock provides low power timing, and also
a way to detect the frequency of the main reference clock. On many
designs without NVRAM and a non-default reference clock, this must
be used or the controller will not function correctly.

Tested-by: Ondrej Jirman <megous@megous.com>
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: hci_bcm: Use "txco" and "extclk" to get clock reference
Chen-Yu Tsai [Mon, 17 Dec 2018 04:04:41 +0000 (12:04 +0800)]
Bluetooth: hci_bcm: Use "txco" and "extclk" to get clock reference

Originally the device tree binding only specified one clock reference,
with the name "extclk". The driver simply retrieves the clock without
bothering to specify a name.

Since we added a second clock to the binding, we need to fetch the
clocks by name now. First we try the new name "txco", then fall back
to the old name "extclk", and finally try retrieving a clock without
using any name, to cover any instances where a bad device tree or
firmware worked by accident.

In the last case, we should take care that we don't get the same
clock twice when we add support for the "lpo" clock.

Tested-by: Ondrej Jirman <megous@megous.com>
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: hci_bcm: Simplify clk_get error handling
Chen-Yu Tsai [Mon, 17 Dec 2018 04:04:40 +0000 (12:04 +0800)]
Bluetooth: hci_bcm: Simplify clk_get error handling

The driver currently checks the clk pointer for an error condition, as
returned by clk_get, before every invocation of the clk consumer API.
This is redundant if the goal is simply to ignore the errors, thereby
making the clk optional. The clk consumer API already checks if the
pointer is NULL or not.

Simplify the code a bit by assigning NULL to the clk pointer if the
error condition is one we want to ignore, which is every error except
deferred probing.

Tested-by: Ondrej Jirman <megous@megous.com>
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: hci_bcm: Handle deferred probing for the clock supply
Chen-Yu Tsai [Mon, 17 Dec 2018 04:04:39 +0000 (12:04 +0800)]
Bluetooth: hci_bcm: Handle deferred probing for the clock supply

On some systems that actually have the bluetooth controller wired up
with an extra clock signal, it's possible the bluetooth controller
probes before the clock provider. clk_get would return a defer probe
error, which was not handled by this driver.

Handle this properly, so that these systems can work reliably.

Tested-by: Ondrej Jirman <megous@megous.com>
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agodt-bindings: net: broadcom-bluetooth: Add BCM4330 compatible string
Chen-Yu Tsai [Mon, 17 Dec 2018 04:04:38 +0000 (12:04 +0800)]
dt-bindings: net: broadcom-bluetooth: Add BCM4330 compatible string

The BCM4330 is a 802.11 a/b/g/n WiFi + Bluetooth 4.0 chip from Broadcom.
It is found in the Ampak AP6330 WiFi+BT module. The partiular one I have
identifies as BCM4330B1 for Bluetooth and BCM4330/4 for WiFi.

It is unclear if the AP6330 module uses this revision of the BCM4330, or
if there are multiple revisions. The module does not have revision
markings. This patch elects to use just BCM4330 for the compatible
string.

Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agodt-bindings: net: broadcom-bluetooth: Add BCM20702A1 compatible string
Chen-Yu Tsai [Mon, 17 Dec 2018 04:04:37 +0000 (12:04 +0800)]
dt-bindings: net: broadcom-bluetooth: Add BCM20702A1 compatible string

The BCM20702A1 is a Bluetooth 4.0 chip from Broadcom. It is found in the
Ampak AP6210 WiFi+BT module, identified from the read verbose config info
command response. However the Bluetooth firmware provided by vendors uses
the name BCM20710. This patch elects to use the chip ID returned by the
chip for the compatible string.

Reviewed-by: Rob Herring <robh@kernel.org>
Tested-by: Ondrej Jirman <megous@megous.com>
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agodt-bindings: net: broadcom-bluetooth: Add VBAT and VDDIO supplies
Chen-Yu Tsai [Mon, 17 Dec 2018 04:04:36 +0000 (12:04 +0800)]
dt-bindings: net: broadcom-bluetooth: Add VBAT and VDDIO supplies

The Broadcom Bluetooth chips have two power inputs, VBAT and VDDIO.
The former provides overall power for the chip, while the latter powers
the I/O pins and buffers.

This patch adds properties for the two so we can describe the power
supply relationships.

Reviewed-by: Rob Herring <robh@kernel.org>
Tested-by: Ondrej Jirman <megous@megous.com>
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agodt-bindings: net: broadcom-bluetooth: Fix external clock names
Chen-Yu Tsai [Mon, 17 Dec 2018 04:04:35 +0000 (12:04 +0800)]
dt-bindings: net: broadcom-bluetooth: Fix external clock names

The Broadcom Bluetooth controllers can take up to two external clocks:
an external frequency reference, substituting the main crystal, and a
LPO clock at 32.768 kHz substituting the internal LPO clock.

In particular, the external LPO clock must be used when the controller
does not have NVRAM connected, and the main reference frequency is not
the default 20 MHz. This is described in detail in the datasheet.

The original "extclk" clock name is ambiguous as to which of these it
refers to, and some designs might even require both.

This patch deprecates the existing name, and adds "txco" and "lpo".

Tested-by: Ondrej Jirman <megous@megous.com>
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agoBluetooth: Change to use DEFINE_SHOW_ATTRIBUTE macro
Yangtao Li [Mon, 5 Nov 2018 14:56:19 +0000 (09:56 -0500)]
Bluetooth: Change to use DEFINE_SHOW_ATTRIBUTE macro

Use DEFINE_SHOW_ATTRIBUTE macro to simplify the code.

Signed-off-by: Yangtao Li <tiny.windzz@gmail.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years ago6lowpan: convert to DEFINE_SHOW_ATTRIBUTE
Yangtao Li [Sat, 15 Dec 2018 07:22:17 +0000 (02:22 -0500)]
6lowpan: convert to DEFINE_SHOW_ATTRIBUTE

Use DEFINE_SHOW_ATTRIBUTE macro to simplify the code.

Signed-off-by: Yangtao Li <tiny.windzz@gmail.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
5 years agotest_rhashtable: remove semaphore usage
Arnd Bergmann [Sun, 16 Dec 2018 19:48:21 +0000 (20:48 +0100)]
test_rhashtable: remove semaphore usage

This is one of only two files that initialize a semaphore to a negative
value. We don't really need the two semaphores here at all, but can do
the same thing in more conventional and more effient way, by using a
single waitqueue and an atomic thread counter.

This gets us a little bit closer to eliminating classic semaphores from
the kernel. It also fixes a corner case where we fail to continue after
one of the threads fails to start up.

An alternative would be to use a split kthread_create()+wake_up_process()
and completely eliminate the separate synchronization.

Acked-by: Phil Sutter <phil@nwl.cc>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: print stack trace in phy_error
Heiner Kallweit [Sun, 16 Dec 2018 18:18:26 +0000 (19:18 +0100)]
net: phy: print stack trace in phy_error

So far phy_error() silently stops the PHY state machine. If the network
driver doesn't inform about a  MDIO error then the user may wonder why
his network is down. Let's print the stack trace to facilitate search
for the root cause of the error.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: phy: improve phy state checking
Heiner Kallweit [Sun, 16 Dec 2018 17:30:14 +0000 (18:30 +0100)]
net: phy: improve phy state checking

Add helpers phy_is_started() and __phy_is_started() to avoid open-coded
checks whether PHY has been started. To make the check easier move
PHY_HALTED before PHY_UP in enum phy_state. Further improvements:

phy_start_aneg():
Return -EBUSY and print warning if function is called from a non-started
state (DOWN, READY, HALTED). Better check because function is exported
and drivers may use it incorrectly.

phy_interrupt():
Return IRQ_NONE also if state is DOWN or READY. We should never receive
an interrupt in one of these states, but better play safe.

phy_stop():
Just return and print a warning if PHY is in a non-started state.
This warning should help to identify drivers with unbalanced calls to
phy_start() / phy_stop().

phy_state_machine():
Schedule state machine run only if PHY is in a started state.
E.g. if state is READY we don't need the state machine, it will be
started by phy_start().

v2:
- don't use __func__ within phy_warn_state
v3:
- use WARN() instead of printing error message to facilitate debugging

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: fec: remove workaround to restart phylib state machine on MDIO timeout
Heiner Kallweit [Sun, 16 Dec 2018 14:00:40 +0000 (15:00 +0100)]
net: fec: remove workaround to restart phylib state machine on MDIO timeout

There's a workaround to restart the phylib state machine in case of a
MDIO access timeout. Seems it was introduced to deal with the
consequences of a too small MDIO timeout. See also commit message of
c3b084c24c8a ("net: fec: Adjust ENET MDIO timeouts") which increased
the timeout value later. Due to the later timeout value fix it seems
to be safe to remove the workaround.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobonding: fix indentation issues, remove extra spaces
Colin Ian King [Sun, 16 Dec 2018 13:33:15 +0000 (13:33 +0000)]
bonding: fix indentation issues, remove extra spaces

There are two statements that are indented too much by one space each,
fix these.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'hns3-next'
David S. Miller [Tue, 18 Dec 2018 20:01:02 +0000 (12:01 -0800)]
Merge branch 'hns3-next'

Peng Li says:

====================
net: hns3: code optimizations & bugfixes for HNS3 driver

This patchset includes bugfixes and code optimizations for the HNS3
ethernet controller driver
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: fix a SSU buffer checking bug
Yunsheng Lin [Tue, 18 Dec 2018 11:37:59 +0000 (19:37 +0800)]
net: hns3: fix a SSU buffer checking bug

When caculating the SSU buffer, it first allocate tx and
rx private buffer, then the remaining buffer is for rx
shared buffer. The remaining buffer size should be at
least bigger than or equal to the shared_std, which is the
minimum shared buffer size required by the driver, but
currently if the remaining buffer size is equal to the
shared_std, it returns failure, which causes SSU buffer
allocation failure problem.

This patch fixes this problem by rounding up shared_std before
checking the the remaining buffer size bigger than or equal to
the shared_std.

Fixes: 46a3df9f9718 ("net: hns3: Add HNS3 Acceleration Engine & Compatibility Layer Support")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: aligning buffer size in SSU to 256 bytes
Yunsheng Lin [Tue, 18 Dec 2018 11:37:58 +0000 (19:37 +0800)]
net: hns3: aligning buffer size in SSU to 256 bytes

The hardware expects the buffer size set to SSU is aligned to
256 bytes, this patch aligns the buffer size to 256 byte using
roundup or rounddown function.

Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: getting tx and dv buffer size through firmware
Yunsheng Lin [Tue, 18 Dec 2018 11:37:57 +0000 (19:37 +0800)]
net: hns3: getting tx and dv buffer size through firmware

This patch adds support of getting tx and dv buffer size through
firmware, because different version of hardware requires different
size of tx and dv buffer.

This patch also add dv_buf_size to tc' private buffer size even if
pfc is not enable for the tc.

Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: synchronize speed and duplex from phy when phy link up
Peng Li [Tue, 18 Dec 2018 11:37:56 +0000 (19:37 +0800)]
net: hns3: synchronize speed and duplex from phy when phy link up

Driver calls phy_connect_direct and registers hclge_mac_adjust_link
to synchronize mac speed and duplex from phy. It is better to
synchronize mac speed and duplex from phy when phy link up.

Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: remove 1000M/half support of phy
Fuyun Liang [Tue, 18 Dec 2018 11:37:55 +0000 (19:37 +0800)]
net: hns3: remove 1000M/half support of phy

Our phy does not support 1000M/half, this patch removes 1000M/half from
PHY_SUPPORTED_FEATURES.

Signed-off-by: Fuyun Liang <liangfuyun1@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: update coalesce param per second
Peng Li [Tue, 18 Dec 2018 11:37:54 +0000 (19:37 +0800)]
net: hns3: update coalesce param per second

coalesce param updates every 100 napi times, it may update a little
late if ping test after a high rate flow, may over napi poll is called
100 times as ping test sends packets every second.

This patch updates coalesce param every second, instead with every
100 napi times. It can not update the param 100% in time, but the
lag time is very short.

Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: fix incomplete uninitialization of IRQ in the hns3_nic_uninit_vector_data()
Huazhong Tan [Tue, 18 Dec 2018 11:37:53 +0000 (19:37 +0800)]
net: hns3: fix incomplete uninitialization of IRQ in the hns3_nic_uninit_vector_data()

In the hns3_nic_uninit_vector_data(), the procedure of uninitializing
the tqp_vector's IRQ has not set affinity_notify to NULL and changes
its init flag. This patch fixes it. And for simplificaton, local
variable tqp_vector is used instead of priv->tqp_vector[i].

Fixes: 424eb834a9be ("net: hns3: Unified HNS3 {VF|PF} Ethernet Driver for hip08 SoC")
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: remove unnecessary configuration recapture while resetting
Huazhong Tan [Tue, 18 Dec 2018 11:37:52 +0000 (19:37 +0800)]
net: hns3: remove unnecessary configuration recapture while resetting

When doing reset, it is unnecessary to get the hardware's default
configuration again, otherwise, the user's configuration will be
overwritten.

Fixes: 4ed340ab8f49 ("net: hns3: Add reset process in hclge_main")
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: update some variables while hclge_reset()/hclgevf_reset() done
Huazhong Tan [Tue, 18 Dec 2018 11:37:51 +0000 (19:37 +0800)]
net: hns3: update some variables while hclge_reset()/hclgevf_reset() done

When hclge_reset() completes successfully, it should update the
last_reset_time, set reset_fail_cnt to 0, and set reset_type of
hnae3_ae_dev to HNAE3_NONE_RESET.

Also when hclgevf_reset() completes successfully, it should update
the last_reset_time, and set reset_type of hnae3_ae_dev to
HNAE3_NONE_RESET.

Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: fix napi_disable not return problem
Huazhong Tan [Tue, 18 Dec 2018 11:37:50 +0000 (19:37 +0800)]
net: hns3: fix napi_disable not return problem

While doing DOWN, the calling of napi_disable() may not return, since the
napi_complete() in the hns3_nic_common_poll() will never be called when
HNS3_NIC_STATE_DOWN is set. So we need to call napi_complete() before
checking HNS3_NIC_STETE_DOWN.

Fixes: ff0699e04b97 ("net: hns3: stop napi polling when HNS3_NIC_STATE_DOWN is set")
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: uninitialize pci in the hclgevf_uninit
Huazhong Tan [Tue, 18 Dec 2018 11:37:49 +0000 (19:37 +0800)]
net: hns3: uninitialize pci in the hclgevf_uninit

In the hclgevf_pci_reset(), it only uninitialize and initialize
the msi, so if the initialization fails, hclgevf_uninit_hdev()
does not need to uninitialize the msi, but needs to uninitialize
the pci, otherwise it will cause pci resource not free.

Fixes: 862d969a3a4d ("net: hns3: do VF's pci re-initialization while PF doing FLR")
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hns3: fix error handling int the hns3_get_vector_ring_chain
Huazhong Tan [Tue, 18 Dec 2018 11:37:48 +0000 (19:37 +0800)]
net: hns3: fix error handling int the hns3_get_vector_ring_chain

When hns3_get_vector_ring_chain() failed in the
hns3_nic_init_vector_data(), it should do the error handling instead
of return directly.

Also, cur_chain should be freed instead of chain and head->next should
be set to NULL in error handling of hns3_get_vector_ring_chain.

This patch fixes them.

Fixes: 73b907a083b8 ("net: hns3: bugfix for buffer not free problem during resetting")
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec...
David S. Miller [Tue, 18 Dec 2018 16:49:48 +0000 (08:49 -0800)]
Merge branch 'master' of git://git./linux/kernel/git/klassert/ipsec-next

Steffen Klassert says:

====================
pull request (net-next): ipsec-next 2018-12-18

1) Add xfrm policy selftest scripts.
   From Florian Westphal.

2) Split inexact policies into four different search list
   classes and use the rbtree infrastructure to store/lookup
   the policies. This is to improve the policy lookup
   performance after the flowcache removal.
   Patches from Florian Westphal.

3) Various coding style fixes, from Colin Ian King.

4) Fix policy lookup logic after adding the inexact policy
   search tree infrastructure. From Florian Westphal.

5) Remove a useless remove BUG_ON from xfrm6_dst_ifdown.
   From Li RongQing.

6) Use the correct policy direction for lookups on hash
   rebuilding. From Florian Westphal.

Please pull or let me know if there are problems.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomac80211: do not advertise HE cap IE if HE disabled
Shaul Triebitz [Sat, 15 Dec 2018 09:03:20 +0000 (11:03 +0200)]
mac80211: do not advertise HE cap IE if HE disabled

When disabling HE due to the lack of HT/VHT, do it
at an earlier stage to avoid advertising HE capabilities IE.
Also, at this point, no need to check if AP supports HE, since
it is already checked earlier (in ieee80211_prep_channel).

Signed-off-by: Shaul Triebitz <shaul.triebitz@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: set STA flag DISABLE_HE if HE is not supported
Shaul Triebitz [Sat, 15 Dec 2018 09:03:19 +0000 (11:03 +0200)]
mac80211: set STA flag DISABLE_HE if HE is not supported

Up until now, the IEEE80211_STA_DISABLE_HE flag was set only based
on whether the AP has advertised HE capabilities.
This flag should be set also if STA does not support HE
(regardless of the AP support).

Signed-off-by: Shaul Triebitz <shaul.triebitz@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: update driver when MU EDCA params change
Shaul Triebitz [Sat, 15 Dec 2018 09:03:15 +0000 (11:03 +0200)]
mac80211: update driver when MU EDCA params change

Similar to WMM IE, if MU_EDCA IE parameters changed (or ceased to exist)
tell the Driver about it.

Signed-off-by: Shaul Triebitz <shaul.triebitz@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: update HE operation fields to D3.0
Shaul Triebitz [Sat, 15 Dec 2018 09:03:05 +0000 (11:03 +0200)]
mac80211: update HE operation fields to D3.0

HE Operation element has changed in 11ax D3.0.  Update the fields
accordingly.

Signed-off-by: Shaul Triebitz <shaul.triebitz@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: propagate the support for TWT to the driver
Emmanuel Grumbach [Sat, 15 Dec 2018 09:03:04 +0000 (11:03 +0200)]
mac80211: propagate the support for TWT to the driver

TWT is a feature that was added in 11ah and enhanced in
11ax. There are two bits that need to be set if we want
to use the feature in 11ax: one in the HE Capability IE
and one in the Extended Capability IE. This is because
of backward compatibility between 11ah and 11ax.

In order to simplify the flow for the low level driver
in managed mode, aggregate the two bits and add a boolean
that tells whether TWT is supported or not, but only if
11ax is supported.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agoieee80211: add bits for TWT in Extended Capabilities IE
Emmanuel Grumbach [Sat, 15 Dec 2018 09:03:03 +0000 (11:03 +0200)]
ieee80211: add bits for TWT in Extended Capabilities IE

These bits are defined in ieee802.11ax to advertise support
for TWT in addition to the bits in the HE IE.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agonl80211: Add support to notify radar event info received from STA
Sriram R [Tue, 4 Dec 2018 12:16:52 +0000 (17:46 +0530)]
nl80211: Add support to notify radar event info received from STA

Currently radar detection and corresponding channel switch is handled
at the AP device. STA ignores these detected radar events since the
radar signal can be seen mostly by the AP as well. But in scenarios where
a radar signal is seen only at STA, notifying this event to the AP which
can trigger a channel switch can be useful.
Stations can report such radar events autonomously through Spectrum
management (Measurement Report) action frame to its AP. The userspace on
processing the report can notify the kernel with the use of the added
NL80211_CMD_NOTIFY_RADAR to indicate the detected event and inturn adding
the reported channel to NOL.

Signed-off-by: Sriram R <srirrama@codeaurora.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: don't build AMSDU from GSO packets
Sara Sharon [Sat, 15 Dec 2018 09:03:07 +0000 (11:03 +0200)]
mac80211: don't build AMSDU from GSO packets

If we build AMSDU from GSO packets, it can lead to
bad results if anyone tries to call skb_gso_segment
on the packets.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: document RCU requirements for ieee80211_tx_dequeue()
Johannes Berg [Sat, 15 Dec 2018 09:03:08 +0000 (11:03 +0200)]
mac80211: document RCU requirements for ieee80211_tx_dequeue()

In the iwlwifi conversion, we sometimes call this from outside
of the wake_tx_queue() method, and in those cases must be in an
RCU critical section. Document this requirement.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: remove superfluous NULL check
Johannes Berg [Sat, 15 Dec 2018 09:03:09 +0000 (11:03 +0200)]
mac80211: remove superfluous NULL check

At the place where this code lives now, the skb can never be
NULL, so we can remove the pointless NULL check.

It seems to exist because this code was moved around a few times
and originally came from a place where it could in fact be NULL.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: never pass NULL params to ieee80211_if_add()
Johannes Berg [Sat, 15 Dec 2018 09:03:11 +0000 (11:03 +0200)]
mac80211: never pass NULL params to ieee80211_if_add()

This isn't really a problem now, but it means that the function
has a few NULL checks that are only relevant when coming from
the initial interface added in mac80211, and that's confusing.
Just pass non-NULL (but equivalently empty) in that case and
remove all the NULL checks.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: Properly handle SKB with radiotap only
Ilan Peer [Sat, 15 Dec 2018 09:03:17 +0000 (11:03 +0200)]
mac80211: Properly handle SKB with radiotap only

The monitor interface Rx handling of SKBs that contain only
radiotap information was buggy as it tried to access the
SKB assuming it contains a frame.

To fix this, check the RX_FLAG_NO_PSDU flag in the Rx status
(indicting that the SKB contains only radiotap information),
and do not perform data path specific processing when the flag
is set.

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agocfg80211: add some missing fall through annotations
Luca Coelho [Sat, 15 Dec 2018 09:03:21 +0000 (11:03 +0200)]
cfg80211: add some missing fall through annotations

There are talks about enabling -Wimplicit-fallthrough warnings in the
mainline and it is already enabled in linux-next.  Add all the
missing annotations to prevent warnings when this happens.

And in one case, remove the extra text from the annotation so that the
compiler recognizes it.

Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agocfg80211: clarify LCI/civic location documentation
Johannes Berg [Sat, 15 Dec 2018 09:03:23 +0000 (11:03 +0200)]
cfg80211: clarify LCI/civic location documentation

The older code and current userspace assumed that this data
is the content of the Measurement Report element, starting
with the Measurement Token. Clarify this in the documentation.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211: ftm responder: remove pointless defensive coding
Johannes Berg [Sat, 15 Dec 2018 09:03:24 +0000 (11:03 +0200)]
mac80211: ftm responder: remove pointless defensive coding

The pointer and corresponding length is always set in pairs
in cfg80211, so no need to have this strange defensive check
that also confuses static checkers. Clean it up.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agowireless: FTM: fix kernel-doc "cannot understand" warnings
Randy Dunlap [Thu, 6 Dec 2018 20:58:30 +0000 (12:58 -0800)]
wireless: FTM: fix kernel-doc "cannot understand" warnings

Fix kernel-doc warnings in FTM due to missing "struct" keyword.

Fixes 109 warnings from <net/cfg80211.h>:
../include/net/cfg80211.h:2838: warning: cannot understand function prototype: 'struct cfg80211_ftm_responder_stats '

and fixes 88 warnings from <net/mac80211.h>:
../include/net/mac80211.h:477: warning: cannot understand function prototype: 'struct ieee80211_ftm_responder_params '

Fixes: 81e54d08d9d8 ("cfg80211: support FTM responder configuration/statistics")
Fixes: bc847970f432 ("mac80211: support FTM responder configuration/statistics")

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Pradeep Kumar Chitrapu <pradeepc@codeaurora.org>
Cc: Johannes Berg <johannes.berg@intel.com>
Cc: David Spinadel <david.spinadel@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agorfkill: gpio: Remove unused include
Linus Walleij [Wed, 12 Dec 2018 09:19:50 +0000 (10:19 +0100)]
rfkill: gpio: Remove unused include

The legacy <linux/gpio.h> header is no longer in use by the
rfkill driver, so drop this include.

Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agomac80211_hwsim: fix overwriting of if_combination
James Prestwood [Wed, 12 Dec 2018 00:52:14 +0000 (16:52 -0800)]
mac80211_hwsim: fix overwriting of if_combination

Moved setting if_combination.num_different_channels/radar_detect_widths
into an else after use_chanctx. In the case of use_chanctx, these two
settings were getting overwritten.

Signed-off-by: James Prestwood <james.prestwood@linux.intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
5 years agonfp: flower: fix cb_ident duplicate in indirect block register
John Hurley [Tue, 18 Dec 2018 03:18:39 +0000 (19:18 -0800)]
nfp: flower: fix cb_ident duplicate in indirect block register

Previously the identifier used for indirect block callback registry and
for block rule cb registry (when done via indirect blocks) was the pointer
to the netdev we were interested in receiving updates on. This worked fine
if a single app existed that registered one callback per netdev of
interest. However, if multiple cards are in place and, in turn, multiple
apps, then each app may register the same callback with the same
identifier to both the netdev's indirect block cb list and to a block's cb
list. This can lead to EEXIST errors and/or incorrect cb deletions.

Prevent this conflict by using the app pointer as the identifier for
netdev indirect block cb registry, allowing each app to register a unique
callback per netdev. For block cb registry, the same app may register
multiple cbs to the same block if using TC shared blocks. Instead of the
app, use the pointer to the allocated cb_priv data as the identifier here.
This means that there can be a unique block callback for each app/netdev
combo.

Fixes: 3166dd07a9cb ("nfp: flower: offload tunnel decap rules via indirect TC blocks")
Reported-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: John Hurley <john.hurley@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agomlxsw: spectrum: Update the supported firmware to version 13.1910.622
Shalom Toledo [Tue, 18 Dec 2018 07:31:31 +0000 (07:31 +0000)]
mlxsw: spectrum: Update the supported firmware to version 13.1910.622

This new firmware contains:
 * New packet traps for discarded packets
 * Secure firmware flash bug fix
 * Fence mechanism bug fix
 * TCAM RMA bug fix

Signed-off-by: Shalom Toledo <shalomt@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoip6mr: Drop mfc6_cache argument to ip6mr_forward2
David Ahern [Mon, 17 Dec 2018 23:36:11 +0000 (15:36 -0800)]
ip6mr: Drop mfc6_cache argument to ip6mr_forward2

mfc6_cache is not needed by ip6mr_forward2 so drop it from the input
argument list.

Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoipmr: Drop mfc_cache argument to ipmr_queue_xmit
David Ahern [Mon, 17 Dec 2018 23:34:48 +0000 (15:34 -0800)]
ipmr: Drop mfc_cache argument to ipmr_queue_xmit

mfc_cache is not needed by ipmr_queue_xmit so drop it from the input
argument list.

Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: dccp: initialize (addr,port) listening hashtable
Peter Oskolkov [Sun, 16 Dec 2018 23:42:48 +0000 (15:42 -0800)]
net: dccp: initialize (addr,port) listening hashtable

Commit d9fbc7f6431f "net: tcp: prefer listeners bound to an address"
removes port-only listener lookups. This caused segfaults in DCCP
lookups because DCCP did not initialize the (addr,port) hashtable.

This patch adds said initialization.

The only non-trivial issue here is the size of the new hashtable.
It seemed reasonable to make it match the size of the port-only
hashtable (= INET_LHTABLE_SIZE) that was used previously. Other
parameters to inet_hashinfo2_init() match those used in TCP.

V2 changes: marked inet_hashinfo2_init as an exported symbol
so that DCCP compiles when configured as a module.

Tested: syzcaller issues fixed; the second patch in the patchset
        tests that DCCP lookups work correctly.

Fixes: d9fbc7f6431f "net: tcp: prefer listeners bound to an address"
Reported-by: syzcaller <syzkaller@googlegroups.com>
Signed-off-by: Peter Oskolkov <posk@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'bnxt_en-next'
David S. Miller [Tue, 18 Dec 2018 07:08:54 +0000 (23:08 -0800)]
Merge branch 'bnxt_en-next'

Michael Chan says:

====================
bnxt_en: Updates for net-next.

Two main changes in this seris plus some miscellaneous changes.

1. Improvements and fixes for resource accounting which are required
for enabling SR-IOV and RDMA on the new 57500 chips.  Only SR-IOV
for 57500 chips is enabled in this series.

2. New statistics counters and improvements to keep the basic
counters and port counters during IFDOWN.

3. Msic. small changes for ETS, returning proper error codes
when flashing NVRAM, and a link speed related fix for ethtool
loopback selftest.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: query force speeds before disabling autoneg mode.
Vasundhara Volam [Sun, 16 Dec 2018 23:46:31 +0000 (18:46 -0500)]
bnxt_en: query force speeds before disabling autoneg mode.

With autoneg enabled, PHY loopback test fails. To disable autoneg,
driver needs to send a valid forced speed to FW. FW is not sending
async event for invalid speeds. To fix this, query forced speeds
and send the correct speed when disabling autoneg mode.

Signed-off-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Do not free port statistics buffer when device is down.
Michael Chan [Sun, 16 Dec 2018 23:46:30 +0000 (18:46 -0500)]
bnxt_en: Do not free port statistics buffer when device is down.

Port statistics which include RDMA counters are useful even when the
netdevice is down.  Do not free the port statistics DMA buffers
when the netdevice is down.  This is keep the snapshot of the port
statistics and counters will just continue counting when the
netdevice goes back up.

Split the bnxt_free_stats() function into 2 functions.  The port
statistics buffers will only be freed when the netdevice is
removed.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Save ring statistics before reset.
Michael Chan [Sun, 16 Dec 2018 23:46:29 +0000 (18:46 -0500)]
bnxt_en: Save ring statistics before reset.

With the current driver, the statistics reported by .ndo_get_stats64()
are reset when the device goes down.  Store a snapshot of the
rtnl_link_stats64 before shutdown.  This snapshot is added to the
current counters in .ndo_get_stats64() so that the counters will not
get reset when the device is down.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Return linux standard errors in bnxt_ethtool.c
Vasundhara Volam [Sun, 16 Dec 2018 23:46:28 +0000 (18:46 -0500)]
bnxt_en: Return linux standard errors in bnxt_ethtool.c

Currently firmware specific errors are returned directly in flash_device
and reset ethtool hooks. Modify it to return linux standard errors
to userspace when flashing operations fail.

Signed-off-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Don't set ETS on unused TCs.
Michael Chan [Sun, 16 Dec 2018 23:46:27 +0000 (18:46 -0500)]
bnxt_en: Don't set ETS on unused TCs.

Currently, the code allows ETS bandwidth weight 0 to be set on unused TCs.
We should not set any DCB parameters on unused TCs at all.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Add ethtool -S priority counters.
Michael Chan [Sun, 16 Dec 2018 23:46:26 +0000 (18:46 -0500)]
bnxt_en: Add ethtool -S priority counters.

Display the CoS counters as additional priority counters by looking up
the priority to CoS queue mapping.  If the TX extended port statistics
block size returned by firmware is big enough to cover the CoS counters,
then we will display the new priority counters.  We call firmware to get
the up-to-date pri2cos mapping to convert the CoS counters to
priority counters.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Add SR-IOV support for 57500 chips.
Michael Chan [Sun, 16 Dec 2018 23:46:25 +0000 (18:46 -0500)]
bnxt_en: Add SR-IOV support for 57500 chips.

There are some minor differences when assigning VF resources on the
new chips.  The MSIX (NQ) resource has to be assigned and ring group
is not needed on the new chips.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Disable MSIX before re-reserving NQs/CMPL rings.
Michael Chan [Sun, 16 Dec 2018 23:46:24 +0000 (18:46 -0500)]
bnxt_en: Disable MSIX before re-reserving NQs/CMPL rings.

When bringing up a device, the code checks to see if the number of
MSIX has changed.  pci_disable_msix() should be called first before
changing the number of reserved NQs/CMPL rings.  This ensures that
the MSIX vectors associated with the NQs/CMPL rings are still
properly mapped when pci_disable_msix() masks the vectors.

This patch will prevent errors when RDMA support is added for the new
57500 chips.  When the RDMA driver shuts down, the number of NQs is
decreased and we must use the new sequence to prevent MSIX errors.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Reserve 1 stat_ctx for RDMA driver.
Vasundhara Volam [Sun, 16 Dec 2018 23:46:23 +0000 (18:46 -0500)]
bnxt_en: Reserve 1 stat_ctx for RDMA driver.

bnxt_en requires same number of stat_ctxs as CP rings but RDMA
requires only 1 stat_ctx.  Also add a new parameter resv_stat_ctxs
to better keep track of stat_ctxs reserved including resources used
by RDMA.  Add a stat_ctxs parameter to all the relevant resource
reservation functions so we can reserve the correct number of
stat_ctxs.

Prior to this patch, we were not reserving the extra stat_ctx for
RDMA and RDMA would not work on the new 57500 chips.

Signed-off-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Do not modify max_stat_ctxs after RDMA driver requests/frees stat_ctxs
Vasundhara Volam [Sun, 16 Dec 2018 23:46:22 +0000 (18:46 -0500)]
bnxt_en: Do not modify max_stat_ctxs after RDMA driver requests/frees stat_ctxs

Calling bnxt_set_max_func_stat_ctxs() to modify max stat_ctxs requested
or freed by the RDMA driver is wrong. After introducing reservation of
resources recently, the driver has to keep track of all stat_ctxs
including the ones used by the RDMA driver.  This will provide a better
foundation for accurate accounting of the stat_ctxs.

Signed-off-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: get rid of num_stat_ctxs variable
Vasundhara Volam [Sun, 16 Dec 2018 23:46:21 +0000 (18:46 -0500)]
bnxt_en: get rid of num_stat_ctxs variable

For bnxt_en driver, stat_ctxs created will always be same as
cp_nr_rings. Remove extra variable that duplicates the value.
Also introduce bnxt_get_avail_stat_ctxs_for_en() helper to get
available stat_ctxs and bnxt_get_ulp_stat_ctxs() helper to return
number of stat_ctxs used by RDMA.

Signed-off-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Add bnxt_get_avail_cp_rings_for_en() helper function.
Michael Chan [Sun, 16 Dec 2018 23:46:20 +0000 (18:46 -0500)]
bnxt_en: Add bnxt_get_avail_cp_rings_for_en() helper function.

The available CP rings are calculated differently on the new 57500
chips, so add this helper to do this calculation correctly.  The
VFs will be assigned these available CP rings.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agobnxt_en: Store the maximum NQs available on the PF.
Michael Chan [Sun, 16 Dec 2018 23:46:19 +0000 (18:46 -0500)]
bnxt_en: Store the maximum NQs available on the PF.

The PF has a pool of NQs and MSIX vectors assigned to it based on
NVRAM configurations.  The number of usable MSIX vectors on the PF
is the minimum of the NQs and MSIX vectors.  Any excess NQs without
associated MSIX may be used for the VFs, so we need to store this
max_nqs value.  max_nqs minus the NQs used by the PF will be the
available NQs for the VFs.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agofou: Prevent unbounded recursion in GUE error handler
Stefano Brivio [Mon, 17 Dec 2018 23:13:17 +0000 (00:13 +0100)]
fou: Prevent unbounded recursion in GUE error handler

Handling exceptions for direct UDP encapsulation in GUE (that is,
UDP-in-UDP) leads to unbounded recursion in the GUE exception handler,
syzbot reported.

While draft-ietf-intarea-gue-06 doesn't explicitly forbid direct
encapsulation of UDP in GUE, it probably doesn't make sense to set up GUE
this way, and it's currently not even possible to configure this.

Skip exception handling if the GUE proto/ctype field is set to the UDP
protocol number. Should we need to handle exceptions for UDP-in-GUE one
day, we might need to either explicitly set a bound for recursion, or
implement a special iterative handling for these cases.

Reported-and-tested-by: syzbot+43f6755d1c2e62743468@syzkaller.appspotmail.com
Fixes: b8a51b38e4d4 ("fou, fou6: ICMP error handlers for FoU and GUE")
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>